globals

Updated October 20, 2026

The globals namespace: 1558 functions.

globals/ColorSequence/deserialize

ColorSequence.deserialize(data: any?) -> ColorSequenceObj

Rebuild a ColorSequence from a {kind = "ColorSequence", keypoints = {...}} payload produced by :serialize(). Used by scene save/load.

Parameters

data any (optional) — The serialized payload.

Returns ColorSequenceObj — A fresh ColorSequenceObj with the deserialized keypoints.

local c = ColorSequence.deserialize(savedData)

globals/ColorSequence/new

ColorSequence.new(...: any?) -> ColorSequenceObj

Construct a ColorSequence from a constant color (3-array {r,g,b} or {r=,g=,b=} record), a two-point lerp from c0 to c1, or a keypoints array. An entry of that array is a named { time =, value =, envelope? = } record, a { time, {r,g,b}, envelope? } pair, or a bare {r,g,b} colour whose time is its place in the list — so a list of colours is a ramp through them. envelope is optional and may be a single number (broadcast across channels) or a 3-array. Up to 64 keypoints; the first must anchor at time = 0, the last at time = 1. NaN / Inf rejected.

Parameters

... any (optional) — (color), (c0, c1), or ({ keypoint, ... }) where a keypoint is {time =, value =, envelope? =}, {time, {r,g,b}, envelope?}, or {r,g,b}.

Returns ColorSequenceObj — A ColorSequenceObj with :evaluate, :sample, :keypoints, :duration, :serialize, :destroy.

local solid = ColorSequence.new({ 1, 0.5, 0.25 })
local fade  = ColorSequence.new({ 1, 1, 1 }, { 0, 0, 0 })
local bow   = ColorSequence.new({ { time = 0, value = {1,0,0} }, { time = 0.5, value = {0,1,0} }, { time = 1, value = {0,0,1} } })
local stops = ColorSequence.new({ { 0, {1,0,0} }, { 1, {0,0,1} } })
local ramp  = ColorSequence.new({ { 1, 0.85, 0.35 }, { 1, 0.35, 0.05 } })

globals/Entity

Entity: any

Entity wrapper namespace — auto-injected by the prelude from @builtin::modules.entity_reflect. Also reachable as the lowercase entity(...) proxy constructor.

globals/Entity/allTypes

Entity.allTypes() -> { string }

List all registered reflectable component types in this engine instance.

Returns { string } — Array of component name strings.

local types = Entity.allTypes()

globals/Entity/distance

Entity.distance(entityIdA: string, entityIdB: string) -> number?

Compute the straight-line distance between two entities' Transform.position fields.

Parameters

entityIdA string — First entity id.
entityIdB string — Second entity id.

Returns number? — The distance, or nil when either entity is missing a position.

local d = Entity.distance("player", "enemy")

globals/Entity/getComponents

Entity.getComponents(entityId: string) -> { string }?

List all reflected component types on an entity.

Parameters

entityId string — The entity id.

Returns { string }? — Array of component name strings, or nil when the entity has no reflected components.

local comps = Entity.getComponents("player")

globals/Entity/getField

Entity.getField(entityId: string, component: string, field: string) -> any

Get a specific component field value. An engine-struct component reads through reflection; a component declared in Luau reads through its component ref, so one call serves both.

Parameters

entityId string — The entity id.
component string — The component name (e.g. "Transform", "Model").
field string — The field name (e.g. "position", "tintBlend").

Returns any — The field value, or nil when the entity/component/field is missing. A component name this engine does not declare raises.

local pos = Entity.getField("player", "Transform", "position")

globals/Entity/getName

Entity.getName(entityId: string) -> string?

Get the name of an entity from its Name component.

Parameters

entityId string — The entity id.

Returns string? — The name string, or nil when the entity has no Name component.

local name = Entity.getName("player")

globals/Entity/getPosition

Entity.getPosition(entityId: string) -> Vec3?

Get the position of an entity — shortcut for getField(id, "Transform", "position").

Parameters

entityId string — The entity id.

Returns Vec3? — The position { x, y, z }, or nil when the entity has no Transform.

local pos = Entity.getPosition("player")

globals/Entity/getRotation

Entity.getRotation(entityId: string) -> Quat?

Get the rotation of an entity — shortcut for getField(id, "Transform", "rotation").

Parameters

entityId string — The entity id.

Returns Quat? — The rotation quaternion { x, y, z, w }, or nil when the entity has no Transform.

local rot = Entity.getRotation("player")

globals/Entity/getScale

Entity.getScale(entityId: string) -> Vec3?

Get the scale of an entity — shortcut for getField(id, "Transform", "scale").

Parameters

entityId string — The entity id.

Returns Vec3? — The scale { x, y, z }, or nil when the entity has no Transform.

local scale = Entity.getScale("player")

globals/Entity/getSchema

Entity.getSchema(componentName: string) -> { ComponentFieldSchema }?

Get the full schema of a component type — field names + types.

Parameters

componentName string — The component name.

Returns { ComponentFieldSchema }? — Array of { name, type } schema entries, or nil when the component is not registered.

local schema = Entity.getSchema("Transform")

globals/Entity/isVisible

Entity.isVisible(entityId: string) -> boolean

Check if an entity is visible — reads the Visible component. Missing component is treated as visible.

Parameters

entityId string — The entity id.

Returns boolean — true when visible (or no Visible component is present), false when explicitly hidden.

local visible = Entity.isVisible("player")

globals/Entity/patch

Entity.patch(entityId: string, component: string, fields: { [string]: any }) -> boolean

Patch multiple fields on a single component at once. Serves an engine-struct component and a component declared in Luau alike.

Parameters

entityId string — The entity id.
component string — The component name.
fields { [string]: any } — { fieldName = value, ... } map of fields to write.

Returns boolean — true when every named field was written. A component name this engine does not declare raises.

Entity.patch("player", "Transform", { position = pos, scale = scl })

globals/Entity/setField

Entity.setField(entityId: string, component: string, field: string, value: any?) -> boolean

Set a specific component field value. An engine-struct component writes through reflection; a component declared in Luau writes through its component ref, so one call serves both.

Parameters

entityId string — The entity id.
component string — The component name.
field string — The field name.
value any (optional) — The new value.

Returns boolean — true when the write succeeded, false otherwise. A component name this engine does not declare raises.

Entity.setField("player", "Transform", "position", { x = 0, y = 1, z = 0 })

globals/Entity/setPosition

Entity.setPosition(entityId: string, pos: Vec3)

Set the position of an entity — shortcut for setField(id, "Transform", "position", pos).

Parameters

entityId string — The entity id.
pos Vec3 — The new position { x, y, z }.

Entity.setPosition("player", { x = 0, y = 1, z = 0 })

globals/Entity/setScale

Entity.setScale(entityId: string, scale: Vec3)

Set the scale of an entity — shortcut for setField(id, "Transform", "scale", scale).

Parameters

entityId string — The entity id.
scale Vec3 — The new scale { x, y, z }.

Entity.setScale("player", { x = 1, y = 1, z = 1 })

globals/Entity/snapshot

Entity.snapshot(entityId: string) -> any

Snapshot all reflected components on an entity into a { ComponentName = { field = value, ... }, ... } map.

Parameters

entityId string — The entity id.

Returns any — The snapshot table, or nil when the entity does not exist.

local snap = Entity.snapshot("player")

globals/Field/alias

Field.alias(target: string | { string }, description: string?) -> FieldDesc<any>

Alias for one or more existing fields. An alias is an ACCEPTED key that is not stored itself — it routes the written value to the real field(s) it points at, so a component answers to a caller's natural key without hand-rolling translation code, and both the runtime and the LSP recognise the key.

The key works everywhere the fields it targets do: as a component.add init key, and as a read and a write on the live component. Reading it returns what the target(s) hold right now.

Two forms:

Field.alias("radius") — rename. The value is written verbatim to the single target field (running that field's normal coercion, so an alias onto an assetRef field resolves the ref), and reads back as that field's value.
Field.alias({ "colorR", "colorG", "colorB" }) — fan-out. The value is destructured across the targets: an array {a, b, c} positionally, or a named {r=, g=, b=} / {x=, y=, z=} table by the target's position (r/x, g/y, b/z, a/w). It reads back as an array in target order, so c.color = c.color round-trips.

An alias never replicates and is never persisted — the fields it targets own their own Sync/NoSync, so no mode argument is taken.

Parameters

target string | { string } — A single target field name, or an array of target field names.
description string (optional) — Documents the alias for the LSP; nil to leave it undocumented.

Returns FieldDesc<any> — FieldDesc descriptor with kind = "alias".

type  = Field.alias("kind")
color = Field.alias({ "colorR", "colorG", "colorB" })

globals/Field/assetRef

Field.assetRef(category: (C & string) | { string }, default: AssetRef<C> | I | nil?, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<AssetRef<C, I>>

Typed asset reference field. C is the asset category — a singleton string type inferred from the category argument ("material", "mesh", "@user/customCategory", etc.). One constructor handles every category, including user-registered ones.

Default accepts a resolved AssetRef<C> handle, an identity string (full @library::path form OR a bare leaf name resolved category-locally via asset.resolve(identity, category)), or nil.

At registration the engine resolves any string default through the same category-aware resolver public_newindex uses for runtime writes, so the first read of public.<field> already returns a resolved envelope — not a raw string.

Parameters

category (C & string) | { string } — Asset category as a string literal ("material", "mesh", etc.), "*" for any category, or a list of categories the field takes any of ({ "importer", "component" }). A string is inferred into C.
default AssetRef<C> | I | nil (optional) — AssetRef envelope, identity string, or nil.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc<AssetRef<C, I>> descriptor.

material = Field.assetRef("material", "@my-library::materials.gold", Sync)
source = Field.assetRef("bundle", nil, Sync)
declarer = Field.assetRef({ "importer", "component" }, nil, NoSync)

globals/Field/bitmask

Field.bitmask(bits: number, default: number?, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<number>

Bit-mask number field. bits declares the width the consumer can address; a default or a write that is not a whole number in 0 .. 2^bits - 1 is rejected, naming the width. Use it wherever a numeric field is read as a set of bits rather than as a quantity — what the field reads back is then a mask, so a read-back is evidence the value took.

Parameters

bits number — How many bits wide the mask is, 1..53.
default number (optional) — Numeric default for public.<field>, or nil to leave it unset.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc descriptor carrying a bitmask constraint.

lightChannels = Field.bitmask(32, 0, Sync)

globals/Field/bool

Field.bool(default: boolean?, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<boolean>

Boolean field. nil leaves the field unset.

Parameters

default boolean (optional) — Boolean default for public.<field>, or nil to leave it unset.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc descriptor.

enabled = Field.bool(true, Sync)

globals/Field/check

Field.check(desc: FieldDesc<any>, value: any?) -> (boolean, { { path: string, message: string } })

Check one value against one Field descriptor: its base kind, its constraint, and every nested subfield or element. Returns whether the value is accepted plus every violation found, each naming the path it sits at and what that field takes. A nil value is accepted — a field the values do not set takes its declared default.

Parameters

desc FieldDesc<any> — The Field descriptor to check against.
value any (optional) — The value to check.

Returns (boolean, { { path: string, message: string } }) — (ok, violations) — violations is empty when ok.

local ok, violations = Field.check(Field.number(0, NoSync), "x")

globals/Field/checkSchema

Field.checkSchema(schema: FieldSchema, values: { [string]: any }) -> (boolean, { { path: string, message: string } })

Check a table of values against a schema of Field descriptors. A key the schema does not declare is a violation naming every declared field.

Parameters

schema FieldSchema — Map of field name to Field descriptor.
values { [string]: any } — The values to check.

Returns (boolean, { { path: string, message: string } }) — (ok, violations) — violations is empty when ok.

local ok, violations = Field.checkSchema(schema, { filter = "nearest" })

globals/Field/color

Field.color(default: color?, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<color>

Color field. Default is a color — either {r = .., g = .., b = .., a = ..?} or {r, g, b, a?}.

Parameters

default color (optional) — color default for public.<field>.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc descriptor.

tint = Field.color({ 1, 1, 1, 1 }, Sync)

globals/Field/componentRef

Field.componentRef(componentType: T & string, default: ComponentRef<T> | nil?, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<ComponentRef<T>>

Typed component reference field. T is the component-type name — a singleton string type inferred from the componentType argument ("Camera", "Transform", "@user/Inventory"). The engine validates the referent exists and is of the declared type at every write.

Parameters

componentType T & string — Component type name as a string literal. Inferred into T.
default ComponentRef<T> | nil (optional) — ComponentRef envelope or nil.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc<ComponentRef> descriptor.

aimCam = Field.componentRef("Camera", nil, NoSync)

globals/Field/dataRef

Field.dataRef(contract: C & string, default: AssetRef<"data"> | string | nil?, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<AssetRef<"data">>

Contract-constrained typed-data reference field. Accepts only .data assets whose dataType contract chain includes contract. Rides the assetRef machinery (category "data") — dependency graph, sync, and rehydration behave exactly like Field.assetRef — with the contract gate enforced through the generic field-constraint hook on every write and on the registration-time default.

Parameters

contract C & string — The required dataType contract identity. Inferred into C.
default AssetRef<"data"> | string | nil (optional) — AssetRef envelope, identity string, or nil.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc<AssetRef<"data">> descriptor.

weapon = Field.dataRef("weapon", nil, Sync)

globals/Field/entityRef

Field.entityRef(default: EntityRef | string | nil?, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<EntityRef>

Entity reference field. Accepts a live entity proxy (EntityRef), a raw entity-id string, or nil (no target). Writes are normalised to the plain id string for storage/replication; reads return a live EntityRef proxy (or nil), so public.<field>:method() and public.<field>.id work directly without re-resolving. As a Field.struct subfield or a Field.list element it stores the same id string and reads back as that id string.

Parameters

default EntityRef | string | nil (optional) — Live entity proxy, entity-id string, or nil.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc descriptor.

target = Field.entityRef(nil, Sync)

globals/Field/enum

Field.enum(values: { string }, default: string?, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<string>

Closed-set string field. values declares every member; a default or a write outside the set is rejected with the whole set named. The editor renders the members as a choice and the LSP completes them.

Parameters

values { string } — The members, as an array of non-empty, distinct strings.
default string (optional) — The default member, or nil to leave the field unset.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc descriptor carrying an enum constraint.

fit = Field.enum({ "exact", "hull" }, "hull", Sync)

globals/Field/instantiableRef

Field.instantiableRef(default: AssetRef<any> | string | nil?, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<AssetRef<any>>

Scene-instantiable asset reference field — accepts ANY asset whose type can be instantiated into a scene, gated by CAPABILITY rather than a hardcoded type list. Rides the assetRef machinery with no category filter (any asset type resolves), and the generic field-constraint hook rejects, on every write and on the registration-time default, any asset whose type defines no instantiate method (ref:canInstantiate() is false). A new scene-instantiable asset type is accepted here the moment it defines the hook — no edit to this field or its consumers. The uniform ref:instantiate(target?, opts?) is how a consumer then instantiates the assigned asset (Asset.component, a viewport drop, a tool argument).

Parameters

default AssetRef<any> | string | nil (optional) — AssetRef envelope, identity string, or nil.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc<AssetRef> descriptor.

source = Field.instantiableRef(nil, Sync)

globals/Field/list

Field.list(element: FieldDesc<any>, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<ListValue>

List field — an array of one repeated element type. The element is the Field constructor descriptor every item conforms to, often a Field.struct for a list of records. The list value is an array of the element's value type. Like Field.struct, the engine descends the element schema to resolve nested asset refs into envelopes, so a stack of structs each holding an asset ref has every ref appear in the asset dependency graph, validates each item, and the LSP type-checks the array. An entity reference element is stored as the entity's id string, whether a proxy or an id was written, so the list saves into a scene and replicates. The default value is an empty list. The element declares its own Sync or NoSync for typing; the list's own mode governs replication of the whole array as a unit.

Parameters

element FieldDesc<any> — The Field constructor descriptor each item conforms to.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc<ListValue> — FieldDesc whose value is an array of the element's values.

globals/Field/number

Field.number(default: number?, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<number>

Number field. nil leaves the field unset, so a component can treat an absent value as "derive this from somewhere else" without a second field recording whether the first one was authored.

Parameters

default number (optional) — Numeric default for public.<field>, or nil to leave it unset.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc descriptor consumed by component registration.

positionX = Field.number(0, Sync)

globals/Field/quat

Field.quat(default: quat?, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<quat>

Quaternion field. Default is a quat — either {x = .., y = .., z = .., w = ..} or {x, y, z, w}.

Parameters

default quat (optional) — quat default for public.<field>.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc descriptor.

rotation = Field.quat({ 0, 0, 0, 1 }, Sync)

globals/Field/range

Field.range(min: number?, max: number?, default: number?, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<number>

Bounded number field. min and max declare the interval the value means something in; a default or a write outside it is rejected with the interval named. Either bound may be nil, leaving that side open. The value the field reads back is one the system consuming it can use, and a number that lands outside is reported where it was written.

Parameters

min number (optional) — Lowest accepted value, or nil to leave the low side open.
max number (optional) — Highest accepted value, or nil to leave the high side open.
default number (optional) — Numeric default for public.<field>, or nil to leave it unset.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc descriptor carrying a range constraint.

volume = Field.range(0, 1, 1, Sync)

globals/Field/resource

Field.resource(category: C & string, default: AssetRef<C> | Handle<C> | string | nil?, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<AssetRef<C> | Handle<C>>

Category-gated RESOURCE field — accepts EITHER a persistent AssetRef<C> OR a live GPU Handle<C>, gated by category. This is the renderer-facing field type (e.g. Model.model, Model.material, material texture slots): content can author a persistent asset OR pass a runtime handle (renderer.<resource>.create(...)); the component bridges either to the GPU resource. The category gate still holds — an AssetRef<audio> or a wrong-category handle (a TextureHandle on a "mesh" slot) is a type error AND a runtime rejection. Use Field.assetRef instead when the field MUST be a persistent asset (handles rejected). With Sync, persistent-asset values replicate to peers; a live GPU handle value is local by construction and stays local — peers keep the last replicated asset value.

Parameters

category C & string — Resource category string literal ("mesh", "texture", "material", ...). Inferred into C.
default AssetRef<C> | Handle<C> | string | nil (optional) — AssetRef<C> / Handle<C> / identity string / nil.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc<AssetRef | Handle> descriptor.

model = Field.resource("mesh", nil, Sync)

globals/Field/settingsOf

Field.settingsOf(declarer: string, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?, default: StructValue?) -> FieldDesc<StructValue>

A struct field whose schema is another declarer's settings schema — an assetType, an importer, a component. The declarer is named by its asset identity and resolved by guid, so a same-named type in another package is never mistaken for it. The subfields are the declarer's own, so a field embedding them never restates them.

Parameters

declarer string — The declarer's asset identity, e.g. "@builtin::assetTypes.texture".
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.
default StructValue (optional) — The group's declared default: a table of the declarer's fields, or nil for none.

Returns FieldDesc<StructValue> — FieldDesc whose value is a table of the declarer's settings.

textureSettings = Field.settingsOf("@builtin::assetTypes.texture", NoSync, "Stamped onto every texture.", { filter = "nearest" })

globals/Field/string

Field.string(default: string?, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<string>

String field. nil leaves the field unset.

Parameters

default string (optional) — String default for public.<field>, or nil to leave it unset.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc descriptor.

label = Field.string("hello", Sync)

globals/Field/struct

Field.struct(schema: FieldSchema, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<StructValue>

Struct field — a table whose keys are themselves typed fields. The schema maps each subfield name to its Field constructor descriptor; the struct value is a table holding one value per subfield. Use this instead of Field.table when the table carries asset references or other typed data: the engine descends the schema to resolve nested asset refs into envelopes at registration and at write time, so they appear in the asset dependency graph, validates writes per subfield, and the LSP type-checks the shape. Each subfield declares its own Sync or NoSync for typing; the struct's own mode governs replication of the whole value as a unit.

Parameters

schema FieldSchema — Map of subfield name to a Field constructor descriptor.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc<StructValue> — FieldDesc whose value is a table of the subfields' values.

globals/Field/table

Field.table(default: T, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<T>

Generic table field. T is the table's shape — usually inferred from the default value, or supplied explicitly via an explicit ascription Field.table({} :: MyShape, mode) when the default doesn't cover every key the runtime will write. The engine accepts any Luau table as a value at write time; per-shape enforcement is opt-in static typing only.

Parameters

default T — Table value to use as the default for public.<field>.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc descriptor.

idMap = Field.table({} :: { [string]: string }, NoSync)

globals/Field/taggedRef

Field.taggedRef(tag: string, default: AssetRef<any> | string | nil?, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<AssetRef<any>>

Tag-constrained asset reference field — accepts any asset carrying tag in its .metadata.tags, whatever its type. This is how a slot states the KIND of asset it takes (a camera behavior, a player visual) without naming the assets themselves: a new asset becomes assignable the moment it is tagged, with no edit here or in the consumer. Rides the assetRef machinery with no category filter — dependency graph, sync and rehydration behave exactly like Field.assetRef — and the generic field-constraint hook rejects an untagged asset on every write and on the registration-time default. asset.list({ fields = { tags = tag } }) enumerates what fits the slot.

Parameters

tag string — The tag an assigned asset must carry.
default AssetRef<any> | string | nil (optional) — AssetRef envelope, identity string, or nil.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc<AssetRef> descriptor.

behavior = Field.taggedRef("cameraBehavior", nil, Sync)

globals/Field/vec2

Field.vec2(default: vec2?, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<vec2>

Parameters

default vec2 (optional)
mode SyncMode
marker (SerializedMode | string | FieldOptions) (optional)

Returns FieldDesc<vec2>

globals/Field/vec3

Field.vec3(default: vec3?, mode: SyncMode, marker: (SerializedMode | string | FieldOptions)?) -> FieldDesc<vec3>

Vec3 field. Default is a vec3 — either {x = .., y = .., z = ..} or the 3-element array form {x, y, z}.

Parameters

default vec3 (optional) — vec3 default for public.<field>.
mode SyncMode — Sync or NoSync — required.
marker (SerializedMode | string | FieldOptions) (optional) — Serialized, a description string, or a { serialized, description } options table; omit for neither.

Returns FieldDesc descriptor.

offset = Field.vec3({ 0, 0, 0 }, Sync)

globals/Material

Material: any

Material helpers namespace — auto-injected by the prelude from @builtin::modules.material_utils.

globals/Material/Apply

Material.Apply(entityId: string, materialRef: MaterialRefOrName) -> boolean

PascalCase back-compat alias for apply.

Parameters

entityId string — Target entity id.
materialRef MaterialRefOrName — Either an AssetRef envelope or a material-name string.

Returns boolean — True on success.

globals/Material/Create

Material.Create(name: string, opts_or_shader: MaterialOpts | string | nil?, props: MaterialOpts?) -> MaterialRef?

PascalCase back-compat alias for create. Accepts the legacy 3-arg form (name, shader_string, opts_table) by folding shader into opts, and the canonical 2-arg form (name, opts).

Parameters

name string — Material name.
opts_or_shader MaterialOpts | string | nil (optional) — Either an options table OR a legacy shader string.
props MaterialOpts (optional) — Optional extra options table — only used when opts_or_shader is a shader string.

Returns MaterialRef? — The AssetRef envelope, or nil on failure.

Material.Create("gold", "pbr", { color = { 1, 0.85, 0.2 } })  -- legacy 3-arg
Material.Create("gold", { shader = "pbr", color = { 1, 0.85, 0.2 } })  -- canonical
Material.Create("neon", { shader = "pbr", emissive = { 0.1, 0.9, 1 }, emissive_intensity = 5 })  -- cyan glow at 5× intensity

globals/Material/Exists

Material.Exists(name: MaterialRefOrName) -> boolean

PascalCase back-compat alias for exists.

Parameters

name MaterialRefOrName — AssetRef envelope or material-name string.

Returns boolean — True when the material is registered.

globals/Material/GetProperty

Material.GetProperty(materialName: MaterialRefOrName, propertyName: string) -> any

PascalCase back-compat alias for getProperty.

Parameters

materialName MaterialRefOrName — AssetRef envelope or material-name string.
propertyName string — Property name.

Returns any — The property value, or nil when not found.

globals/Material/GetPropertyNames

Material.GetPropertyNames(materialName: MaterialRefOrName) -> { string }?

PascalCase back-compat alias for getPropertyNames.

Parameters

materialName MaterialRefOrName — AssetRef envelope or material-name string.

Returns { string }? — Array of property names, or nil when the material is not found.

globals/Material/SetProperty

Material.SetProperty(materialName: MaterialRefOrName, property: string, value: any?) -> any

PascalCase alias. Writes the property on the material ASSET by name/ref (a runtime change every entity using it takes; matRef:saveDefinition() writes it into mat.yaml) — distinct from M.setProperty, which targets the material on one entity's model.

Parameters

materialName MaterialRefOrName — AssetRef envelope or material-name string.
property string — Property name.
value any (optional) — New value.

Returns any — True on success.

Material.SetProperty("gold", "roughness", 0.1)

globals/Material/SetTexture

Material.SetTexture(materialName: MaterialRefOrName, slot: string, textureRef: string) -> boolean

PascalCase back-compat alias for setTexture.

Parameters

materialName MaterialRefOrName — AssetRef envelope or material-name string.
slot string — Texture slot name ("albedo", "normal", etc.).
textureRef string — Texture reference string.

Returns boolean — True on success.

globals/Material/Update

Material.Update(target: any?, props: { [string]: any }) -> number

PascalCase alias for update.

Parameters

target any (optional) — Material name / AssetRef, OR an entity — a proxy, an entity-id, or the display name the entity carries.
props { [string]: any } — Table of { [propertyName] = value } pairs.

Returns number — Number of properties applied.

Material.Update("gold", { base_color = { 1, 0, 0 }, roughness = 0.15 })

globals/Material/apply

Material.apply(entity: any?, materialRef: MaterialRefOrName) -> boolean

Apply a material to an entity's Model / SkinnedModel component by setting its material field. Accepts either an AssetRef envelope (from Material.create) or a bare material-name string. Errors when the entity has no Model or SkinnedModel — a material only renders where there is a mesh.

Parameters

entity any (optional) — The entity to apply to — an entity proxy (recommended: a validated handle to a real entity), an entity-id string, or the display name the entity carries.
materialRef MaterialRefOrName — Either an AssetRef envelope or a material-name string.

Returns boolean — True on success.

Material.apply(entityId, "gold")
local mat = assert(Material.create("gold", { color = { 1, 0.8, 0.3, 1 }, metallic = 1 })); Material.apply(entityId, mat)

globals/Material/create

Material.create(name: string, opts: MaterialOpts?) -> MaterialRef?

Create a named material in the MaterialRegistry. Returns the canonical AssetRef envelope ({ __ref, type="material", name, guid }) — pass directly to Material.apply, the Model / SkinnedModel material field, or any AssetRef<material> consumer.

Parameters

name string — Material name. Must be a non-empty string.
opts MaterialOpts (optional) — Optional material options. shader?, color?, roughness?, metallic?, emissive? (the glow colour), emissive_intensity? (the HDR multiplier over it), textures?.

Returns MaterialRef? — The AssetRef envelope on success, or nil on failure.

local gold = Material.create("gold", { color = { 1, 0.85, 0.2 }, metallic = 1 })

globals/Material/exists

Material.exists(name: MaterialRefOrName) -> boolean

Check whether a material exists in the registry. Accepts an AssetRef envelope or a bare material name.

Parameters

name MaterialRefOrName — AssetRef envelope or material-name string.

Returns boolean — True when the material is registered.

if Material.exists("gold") then ... end

globals/Material/getProperty

Material.getProperty(materialName: MaterialRefOrName, propertyName: string) -> any

Read the current value of a material property.

Parameters

materialName MaterialRefOrName — AssetRef envelope or material-name string.
propertyName string — Property name ("roughness", "metallic", "base_color", etc.).

Returns any — The property value, or nil when not found.

local r = Material.getProperty("gold", "roughness")

globals/Material/getPropertyNames

Material.getPropertyNames(materialName: MaterialRefOrName) -> { string }?

List the property names exposed by a registered material.

Parameters

materialName MaterialRefOrName — AssetRef envelope or material-name string.

Returns { string }? — Array of property names, or nil when the material is not found.

local props = Material.getPropertyNames("gold")

globals/Material/setProperties

Material.setProperties(target: any?, props: { [string]: any }) -> number

Set many material properties in one call. Addresses the target the same way setProperty does: a material name / AssetRef writes the material ASSET (affecting every entity using it), an entity proxy / entity-id / entity name writes the material bound to that entity's Model / SkinnedModel. Each key resolves against the shader's declared vocabulary, so the spellings create accepts reach the same uniforms; a key the shader does not expose is skipped, which lets one patch table serve materials built on different shaders.

Parameters

target any (optional) — Material name / AssetRef, OR an entity — a proxy, an entity-id, or the display name the entity carries.
props { [string]: any } — Table of { [propertyName] = value } pairs.

Returns number — Number of properties applied.

Material.setProperties("gold", { roughness = 0.2, metallic = 0.9 })
Material.setProperties(entityId, { base_color = { 1, 0, 0 } })

globals/Material/setProperty

Material.setProperty(target: any?, property: string, value: any?)

Set a material property. Addresses the target the same way getProperty does: pass a material name / AssetRef to write the material ASSET (affecting every entity using it), or an entity — a proxy, an id, or a display name — to write the material bound to that entity's Model / SkinnedModel. Errors when an entity target has no Model or SkinnedModel.

Parameters

target any (optional) — Material name / AssetRef, OR an entity — a proxy, an entity-id, or the display name the entity carries.
property string — Property name.
value any (optional) — New value (type depends on the property).

Material.setProperty("gold", "roughness", 0.4)     -- by material name
Material.setProperty(entityId, "roughness", 0.4)   -- by entity

globals/Material/setTexture

Material.setTexture(materialName: MaterialRefOrName, slot: string, textureRef: any?) -> boolean

Set a texture slot on a named material.

Parameters

materialName MaterialRefOrName — AssetRef envelope or material-name string.
slot string — Texture slot name ("albedo", "normal", etc.).
textureRef any (optional) — Texture reference. Formats: "color:r,g,b,a", "@builtin::textures.foo", or a render-output guid (camera target, video handle).

Returns boolean — True on success.

Material.setTexture("gold", "albedo", "@builtin::textures.gold")

globals/Material/update

Material.update(target: any?, props: { [string]: any }) -> number

Change an existing material from a property table — the counterpart to create, taking the same table shape. Addresses its target and counts its writes the way setProperties does.

Parameters

target any (optional) — Material name / AssetRef, OR an entity — a proxy, an entity-id, or the display name the entity carries.
props { [string]: any } — Table of { [propertyName] = value } pairs.

Returns number — Number of properties applied.

Material.update("gold", { color = { 1, 0.85, 0.2 }, roughness = 0.15 })

globals/NumberRange/new

NumberRange.new(min: number, max: number?) -> any

Construct a NumberRange. Pass one number for a constant range (min == max); pass two for a uniform random range. Reversed arguments are normalized to ascending order.

Parameters

min number — Lower bound.
max number (optional) — Upper bound; defaults to min.

Returns any

local lifetime = NumberRange.new(1.0)        -- always 1.0
local speed    = NumberRange.new(0.5, 2.0)   -- random

globals/NumberSequence/deserialize

NumberSequence.deserialize(data: any?) -> NumberSequenceObj

Rebuild a NumberSequence from a {kind = "NumberSequence", keypoints = {...}} payload produced by :serialize(). Used by scene save/load.

Parameters

data any (optional) — The serialized payload.

Returns NumberSequenceObj — A fresh NumberSequenceObj with the deserialized keypoints.

local s = NumberSequence.deserialize(savedData)

globals/NumberSequence/new

NumberSequence.new(...: any?) -> NumberSequenceObj

Construct a NumberSequence from one of three signatures: a constant value, a two-point lerp from v0 to v1, or a keypoints array of { time, value, envelope? } records. Envelope defaults to 0 when omitted. Up to 64 keypoints; the first must anchor at time = 0, the last at time = 1. NaN / Inf in time / value / envelope is rejected.

Parameters

... any (optional) — (v), (v0, v1), or ({ {time, value, envelope?}, ... }).

Returns NumberSequenceObj — A NumberSequenceObj with :evaluate, :sample, :keypoints, :duration, :serialize, :destroy.

local fade = NumberSequence.new(1.0)
local fadeOut = NumberSequence.new(1.0, 0.0)
local size = NumberSequence.new({{time=0,value=0.5,envelope=0.1},{time=0.5,value=1.5},{time=1,value=0}})

globals/Physics

Physics: any

Physics namespace — collisions, joints, raycasts, body manipulation. Auto-injected by the prelude from @builtin::modules.api.engine.physics. Same table as the lowercase physics alias.

globals/Physics/COLLIDER_COMPONENTS

Physics.COLLIDER_COMPONENTS()

Every collider component, in the order a lookup walks them. Shape is the component's identity, so code that works on "whatever collider this entity has" walks this list rather than guessing a shape.

for _, name in ipairs(Physics.COLLIDER_COMPONENTS) do ... end

globals/Physics/addCollider

Physics.addCollider(entityId: string | entityRef, component: string, config: table?)

Add a collider component to an entity, naming the shape you want.

Parameters

entityId string | entityRef — Target entity id.
component string — One of Physics.COLLIDER_COMPONENTS.
config table (optional) — The component's own fields, e.g. { radius = 0.5 } for a sphere.

Physics.addCollider(id, "SphereCollider", { radius = 0.5 })

globals/Physics/addConstraint

Physics.addConstraint(entityId: string | entityRef, opts: table?)

Add a transform constraint to an entity.

Parameters

entityId string | entityRef — Target entity id.
opts table (optional) — Optional constraint description (targetEntityId, position, rotation, scale, lookAt, targetPosition, axes, weight).

Physics.addConstraint(id, { targetEntityId = parent, position = true })

globals/Physics/addJoint

Physics.addJoint(entityIdA: string | entityRef, entityIdB: string | entityRef, opts: table?)

Add a Joint component connecting two entities. Accepts either vec3-style anchor inputs (localAnchor = {x,y,z}) or pre-split scalar keys (localAnchorX/Y/Z).

Parameters

entityIdA string | entityRef — Entity that hosts the Joint component.
entityIdB string | entityRef — Connected entity.
opts table (optional) — Optional joint description (kind, anchors, axis, stiffness, damping, restLength, maxDistance, breakForce, breakTorque, region, remoteRegion). Either end may be a soft body (an entity with a SoftBody component): region anchors this end on a region of entity A's soft body and remoteRegion the other end on a region of B's, each as Physics.softRegion takes it ({ box = { min, max } } or { nearest = point }); an end with no region is anchored on the soft body as a whole. The anchor of a region end is measured from the centre of its particles.

Physics.addJoint(a, b, { kind = "fixed" })
Physics.addJoint(a, b, { kind = "hinge", axis = {x=0,y=1,z=0} })
Physics.addJoint(a, b, { kind = "rope", maxDistance = 8 })
Physics.addJoint(a, b, { kind = "fixed", breakForce = 1200, breakTorque = 800 })
Physics.addJoint(bag, hook, { kind = "ball", region = { nearest = {0, 0.5, 0} } })
Physics.addJoint(jellyA, jellyB, { kind = "fixed", region = { box = { min = {0.4, -0.5, -0.5}, max = {0.5, 0.5, 0.5} } }, remoteRegion = { box = { min = {-0.5, -0.5, -0.5}, max = {-0.4, 0.5, 0.5} } } })

globals/Physics/addVelocity

Physics.addVelocity(a: string | entityRef | number | vec3, b: (number | vec3)?, c: number?, d: number?)

Add to the linear velocity of an entity. Same call shapes as setVelocity.

Parameters

a string | entityRef | number | vec3 — dx, a {x, y, z} delta vector, or an entity id (explicit target).
b (number | vec3) (optional) — dy, dx, or the delta vector depending on call form.
c number (optional) — dz or dy depending on call form.
d number (optional) — Optional dz when targeting an explicit entity.

Physics.addVelocity(0, 5, 0)
Physics.addVelocity(entityId, 0, 5, 0)
Physics.addVelocity(entityId, {x=0, y=5, z=0})

globals/Physics/addWheelCollider

Physics.addWheelCollider(entityId: string | entityRef, config: table?)

Add a WheelCollider to an entity. The entity must be a child (or descendant) of a rigid body — the system walks up the hierarchy to find the Physics component. Every wheel under one body is a wheel of that body's ray-cast vehicle.

Parameters

entityId string | entityRef — Target entity id.
config table (optional) — Optional wheel configuration (radius?, suspensionDistance?, springRate?, damperRate?, targetPosition?, maxSuspensionForce?, motorTorque?, brakeTorque?, steerAngle?, forwardFriction?, sidewaysFriction?, frictionSlip?, sideFrictionStiffness?, rollingResistance?, width?, is2D?, visual?).

Physics.addWheelCollider(id, { radius = 0.35, motorTorque = 500 })

globals/Physics/applyForce

Physics.applyForce(entityIdOrForce: string | entityRef | vec3, force: vec3?, region: PhysicsSoftRegion?)

Apply a force to an entity's rigid or soft body for the next physics step — call every frame for continuous thrust. With one argument the script-context entity is targeted; with two args the explicit entity id wins. A soft body takes the force spread over its particles by mass, so it accelerates as a whole the way a rigid body of its mass would; region puts it on part of the body instead ({ box = { min, max } } or { nearest = point }, as Physics.softRegion takes it), which accelerates that part alike and leaves the rest to follow.

Parameters

entityIdOrForce string | entityRef | vec3 — Entity id (when paired with force) OR a force vector for the script-context entity.
force vec3 (optional) — Optional force vector when targeting an explicit entity.
region PhysicsSoftRegion (optional) — Part of a soft body to push, as Physics.softRegion takes it; omit for the whole body.

Physics.applyForce({x=0, y=10, z=0})
Physics.applyForce(entityId, {x=0, y=10, z=0})
Physics.applyForce(cloth, {x=0, y=0, z=4}, { box = { min = {-0.5, -1, 0.4}, max = {0.5, 1, 0.5} } })

globals/Physics/applyForceAtPoint

Physics.applyForceAtPoint(entityId: string | entityRef, force: vec3, point: vec3, opts: PhysicsPushOptions?)

Apply a force at a specific world-space point for the next physics step. On a rigid body it generates the matching torque from the lever arm. On a soft body it pushes the particles around the point, weighted from the whole at the point to nothing at opts.radius (by default twice the spacing of the body's particles), so the body gives way there first. It always reaches at least the four particles nearest the point, the smallest patch a body can dent, so a radius finer than the particles are spaced spreads over that patch.

Parameters

entityId string | entityRef — Target entity id.
force vec3 — Force vector, in newtons.
point vec3 — World-space application point.
opts PhysicsPushOptions (optional) — { radius: number? }: how far into a soft body the force reaches, in metres.

Physics.applyForceAtPoint(id, {x=0,y=10,z=0}, {x=1,y=0,z=0})
Physics.applyForceAtPoint(jelly, {x=0,y=-40,z=0}, {x=0,y=1.5,z=0}, { radius = 0.3 })

globals/Physics/applyImpulse

Physics.applyImpulse(entityIdOrImpulse: string | entityRef | vec3, impulse: vec3?, region: PhysicsSoftRegion?)

Apply an instantaneous impulse (one-shot velocity change). With one argument the script-context entity is targeted; with two args the explicit entity id wins. A soft body takes the impulse spread over its particles by mass, so it moves off as a whole the way a rigid body of its mass would; region puts it on part of the body instead (as Physics.softRegion takes it), which sets that part moving and leaves the rest to follow. applyImpulseAtPoint aims one at a point.

Parameters

entityIdOrImpulse string | entityRef | vec3 — Entity id (with impulse) OR an impulse vector for the script-context entity.
impulse vec3 (optional) — Optional impulse vector when targeting an explicit entity.
region PhysicsSoftRegion (optional) — Part of a soft body to push, as Physics.softRegion takes it; omit for the whole body.

Physics.applyImpulse({x=0, y=5, z=0})
Physics.applyImpulse(entityId, {x=0, y=5, z=0})
Physics.applyImpulse(jelly, {x=0, y=-2, z=0}, { box = { min = {-0.5, 0.4, -0.5}, max = {0.5, 0.5, 0.5} } })

globals/Physics/applyImpulseAtPoint

Physics.applyImpulseAtPoint(entityId: string | entityRef, impulse: vec3, point: vec3, opts: PhysicsPushOptions?)

Apply an instantaneous impulse at a world-space point: a hit, a kick, a crash. On a rigid body it changes the body's velocity and spin as an impulse at that point of it would. On a soft body it pushes the particles around the point, weighted from the whole at the point to nothing at opts.radius (by default twice the spacing of the body's particles), so the body dents there and the rest of it is carried along through its elastic elements. It always reaches at least the four particles nearest the point, the smallest patch a body can dent, so a radius finer than the particles are spaced spreads over that patch; a finer dent takes a finer body (a higher resolution). The impulse is the momentum the body gains, in newton-seconds, wherever it lands.

Parameters

entityId string | entityRef — The body's entity id or proxy.
impulse vec3 — The impulse, in newton-seconds.
point vec3 — Where it lands, in world space: a contact point.
opts PhysicsPushOptions (optional) — { radius: number? }: how far into a soft body the impulse reaches, in metres.

Physics.applyImpulseAtPoint(crate, {x=0, y=0, z=-30}, hit.point)
local contact = Physics.contacts(shell)[1]; if contact then Physics.applyImpulseAtPoint(shell, {x=-60, y=0, z=0}, contact.points[1].point, { radius = 0.4 }) end

globals/Physics/applyRadialImpulse

Physics.applyRadialImpulse(entityId: string | entityRef, center: vec3, magnitude: number, opts: PhysicsPushOptions?)

Push a body straight away from a world point, like a blast: the push is magnitude newton-seconds at the centre, weighted to nothing at opts.radius (omitted, it reaches any distance at full strength). A rigid body is pushed by the distance to its centre of mass; a soft body particle by particle, each away from the centre by its own distance, so a blast beside a balloon pushes it off and one inside a body spreads it out. Each moves off at the speed magnitude gives a body of its mass.

Parameters

entityId string | entityRef — The body's entity id or proxy.
center vec3 — The blast's world position.
magnitude number — The impulse at the centre, in newton-seconds.
opts PhysicsPushOptions (optional) — { radius: number? }: how far the push reaches, in metres.

Physics.applyRadialImpulse(balloon, {x=0, y=1, z=-1}, 8, { radius = 3 })
for _, id in pile do Physics.applyRadialImpulse(id, blast, 20, { radius = 4 }) end

globals/Physics/applyTorque

Physics.applyTorque(entityIdOrTorque: string | entityRef | vec3, torque: vec3?)

Apply a torque to an entity's rigid body for the next physics step — call every frame for continuous spin-up. With one argument the script-context entity is targeted; with two args the explicit entity id wins.

Parameters

entityIdOrTorque string | entityRef | vec3 — Entity id (with torque) OR a torque vector for the script-context entity.
torque vec3 (optional) — Optional torque vector when targeting an explicit entity.

Physics.applyTorque({x=0, y=1, z=0})
Physics.applyTorque(entityId, {x=0, y=1, z=0})

globals/Physics/attachSoftRegion

Physics.attachSoftRegion(softId: string | entityRef, targetId: string | entityRef, region: PhysicsSoftRegion?) -> number

Hold a region of a soft body to a rigid body, two-way: the region follows the rigid body wherever it goes, and the rigid body carries the region's weight and pull. Each particle in the region is tied to the point of the rigid body it stands at when the attachment is made. The attachment is an entry of the soft body's SoftBody component (attachments), so it is saved with the scene, reaches every peer, and is made again on a rebuilt body; it is released when either side is removed. Its load reads back from Physics.softBodyState(softId).attachments. A cloth's edge on a flagpole, a soft panel on a car's chassis, a jelly glued to a moving platform. Where the rigid body's colliders touch the soft body (a shell resting on its chassis), the contact and the hold push on each other: keep them clear, or Physics.ignoreCollision(softId, targetId).

Parameters

softId string | entityRef — The soft body's entity (it carries a SoftBody component).
targetId string | entityRef — The rigid body's entity (it carries Physics), by id, name or proxy.
region PhysicsSoftRegion (optional) — Which particles to hold, as Physics.softRegion takes it: { box = { min, max } }, { nearest = point }, with space = "world" for world coordinates; nil holds the whole body.

Returns number — The attachment's index in the component's attachments.

Physics.attachSoftRegion(flag, pole, { box = { min = {-0.5, -1, -1}, max = {-0.45, 1, 1} } })
Physics.attachSoftRegion(shell, chassis, { box = { min = {-0.5, -0.5, -0.5}, max = {0.5, -0.4, 0.5} } })

globals/Physics/bodyState

Physics.bodyState(entityId: string | entityRef) -> PhysicsBodyState?

Everything the solver holds for one body — its type, mass, centre of mass, inertia, gravity scale, damping, lock flags, CCD, collision groups, sleep state, velocities, the force and torque queued for the next step, its colliders, contacts, joints and transform constraints, and why it is not moving.

Parameters

entityId string | entityRef — Entity id or proxy.

Returns PhysicsBodyState? — A PhysicsBodyState (with exists = false for an entity that carries no rigid body, and waitingFor naming what its collider needs while that collider cannot be built), or nil when nothing in the scene answers to that id.

local b = Physics.bodyState(id); print(b.bodyType, b.mass, b.stillness)
local b = Physics.bodyState(id); if not b.exists then print(b.waitingFor or "no body was built") end

globals/Physics/boxCast

Physics.boxCast(origin: vec3, halfExtents: vec3, direction: vec3, maxDistance: number?, exclude: (string | {string})?) -> table?

Cast a box along a direction and return the first hit. A trigger (isTrigger) holds nothing up, so the sweep passes through it.

Parameters

origin vec3 — Box center at the start of the cast.
halfExtents vec3 — Half the size of the box on each axis.
direction vec3 — Cast direction, of any length: the engine normalises it, so maxDistance and distance are world units.
maxDistance number (optional) — Optional distance limit.
exclude (string | {string}) (optional) — Optional entity id, or array of entity ids, to exclude from hits. Applied while sweeping rather than to the answer, so a cast that starts inside an excluded collider reports what is behind it.

Returns table? — Hit table { entityId, point, normal, distance, startedInside, softBody? }, or nil if nothing hit. startedInside is false for a surface the query crossed on its way there, and true when the query's own start already lay inside that collider — where distance is 0, point is the start itself, and normal is the shortest way out of the collider. normal is a unit vector either way. softBody is set when what was hit is a soft body, whose surface the query meets as it is deformed now: a PhysicsSoftHit naming the surface triangle, the particles carrying it with the hit's weight on each, the particle nearest the hit, and region, the hit as a soft region to hand to Physics.applyImpulse, Physics.pinSoftRegion or Physics.softParticles. Physics.applyImpulseAtPoint takes the hit's point as it is.

local hit = Physics.boxCast(o, {x=0.5,y=0.5,z=0.5}, dir, 10)
local hit = Physics.boxCast(o, {x=0.5,y=0.5,z=0.5}, dir, 10, { selfId, carriedId })

globals/Physics/capsuleCast

Physics.capsuleCast(origin: vec3, radius: number, halfHeight: number, direction: vec3, maxDistance: number?, exclude: (string | {string})?) -> table?

Cast an upright capsule along a direction and return the first hit. This is the sweep that answers whether a body of that shape fits through a passage: a capsule of radius r reports a hit on anything that leaves it less than 2 * r of clearance. A trigger (isTrigger) holds nothing up, so the sweep passes through it.

Parameters

origin vec3 — Capsule centre at the start of the cast.
radius number — Capsule radius.
halfHeight number — Distance from the centre to either cap centre. The capsule stands halfHeight + radius tall in each direction.
direction vec3 — Cast direction, of any length: the engine normalises it, so maxDistance and distance are world units.
maxDistance number (optional) — Optional distance limit.
exclude (string | {string}) (optional) — Optional entity id, or array of entity ids, to exclude from hits. Applied while sweeping rather than to the answer, so a cast that starts inside an excluded collider reports what is behind it.

local hit = Physics.capsuleCast(o, 0.3, 0.6, dir, 0.5)
local hit = Physics.capsuleCast(o, 0.3, 0.6, dir, 0.5, selfId)

globals/Physics/colliderCount

Physics.colliderCount() -> number

How many colliders the physics world holds. Zero means no ray, cast or overlap fired into this world can hit anything, so it is what separates a query that MISSED from a query fired into a world that holds nothing to hit. Read off the collider set itself, so it costs the same whatever the world holds.

Returns number — colliders across the whole physics world.

if Physics.colliderCount() == 0 then print("nothing here is solid") end

globals/Physics/colliderGeometry

Physics.colliderGeometry(options: table?) -> table?

Read the physics world as drawable triangles: every collider triangulated in world space into one indexed mesh, in GPU buffers ready to draw.

Box, sphere, capsule, cylinder, cone, convex, triangle-mesh and heightfield colliders return their real surface, and a compound returns its children folded together; a shape with no triangulation returns its bounding box and reports exact = false.

options.colors is POSITIONAL over colliderManifest() — entry i colours collider i — so you can colour by role, shape, entity or anything else you read there. A position you leave out takes options.defaultColor.

The returned buffers are yours: destroy them when you replace them.

Parameters

options table (optional) — { tessellation = "low"|"medium"|"high", colors = { {r,g,b,a}, ... }, defaultColor = {r,g,b,a} }.

Returns table? — { vertices, indices, vertexCount, indexCount, colliders } where each entry of colliders is { entity, colliderName?, shapeType, role, exact, firstIndex, indexCount }.

local geo = Physics.colliderGeometry({ tessellation = "high" })

globals/Physics/colliderManifest

Physics.colliderManifest() -> table

List every physics collider in the world with what it is and what it takes part in — no geometry, so it is the cheap read to make before deciding what to do with each one.

role is one of static, dynamic, kinematic, sensor. A sensor is a collider the simulation holds as one, reported ahead of the body type behind it, and a collider with no rigid body is static. exact says whether colliderGeometry would return this collider's true surface or its bounding box.

Every collider of one entity shares its entity, so this is what to key per-object decisions on. The order is stable across calls over an unchanged world, which is what makes colliderGeometry's positional colours usable.

Returns table — Array of { entity, colliderName?, shapeType, role, exact }.

for _, c in ipairs(Physics.colliderManifest()) do print(c.entity, c.role) end

globals/Physics/colliderOn

Physics.colliderOn(entityId: string | entityRef) -> string?

Which collider component an entity carries, or nil when it carries none.

Parameters

entityId string | entityRef — Target entity id.

Returns string? The component name, e.g. "SphereCollider".

local which = Physics.colliderOn(id)

globals/Physics/colliderShapes

Physics.colliderShapes(entityId: string | entityRef) -> table

Read an entity's resolved physics collider shape(s) as the physics engine sees them, including auto-sized colliders.

shapeType is one of box, sphere, capsule, convex, mesh, heightfield, compound, other — the shape the simulation is running, so a mesh collider reads mesh.

params carries half-extents for a box, radius for a sphere, radius and half-height for a capsule, and the collider's bounding half-extents for the shapes that have no parametric description. A convex collider reports its outline in linePoints instead.

Parameters

entityId string | entityRef — Target entity id.

Returns table — Array of resolved collider shapes (empty if none): { shapeType, position, rotation, params, linePoints, name? }.

local shapes = Physics.colliderShapes(id)

globals/Physics/contacts

Physics.contacts(entityId: string | entityRef, otherId: (string | entityRef)?) -> { PhysicsContact }

Every contact one body's colliders are in right now, with the other entity, the normal, how deeply the two interpenetrate, the impulse the last step applied, and each contact point. A pair is listed while its surfaces touch: a pair kept apart (Physics.ignoreCollision, collision groups) or no longer touching is not. For a soft body these are the contacts of its surface with every other collider, another soft body's surface included. A contact with a soft body on either side names the particles of it that are touching (particles for this side, otherParticles for the other), so a hit can be answered where it landed. Name otherId to read only the contacts with that one entity: a body resting on the ground holds a contact point per surface element that touches it, and a read of one pair costs that pair alone.

Parameters

entityId string | entityRef — Entity id or proxy.
otherId (string | entityRef) (optional) — Optional entity id or proxy: report only the contacts with it.

Returns { PhysicsContact } — An array of PhysicsContact — empty when the body touches nothing, or when the entity carries neither a rigid body nor a soft body.

for _, c in Physics.contacts(id) do print(c.other, c.deepestPenetration) end
for _, c in Physics.contacts(truck, pole) do if c.totalImpulse > 5 then print("dented at", #c.particles, "particles") end end

globals/Physics/cutSoftBody

Physics.cutSoftBody(softId: string | entityRef, blade: { vec3 } | { point: vec3, normal: vec3 })

Cut a soft body along a blade, every element of the body the blade meets opening along it: { point = p, normal = n } cuts along the whole plane through p facing n, splitting the body in two wherever it lies on both sides; { a, b, c } cuts along that world-space triangle, which cuts only as far as the triangle reaches. Applied straight after the next physics step; a piece the cut separates becomes an entity of its own (see onSoftBodyTear), and the cut reports as a tear of kind "cut". A plane is measured against the body where that step leaves it, so it cuts a moving body wherever it has gone; a cut that opens nothing is reported in the log. Raises when the entity carries no SoftBody, and in edit mode, where nothing simulates.

Parameters

softId string | entityRef — The soft body's entity.
blade { vec3 } | { point: vec3, normal: vec3 } — { point = vec3, normal = vec3 } for a plane, or the triangle's three world-space corners { a, b, c }.

Physics.cutSoftBody(jelly, { point = vector.create(0, 1, 0), normal = vector.create(1, 0, 0) })   -- in two along x = 0
Physics.cutSoftBody(jelly, { vector.create(0, -5, -5), vector.create(0, 5, -5), vector.create(0, 0, 10) })
-- slice off the part of a body 0.2 m above its centre
local c = Physics.softBodyState(jelly).centerOfMass
Physics.cutSoftBody(jelly, { point = vector.create(c.x, c.y + 0.2, c.z), normal = vector.create(0, 1, 0) })

globals/Physics/debugLines

Physics.debugLines(opts: table?) -> (PhysicsDebugLines?, string?)

Draw what the physics world simulates beyond collider surfaces as world-space lines, with what each object's lines picture as data. This is the read the physics debug views draw from: capture { pass = "physics" } and the edit-mode collider overlay (debugViz category colliders) both show these lines. Layers: cage draws a soft body's elements coloured by load: slack blue, yellow half way, red at the load it tears (see PhysicsSoftBodyDiagnostics for how a load is read); particles draws its particles: free ones as small grey crosses, pinned ones as green octahedra, ones held to a rigid body as magenta octahedra; contacts draws the contacts soft surfaces hold against soft surfaces (their own included): the segment joining the two witness points in orange, the normal in cyan; wheels draws each WheelCollider wheel: its suspension ray as far as it reaches, its whole travel and the tyre's radius below the mount (green on ground, red in the air, as the wheel's isGrounded reads), the travel from its mount to the tyre's centre coloured by how much of it is taken up on the same ramp, its tyre circle in white across its steered axle, and its ground point (yellow) and normal (cyan). Answers in edit mode as well as play mode; in edit mode a vehicle stands at rest with its wheels where they are authored.

Parameters

opts table (optional) — { layers: {string}?, entities: {(string | entityRef)}? }: layers names the layers to draw (every layer when omitted; an unknown name is an error naming the layers), entities keeps to those entities' objects (a wheel is kept by its own entity or its chassis').

Returns (PhysicsDebugLines?, string?) — A PhysicsDebugLines, or nil and the reason on a VM that holds no physics world (a worker).

local d = Physics.debugLines(); print(d.lineCount, #d.objects)
for _, o in Physics.debugLines({ layers = { "particles" } }).objects do print(o.entity, o.pinned, o.attached) end
local w = Physics.debugLines({ layers = { "wheels" }, entities = { truck } }); for _, o in w.objects do print(o.index, o.grounded, o.compression) end

globals/Physics/detachSoftRegion

Physics.detachSoftRegion(softId: string | entityRef, targetId: (string | entityRef)?) -> number

Release regions of a soft body held with attachSoftRegion (or authored in its SoftBody component's attachments).

Parameters

softId string | entityRef — The soft body's entity.
targetId (string | entityRef) (optional) — Release only the regions held to this rigid body; omit to release every one.

Returns number — How many attachments were released.

Physics.detachSoftRegion(flag, pole)
Physics.detachSoftRegion(shell)

globals/Physics/drawnPose

Physics.drawnPose(entityId: string | entityRef) -> PhysicsDrawnPose?

Where an entity is drawn this frame, and where its transform puts it. The simulation advances in fixed ticks (physics.tick_rate, 30 a second by default) while frames are drawn at the display's rate, so a body with Physics.interpolation on is drawn blended between its last two ticks, and everything parented under it is drawn moved with it. entity.position reports the simulated pose, which moves only when a tick runs; this read reports the drawn one as well. Read it from a component's update to place anything that has to stay with a moving body on screen: a camera following it, a label over it, a measurement of how smoothly it moves.

Parameters

entityId string | entityRef — Entity id or proxy. Any entity answers; one the simulation does not blend is drawn at its simulated pose.

Returns PhysicsDrawnPose? — A PhysicsDrawnPose, or nil when nothing in the scene answers to that id.

local d = Physics.drawnPose(car); print(d.position.z, d.simulatedPosition.z, d.interpolated)
local d = Physics.drawnPose(tyre); print(d.interpolated, d.interpolatedBy)

globals/Physics/getAngularVelocity

Physics.getAngularVelocity(entityId: (string | entityRef)?) -> vec3?

Read the angular velocity of an entity's rigid body.

Parameters

entityId (string | entityRef) (optional) — Target entity id or proxy; resolves from script context when omitted.

Returns vec3? — Angular velocity in rad/s, or nil if the entity has no rigid body.

local w = Physics.getAngularVelocity(id)

globals/Physics/getGravity

Physics.getGravity() -> vec3

Read the current world gravity vector.

Returns vec3 — Gravity vector in m/s² (negative y is "down" in the default world).

local g = Physics.getGravity()

globals/Physics/getVelocity

Physics.getVelocity(entityId: (string | entityRef)?) -> vec3?

Read the linear velocity of an entity's rigid body, or of its soft body as a whole (its particles' velocities averaged by mass).

Parameters

entityId (string | entityRef) (optional) — Target entity id or proxy; resolves from script context when omitted.

Returns vec3? — Velocity in m/s, or nil if the entity has no rigid body.

local v = Physics.getVelocity(id)

globals/Physics/getWheelState

Physics.getWheelState(entityId: string | entityRef) -> table?

Read a wheel collider's runtime state. Reads the native component the wheel system writes after each physics step.

Parameters

entityId string | entityRef — Target entity id (must carry a WheelCollider component).

Returns table? — { isGrounded, compression, angularVelocity, rotation, skid, sideImpulse, forwardImpulse, suspensionForce, stopImpulse, center, hitPoint, hitNormal }, or nil if the component is absent. compression runs from 0 at the spring's rest up to suspensionDistance, where the bump stop is; rotation is the wheel's angle about its axle in radians, growing as it rolls forward; skid is how fast the tyre slides sideways (m/s) while its grip is used up, 0 while it holds; the impulses are what the tyre applied to the chassis this tick (N·s), and stopImpulse what the bump stop applied once the suspension was used up; center is the wheel centre in world space.

local state = Physics.getWheelState(id)

globals/Physics/hasLineOfSight

Physics.hasLineOfSight(fromId: string, toId: string) -> boolean

Check whether two entities have line-of-sight between their origins.

Parameters

fromId string — Viewer entity id.
toId string — Target entity id.

Returns boolean — true when no collider sits between them (including coincident origins), false otherwise.

if Physics.hasLineOfSight(a, b) then ... end

globals/Physics/ignoreCollision

Physics.ignoreCollision(entityIdA: string | entityRef, entityIdB: string | entityRef, ignore: boolean?)

Keep two entities from colliding, or let them collide again. Covers every collider each one owns (its rigid body with the shapes composed onto it, its standalone collider, its ragdoll's bones, its soft body's surface), so it keeps a soft shell off the chassis it is held to as well as two rigid bodies apart. An ignored pair holds no contact: the solver pushes neither one and Physics.contacts lists neither to the other. It applies from the next physics step, to bodies already touching as well, and lasts until it is turned off, either entity is despawned, or the mode changes: entering or leaving play starts without it, so make the call where the bodies are set up for play (a component's onLoad). It stands beside collision groups: a pair either one excludes does not collide.

Parameters

entityIdA string | entityRef — First entity id.
entityIdB string | entityRef — Second entity id.
ignore boolean (optional) — When true (default) the pair is kept apart; false lets it collide again.

Physics.ignoreCollision(a, b, true)

globals/Physics/isSleeping

Physics.isSleeping(entityId: (string | entityRef)?) -> boolean?

Whether an entity's rigid or soft body is currently asleep (at rest and not simulating). A body sleeps once it stops moving, to save simulation cost.

Parameters

entityId (string | entityRef) (optional) — Target entity id or proxy; resolves from script context when omitted.

Returns boolean? — true if asleep, false if awake, or nil if the entity has neither a rigid nor a soft body.

if Physics.isSleeping(id) then Physics.wakeUp(id) end

globals/Physics/jointBreaks

Physics.jointBreaks() -> table

Every joint that has broken since the last call to this function. A joint breaks when the reaction it carries exceeds the breakForce (newtons of linear reaction) or breakTorque (the angular row of the same reaction) its joint was given; each joint reports once and its constraint is already released when the record arrives. The 256 most recent are kept: a structure that comes apart while nothing reads them drops the oldest beyond that, as the engine's own queue does beyond 1024.

Returns table — Array of { entityId, connectedEntityId, kind, impulse, angularImpulse, force, torque, position }, oldest first.

for _, e in ipairs(Physics.jointBreaks()) do print(e.entityId, e.force) end

globals/Physics/jointReaction

Physics.jointReaction(entityId: string | entityRef) -> table?

The load an entity's joint is carrying right now, as the constraint solver resolved it on the last physics step. This is the same quantity a break threshold is measured against, so it is what to size breakForce and breakTorque from. A joint with a soft body at either end reads the same way.

Parameters

entityId string | entityRef — Entity carrying the Joint component.

Returns table? — { impulse, angularImpulse, force, torque, position }, or nil when the entity owns no joint.

local r = Physics.jointReaction(id); print(r and r.force)

globals/Physics/observe

Physics.observe(entityId: (string | entityRef)?, opts: table?) -> PhysicsObservation?

Read the solver's own state — the world's accounting, and what it holds for each body plus why it is not moving one. Every value comes off the simulation rather than the Physics component, so a write the solver refused or clamped reads back as what it kept. Answers in edit mode as well as play mode.

Parameters

entityId (string | entityRef) (optional) — Report on this one entity. Omit for every body in the world.
opts table (optional) — { bodies: boolean?, contactPoints: boolean? } — bodies = false builds the world accounting alone, and contactPoints = false keeps each contact pair's normal, depth, impulse and point count while leaving out the individual points. Both default to true.

Returns PhysicsObservation? — A PhysicsObservation, or nil when entityId names nothing in the scene. bodies is an array, not a table keyed by entity id — each entry names its own entity in entity.

local o = Physics.observe(); for _, b in o.bodies do print(b.entity, b.stillness) end
local o = Physics.observe(id); print(o.bodies[1].stillness, o.bodies[1].stillnessDetail)

globals/Physics/onJointBreak

Physics.onJointBreak(fn: (table) -> ()) -> () -> ()

Call fn for every joint that breaks from now on, with the same record jointBreaks returns.

Parameters

fn (table) -> () — Receives one break record per broken joint.

Returns () -> () — A function that removes this listener.

local off = Physics.onJointBreak(function(e) print(e.kind, e.force, e.position) end)

globals/Physics/onSoftBodyTear

Physics.onSoftBodyTear(fn: (PhysicsSoftBodyTear) -> ()) -> () -> ()

Call fn for every tear of a soft body from now on: a body loaded past its tearStrain or tearForce (fields of its SoftBody component), cut by Physics.cutSoftBody or torn by Physics.tearSoftBody. The torn entity keeps the largest piece; every piece split off is an entity of its own by the time fn runs, standing at the root of the scene and carrying a SoftBody with the torn body's fields, its share of the model and the material, the velocity, dents and pins that fell on its side. Its body joins it on the next frame, and an impulse applied to it in fn reaches the body then. A piece is play state: it is left out of every save, and leaving play takes it away and gives the torn entity back its whole body.

Parameters

fn (PhysicsSoftBodyTear) -> () — Receives one PhysicsSoftBodyTear per tear: { entity, pieces, position, kind, particles }.

Returns () -> () — A function that removes this listener.

local off = Physics.onSoftBodyTear(function(e) print(e.entity, #e.pieces, e.position) end)
Physics.onSoftBodyTear(function(e) for _, id in e.pieces do Physics.applyImpulse(id, { x = 0, y = 2, z = 0 }) end end)

globals/Physics/overlapSphere

Physics.overlapSphere(center: vec3, radius: number) -> table

Find every entity id whose colliders overlap a sphere. A trigger (isTrigger) holds nothing up, so it is left out.

Parameters

center vec3 — Sphere center in world space.
radius number — Sphere radius.

Returns table — Array of overlapping entity ids.

local ids = Physics.overlapSphere({x=0,y=0,z=0}, 5)

globals/Physics/pinSoftRegion

Physics.pinSoftRegion(softId: string | entityRef, region: PhysicsSoftRegion?, opts: { name: string?, target: vec3? }?) -> string

Pin a region of a soft body: its particles are held kinematic, where they stand now, and the rest of the body hangs from them, rests on them or swings about them through its elastic elements. A cloth pinned by two corners hangs; a jelly pinned along its base stays put while its top wobbles. Give the pin a target (or set one later with Physics.setPinTarget) and the pinned particles move together so their centre stands there, dragging the body after them: a hand grabbing a corner, a press driven down into a block. The pin is an entry of the soft body's SoftBody component (pins), so it is saved with the scene, reaches every peer and is made again on a rebuilt body; it reads back from Physics.softBodyState(softId).pins.

Parameters

softId string | entityRef — The soft body's entity (it carries a SoftBody component).
region PhysicsSoftRegion (optional) — Which particles to pin, as Physics.softRegion takes it: { box = { min, max } } or { nearest = point } in the entity's unit box (-0.5 to 0.5 on each axis), space = "world" for world coordinates; nil pins the whole body.
opts { name: string?, target: vec3? } (optional) — { name: string?, target: vec3? }: the name the pin is addressed by (by default pin1, pin2, ...), and a world position to move the pinned particles' centre to.

Returns string — The pin's name.

local corner = Physics.pinSoftRegion(cloth, { nearest = {-0.5, 0, 0.5} })
Physics.pinSoftRegion(cloth, { box = { min = {-0.5, -1, 0.45}, max = {0.5, 1, 0.5} } }, { name = "top" })
local grab = Physics.pinSoftRegion(jelly, { nearest = hit.point, space = "world" }, { target = hit.point })

globals/Physics/pumpJointBreaks

Physics.pumpJointBreaks()

Deliver every joint break the simulation has recorded to the registered listeners. An enabled Joint component calls this each tick, so listeners fire on their own wherever joints come from that component. A joint made by writing ecs.PhysicsJoint directly has no such tick behind it — call this each frame, or poll jointBreaks, to deliver its breaks.

Physics.pumpJointBreaks()

globals/Physics/pumpSoftBodyTears

Physics.pumpSoftBodyTears()

Give every piece the soft bodies tore off since the last call its entity, and deliver each tear to the onSoftBodyTear listeners. Every SoftBody component calls this each frame, so tears are handled on their own; the work is done once per frame however many call it.

Physics.pumpSoftBodyTears()

globals/Physics/raycast

Physics.raycast(origin: vec3, direction: vec3, maxDistance: number?, exclude: (string | {string})?) -> table?

Cast a ray and return the first hit. Answers from COLLIDERS ALONE: a mesh that renders but carries no collider is not in the physics world, so a ray fired through it reports the same nil a ray through open air does. renderer.raycast answers the same ray against the geometry the renderer DRAWS, which is what reads the surface of a terrain, a procedurally generated mesh, or any plain Model. A trigger (isTrigger) holds nothing up, so the ray passes through it.

Parameters

origin vec3 — Ray origin in world space.
direction vec3 — Ray direction, of any length: the engine normalises it, so maxDistance and distance are world units.
maxDistance number (optional) — Maximum distance along the ray (defaults to 1000).
exclude (string | {string}) (optional) — Optional entity id, or array of entity ids, to exclude from hits.

local hit = Physics.raycast({x=0,y=2,z=0}, {x=0,y=-1,z=0})
local hit = Physics.raycast(origin, dir, 50, { selfId, carriedId })
local hit = Physics.raycast(muzzle, aim, 100); if hit and hit.softBody then Physics.applyImpulseAtPoint(hit.entityId, {x=0, y=0, z=-20}, hit.point, { radius = 0.3 }) end
if Physics.colliderCount() == 0 then hit = renderer.raycast(eye, down, 200) end

globals/Physics/raycastAll

Physics.raycastAll(origin: vec3, direction: vec3, maxDistance: number?, maxHits: number?, exclude: (string | {string})?) -> table

Cast a ray and return every hit up to maxHits. Answers from COLLIDERS ALONE, so a rendered mesh with no collider is absent from the result; renderer.raycastAll answers the same ray against the geometry the renderer draws. A trigger (isTrigger) holds nothing up, so the ray passes through it.

Parameters

origin vec3 — Ray origin in world space.
direction vec3 — Ray direction, of any length: the engine normalises it, so maxDistance and distance are world units.
maxDistance number (optional) — Optional distance limit along the ray.
maxHits number (optional) — Optional cap on the number of hits returned.
exclude (string | {string}) (optional) — Optional entity id, or array of entity ids, to exclude from hits.

Returns table — Array of hit tables { entityId, point, normal, distance, startedInside, softBody? } — empty when nothing was hit. startedInside is false for a surface the query crossed on its way there, and true when the query's own start already lay inside that collider — where distance is 0, point is the start itself, and normal is the shortest way out of the collider. normal is a unit vector either way. softBody is set when what was hit is a soft body, whose surface the query meets as it is deformed now: a PhysicsSoftHit naming the surface triangle, the particles carrying it with the hit's weight on each, the particle nearest the hit, and region, the hit as a soft region to hand to Physics.applyImpulse, Physics.pinSoftRegion or Physics.softParticles. Physics.applyImpulseAtPoint takes the hit's point as it is.

local hits = Physics.raycastAll(origin, dir, 50, 4)

globals/Physics/raycastBetween

Physics.raycastBetween(fromId: string, toId: string, maxDistance: number?) -> table?

Cast a ray from one entity toward another and return the first hit.

Parameters

fromId string — Origin entity id.
toId string — Target entity id.
maxDistance number (optional) — Optional distance cap (default 1000).

Returns table? — Hit table, or nil if the entities are coincident or nothing was hit.

local hit = Physics.raycastBetween(a, b)

globals/Physics/raycastScreen

Physics.raycastScreen(sx: number, sy: number, maxDistance: number?, exclude: (string | {string})?) -> table?

Cast a ray from a screen pixel into the scene and return the first hit. Unprojects the pixel with screenToRay, then casts with raycast.

Parameters

sx number — Screen X in viewport-local pixels (the space of input.mouse_position and screenToRay).
sy number — Screen Y in viewport-local pixels.
maxDistance number (optional) — Maximum distance along the ray (defaults to 1000, matching raycast).
exclude (string | {string}) (optional) — Optional entity id, or array of entity ids, to exclude from hits.

Returns table? — Hit table { entityId, point, normal, distance, startedInside, softBody? }, or nil on a miss or when no camera has rendered yet. startedInside is false for a surface the query crossed on its way there, and true when the query's own start already lay inside that collider — where distance is 0, point is the start itself, and normal is the shortest way out of the collider. normal is a unit vector either way. softBody is set when what was hit is a soft body, whose surface the query meets as it is deformed now: a PhysicsSoftHit naming the surface triangle, the particles carrying it with the hit's weight on each, the particle nearest the hit, and region, the hit as a soft region to hand to Physics.applyImpulse, Physics.pinSoftRegion or Physics.softParticles. Physics.applyImpulseAtPoint takes the hit's point as it is.

local Zin = require("@builtin::modules.zinput")
local x, y = Zin.state.mousePosition(); local hit = Physics.raycastScreen(x, y)

globals/Physics/releasePin

Physics.releasePin(softId: string | entityRef, name: string?) -> number

Release a pin of a soft body, or every pin: its particles take their mass back, keep the velocity they had, and move with the body again.

Parameters

softId string | entityRef — The soft body's entity.
name string (optional) — The pin to release; omit to release every one.

Returns number — How many pins were released.

Physics.releasePin(cloth, corner)
Physics.releasePin(cloth)

globals/Physics/removeCollider

Physics.removeCollider(entityId: string | entityRef) -> string?

Remove whichever collider component an entity carries.

Parameters

entityId string | entityRef — Target entity id.

Returns string? The component that was removed, or nil when there was none.

Physics.removeCollider(id)

globals/Physics/removeConstraint

Physics.removeConstraint(entityId: string | entityRef, index: number?)

Remove transform constraints from an entity (if any are present).

Parameters

entityId string | entityRef — Target entity id.
index number (optional) — Optional constraint index (currently ignored — the whole component is removed).

Physics.removeConstraint(id)

globals/Physics/removeJoint

Physics.removeJoint(entityId: string | entityRef)

Remove the Joint component from an entity (if present).

Parameters

entityId string | entityRef — Target entity id.

Physics.removeJoint(id)

globals/Physics/removeWheelCollider

Physics.removeWheelCollider(entityId: string | entityRef)

Remove the WheelCollider component from an entity (if present).

Parameters

entityId string | entityRef — Target entity id.

Physics.removeWheelCollider(id)

globals/Physics/repairSoftBody

Physics.repairSoftBody(softId: string | entityRef) -> number

Restore the shape a soft body was built in: every permanent dent, crumple and crease it took past its plasticYield is undone, and the body is put back in that shape where it stands, keeping its motion as a whole. The repair is the SoftBody component's repairs count raised by one, so it reaches every peer. Physics.softBodyState(softId).plastic reads 0 once it is applied (at the start of the next frame).

Parameters

softId string | entityRef — The soft body's entity (it carries a SoftBody component).

Returns number — The component's new repairs count.

Physics.repairSoftBody(shell)

globals/Physics/setAngularDamping

Physics.setAngularDamping(entityIdOrDamping: string | entityRef | number, damping: number?)

Set angular damping on an entity's rigid body. One-arg form targets the script-context entity.

Parameters

entityIdOrDamping string | entityRef | number — Entity id (with damping) OR damping value (script-context entity).
damping number (optional) — Optional explicit damping when targeting another entity.

Physics.setAngularDamping(0.1)
Physics.setAngularDamping(entityId, 0.1)

globals/Physics/setAngularVelocity

Physics.setAngularVelocity(a: string | entityRef | number | vec3, b: (number | vec3)?, c: number?, d: number?)

Set the angular velocity of an entity (radians/sec). Same call shapes as setVelocity.

Parameters

a string | entityRef | number | vec3 — x-component, a {x, y, z} vector, or an entity id (explicit target).
b (number | vec3) (optional) — y-component, x-component, or the vector depending on call form.
c number (optional) — z-component or y-component depending on call form.
d number (optional) — Optional z-component when targeting an explicit entity.

Physics.setAngularVelocity(0, 0, 1)
Physics.setAngularVelocity(entityId, 0, 0, 1)
Physics.setAngularVelocity(entityId, {x=0, y=0, z=1})

globals/Physics/setBodyType

Physics.setBodyType(entityId: string | entityRef, bodyType: string)

Change a rigid body's type at runtime. Mass, colliders, and joints are preserved — only the body's response to forces and position writes changes.

Parameters

entityId string | entityRef — Target entity id.
bodyType string — One of "dynamic", "kinematic", "static".

Physics.setBodyType(entityId, "kinematic")

globals/Physics/setCcdEnabled

Physics.setCcdEnabled(entityIdOrEnabled: string | entityRef | boolean, enabled: boolean?)

Enable or disable continuous collision detection on an entity's rigid body. One-arg form targets the script-context entity.

Parameters

entityIdOrEnabled string | entityRef | boolean — Entity id (with enabled) OR boolean (script-context entity).
enabled boolean (optional) — Optional explicit boolean when targeting another entity.

Physics.setCcdEnabled(true)
Physics.setCcdEnabled(entityId, true)

globals/Physics/setCollisionGroups

Physics.setCollisionGroups(entityId: string | entityRef, membership: number, filter: number)

Set the collision-group membership and filter bitmasks on an entity's colliders. Adds a CollisionGroup component if missing; the component holds the groups, so they apply to the colliders of the entity's rigid body or standalone collider by the next physics step, and a body built or rebuilt later takes them as it is built, in edit mode as well as play. A soft body's groups are its SoftBody component's collisionMemberships and collisionFilter.

Parameters

entityId string | entityRef — Target entity id.
membership number — Bitmask: which groups this collider belongs to.
filter number — Bitmask: which groups this collider can collide with.

Physics.setCollisionGroups(id, 0x0001, 0xFFFF)

globals/Physics/setGravity

Physics.setGravity(gravity: vec3)

Replace the world gravity vector.

Parameters

gravity vec3 — New gravity vector in m/s².

Physics.setGravity({x=0, y=-9.81, z=0})

globals/Physics/setGravityScale

Physics.setGravityScale(entityIdOrScale: string | entityRef | number, scale: number?)

Set the per-entity gravity scale (1.0 = normal, 0.0 = no gravity). One-arg form targets the script-context entity.

Parameters

entityIdOrScale string | entityRef | number — Entity id (with scale) OR scale value (script-context entity).
scale number (optional) — Optional explicit scale when targeting another entity.

Physics.setGravityScale(0.5)
Physics.setGravityScale(entityId, 0.5)

globals/Physics/setJointMotor

Physics.setJointMotor(entityId: string | entityRef, targetVelocity: number, maxForce: number)

Set a motor on an entity's joint.

Parameters

entityId string | entityRef — Target entity id (must carry a Joint component).
targetVelocity number — Desired joint velocity.
maxForce number — Maximum force the motor can apply.

Physics.setJointMotor(id, 5.0, 1000)

globals/Physics/setLinearDamping

Physics.setLinearDamping(entityIdOrDamping: string | entityRef | number, damping: number?)

Set linear damping on an entity's rigid body (0 = no damping). One-arg form targets the script-context entity.

Parameters

entityIdOrDamping string | entityRef | number — Entity id (with damping) OR damping value (script-context entity).
damping number (optional) — Optional explicit damping when targeting another entity.

Physics.setLinearDamping(0.05)
Physics.setLinearDamping(entityId, 0.05)

globals/Physics/setMass

Physics.setMass(entityIdOrMass: string | entityRef | number, mass: number?)

Set the mass of an entity's rigid body (kg). One-arg form targets the script-context entity.

Parameters

entityIdOrMass string | entityRef | number — Entity id (with mass) OR mass value (script-context entity).
mass number (optional) — Optional explicit mass when targeting another entity.

Physics.setMass(10)
Physics.setMass(entityId, 10)

globals/Physics/setPinTarget

Physics.setPinTarget(softId: string | entityRef, name: string, position: vec3?)

Drive a pin of a soft body to a world position: its particles move together over the next step so the centre they had when the pin was made stands at position, with the velocity that carries them there, and the rest of the body follows. Set it every frame to drag the region along a path. nil stops driving it and holds it where it stands. The target is the pin's entry in the SoftBody component's pins, so it reaches every peer.

Parameters

softId string | entityRef — The soft body's entity.
name string — The pin's name, as pinSoftRegion returned it.
position vec3 (optional) — The world position for the pinned particles' centre; nil holds them where they stand.

Physics.setPinTarget(cloth, corner, {x=2, y=3, z=0})

globals/Physics/setRotationLocks

Physics.setRotationLocks(entityId: string | entityRef, x: boolean, y: boolean, z: boolean)

Lock or unlock rotation on specific axes.

Parameters

entityId string | entityRef — Target entity id.
x boolean — Lock rotation about the world X axis.
y boolean — Lock rotation about the world Y axis.
z boolean — Lock rotation about the world Z axis.

Physics.setRotationLocks(id, false, true, false)

globals/Physics/setTranslationLocks

Physics.setTranslationLocks(entityId: string | entityRef, x: boolean, y: boolean, z: boolean)

Lock or unlock translation on specific axes.

Parameters

entityId string | entityRef — Target entity id.
x boolean — Lock translation along the world X axis.
y boolean — Lock translation along the world Y axis.
z boolean — Lock translation along the world Z axis.

Physics.setTranslationLocks(id, false, false, true)

globals/Physics/setVelocity

Physics.setVelocity(a: string | entityRef | number | vec3, b: (number | vec3)?, c: number?, d: number?)

Set the linear velocity of an entity. Accepts (x, y, z) or a {x, y, z} vector for the script-context entity, or the same prefixed with an explicit entityId.

Parameters

a string | entityRef | number | vec3 — x-component, a {x, y, z} vector, or an entity id (explicit target).
b (number | vec3) (optional) — y-component, x-component, or the vector depending on call form.
c number (optional) — z-component or y-component depending on call form.
d number (optional) — Optional z-component when targeting an explicit entity.

Physics.setVelocity(0, 10, 0)
Physics.setVelocity(entityId, 0, 10, 0)
Physics.setVelocity(entityId, {x=0, y=10, z=0})

globals/Physics/softBodySettings

Physics.softBodySettings(changes: table?) -> PhysicsSoftBodySettings

The settings every soft body of the world is simulated under, and the way to change them: how many extra substeps a hard hit may run, how stiff soft contacts are, which tangle and overlap recovery runs, and how far the FEM solver (SoftBody.solver = "fem") solves each step. Each one trades simulation quality for cost; Physics.stepCost().softBodies shows what each body costs under them. They are world settings (physics.soft_body_* in .world_settings) saved with the world: a change reads back at once and the solver runs under it from the next physics tick. Written during play, they land on the play shadow like any other /source write: the solver runs under them at once, and they are gone when the play edits are discarded. Write them in edit mode (or keep them with vfs.promotePlayShadow("/zero/source/.world_settings")) for the world to keep them.

Parameters

changes table (optional) — Optional PhysicsSoftBodySettings fields to change; the fields left out keep their values. A field the settings do not have, or a value of the wrong type or out of range, is refused by name.

Returns PhysicsSoftBodySettings — The settings with changes applied: what the next tick runs under.

print(Physics.softBodySettings().maxExtraSubsteps)
Physics.softBodySettings({ maxExtraSubsteps = 2 })   -- cheaper hard hits
Physics.softBodySettings({ tangleDetection = false, overlapConstraints = false })
Physics.softBodySettings({ femMaxIterations = 60 })   -- FEM bodies converge further each substep

globals/Physics/softBodyState

Physics.softBodyState(entityId: string | entityRef, opts: table?) -> PhysicsSoftBodyState?

What the solver holds for one soft body (an entity with a SoftBody component): its particle, cell and edge counts, the volume it encloses against its rest volume, its centre of mass and frame, the box around its particles, how fast they move, whether it sleeps, and what it touches, and in render how the entity's model is drawn carried by the body. plastic is how deep the deformation that stayed is, in metres (a dent past the component's plasticYield; plasticMean the mean over the particles), 0 for a body that only deformed elastically. diverged is true while a step has carried the body into a state no material reaches (its particles spread, or its volume grown, past ten times the shape it is held at, or a particle carried to a value that is not a number), and divergence says what was measured; the engine log names the body the step it diverges. tearStrain and tearForce are the thresholds it tears past (0 tears never), and piece names the torn piece a body split off another continues (nil for a body built from its fields). attachments lists the regions held to rigid bodies (attachSoftRegion) with the force each pulls with, attachedParticles counts the particles tied, and clusters counts the body's frame plus one per joint end anchored on a region of it. pins lists the regions pinned (pinSoftRegion) with how many particles each holds, its target and where its particles' centre stands, and pinnedParticles counts the particles held kinematic. diagnostics is what the physics debug views picture of the body as numbers: inverted cells, self-crossings, element loads (maxLoad, loadedElements), pinned and attached particles (see PhysicsSoftBodyDiagnostics; Physics.debugLines draws them). replication, on a synced entity in a multiplayer session, says which peer simulates the body (authority "local" or "remote", owner) and what it costs on the wire here (lastSync in seconds, bytesPerSecond). status says where the body stands: "built", "pending" (the component stands and the body is not built yet, waitingFor naming what the build waits on), "refused" (refusal says why the component was not built) or "none" (no SoftBody component). A body is built at the start of the frame after its component is added, so a read in the call that added it answers "pending"; wait on status rather than reading exists = false as a refusal. Answers in edit mode as well as play mode.

Parameters

entityId string | entityRef — Entity id or proxy.
opts table (optional) — { positions: boolean?, contactPoints: boolean?, contactsWith: (string | entityRef)? }: positions = true adds every particle's world position; contactPoints = true adds each contact's individual points (each contact names its count, normal, depth, impulse and particles either way); contactsWith reports only the contacts with that one entity, so a read about one pair costs that pair alone.

Returns PhysicsSoftBodyState? — A PhysicsSoftBodyState, or nil when nothing in the scene answers to that id.

local s = Physics.softBodyState(jelly); print(s.particleCount, s.volumeRatio, s.sleeping)
print(Physics.softBodyState(shell).plastic)   -- metres of dent the shell kept
local s = Physics.softBodyState(shell); if s.diverged then print(s.divergence) end
for _, p in Physics.softBodyState(jelly, { positions = true }).positions do print(p.y) end
local r = Physics.softBodyState(jelly).render; print(r.drawn, r.vertexCount, r.uploadedThisFrame)
local s = Physics.softBodyState(jelly); if s.status == "pending" then print("waiting for", s.waitingFor) end

globals/Physics/softBodyTears

Physics.softBodyTears() -> { PhysicsSoftBodyTear }

Every soft body tear since the last call to this function, oldest first, with the entities of the pieces each split off: the same records onSoftBodyTear delivers. The 256 most recent are kept.

Returns { PhysicsSoftBodyTear } — Array of PhysicsSoftBodyTear.

for _, t in Physics.softBodyTears() do print(t.kind, t.entity, #t.pieces) end

globals/Physics/softParticles

Physics.softParticles(softId: string | entityRef, region: PhysicsSoftRegion?) -> PhysicsSoftParticles?

Read the particles of a soft body that a region selects: where each stands, how fast it moves and whether it is pinned, with the selection's centre of mass and mean velocity. The way to see a dent where a hit landed, how far a pinned corner was dragged, or how a region moves. indices name the particles for comparing two reads of the same body. The region is resolved at each call against where the particles stand then (a local-space region in the body's frame as it stands now), so a read after the body moved can select other particles than an earlier read of the same region; a pin's own particles read back through Physics.softBodyState(softId).pins (center) and the pinned flags.

Parameters

softId string | entityRef — The soft body's entity.
region PhysicsSoftRegion (optional) — Which particles to read, as Physics.softRegion takes it; nil reads the whole body.

Returns PhysicsSoftParticles? — A PhysicsSoftParticles report, or nil when nothing in the scene answers to that id.

local top = Physics.softParticles(jelly, { box = { min = {-0.5, 0.4, -0.5}, max = {0.5, 0.5, 0.5} } }); print(top.count, top.center.y)
local near = Physics.softParticles(shell, { box = { min = {1.6, 0, -0.3}, max = {2.2, 1, 0.3} }, space = "world" })

globals/Physics/softRegion

Physics.softRegion(region: PhysicsSoftRegion?) -> { [string]: any }

The engine form of a soft body region: the record the SoftBody component's attachments and a Joint's soft body ends carry. A region names part of a soft body by where its particles stand: { box = { min, max } } is every particle between the two corners, { nearest = point } the one particle nearest the point, and nil (or {}) the whole body. Coordinates are in the soft body entity's local space, the unit box it is built in (-0.5 to 0.5 on each axis, before its scale), so a region stays the same part of the body wherever the body is; space = "world" reads them in world space instead. A box counts a particle whose contact skin reaches into it: a generated body's outer particles stand that skin inside the unit box, so a box flat on one of its faces (min = {-0.5, -1, -1}, max = {-0.5, 1, 1}) holds that face's particles. Errors naming the key that is wrong.

Parameters

region PhysicsSoftRegion (optional) — { box = { min = v, max = v } | { v, v }, nearest = v, space = "local" | "world" }, or nil for the whole body.

Returns { [string]: any } — { kind = "Whole" | "Box" | "Nearest", min, max, point, world }.

Physics.softRegion({ box = { min = {-0.5, 0.45, -0.5}, max = {0.5, 0.5, 0.5} } })
Physics.softRegion({ nearest = {0, 0.5, 0} })

globals/Physics/sphereCast

Physics.sphereCast(origin: vec3, radius: number, direction: vec3, maxDistance: number?, exclude: (string | {string})?) -> table?

Cast a sphere along a direction and return the first hit. A trigger (isTrigger) holds nothing up, so the sweep passes through it.

Parameters

origin vec3 — Sphere center at the start of the cast.
radius number — Sphere radius.
direction vec3 — Cast direction, of any length: the engine normalises it, so maxDistance and distance are world units.
maxDistance number (optional) — Optional distance limit.
exclude (string | {string}) (optional) — Optional entity id, or array of entity ids, to exclude from hits. Applied while sweeping rather than to the answer, so a cast that starts inside an excluded collider reports what is behind it.

local hit = Physics.sphereCast(o, 0.5, dir, 10)
local hit = Physics.sphereCast(o, 0.5, dir, 10, selfId)

globals/Physics/stepCost

Physics.stepCost() -> PhysicsStepCost?

What the last physics step cost, stage by stage — the same figures worldState().step carries, for a caller that wants only these. Each covers that one step rather than a window of them, and consecutive steps over the same resting scene vary by tens of percent, so several samples averaged is the honest read of what a step costs. softBodyMs is the soft body part of the step and softBodies each soft body's part of it. A frame can run several steps; recent lists the last 64 with their ordinals (step), so a loop reading once a frame counts every step.

Returns PhysicsStepCost? — A PhysicsStepCost, or nil on a frame where the pipeline did not step — a paused simulation, or a world still bootstrapping.

local c = Physics.stepCost(); if c then print(c.stepMs, c.narrowPhaseMs) end
local c = Physics.stepCost(); if c then for _, b in c.softBodies do print(b.entity, b.ms, b.particles) end end
local seen, total = 0, 0
-- once a frame: every step since the last read, however many a frame ran
local c = Physics.stepCost(); if c then for _, s in c.recent do if s.step > seen then total += s.softBodyMs; seen = s.step end end end

globals/Physics/stillnessReasons

Physics.stillnessReasons() -> { string }

Every reason whyStill can answer with, in the order the engine considers them. Read from the engine, so the list is the one the answers come from.

Returns { string } — An array of reason names.

for _, reason in Physics.stillnessReasons() do print(reason) end

globals/Physics/tearSoftBody

Physics.tearSoftBody(softId: string | entityRef, region: PhysicsSoftRegion)

Tear a soft body at a region: every edge and cell holding a particle of the region comes apart there, opening a hole or, when the region runs across the body, splitting it. Applied straight after the next physics step; a piece the tear separates becomes an entity of its own (see onSoftBodyTear), and it reports as a tear of kind "tear". Raises when the entity carries no SoftBody, for a region naming no box and no point (the whole body), and in edit mode, where nothing simulates.

Parameters

softId string | entityRef — The soft body's entity.
region PhysicsSoftRegion — Which particles to tear through, as Physics.softRegion takes it: { box = { min, max } } or { nearest = point } in the entity's unit box (-0.5 to 0.5 on each axis), space = "world" for world coordinates.

Physics.tearSoftBody(cloth, { nearest = hit.point, space = "world" })
Physics.tearSoftBody(sheet, { box = { min = {-0.02, -1, -0.5}, max = {0.02, 1, 0.5} } })

globals/Physics/touching

Physics.touching(entityId: string | entityRef, otherId: string | entityRef) -> (boolean, number, { PhysicsContactPoint })

Whether two entities are touching, and how deeply.

Parameters

entityId string | entityRef — Entity id or proxy.
otherId string | entityRef — The other entity id or proxy.

Returns (boolean, number, { PhysicsContactPoint }) — (touching, deepestPenetration, points) — deepestPenetration is in metres and 0 for surfaces that meet without overlapping.

local hit, depth = Physics.touching(a, b); print(hit, depth)

globals/Physics/vehicleState

Physics.vehicleState(chassisId: string | entityRef) -> table?

Read a vehicle: the chassis body's forward speed and the state of every wheel under it (the wheels whose nearest rigid-body ancestor is this chassis).

Parameters

chassisId string | entityRef — The chassis entity (the rigid body the wheels sit under).

Returns table? — { speed, groundedWheels, wheels }, or nil when the chassis has no WheelCollider under it. speed is metres per second along the chassis' forward (-Z), negative in reverse; wheels lists getWheelState tables, each with its entityId, in hierarchy order.

local car = Physics.vehicleState(carId); print(car.speed, car.groundedWheels)

globals/Physics/wakeUp

Physics.wakeUp(entityId: (string | entityRef)?)

Wake an entity's sleeping rigid or soft body so it resumes simulating. The motion setters (applyImpulse, setVelocity, setAngularVelocity) wake the body for you; call this to wake one explicitly.

Parameters

entityId (string | entityRef) (optional) — Target entity id or proxy; resolves from script context when omitted.

Physics.wakeUp(id)

globals/Physics/whyStill

Physics.whyStill(entityId: string | entityRef) -> (string?, string?)

Why the solver is not moving a body. Returns nil when it IS moving it, and otherwise one of noBody, simulationNotStepping, disabled, static, kinematic, infiniteMass, translationLocked, gravityDisabled, asleep, outsideIsland, resting, aboutToMove — the nearest cause, so the answer names the thing to change. A second return carries the detail: which collider it rests on and how deeply, what its effective gravity works out to, and so on.

Parameters

entityId string | entityRef — Entity id or proxy.

Returns (string?, string?) — (reason, detail).

local why, detail = Physics.whyStill(id); if why then print(why, detail) end

globals/Physics/worldState

Physics.worldState() -> PhysicsWorldState

How many bodies, colliders, joints and contacts the simulation holds right now, with world gravity, the timestep, whether the pipeline is stepping at all, and what the last step cost. Counted off the solver, so a body that failed to build is absent here while its Physics component still exists.

Returns PhysicsWorldState — A PhysicsWorldState.

local w = Physics.worldState(); print(w.bodies.awake .. "/" .. w.bodies.total .. " awake")
print(Physics.worldState().contacts.touchingPairs .. " pairs touching")

globals/Random/fromState

Random.fromState(state: string) -> RandomGenerator

Make a generator that continues a sequence from a position rng:state() recorded. It draws exactly what the recorded generator drew next, so a save holds the string and a load picks the stream up.

Parameters

state string — The 32-character hex string rng:state() returned.

Returns RandomGenerator — A generator at that position.

local resumed = Random.fromState(Random.new(7):state())

globals/Random/new

Random.new(seed: number?) -> RandomGenerator

Make a generator with its own stream. The same seed gives the same sequence on every platform and build; any number of generators run side by side without moving one another or math.random. With no seed it takes one from the clock, and rng:state() records where it stands.

Parameters

seed number (optional) — Any finite number; every distinct number is a distinct stream. Omitted, the clock seeds it.

Returns RandomGenerator — A generator answering :nextNumber(), :nextNumber(min, max), :nextInteger(min, max), :clone() and :state().

local roll = Random.new(42):nextInteger(1, 6)
local jitter = Random.new(42):nextNumber(-0.5, 0.5)
local clockSeeded = Random.new()

globals/Transform

Transform: any

Transform helpers namespace — auto-injected by the prelude from @builtin::modules.transform. Positions, distances, directions, quaternion constructors, and the look-at builders lookRotation (pure) and lookAt (aims an entity). Same table as the lowercase transform alias.

globals/Transform/direction

Transform.direction(fromX: number, fromY: number, fromZ: number, toX: number, toY: number, toZ: number) -> (number, number, number)

Normalized direction vector from point A to point B. Returns zeros when the two points coincide (within ~0.001 units).

Parameters

fromX number — From x.
fromY number — From y.
fromZ number — From z.
toX number — To x.
toY number — To y.
toZ number — To z.

Returns (number, number, number) — Three numbers dx, dy, dz — the unit direction.

local dx, dy, dz = Transform.direction(0, 0, 0, 1, 0, 0)

globals/Transform/directionBetween

Transform.directionBetween(entityA: string | EntityRef, entityB: string | EntityRef) -> (number, number, number)

Normalized world-space direction from one entity to another, read from their world positions. Returns zeros if either entity can't be resolved.

Parameters

entityA string | EntityRef — Source entity (id string or proxy).
entityB string | EntityRef — Target entity (id string or proxy).

Returns (number, number, number) — Three numbers dx, dy, dz — the unit direction.

local dx, dy, dz = Transform.directionBetween("cam", "target")

globals/Transform/distance

Transform.distance(x1: number, y1: number, z1: number, x2: number, y2: number, z2: number) -> number

Euclidean distance between two world-space positions.

Parameters

x1 number — First point x.
y1 number — First point y.
z1 number — First point z.
x2 number — Second point x.
y2 number — Second point y.
z2 number — Second point z.

Returns number — The Euclidean distance.

local d = Transform.distance(0, 0, 0, 1, 1, 1)

globals/Transform/distanceBetween

Transform.distanceBetween(entityA: string | EntityRef, entityB: string | EntityRef) -> number?

Distance between two entities in world space. Each entity's world position is what is measured, so a parent's offset counts toward the distance the way the scene shows it.

Parameters

entityA string | EntityRef — First entity (id string or proxy).
entityB string | EntityRef — Second entity (id string or proxy).

Returns number? — The Euclidean distance, or nil when either entity can't be resolved.

local d = Transform.distanceBetween("cam", "box")

globals/Transform/euler

Transform.euler(qx: number, qy: number, qz: number, qw: number) -> (number, number, number)

Convert quaternion to euler angles (yaw, pitch, roll) in radians.

Parameters

qx number — Quaternion x.
qy number — Quaternion y.
qz number — Quaternion z.
qw number — Quaternion w.

Returns (number, number, number) — Three numbers yaw, pitch, roll (Y, X, Z rotations).

local yaw, pitch, roll = Transform.euler(0, 0, 0, 1)

globals/Transform/eulerToQuat

Transform.eulerToQuat(yaw: number, pitch: number?, roll: number?) -> (number, number, number, number)

Identity-aware overload of euler-to-quaternion. Uses the negative-yaw convention shared with quatFromYaw, quatFromYawPitch, lookAtQuat, and T.euler extraction — so T.euler(T.eulerToQuat(y, p, r)) returns (y, p, r). Order is yaw (Y) then pitch (X) then roll (Z).

Parameters

yaw number — Y-axis rotation in radians.
pitch number (optional) — X-axis rotation in radians. Defaults to 0.
roll number (optional) — Z-axis rotation in radians. Defaults to 0.

Returns (number, number, number, number) — Four numbers qx, qy, qz, qw.

local qx, qy, qz, qw = Transform.eulerToQuat(math.pi / 2)

globals/Transform/lerp

Transform.lerp(ax: number, ay: number, az: number, bx: number, by: number, bz: number, t: number) -> (number, number, number)

Linearly interpolate between two positions.

Parameters

ax number — Start x.
ay number — Start y.
az number — Start z.
bx number — End x.
by number — End y.
bz number — End z.
t number — Interpolation factor [0, 1].

Returns (number, number, number) — Three numbers — the interpolated position.

local x, y, z = Transform.lerp(0, 0, 0, 1, 1, 1, 0.5)

globals/Transform/lerp1

Transform.lerp1(a: number, b: number, t: number) -> number

Linearly interpolate two scalars.

Parameters

a number — Start value.
b number — End value.
t number — Interpolation factor [0, 1].

Returns number — The interpolated scalar.

local v = Transform.lerp1(0, 10, 0.5)

globals/Transform/lerpAngle

Transform.lerpAngle(a: number, b: number, t: number) -> number

Lerp between two angles via the shortest arc; returns a value in [-pi, pi].

Parameters

a number — Start angle in radians.
b number — End angle in radians.
t number — Interpolation factor [0, 1].

Returns number — The interpolated angle, normalized to [-pi, pi].

local a = Transform.lerpAngle(0, math.pi, 0.5)

globals/Transform/localToWorld

Transform.localToWorld(px: number, py: number, pz: number, pqx: number, pqy: number, pqz: number, pqw: number, lx: number, ly: number, lz: number) -> (number, number, number)

Transform a local-space position into world space using a parent pose.

Parameters

px number — Parent position x.
py number — Parent position y.
pz number — Parent position z.
pqx number — Parent rotation x.
pqy number — Parent rotation y.
pqz number — Parent rotation z.
pqw number — Parent rotation w.
lx number — Local x.
ly number — Local y.
lz number — Local z.

Returns (number, number, number) — Three numbers wx, wy, wz — the world position.

local wx, wy, wz = Transform.localToWorld(px, py, pz, pqx, pqy, pqz, pqw, lx, ly, lz)

globals/Transform/lookAt

Transform.lookAt(entityOrId: string | EntityRef, txOrTarget: any?, ty: any?, tz: number?, up: any?) -> (boolean, string?)

Make an entity face a world position. The target slot accepts three explicit coordinates, one point as { x, y, z } / { x =, y =, z = } / a vector, or an entity — an id string, an entity NAME, or a proxy — whose WORLD position is resolved. A table carrying an entity id reads as that entity; any other table reads as the point it spells. The subject slot takes the three entity spellings. Everything here is world space: the subject and the target are read as entity(id).position and the aim is written as entity(id).rotation, so a parent under either one moves the entity and the aim still lands on the point named. Returns whether the rotation was written, so a caller that named an entity the scene does not carry learns the aim did not happen instead of reading a stale orientation back as the answer.

Parameters

entityOrId string | EntityRef — Entity id, name, or proxy for the entity to rotate.
txOrTarget any (optional) — A number (world x), a point table, or an entity id / name / proxy whose world position is resolved as the look-at target.
ty any (optional) — World y of the target. Omitted when txOrTarget is a point or an entity.
tz number (optional) — World z of the target. Omitted when txOrTarget is a point or an entity.
up any (optional) — Optional world up hint deciding the roll — { x, y, z }, { x =, y =, z = } or a vector. World +Y when omitted. It never bends the aim; it only says which way is up around it. When the target slot is an entity or a point this is the third argument, and when it is coordinates the fifth.

Returns (boolean, string?) — True when the entity's world rotation was written, and nil for the second value. The target and up slots take any value, because naming which of the shapes arrived is this call's own job: a value that is none of them comes back as a reason rather than as an error raised out of the argument check. False plus a reason otherwise: "unresolved" when a reference names no entity, "no-transform" when one carries no transform, "incomplete-target" when the target spells no point — coordinates with a y or z missing, or a table carrying neither three numbers nor x/y/z, "incomplete-up" when the up hint spells none either, "degenerate" when the two points coincide so no facing direction exists.

Transform.lookAt("cam", 0, 1, 0)
Transform.lookAt("cam", "box")  -- resolve target entity position
Transform.lookAt(cam, box)      -- entity proxies for both
Transform.lookAt("cam", { 0, 1, 0 })         -- one point table
Transform.lookAt("cam", "box", { 0, 0, 1 })  -- rolled to a +Z up

globals/Transform/lookAtQuat

Transform.lookAtQuat(fx: number, fy: number, fz: number, tx: number, ty: number, tz: number) -> (number?, number?, number?, number?)

Compute quaternion to look from origin position toward a target. Returns four components (qx, qy, qz, qw), or nil when the from and to points are too close to derive a meaningful direction.

Parameters

fx number — Origin x.
fy number — Origin y.
fz number — Origin z.
tx number — Target x.
ty number — Target y.
tz number — Target z.

Returns (number?, number?, number?, number?) — Four numbers qx, qy, qz, qw — the look-at quaternion. Nil when degenerate.

local qx, qy, qz, qw = Transform.lookAtQuat(0, 0, 0, 1, 0, 1)

globals/Transform/lookRotation

Transform.lookRotation(fx: number, fy: number, fz: number, tx: number, ty: number, tz: number, ux: number?, uy: number?, uz: number?) -> (number?, number?, number?, number?)

The rotation that aims an entity standing at one world point at another, with a world up hint deciding the roll. Where lookAtQuat derives the aim from yaw and pitch alone — clamping the pitch just short of vertical, so a point directly overhead comes back a twentieth of a degree off — this builds all three axes, so the aim lands on the point at any elevation and straight up and straight down are ordinary cases. The aimed axis is the entity's local -Z, the same forward quatFromBasis, Transform.lookAt and entity(id):lookAt state and the direction entity(id).transform.forward reads back. The up hint is a world direction the entity's own +Y is turned toward as far as the aim allows; it never bends the forward axis. A hint parallel to the aim leaves the roll undetermined, and a hint of no length names no direction — both fall back to a stable roll rather than a NaN.

Parameters

fx number — Eye x — where the entity stands.
fy number — Eye y.
fz number — Eye z.
tx number — Target x — the world point it faces.
ty number — Target y.
tz number — Target z.
ux number (optional) — Up hint x. World +Y when the hint is omitted.
uy number (optional) — Up hint y.
uz number (optional) — Up hint z.

Returns (number?, number?, number?, number?) — Four numbers qx, qy, qz, qw. Nil when the eye and the target coincide, so no facing direction exists.

local qx, qy, qz, qw = Transform.lookRotation(0, 2, 10, 0, 1, 0)
entity("cam").rotation = { Transform.lookRotation(0, 2, 10, 0, 1, 0) }
-- a dutch tilt: the same aim, rolled by leaning the up hint
local q = { Transform.lookRotation(0, 2, 10, 0, 1, 0, 0.2, 1, 0) }

globals/Transform/normalizeAngle

Transform.normalizeAngle(a: number) -> number

Normalize an angle into [-pi, pi].

Parameters

a number — The angle in radians.

Returns number — The same angle wrapped into [-pi, pi].

local a = Transform.normalizeAngle(3 * math.pi)

globals/Transform/orbit

Transform.orbit(centerX: number, centerY: number, centerZ: number, radius: number, height: number, angle: number) -> (number, number, number, number, number, number, number)

Position + rotation for orbiting around a center point. Returns the world position followed by the orientation that faces the center.

Parameters

centerX number — Center x.
centerY number — Center y.
centerZ number — Center z.
radius number — Horizontal distance from the center.
height number — Vertical offset from centerY.
angle number — Orbital angle in radians.

Returns (number, number, number, number, number, number, number) — Seven numbers x, y, z, qx, qy, qz, qw.

local x, y, z, qx, qy, qz, qw = Transform.orbit(0, 1, 0, 5, 2, t)

globals/Transform/quatFromAxisAngle

Transform.quatFromAxisAngle(ax: number, ay: number, az: number, angle: number) -> (number, number, number, number)

Create quaternion from axis and angle (radians). Returns the identity quaternion when the axis is degenerate (length < 0.001).

Parameters

ax number — Axis x.
ay number — Axis y.
az number — Axis z.
angle number — Rotation angle in radians.

Returns (number, number, number, number) — Four numbers qx, qy, qz, qw.

local qx, qy, qz, qw = Transform.quatFromAxisAngle(0, 1, 0, math.pi)

globals/Transform/quatFromBasis

Transform.quatFromBasis(rx: number, ry: number, rz: number, ux: number, uy: number, uz: number, fx: number, fy: number, fz: number) -> (number, number, number, number)

Build the rotation whose right, up and forward ARE the given axes. Where lookAtQuat derives a rotation from a direction alone — yaw and pitch, with pitch clamped just short of straight up or down and no say in the roll — this states all three axes, so a view straight down has a defined image-up instead of whatever the yaw implied. The axes are expected orthonormal and are used as given: right and up are the entity's local +X and +Y, forward its local -Z (the direction it faces).

Parameters

rx number — Right axis x.
ry number — Right axis y.
rz number — Right axis z.
ux number — Up axis x.
uy number — Up axis y.
uz number — Up axis z.
fx number — Forward axis x.
fy number — Forward axis y.
fz number — Forward axis z.

Returns (number, number, number, number) — x, y, z, w of the rotation quaternion.

local qx, qy, qz, qw = Transform.quatFromBasis(1,0,0, 0,1,0, 0,0,-1) -- identity
-- looking straight down with the subject's front toward the top of frame
local qx, qy, qz, qw = Transform.quatFromBasis(1,0,0, 0,0,-1, 0,-1,0)

globals/Transform/quatFromYaw

Transform.quatFromYaw(yaw: number) -> (number, number, number, number)

Create quaternion from yaw (Y-axis rotation) in radians. Uses the negative-yaw convention shared with quatFromYawPitch, lookAtQuat, and T.euler extraction — so T.euler(T.quatFromYaw(y)) round-trips to y.

Parameters

yaw number — Rotation in radians around the Y axis.

Returns (number, number, number, number) — Four numbers qx, qy, qz, qw.

local qx, qy, qz, qw = Transform.quatFromYaw(math.pi / 2)

globals/Transform/quatFromYawPitch

Transform.quatFromYawPitch(yaw: number, pitch: number) -> (number, number, number, number)

Create quaternion from yaw and pitch in radians.

Parameters

yaw number — Y-axis rotation in radians.
pitch number — X-axis rotation in radians.

Returns (number, number, number, number) — Four numbers qx, qy, qz, qw.

local qx, qy, qz, qw = Transform.quatFromYawPitch(0, math.pi / 4)

globals/Transform/quatIdentity

Transform.quatIdentity() -> (number, number, number, number)

Identity quaternion (0, 0, 0, 1).

Returns (number, number, number, number) — Four numbers qx, qy, qz, qw — the identity.

local qx, qy, qz, qw = Transform.quatIdentity()

globals/Transform/quatInverse

Transform.quatInverse(qx: number, qy: number, qz: number, qw: number) -> (number, number, number, number)

Quaternion inverse. Equal to the conjugate for unit quaternions.

Parameters

qx number — Quaternion x.
qy number — Quaternion y.
qz number — Quaternion z.
qw number — Quaternion w.

Returns (number, number, number, number) — Four numbers qx, qy, qz, qw — the inverse.

local ix, iy, iz, iw = Transform.quatInverse(qx, qy, qz, qw)

globals/Transform/quatMul

Transform.quatMul(ax: number, ay: number, az: number, aw: number, bx: number, by: number, bz: number, bw: number) -> (number, number, number, number)

Quaternion multiplication: returns qa * qb (composition: rotate by qb then qa).

Parameters

ax number — Left quat x.
ay number — Left quat y.
az number — Left quat z.
aw number — Left quat w.
bx number — Right quat x.
by number — Right quat y.
bz number — Right quat z.
bw number — Right quat w.

Returns (number, number, number, number) — Four numbers qx, qy, qz, qw — the composed quaternion.

local qx, qy, qz, qw = Transform.quatMul(ax, ay, az, aw, bx, by, bz, bw)

globals/Transform/quatRotateVec

Transform.quatRotateVec(qx: number, qy: number, qz: number, qw: number, vx: number, vy: number, vz: number) -> (number, number, number)

Rotate a 3-vector by a quaternion.

Parameters

qx number — Quaternion x.
qy number — Quaternion y.
qz number — Quaternion z.
qw number — Quaternion w.
vx number — Vector x.
vy number — Vector y.
vz number — Vector z.

Returns (number, number, number) — Three numbers — the rotated vector.

local rx, ry, rz = Transform.quatRotateVec(qx, qy, qz, qw, 1, 0, 0)

globals/Transform/quatToEuler

Transform.quatToEuler(qx: number, qy: number, qz: number, qw: number) -> (number, number, number)

Convert quaternion to (yaw, pitch, roll). Alias of euler with the explicit name so callers don't have to remember the order.

Parameters

qx number — Quaternion x.
qy number — Quaternion y.
qz number — Quaternion z.
qw number — Quaternion w.

Returns (number, number, number) — Three numbers yaw, pitch, roll (Y, X, Z rotations).

local yaw, pitch, roll = Transform.quatToEuler(qx, qy, qz, qw)

globals/Transform/readVec3

Transform.readVec3(value: Vec3Input, label: string?) -> { number }

Normalize a vector a caller wrote to a plain { x, y, z } array. Accepts a positional array {1, 2, 3}, a keyed table {x =, y =, z =}, or a live vec handle. Missing components read as 0. Raises when the value is not a vector; label names the caller in that error.

Parameters

value Vec3Input — The vector to normalize.
label string (optional) — Name reported in the error when the value is not a vector. Defaults to "Transform".

Returns { number } — A three-element array { x, y, z }.

local v = Transform.readVec3({ x = 1, y = 2, z = 3 })

globals/Transform/slerp

Transform.slerp(ax: number, ay: number, az: number, aw: number, bx: number, by: number, bz: number, bw: number, t: number) -> (number, number, number, number)

Spherical linear interpolation between two quaternions. Picks the shortest path (flips sign if dot < 0). Falls back to lerp+normalize when the two quats are very close (avoids div-by-zero on near-parallel inputs).

Parameters

ax number — Start quaternion x.
ay number — Start quaternion y.
az number — Start quaternion z.
aw number — Start quaternion w.
bx number — End quaternion x.
by number — End quaternion y.
bz number — End quaternion z.
bw number — End quaternion w.
t number — Interpolation factor [0, 1].

Returns (number, number, number, number) — Four numbers qx, qy, qz, qw — the interpolated unit quaternion.

local qx, qy, qz, qw = Transform.slerp(0, 0, 0, 1, 1, 0, 0, 0, 0.5)

globals/Transform/snapVec3

Transform.snapVec3(v: { number }, step: number | Vec3Input) -> { number }

Quantize each component of a vector to the nearest multiple of step — a number for uniform steps, or a vector for per-axis steps. A step of 0 on an axis leaves that axis at its exact value.

Parameters

v { number } — The vector to quantize, as { x, y, z }.
step number | Vec3Input — Uniform step size, or a per-axis vector of step sizes.

Returns { number } — A three-element array { x, y, z } snapped to the step grid.

local v = Transform.snapVec3({ 1.4, 2.6, -0.4 }, 1)

globals/Transform/toQuaternion

Transform.toQuaternion(rotation: any?, label: string?) -> { number }

Normalize a rotation a caller wrote to a { qx, qy, qz, qw } quaternion. Accepts a quaternion ({x,y,z,w} or {x=,y=,z=,w=}) or euler DEGREES ({pitch,yaw,roll} or {pitch=,yaw=,roll=}), so one call site takes whichever form the caller finds natural. This is the reading every rotation-taking surface in the engine shares, so a quaternion and euler degrees mean the same thing at all of them. Raises when the value matches no form; label names the caller in that error, and a value that is one of the shapes a quaternion helper returns is named as such along with the packing it goes in as.

Parameters

rotation any (optional) — The rotation to normalize, in any form of the RotationInput union.
label string (optional) — Name reported in the error when the value is not a rotation. Defaults to "Transform".

Returns { number } — A four-element array { qx, qy, qz, qw }.

local q = Transform.toQuaternion({ pitch = 0, yaw = 90, roll = 0 })

globals/Transform/tryQuaternion

Transform.tryQuaternion(rotation: any?, label: string?) -> ({ number }?, string?)

Read a rotation a caller wrote WITHOUT raising: returns the canonical { qx, qy, qz, qw }, or nil and the message describing what arrived. The forms are the RotationInput union — a quaternion ({x,y,z,w} or {x=,y=,z=,w=}) or euler DEGREES ({pitch,yaw,roll} or {pitch=,yaw=,roll=}). Takes any value because reporting on a value that is none of those forms is the whole job; a setter built on this raises the returned message itself, so the error points at the line that wrote the value rather than at the reading.

Parameters

rotation any (optional) — The value to read as a rotation.
label string (optional) — Name reported in the message. Defaults to "Transform".

Returns ({ number }?, string?) — The quaternion { qx, qy, qz, qw }, or nil and the message.

local q, why = Transform.tryQuaternion(value, "myTool")

globals/Transform/vec/add

Transform.vec.add(ax: number, ay: number, az: number, bx: number, by: number, bz: number) -> (number, number, number)

Component-wise vec3 addition.

Parameters

ax number — First vector x.
ay number — First vector y.
az number — First vector z.
bx number — Second vector x.
by number — Second vector y.
bz number — Second vector z.

Returns (number, number, number) — Three numbers — the sum.

local x, y, z = Transform.vec.add(1, 2, 3, 4, 5, 6)

globals/Transform/vec/cross

Transform.vec.cross(ax: number, ay: number, az: number, bx: number, by: number, bz: number) -> (number, number, number)

Cross product a x b.

Parameters

ax number — First vector x.
ay number — First vector y.
az number — First vector z.
bx number — Second vector x.
by number — Second vector y.
bz number — Second vector z.

Returns (number, number, number) — Three numbers cx, cy, cz — the cross product.

local cx, cy, cz = Transform.vec.cross(1, 0, 0, 0, 1, 0)

globals/Transform/vec/dot

Transform.vec.dot(ax: number, ay: number, az: number, bx: number, by: number, bz: number) -> number

Dot product of two vec3s.

Parameters

ax number — First vector x.
ay number — First vector y.
az number — First vector z.
bx number — Second vector x.
by number — Second vector y.
bz number — Second vector z.

Returns number — The scalar dot product.

local d = Transform.vec.dot(1, 0, 0, 0, 1, 0)

globals/Transform/vec/length

Transform.vec.length(x: number, y: number, z: number) -> number

Euclidean length of a vec3.

Parameters

x number — Vector x.
y number — Vector y.
z number — Vector z.

Returns number — The length.

local len = Transform.vec.length(1, 2, 3)

globals/Transform/vec/normalize

Transform.vec.normalize(x: number, y: number, z: number) -> (number, number, number)

Normalize a vec3. Returns zeros when the input is degenerate (length < 1e-8).

Parameters

x number — Vector x.
y number — Vector y.
z number — Vector z.

Returns (number, number, number) — Three numbers — the unit-length vec3.

local nx, ny, nz = Transform.vec.normalize(0, 5, 0)

globals/Transform/vec/scale

Transform.vec.scale(x: number, y: number, z: number, s: number) -> (number, number, number)

Component-wise scalar multiplication of a vec3.

Parameters

x number — Vector x.
y number — Vector y.
z number — Vector z.
s number — Scalar factor.

Returns (number, number, number) — Three numbers — the scaled vec3.

local x, y, z = Transform.vec.scale(1, 2, 3, 2)

globals/Transform/vec/sub

Transform.vec.sub(ax: number, ay: number, az: number, bx: number, by: number, bz: number) -> (number, number, number)

Component-wise vec3 subtraction (a - b).

Parameters

ax number — First vector x.
ay number — First vector y.
az number — First vector z.
bx number — Second vector x.
by number — Second vector y.
bz number — Second vector z.

Returns (number, number, number) — Three numbers — the difference.

local x, y, z = Transform.vec.sub(4, 5, 6, 1, 2, 3)

globals/Transform/worldToLocal

Transform.worldToLocal(px: number, py: number, pz: number, pqx: number, pqy: number, pqz: number, pqw: number, wx: number, wy: number, wz: number) -> (number, number, number)

Transform a world-space position into a parent's local space.

Parameters

px number — Parent position x.
py number — Parent position y.
pz number — Parent position z.
pqx number — Parent rotation x.
pqy number — Parent rotation y.
pqz number — Parent rotation z.
pqw number — Parent rotation w.
wx number — World x.
wy number — World y.
wz number — World z.

Returns (number, number, number) — Three numbers lx, ly, lz — the local position.

local lx, ly, lz = Transform.worldToLocal(px, py, pz, pqx, pqy, pqz, pqw, wx, wy, wz)

globals/animation/animating

animation.animating() -> { AnimationBody }

The bodies the engine measured a changing pose on — what is animating right now.

Returns { AnimationBody } — An array of AnimationBody.

for _, b in animation.animating() do print(b.entity, b.clips[1] and b.clips[1].name) end

globals/animation/bodies

animation.bodies() -> { AnimationBody }

Every body the engine holds animation state for.

Returns { AnimationBody } — An array of AnimationBody.

for _, b in animation.bodies() do print(b.entity, b.matched .. "/" .. b.total) end

globals/animation/body

animation.body(entityId: string | EntityRef) -> AnimationBody?

The report for one body, or nil when the engine holds no animation state for it. Accepts the body itself or any ancestor of it, so a character root answers for the skinned body underneath it.

Parameters

entityId string | EntityRef — The entity's stable id, or an EntityRef.

Returns AnimationBody? — An AnimationBody, or nil.

local b = animation.body(hero.id); print(b and b.reason)

globals/animation/clips

animation.clips(entityId: string | EntityRef) -> { AnimationClip }

The clips contributing to a body's pose right now, with their playheads and their retarget coverage.

Parameters

entityId string | EntityRef — The entity's stable id, or an EntityRef.

Returns { AnimationClip } — An array of AnimationClip.

for _, c in animation.clips(hero.id) do print(c.name, c.time, c.matched) end

globals/animation/coverage

animation.coverage(entityId: string | EntityRef) -> (number, number)

How many of a body's bones the clips driving it actually reach. Returns (matched, total). A clip that retargets onto nothing reads (0, 50) while its playhead advances; a partial retarget reads its own count, so 3 of 50 is as visible as none.

Parameters

entityId string | EntityRef — The entity's stable id, or an EntityRef.

Returns (number, number) — (matched, total).

local m, t = animation.coverage(hero.id); print(m .. "/" .. t)

globals/animation/declare

animation.declare(entityId: string, facts: { [string]: any })

Publish what an animator is running on a body, so the observation names its clips, playheads and retarget coverage beside the pose the engine measures. The shipped animators declare through AnimGraph:publish; a custom animator calls this itself, once per frame it runs.

Parameters

entityId string — The body the animator drives.
facts { [string]: any } — { driver, bound, playing, outputKind, failure, clips }, where each clip is { name, nodeKind, time, duration, playing, finished, looping, weight, matched, total, unmatched }.

animation.declare(body.id, { driver = "MyAnimator", playing = true, clips = {} })

globals/animation/forget

animation.forget(entityId: string)

Drop the declaration and the pose evidence the engine holds for one body. An animator calls this when it releases a body, so the observation reports the body as undriven from the next frame.

Parameters

entityId string — The body to drop.

animation.forget(body.id)

globals/animation/observe

animation.observe() -> AnimationObservation

Report what the engine is posing right now and why a body is not moving. One read covering every body the engine holds animation state for, each with the pose evidence the engine measured on its armature beside the clips the animator driving it declared. Answers in edit mode as well as play mode.

Returns AnimationObservation — An AnimationObservation.

local a = animation.observe(); print(a.animatingCount, a.riggedBodyCount)
for _, b in animation.observe().bodies do print(b.entity, b.animating, b.reason) end

globals/animation/whyStill

animation.whyStill(entityId: string | EntityRef) -> (string?, string?)

Why the body on an entity is not animating. Returns nil when it IS animating, and otherwise one of deactivated, noRiggedSkeleton, noGraph, clipUnreadable, noOutputNode, retargetPending, retargetMatchedNoRoles, stopped, finished, paused, poseNotApplied, poseUnchanged — the nearest cause, so the answer names the thing to change. A second return carries the animator's own words when it could not build a graph.

An entity the engine holds no animation state for is answered from the entity itself, in the same order the engine resolves a body it does hold: one carrying no rigged Skeleton is noRiggedSkeleton, and a rigged one nothing drives is noGraph. An id no entity carries is neither — the reason is nil and the detail says so.

Parameters

entityId string | EntityRef — The entity's stable id, or an EntityRef.

Returns (string?, string?) — (reason, detail).

local why, detail = animation.whyStill(hero.id); if why then print(why, detail) end

globals/assert

assert(cond, msg?) -> value

Assert condition is truthy. Returns value if true.

globals/asset/add_tag

asset.add_tag(ref: RefArg, tag: string)

Add a tag to the asset's .metadata.tags. Idempotent. Creates the sidecar and the tags array if missing.

Parameters

ref RefArg — Any name the asset has.
tag string — Tag to add.

asset.add_tag("brick", "wip")

globals/asset/alias

asset.alias(ref: RefArg, alias: string) -> boolean

Add a name the asset answers to. asset.resolve, leaf shorthand, and the typed-argument coercion that content refs travel through all reach the asset by the alias from here on, exactly as they do by its identity — so a material naming a shader by its alias resolves, and content written against an older name keeps working after a rename. The name survives writes to the asset's own files. Raises when the name already resolves to a DIFFERENT asset — an alias extends the identity namespace and never takes a name out of another asset's hands — and when it is shaped like a guid or a VFS path, forms that resolve before identity lookup, so an alias in that shape could never answer.

Parameters

ref RefArg — The asset gaining the name.
alias string — The additional name. Any identity form: a bare leaf (standard) or a scope-qualified path (@builtin::shaders.legacy).

Returns boolean — True when newly added, false when the asset already answered to it.

asset.alias("@builtin::shaders.pbr", "standard")

globals/asset/aliases

asset.aliases(ref: RefArg) -> { string }

The additional names this asset answers to, beyond its own identity — what asset.alias registered, plus the package-relative ~pkg.tail form when the asset lives inside a package.

Parameters

ref RefArg — Any name the asset has.

Returns { string } — Array of alias names in canonical identity form.

for _, n in asset.aliases("pbr") do print(n) end

globals/asset/canCreate

asset.canCreate(typeName: string) -> boolean

Whether asset.create can instance typeName: the type declares creation logic (a behavior.luau onCreate hook) or ships a template/ skeleton the hookless fallback clones. A type with neither — one whose instances only arrive by import — answers false. The query a creation UI derives its offering from, so what it offers is what asset.create accepts.

Parameters

typeName string — Registered asset type (e.g. "material", "scene").

Returns boolean — true when asset.create(typeName, …) can produce one.

if asset.canCreate(kind) then asset.create(kind, name) end

globals/asset/categories

asset.categories() -> { string }

List every asset category the engine currently recognises. Use to discover valid type argument values for the rest of asset.*.

Returns { string } — Array of category names.

for _, c in asset.categories() do print(c) end

globals/asset/containing

asset.containing(path: string) -> AssetRef?

Walk path and its ancestors and return an AssetRef handle for the INNERMOST registered category-folder containing it — the typed asset that directly owns the path (e.g. main.scene for "/source/scenes/main.scene/scene.json"; for a folder that is itself a registered asset, that asset). Returns nil for paths outside any registered asset.

Parameters

path string — VFS path to inspect.

Returns AssetRef? — AssetRef handle, or nil.

local a = asset.containing("/source/scenes/main.scene/scene.json")

globals/asset/cpuResident

asset.cpuResident(ref: RefArg, typeName: string?) -> boolean

True when the asset is CPU-resident — a live script-component context holds it (a component's assetRef field, or an imperative asset.resolve/ref made while a component is the caller), which is what warms its bytes into memory. The CPU pool is a different pool from the device's: asset.observe().cpu lists it, asset.observe().textures / .meshes list what the device holds, and an asset can be in one and not the other.

Parameters

ref RefArg — Any name the asset has — handle, identity, guid, name or path.
typeName string (optional) — Category to restrict the match to. Omit to search every category.

Returns boolean — true when a live context holds it.

if asset.cpuResident(ref) then print("bytes are warm") end

globals/asset/create

asset.create(typeName: string, name: string, opts: { [string]: any }?) -> (AssetRef, { durable: boolean, survivesRestart: boolean?, warning: string?, playShadow: string?, shadowed: { string }?, scratch: { string }? })

Instance a new asset of an existing type. Runs the type's behavior.luau onCreate(name, opts) hook to produce the asset's files, then writes them under /source/<folder>/<name>.<type>/ (/source/<name>.<type>/ when no folder is given). This is the single generic asset-creation API. Refuses to clobber an existing edit-mode asset unless opts.overwrite = true, which re-authors it in place and keeps the existing guid (only the checksum changes). Pairs with asset.exists for content generators that re-run over the same names.

A create made from a script component's callback or a scene entrypoint is output the world reproduces on every load, so it is filed in the ephemeral /runtime/assets/ store instead, where the saved manifest never carries a second copy of it. name and folder spell the same IDENTITY in either store, so a reference written against that identity resolves the asset wherever the call filed it, and one generator run from an execute and from a component names one asset.

Parameters

typeName string — Registered asset type to instance (e.g. "material", "texture").
name string — Destination asset name (becomes /source/<name>.<typeName>). A bare identity — pass opts.folder to place it in a subfolder rather than spelling a path here. The accepted shape is the type's to declare: ^[A-Za-z][A-Za-z0-9_]*$ unless its behavior.luau exports a namePattern, as guide does to take getting-started and 01-overview. This call names the category FIRST and the asset second. Every other asset.* call taking both names them the other way round — asset.exists(name, category), asset.tryResolve(ref, category) — so a create-then-check pair reads if not asset.exists(n, t) then asset.create(t, n, opts) end. A call whose two arguments are read into each other, at either end of that pair, is refused and told which way round the call reads.
opts { [string]: any } (optional) — Optional table forwarded to the type's onCreate hook, minus five framework keys consumed here and never seen by the hook: folder (a relative subfolder path under /source to author the asset in, with / between its folders as in game/materials, so generated content groups instead of accumulating at the source root; those folders joined by dots prefix the asset's identity, game.materials.<name>), into (author INSIDE a resolved container ref), dest (an absolute destination path), overwrite (re-author in place, keeping the guid), and durable (file every content file this create writes on canonical /source with play still running, in place of the play shadow, and raise with the reason when they cannot become source, the same key, meaning the same thing, that vfs.write takes). Every type accepts all five whatever its own contract declares; asset.describe(typeName).frameworkParams names them, and a rejection prints them under the type's own parameters.

Returns (AssetRef, { durable: boolean, survivesRestart: boolean?, warning: string?, playShadow: string?, shadowed: { string }?, scratch: { string }? }) — The created asset's AssetRef, the SAME interned instance asset.resolve returns. ref.path is the VFS path the asset was written to (its folder: /zero/source/<folder>/<name>.<type>, the dest or into destination when one is given, /zero/runtime/assets/<identity>.<type> for a create filed in the ephemeral store), ref.primary the content file inside it (a shader's shader.wgsl, a module's init.luau), and both read and write with vfs.read / vfs.write or the read_file / write_file tools as they stand. ref.identity is the dotted name references use (<folder>.<name>, folder separators spelled .) and is what tostring(ref) prints. ref.guid is the stable id, and the type's ref methods (:getBytes, :ensureHandle, :serialize, ...) ride on it. Returned from execute, the ref reads back with all of these fields. Disk-only: nothing is uploaded to CPU/GPU. Where the bytes this call landed went, as a SECOND return value: { durable = true } when the files are where the call filed them, and { durable = false, survivesRestart = …, warning = …, playShadow = …, shadowed = { … } } when the play shadow took them: live in this session and undone by a guarded play-exit unless kept, with survivesRestart saying whether a restart of this engine keeps them, playShadow naming the routes that keep them and shadowed the paths. A create under /source/tmp/ answers durable = false with survivesRestart = false, a warning opening SCRATCH and scratch naming the paths: scratch reads back while the engine runs and is gone when it stops or restarts. durable is present whatever the answer is, so its absence is never a reading, and one call answers once for every file it wrote. It is the same answer, off the same shadow set and in the same words, that the MCP write_file / edit_file / capture tools attach to their own results — ask vfs.durability for it directly at any other write site.

local m = asset.create("material", "brick", { shader = "pbr" })
local s = asset.create("shader", "glow", { folder = "fx" }) -- s.path: /zero/source/fx/glow.shader, s.primary: /zero/source/fx/glow.shader/shader.wgsl
local wgsl = vfs.read(s.primary) -- the created asset's content file, read by the path the ref names
local mesh = asset.create("mesh", "tree", { positions = {...}, indices = {...} })
local mesh = asset.create("mesh", "rock", { positions = {...}, indices = {...}, folder = "terrain/props" })
local mesh = asset.create("mesh", name, { positions = p, indices = i, overwrite = true }) -- idempotent rebuild
local clip, d = asset.create("soundClip", "hit", { pcm = buf, sampleRate = 48000, channels = 1 })
if not d.durable then log.warn(d.warning) end
local c = asset.create("soundClip", "step", { pcm = buf, sampleRate = 48000, channels = 1, durable = true }) -- lands on source during play

globals/asset/declareReferenceArg

asset.declareReferenceArg(call: string, position: number, assetType: string)

Declare that call's argument at 1-based position names an asset of type, so a string literal written there is recorded as a reference. The positional counterpart of asset.declareReferenceField, for a call that takes its asset as a plain argument — including a world's own spawn helper, which is where a name most often stops being visible to the reference graph. A lookup whose asset is its FIRST argument (asset.resolve and its siblings) is already read and needs no declaration. Only a literal — or a name the file holds in a top-level string constant — is recorded; anything computed is reported as a dynamic resolve. An asset type states the same declaration as content, in the references.args of its type.yaml, which every reference table reads from the moment the type registers.

Parameters

call string — The callee as it is written at a call site.
position number — Which argument holds the name, counting from 1.
assetType string

asset.declareReferenceArg("spawnModel", 3, "mesh")

globals/asset/declareReferenceField

asset.declareReferenceField(call: string, field: string, assetType: string)

Declare that call's options table names an asset of type in its field, so a string literal written there is recorded as a reference by whatever writes the file. This is what puts an API that takes an asset BY NAME into the reference graph: the named asset becomes a dependency, travels with the content that names it into a pack or a pull, and a name nothing answers to becomes an unresolved dependency worldValidation reports and the push gate refuses. A field holding a TABLE of names — a material's textures — records every name in it. Declare once, beside the API; a call taking its asset as the FIRST positional argument is already read and needs no declaration. Only a literal is recorded; a computed name resolves at runtime and is reported as a dynamic resolve. An asset type states the same declaration as content, in the references.fields of its type.yaml, which every reference table reads from the moment the type registers.

Parameters

call string — The callee as it is written at a call site.
field string — The options-table field holding the name, read at the table's own level.
assetType string

asset.declareReferenceField("fx.beam", "material", "material")

globals/asset/declareReferenceKey

asset.declareReferenceKey(assetType: string, key: string, refType: string, via: ("asset_ref" | "settings_of")?)

Declare that, in a data file belonging to an assetType asset, the top-level key names an asset of type ("*" for any type) — a .material's mat.yaml naming the shader it draws with and the textures it binds. The names a format holds are references as surely as ones written in code: recording them carries a material's shader along with the material into a pack or a pull, and turns a name nothing answers to into an unresolved dependency instead of a surface that renders as the magenta error material. A key holding a table of names records one per entry; a { __ref = <guid> } envelope names its guid. via is the provenance the file's .refs records: "asset_ref" (default) for an asset the file uses, "settings_of" for the declarer whose settings schema the file's values follow. A type states the keys of its own files as content, in the references.keys of its type.yaml, which every reference table reads from the moment the type registers.

Parameters

assetType string — The category owning the file, e.g. "material".
key string — The top-level key holding the name(s).
refType string
via ("asset_ref" | "settings_of") (optional) — "asset_ref" or "settings_of".

asset.declareReferenceKey("material", "shader", "shader")
asset.declareReferenceKey("preset", "target", "*", "settings_of")

globals/asset/deprecation

asset.deprecation(ref: RefArg) -> Deprecation.Deprecation?

Whether an asset is deprecated, why, and what replaces it. An asset is deprecated when its own .metadata carries "deprecated": { "reason": "<one line why>", "replacedBy": "<identity>" }, when a container holding it at any depth does (a package, a toolbox, a folder asset), or when its asset type does. A deprecated asset still resolves and loads by identity and guid; search surfaces do not offer it, and resolving it warns once per session.

Parameters

ref RefArg — Asset handle, identity, guid or VFS path.

Returns Deprecation.Deprecation? — { reason, replacedBy, via, source }, where via is "asset", "container" or "type" and source is the folder holding the marker; nil when the asset is not deprecated.

local d = asset.deprecation("@builtin::deprecated.deprecatedParticles.DeprecatedParticleEmitter")

globals/asset/deps

asset.deps(ref: RefArg, type: string?) -> DepsResult

Return the asset's outbound dependency graph — every other asset recorded as a content dependency of it.

Parameters

ref RefArg — Any name the asset has.
type string (optional) — Category hint (optional).

Returns DepsResult — { deps = { { asset_guid, origin, literal, via, ... } }, worlds = { { world, name, commit, sites } }, unresolved_deps, problems }.

for _, d in asset.deps("main.scene").deps do print(d.asset_guid) end

globals/asset/describe

asset.describe(typeName: string) -> DescribeResult

The creation contract for an asset type: the parameters its onCreate(name, opts) hook accepts, as data. kind is "schema" (typed contract), "legacy" (untyped opts — anything passes), "none" (template scaffold — takes no opts), or "error" (the type's schema failed to parse; error says why). contract is the human-readable rendering validation errors print. frameworkParams and frameworkContract carry the options asset.create consumes itself: folder, into, dest, overwrite, durable, which every type accepts and none declares.

Parameters

typeName string — Registered asset type to describe (e.g. "texture").

Returns DescribeResult — the creation contract.

local contract = asset.describe("texture").contract

globals/asset/diagnose

asset.diagnose(ref: RefArg) -> any

Why one asset can or cannot be used, read from the engine rather than from what the caller asked for. Always carries usable; when false, reason is one of asset.unusableReasons() and detail is the engine's own message. primary names the file the type's declared primary list resolved to, the file a load of the asset reads. The payload's bytes are read by the engine's own decoder wherever it has one for that container, so usable is the verdict a load would reach and the call costs that decode.

Parameters

ref RefArg — Any name the asset has — handle, identity, guid, name or path.

Returns any — DiagnoseRecord

local d = asset.diagnose("myTex") if not d.usable then print(d.reason, d.detail) end

globals/asset/exists

asset.exists(name: string, typeName: string) -> boolean

Parameters

name string
typeName string

Returns boolean

globals/asset/get_field

asset.get_field(ref: RefArg, key: string) -> any

Read one top-level field from the asset's .metadata.

Parameters

ref RefArg — Any name the asset has.
key string — Field name.

Returns any — Field value or nil.

local author = asset.get_field("brick", "author")
local tags = asset.get_field("tree", "tags") -- → a Luau table

globals/asset/gpuResident

asset.gpuResident(ref: RefArg) -> boolean

True when the device holds a texture or mesh under this asset's guid, read off the inventory the renderer publishes.

Parameters

ref RefArg — Any name the asset has — handle, identity, guid, name or path.

Returns boolean — true when the device holds it.

print(asset.gpuResident("@builtin::models.Sample.DamagedHelmet"))

globals/asset/guid

asset.guid(ref: RefArg, type: string?) -> string

Return the guid for an asset.

Parameters

ref RefArg — Any name the asset has.
type string (optional) — Category hint (optional).

Returns string — Guid.

local g = asset.guid("@builtin::components.Camera")

globals/asset/has_field

asset.has_field(ref: RefArg, key: string) -> boolean

True when the asset's .metadata carries the named field.

Parameters

ref RefArg — Any name the asset has.
key string — Field name.

Returns boolean — True when present.

if asset.has_field("brick", "author") then end

globals/asset/has_tag

asset.has_tag(ref: RefArg, tag: string) -> boolean

True when the asset's .metadata.tags contains tag.

Parameters

ref RefArg — Any name the asset has.
tag string — Tag to check for.

Returns boolean — True when present.

if asset.has_tag("brick", "wip") then end

globals/asset/identity

asset.identity(ref: RefArg, type: string?) -> string

Return the canonical identity for an asset.

Parameters

ref RefArg — Any name the asset has.
type string (optional) — Category hint (optional).

Returns string — Canonical identity.

local id = asset.identity("brick")

globals/asset/import

asset.import(path: string, opts: { importerSettings: { [string]: { [string]: any } }? }?) -> string?

Import a raw source file NOW and return the produced asset path (a .bundle for a model, .texture for an image, .audio for a sound, …), or nil if no importer claims it. While play runs and holds the source where it stands (vfs.playHoldsInPlace), the import is owed: this returns nil, the source's job in tools.use("importers", "status") reads owed, and the import runs when play ends or pauses. This is the deterministic, on-demand counterpart to the engine's automatic import-on-write: it runs in the calling task and returns only when the import is complete. Pair it with a quiet write — vfs.write(path, bytes, { quiet = true }) lands the raw bytes without firing the automatic importer, then asset.import(path) imports them under your control, so you can act on the result instead of polling for the import to appear.

Parameters

path string — The raw source VFS path to import (e.g. a just-written .glb).
opts { importerSettings: { [string]: { [string]: any } }? } (optional) — Optional { importerSettings = { <importer name> = { <setting> = value } } }: settings for this import only, laid over the claiming importer's own. A setting that importer does not declare is an error.

Returns string? — The produced asset path, or nil when nothing claimed it.

local bundle = asset.import("/zero/source/generated/chest.glb")
local bundle = asset.import("/zero/source/generated/chest.glb", { importerSettings = { model = { keepSource = false } } })

globals/asset/inspect

asset.inspect(ref: RefArg, type: string?) -> InspectRecord

Everything known about one asset in a single record: identity, guid, source, type, scope and origin, its description and tags, the ref methods its type exposes, and the type's own inspect detail when it declares one. The read-everything counterpart to asset.resolve, which hands back a ref.

Parameters

ref RefArg — An AssetRef, an identity string, or a path.
type string (optional) — Narrow the resolve to one asset type when assets of several categories answer to the same bare name.

Returns InspectRecord — The inspect record.

local rec = asset.inspect("@builtin::materials.default")
local rec = asset.inspect(name, "mesh"); print(rec.guid, #rec.tags)

globals/asset/interface

asset.interface(ref: RefArg, category: string?) -> AssetInterface?

What the asset declares it takes and gives, read from its source without running it: the interface its type's interface: block says its sources declare (a .procNode declares one). description is what it does; each of inputs carries its name, type, default, min, max, options, optional and description as the declaration states them (a vector default reads as { ["$vector"] = { x, y, z } }); each of outputs its name and from, what makes it. problems lists each part of the declaration that is not written as literals, which a push refuses. The same data travels with the source's reference table to ZeroMind.

Parameters

ref RefArg — Any name the asset has.
category string (optional) — Category hint (optional).

Returns AssetInterface? — { form, id?, description?, inputs, outputs, problems }, or nil when the asset's type declares no interface or its source hands no builder.

local iface = asset.interface("@builtin::systems.procgen.nodes.points.disc", "procNode")

globals/asset/isPrivateSubasset

asset.isPrivateSubasset(path: string) -> boolean

Whether the asset at path is a private subasset: nested inside another asset whose type does not broadcast its subassets — a .bundle's interior meshes, materials and textures, an effect's own materials. A private subasset is part of its parent: it is not listed, searched or installed on its own, and it is excused from the files that describe an asset. Whether nested assets are private is the enclosing asset's type's to say: broadcasts_subassets: true in its type.yaml (a .package) makes its members public, and an asset nested in no other asset is never private. The same rule asset.list applies, where subassets = true is what reaches private subassets.

Parameters

path string — The asset's VFS folder path.

Returns boolean — True for a private subasset; false for any other asset, and for a path that names no asset.

asset.isPrivateSubasset("/zero/source/models/tree.bundle/material/Bark.material") -- true

globals/asset/list

asset.list(type_or_opts: (AssetCategory | ListOpts)?, scope: string?, opts: ListOpts?) -> ListResult

Query registered assets, returning each match as a resolved AssetRef handle. Every filter narrows the same enumeration and they compose: path selects a VFS subtree (the folder and everything under it), type keeps only those asset types within it, scope keeps only that scope, and fields keeps only assets whose .metadata matches. type and path each take one value or a list matching any of its entries, and all / any / none group whole filters — none excludes what it matches. order, limit, and offset shape the result: matches come back ordered by identity unless order names another field (name / path / type / guid). subassets decides how deep the enumeration reaches into composite assets. An enumeration returns the assets each container broadcasts, which its type declares: a .package and a .toolbox broadcast theirs, so their members are listed on their own, and a .bundle keeps its interior meshes, materials and rigs to itself, so they are reached through the bundle and a list of the folder holding it comes back without them. subassets = true enumerates those private interiors as well — what an asset is built out of, beside what the world offers. Each entry is the same envelope asset.resolve returns (__ref / type / name / guid / identity / path), so it can be passed anywhere an AssetRef is accepted, and the result carries :first() / :random() / :filter() / :sort() and friends. type takes the same values asset.categories() lists. The first positional argument is a path when it is absolute, a type otherwise. A type naming none of those values raises, naming it and any near spelling, so an empty result means a type the engine has. A type-group name stands for a family of those types rather than naming one (model for glb/gltf/fbx/obj, and image, text, config and script for theirs), so it lists whichever members the engine carries, and raises naming them when it carries none. An unknown key raises, as does a table setting both type and its older spelling category. A static (literal) type or path makes the enumeration part of the calling file's content dependencies when it is saved — the set travels with published content, so consumers get at-least the authoring world's assets.

Parameters

type_or_opts (AssetCategory | ListOpts) (optional) — Type or VFS path filter (a static literal so the enumeration can be captured for publish), or the full query table.
scope string (optional) — Scope filter (when first arg is a type).
opts ListOpts (optional) — The query table — see ListOpts.

Returns ListResult — The matched AssetRef handles, as a result carrying query methods.

local hero = asset.list({ path = "/zero/source/props", type = "mesh", fields = { tags = "hero" } }):first()
local parts = asset.list({ path = "/zero/source/models", type = "mesh", subassets = true }) -- a bundle's interior meshes too

globals/asset/listRevision

asset.listRevision() -> (number, number)

Two numbers that together move whenever what asset.list could answer changes: an asset made, removed, moved or renamed, a name it answers to added, a file created in or removed from an asset's folder (a new entity in the open scene writes one), or the index rebuilt. Rewriting a file that already exists leaves them still, and so does a write to a path no asset can sit at. The copies play makes resident in the runtime-asset layer move neither number. Hold the pair beside something derived from a listing and list again only when either differs, so a per-frame reader lists only on the frames where the answer can have changed.

Returns (number, number) — The index epoch and the generation within it.

local e, g = asset.listRevision() if e ~= held.e or g ~= held.g then held.list = asset.list("texture") end

globals/asset/list_field_values

asset.list_field_values(key: string) -> { any }

Distinct values seen for the named field across every asset's .metadata.

Parameters

key string — Field name.

Returns { any } — Array of distinct values.

local authors = asset.list_field_values("author")

globals/asset/list_fields

asset.list_fields() -> { string }

Distinct top-level field keys observed across every asset's .metadata. Useful for tooling discovering custom keys in use.

Returns { string } — Array of field names.

for _, k in asset.list_fields() do print(k) end

globals/asset/meta

asset.meta(ref: RefArg, type: string?) -> AssetMeta

Read the asset's engine-owned identity record (guid / checksum). Distinct from .metadata (agent-editable); for that use asset.metadata.

Parameters

ref RefArg — Any name the asset has.
type string (optional) — Category hint (optional).

Returns AssetMeta — Metadata table.

local m = asset.meta("brick") -- { guid = ..., checksum = ... }

globals/asset/metadata

asset.metadata(ref: RefArg, type: string?) -> AssetMeta

Read the asset's agent-editable .metadata sidecar as a Lua table. Missing sidecar returns {}. Distinct from asset.meta (engine-owned).

Parameters

ref RefArg — Any name the asset has.
type string (optional) — Category hint (optional).

Returns AssetMeta — JSON-shaped table; empty when no sidecar exists.

local md = asset.metadata("brick")

globals/asset/observe

asset.observe() -> any

What the engine is holding for content right now, in one reading: textures and meshes (one row per resource the device holds, each with the bytes it costs, its dimensions or buffer split, and where it came from), cpu (one row per asset a live script-component context holds), offDevice (one row per mesh nothing displays, whose GPU buffers were given back: held is "definition" when the engine keeps the mesh's whole CPU definition and "source" when it keeps only the stored file source it reads the mesh from again, and bytes is the CPU memory that costs), and totals: the aggregates those rows sum to, so the listing reconciles against renderer.textureMemory() and renderer.gpuMemory() (offDeviceBytes sums the offDevice rows).

Each pool is named because they are different pools: an asset can be on the device and not CPU-resident, or the reverse. devicePublished is false when no renderer has published an inventory and cpuPublished when the scripting VM has not published its pool — an engine that cannot answer reads differently from one answering with nothing resident.

Returns any — ResidencyReading

local r = asset.observe() print(#r.textures, r.totals.textureBytes)

globals/asset/pathHolding

asset.pathHolding(guid: string, checksum: string) -> string?

The VFS path holding the asset guid at exactly the version whose content checksum is checksum (the pair asset.meta reports): the guid's home when its bytes are that version, else a copy of that version held beside it, such as an installed dependency's. Returns nil when this world holds that asset at no such version.

Parameters

guid string — The asset's guid.
checksum string — The content checksum of the version wanted.

Returns string? — The path holding that version, or nil.

local m = asset.meta(asset.resolve("@builtin::shaders.pbr", "shader")); local path = asset.pathHolding(m.guid, m.checksum)

globals/asset/preview

asset.preview(ref: RefArg, opts: { [string]: any }?, type: string?) -> { [string]: any }

Render a preview of an asset. Resolves the ref and dispatches to its type's preview ref-method when present; otherwise returns the { available = false } sentinel ("no preview available for this type").

Parameters

ref RefArg — Any name the asset has.
opts { [string]: any } (optional) — Optional { size = { width, height }, angle = { yaw, pitch } }.
type string (optional) — Category hint (optional).

Returns { [string]: any } — { available, imageBase64?, width?, height?, bounds?, stats?, reason? }.

local p = asset.preview("@builtin::materials.gold", { size = { width = 512, height = 512 } })

globals/asset/primaryFile

asset.primaryFile(ref: RefArg) -> any

The file the asset type's declared primary list resolves to inside this asset, as the loader itself resolves it. A *.<ext> entry resolves the lowest-named file with that extension, and never the asset's preview.png, the picture rendered from it. resolved is false when no declaration matched and path is then absent; declared is the type's own primary list in match order.

Parameters

ref RefArg — Any name the asset has — handle, identity, guid, name or path.

Returns any — { path: string?, resolved: boolean, isFolder: boolean, declared: { string } }

print(asset.primaryFile("myTex").path)

globals/asset/ref

asset.ref(ref: RefArg, type: string?) -> AssetRef

Build a reference handle for an asset — the canonical ref envelope constructor. Identical shape to asset.resolve; preferred name for the author-side use case (embedding refs in YAML / JSON / Luau output).

Naming the asset here reads exactly as naming it in asset.resolve, down to raising on a miss: a literal is recorded as this source's dependency, a computed name is a dynamic resolve — free in a tool, refused on the gameplay path. asset.exists(name, type) is the probe a computed name can ask.

Parameters

ref RefArg — The asset to reference — an identity, a guid, a VFS path, or a handle.
type string (optional) — Category hint (optional).

Returns AssetRef — Ref handle. Raises on a miss or ambiguity, as asset.resolve does.

local r = asset.ref("animations.idle", "animation")

globals/asset/reloadPending

asset.reloadPending(ref: RefArg, typeName: string?) -> boolean

True while a write to this asset still owes it a reload — the write is inside the settle window that collects one authoring step's writes, or its reload is queued and the engine has not run it yet. False means every content change written so far has reached its subscribers, so a consumer bound to the asset now cannot be interrupted by a reload the earlier writes already earned. The recording is synchronous with the write, so a call made right after one already reads true.

Parameters

ref RefArg — Any name the asset has — handle, identity, guid, name or path.
typeName string (optional) — Category to restrict the match to. Omit to search every category.

Returns boolean — true while a content-change reload is still owed.

repeat task.wait() until not asset.reloadPending(ref)

globals/asset/reloadSeq

asset.reloadSeq(ref: RefArg, typeName: string?) -> number

How many content-change reloads this asset has been through — the count of onAssetReload dispatches the engine has RUN for it. A write to a file inside an asset does not reload it on the spot: the writes of one authoring step are collected for a settle window and the reload runs on a later frame. Read this, write, then poll for a larger number to learn the write's reload has actually reached subscribers. Monotonic per asset and session-scoped; 0 for an asset whose content has not changed since boot.

Parameters

ref RefArg — Any name the asset has — handle, identity, guid, name or path.
typeName string (optional) — Category to restrict the match to. Omit to search every category.

Returns number — content-change reloads dispatched for this asset.

local at = asset.reloadSeq(ref)
vfs.write(asset.source(ref) .. "/graph.json", encoded)
repeat task.wait() until asset.reloadSeq(ref) > at

globals/asset/remove_field

asset.remove_field(ref: RefArg, key: string)

Remove one top-level field from the asset's .metadata. No-op when the field isn't present.

Parameters

ref RefArg — Any name the asset has.
key string — Field name.

asset.remove_field("brick", "author")

globals/asset/remove_tag

asset.remove_tag(ref: RefArg, tag: string)

Remove a tag from the asset's .metadata.tags. No-op when the tag isn't present.

Parameters

ref RefArg — Any name the asset has.
tag string — Tag to remove.

asset.remove_tag("brick", "wip")

globals/asset/resolve

asset.resolve(ref: RefArg, type: (C & string)?) -> AssetRef<C>

Find an asset. The returned handle carries every name form the asset has (guid, identity, path, type) so downstream code can read any one of them without calling resolve again. Raises when ref resolves to no asset — or, with a type, to no asset of that type — and when ref reaches more than one asset, where it names the candidates for you to pick from instead of picking one of them. A <scope>::-qualified identity reaches exactly one: @root::name for the asset this world holds at its source root, the library identity (@builtin::…) for a library's. For the same lookup answering a miss with nil, use asset.tryResolve(ref, type).

A name written as a string LITERAL is recorded as this source's dependency on that asset, so the asset travels with the content and still resolves once someone installs it in another world. A COMPUTED name cannot be written down, so nothing pins what it reaches: that is a dynamic resolve — free in a tool, refused on the gameplay path (a component or scene entrypoint). asset.tryResolve, asset.ref and asset.source read the name they are given exactly this way too, so which of the four you reach for changes neither answer. To ask whether a computed name has files without reaching a handle, use asset.exists(name, type).

Parameters

ref RefArg — The asset to find — an identity, a guid, a VFS path, or a handle.
type (C & string) (optional) — Category to restrict the match to (optional). Separates a bare name that assets of different categories share (asset.resolve("cube", "mesh")); where several assets of the SAME category answer to it, the scope-qualified identity is what separates them. A reference naming a file an importer has since promoted (wall.png after the texture importer turned it into wall.texture) resolves to the promoted asset, and says so in the log once per reference.

Returns AssetRef<C> — Asset handle, carrying the category when one was named — so the methods that category defines are checked on the result. Raises (rather than returning nil) on a miss, and on a name that reaches more than one asset.

local a = asset.resolve("@builtin::components.Camera")

globals/asset/schema

asset.schema(ref: RefArg) -> any?

The typed settings schema governing this asset — the Field descriptors its type declares for every asset of that type, or the ones the asset declares for itself. Each descriptor carries its kind, default, constraint and description, so an editor renders it and an agent reads what a setting takes without guessing.

Parameters

ref RefArg — The asset (ref, identity, guid or path).

Returns any? — The schema table, or nil when this asset carries no settings.

local schema = asset.schema(texRef); print(schema.filter.constraint.values)

globals/asset/setSettings

asset.setSettings(ref: RefArg, patch: { [string]: any })

Write a partial settings patch. The patch is merged over the asset's current values, the type's cross-field constraints are applied — a field the patch does not name moves to the only value the rest allows (a texture given a high-precision compression stops generating mipmaps), and a field the patch names that cannot hold refuses the write, saying why — and the result is checked against its schema; a violation refuses the whole write, naming every problem, and leaves the asset on its last valid settings. A valid write re-runs the type's onChange, which is where the type re-cooks whatever the settings describe. Older field names are accepted and stored under the current ones, and a field whose default follows another moves with it when the write does not name it.

Parameters

ref RefArg — The asset (ref, identity, guid or path).
patch { [string]: any } — Any subset of the schema's fields.

asset.setSettings(texRef, { filter = "nearest" })

globals/asset/set_field

asset.set_field(ref: RefArg, key: string, value: any?)

Set one field in the asset's .metadata, creating the sidecar if missing. Sibling fields are preserved. When the new value AND the existing value are both maps (objects), the new value DEEP-MERGES into the existing one, so writing one sub-key never drops the others — set_field(ref, "provenance", { bakedBy = "agent" }) keeps every other provenance key. Arrays and scalars replace. Clear a whole field with asset.remove_field; replace the entire sidecar with asset.set_metadata.

Parameters

ref RefArg — Any name the asset has.
key string — Field name.
value any (optional) — Field value (any JSON-serialisable Lua value).

asset.set_field("brick", "author", "me")
asset.set_field("tree", "provenance", { bakedBy = "agent" }) -- merges; sibling keys kept

globals/asset/set_metadata

asset.set_metadata(ref: RefArg, data: AssetMeta)

Replace the asset's .metadata sidecar with the given table. Pass an empty table to clear all fields.

Parameters

ref RefArg — Any name the asset has.
data AssetMeta — Full JSON-shaped contents for the sidecar.

asset.set_metadata("brick", { author = "me", tags = { "wip" } })

globals/asset/settings

asset.settings(ref: RefArg) -> any?

This asset's settings values, with every declared default filled in.

Parameters

ref RefArg — The asset (ref, identity, guid or path).

Returns any? — The values table, or nil when this asset carries no settings.

if asset.settings(texRef).filter == "nearest" then ... end

globals/asset/source

asset.source(ref: RefArg, type: string?) -> string

Return the VFS source path for an asset.

It reaches the asset, so the name carries the same reference contract asset.resolve's does: a literal is recorded as this source's dependency, a computed name is a dynamic resolve — free in a tool, refused on the gameplay path. asset.exists(name, type) is the probe a computed name can ask.

Parameters

ref RefArg — Any name the asset has.
type string (optional) — Category hint (optional).

Returns string — VFS source path.

local p = asset.source("brick") -- "/source/brick.material"

globals/asset/tags

asset.tags(ref: RefArg) -> { string }

Convenience read of the .metadata.tags array.

Parameters

ref RefArg — Any name the asset has.

Returns { string } — Array of tag strings.

for _, t in asset.tags("brick") do print(t) end

globals/asset/tryResolve

asset.tryResolve(ref: RefArg, type: string?, base: string?) -> AssetRef?

The same lookup asset.resolve performs, answering a miss with nil instead of raising. Every name form, the same type narrowing, and the same handle on success — so "use it if it is there" needs no pcall around a call whose failure would otherwise be indistinguishable from a real error.

This consults the asset REGISTRY, so it sees registered assets wherever their files live, @builtin:: ones included. asset.exists(name, type) answers the narrower question of whether an asset's files are present in the current mode's store.

A name that reaches more than one asset of type (or more than one asset, with no type) still raises: that is a question about the reference, and a nil there would report absence for content that is present twice. A name that only assets of other types answer to reaches no asset of type, and answers nil.

Parameters

ref RefArg — The asset to look up — an identity, a guid, a VFS path, or a handle.
type string (optional) — Category to restrict the match to (optional). A reference naming a file an importer has since promoted resolves to the promoted asset, the same as asset.resolve.
base string (optional) — Optional VFS path of the asset the reference is written in, or of its file: a data asset resolving the references it holds, such as a material resolving its mat.yaml. A ~ / ~.tail ref expands against the root enclosing it, and a ref its .refs recorded reaches the asset recorded there. Without it, both are read at the calling Luau source, as asset.resolve reads them.

Returns AssetRef? — Asset handle, or nil when the reference resolves to no asset.

local mat = asset.tryResolve(name, "material")

globals/asset/typeRef

asset.typeRef(target: RefArg) -> string?

Return the pinned asset_type reference (the type's guid) that the asset is an instance of. Resolve the full type with asset.resolve(asset.typeRef(target)). Returns nil for loose files / assets with no pinned type.

Parameters

target RefArg — Asset handle / identity / guid / VFS path.

Returns string? — Pinned type's guid, or nil.

local t = asset.resolve(assert(asset.typeRef("brick")))

globals/asset/unusableReasons

asset.unusableReasons() -> { string }

Every reason asset.diagnose can report an asset unusable for, sorted.

Returns { string } — Array of reason names.

for _, r in ipairs(asset.unusableReasons()) do print(r) end

globals/asset/validate

asset.validate(ref: RefArg, type: string?) -> ValidateResult

Validate an asset folder against its type's type.yaml, plus the type's own semantic validation. Structural problems come from type.yaml — missing required files, unsatisfied one_of_group alternatives, and (when allow_unlisted: false) unexpected children. validated = false when no type.yaml is registered — nothing structural to check. On top of that, when the asset's type ships a behavior.luau exporting a top-level validate(assetRef) -> { { code, message, severity? } }, its reported problems (severity defaults to "error") are appended to problems; error-severity problems flip ok to false, warnings leave it untouched. A hook that raises or returns a non-table is itself reported as a validate.hook_failed error problem — a broken hook blocks. A type with no validate export behaves exactly as the structural check alone. world.push calls this per user asset, so a type's semantic validation is enforced at publish time with no further wiring.

Parameters

ref RefArg — Any name the asset has.
type string (optional) — Category hint (optional).

Returns ValidateResult — { ok, typeName, validated, problems }.

local v = asset.validate("@builtin::components.Camera")

globals/asset/warmup

asset.warmup(ref: RefArg, opts: WarmupOpts?) -> WarmupResult

Warm an asset's bytes into CPU memory and follow its declared content dependencies to each referenced asset, deduped by guid. Type-agnostic (reads the generic ref graph) and CPU-only — never touches the GPU. Side-effect-free name resolution (uses asset.guid/asset.deps, not asset.resolve).

Parameters

ref RefArg — Any name the root asset has — handle, identity, guid, or path.
opts WarmupOpts (optional) — Optional { vias, max } — restrict ref-edge kinds / cap closure size.

Returns WarmupResult — { closure, count } — the deduped guid closure warmed and its size.

local w = asset.warmup("@builtin::scenes.test_arena")

globals/audio/decode

audio.decode(zaud: buffer | string) -> (string?, number?, number?)

Decode a ZAUD payload into interleaved f32 PCM. A PCM payload comes back as the frames its header accounts for, held to the whole frames the bytes behind it fill, so the sample count is always a whole number of channels and a consumer walking it channels at a time ends on a frame.

Parameters

zaud buffer | string — A ZAUD payload — a buffer or a binary string.

Returns (string?, number?, number?) — (pcm, sampleRate, channels), or (nil, err).

globals/audio/device

audio.device() -> AudioDeviceStatus

What the engine's audio output is doing: state is "open" while a stream is running on an output device and "silent" while none is, and device names the device an open stream runs on. The counters record what the engine has been through keeping one open — faults a live stream reported, changes of the host's default output, reopens the engine made, failedOpens the platform refused, and the glitches a listener heard as dropouts, with lastError carrying what the platform said. A device that goes away leaves the mixer running and the engine opening a stream again as soon as one is there.

Returns AudioDeviceStatus — An AudioDeviceStatus.

local d = audio.device(); print(d.state, d.device, d.reopens)

globals/audio/encode

audio.encode(sourceBytes: buffer | string, opts: { [string]: any }?) -> (string?, string?)

Encode container audio bytes (ogg / mp3 / wav / flac) into a ZAUD payload. Every decoded sample must be finite; a source whose samples carry a NaN or an infinity comes back as (nil, err) naming how many fail and where the first one sits.

Parameters

sourceBytes buffer | string — Encoded source audio bytes — a buffer or a binary string.
opts { [string]: any } (optional) — { codec: "opus"|"pcm"?, bitrateKbps: number?, vbr: boolean?, sampleRate: number?, forceMono: boolean?, loopStart: number?, loopEnd: number? }

Returns (string?, string?) — The ZAUD bytes, or (nil, err).

local zaud = audio.encode(oggBytes, { bitrateKbps = 96 })

globals/audio/encodePcm

audio.encodePcm(pcm: any?, sampleRate: number, channels: number, opts: { [string]: any }?) -> (string?, string?)

Encode raw interleaved f32 PCM into a ZAUD payload. Every sample must be finite; a buffer carrying a NaN or an infinity comes back as (nil, err) naming how many fail and where the first one sits, so a filter that diverged over part of a bake is caught before it is written. The sample count is a whole number of channels: a buffer with a tail over comes back as (nil, err) naming the whole frames it holds and the samples past them.

Parameters

pcm any (optional) — Interleaved f32 samples — a buffer or a binary string of little-endian f32, the shape microphone.samples and audio.decode hand back, or a flat number array. A byte payload's samples are its 4-byte lanes, and a length that stops partway through one comes back as (nil, err) naming the whole samples it holds and the bytes past them.
sampleRate number — Source sample rate in Hz.
channels number — 1 or 2, and a divisor of the sample count.
opts { [string]: any } (optional) — Same shape as audio.encode.

Returns (string?, string?) — The ZAUD bytes, or (nil, err).

local s = microphone.status(); local zaud = audio.encodePcm(microphone.samples(), s.sampleRate, 1)

globals/audio/info

audio.info(zaud: buffer | string) -> (AudioInfo?, string?)

Read a ZAUD payload's header. A PCM payload's samples are its bytes, and the header is read against them: a sample count differing from frames * channels comes back as (nil, err) naming both counts, and a sample carrying a NaN or an infinity comes back as (nil, err) naming how many fail and where the first one sits, so the header handed back describes a clip that is as long as it says and can sound. The header describes the clip's shape — rate, channels, frames, duration, codec, loop points. What the samples do where a whole-clip loop wraps is a reading of its own, audio.loopSeam, which is the call that answers whether a bed cycles without a click.

Parameters

zaud buffer | string — A ZAUD payload — a buffer or a binary string.

Returns (AudioInfo?, string?) — An AudioInfo table, or (nil, err).

local info = audio.info(zaud); print(info.durationMs)

globals/audio/levels

audio.levels() -> AudioLevels

The master mix's peak and RMS over the meter's most recent closed window, measured without recording anything.

Returns AudioLevels — An AudioLevels.

local l = audio.levels(); print(l.peak, l.rms, l.windowMs)

globals/audio/listener

audio.listener() -> AudioListenerState

Where the scene is heard from, how many active listeners exist, and which entity's listener drives the ears.

Returns AudioListenerState — An AudioListenerState.

local l = audio.listener(); print(l.present, l.count, l.entity)

globals/audio/loopSeam

audio.loopSeam(zaud: buffer | string) -> (AudioLoopSeam?, string?)

Measure what a clip's samples do where a whole-clip loop wraps, so a bed can be judged before anyone hears it tick. The wrap's own step (|x[1] - x[frames]|) is reported against the step the signal ordinarily makes between neighbouring samples, as ratio = step / meanStep — a figure in the units the signal itself moves in, so a quiet ambience and a loud drone are read the same way. A bed whose partials wrap reads near 1; one carrying a strike at its head and silence at its tail reads in the tens. ratio and the step / meanStep / maxStep beside it belong to the worst channel, channel names it, and channels carries every channel's own reading. seamless is ratio <= threshold, the same threshold asset.create("soundClip", ...) warns past. The reading is taken on the DECODED samples, so it answers for what the codec left behind and for a clip that arrived already encoded and whose source buffer nobody holds. Costs a decode of the whole payload; audio.info reads a header without one.

Parameters

zaud buffer | string — A ZAUD payload — a buffer or a binary string.

Returns (AudioLoopSeam?, string?) — An AudioLoopSeam table, or (nil, err).

local seam = audio.loopSeam(clipRef:getBytes()); print(seam.ratio, seam.seamless)

globals/audio/mixer

audio.mixer() -> AudioMixerLevels

The levels the mixer is applying to the mix right now: the master level, whether the mix is muted, and the level of every channel one has been set on. A channel absent from channels plays at unity, so a source naming it is heard at the volume it asks for.

Returns AudioMixerLevels — An AudioMixerLevels.

local m = audio.mixer(); print(m.master, m.muted, m.channels.music)

globals/audio/observe

audio.observe() -> AudioObservation

Report what the mixer is making audible right now, and why a source is not. One read covering every live voice with the mixer's own playback state and effective gain, the master mix's level, the mixer's voice accounting, the listener, the output device the mix is reaching, and what the subsystem costs. Answers in edit mode as well as play mode.

Returns AudioObservation — An AudioObservation.

local a = audio.observe(); print(a.audibleCount, a.levels.rms)
for _, v in audio.observe().voices do print(v.entity, v.mixerState, v.silence) end

globals/audio/peakSince

audio.peakSince(window: number) -> number?

The loudest peak the master mix reached across the meter's windows that closed after its windows count stood at window. audio.levels() carries the window that closed last, so a reader sees the windows its own frames happen to land on; this spans all of them, which is what measuring a sound shorter than the gap between two reads takes. Take the mark from audio.levels().windows before the sound starts, wait until windows has advanced past the sound's length, then read the span.

Parameters

window number — A windows count taken from audio.levels() earlier.

Returns number? — The loudest window peak in the span, or nil when the meter holds no peak for it — nothing has closed since window, or the span reaches further back than the meter's history of recent windows, so a reader that came back too late learns that instead of reading the maximum of the part that survived.

local mark = audio.levels().windows
local peak = audio.peakSince(mark)

globals/audio/peaks

audio.peaks(zaud: buffer | string, buckets: number) -> ({ number }?, string?)

The loudest sample in each of buckets equal spans of a clip, across its channels, as an absolute value held to 1: the shape a waveform is drawn from. A span is frames // buckets frames (at least one), and the frames past buckets * span fall in no bucket; a clip shorter than buckets frames fills one bucket per frame and leaves the rest at 0. The payload is decoded a block at a time, so the work never holds the clip's whole PCM, however long it runs.

Parameters

zaud buffer | string — A ZAUD payload — a buffer or a binary string.
buckets number — How many spans to divide the clip into: an integer from 1 to 4096.

Returns ({ number }?, string?) — One peak per bucket, or (nil, err).

local bars = audio.peaks(clipRef:getBytes(), 96)

globals/audio/profile

audio.profile() -> AudioProfile

What the audio subsystem has cost since the profiling window opened — the streaming pump, clip decode, clip encode, voice starts, and building the observation itself. Every total is a SUM across that window rather than a per-frame figure, and the window runs from the last audio.resetProfile() or from engine start. For what a frame costs now, reset, let frames pass, then divide by the frames the window reports.

Returns AudioProfile — An AudioProfile.

audio.resetProfile(); task.wait(1); local p = audio.profile()
print("per frame:", (p.pump.totalMs + p.observe.totalMs) / p.frames)

globals/audio/resetProfile

audio.resetProfile()

Open a new audio profiling window, discarding what the previous one measured. Call this before timing a stretch of frames: without it audio.profile() reports totals reaching back to engine start.

audio.resetProfile()

globals/audio/setChannelVolume

audio.setChannelVolume(channel: string, volume: number)

Set the level of one mixer channel — the channel an Audio component names, such as "sfx", "music" or "ambient", or any name the scene invents. It scales every voice on that channel and nothing else, reaches voices that are already playing, and comes back per voice as gain.channel. A channel no level has been set on plays at unity.

Parameters

channel string — The channel name, matching Audio.channel.
volume number — Channel level, 0..1.

audio.setChannelVolume("music", 0.3)
for _, v in audio.voices() do print(v.channel, v.gain.channel) end

globals/audio/setMasterVolume

audio.setMasterVolume(volume: number)

Set the master level of the mix, on the engine's 0..1 amplitude scale. It scales every voice whatever channel it plays on, reaches voices that are already playing, and comes back per voice as gain.master.

Parameters

volume number — Master level, 0..1.

audio.setMasterVolume(0.5)

globals/audio/setMuted

audio.setMuted(muted: boolean)

Silence or unsilence the whole mix. A muted mix sounds nothing whatever its master and channel levels read, every voice reports masterSilent, and unmuting hands the levels back untouched.

Parameters

muted boolean — Whether the mix is silenced.

audio.setMuted(true)

globals/audio/voice

audio.voice(entityId: string) -> AudioVoice?

The voice on one entity, or nil when that entity carries no audio source. An entity despawned since the frame the observation describes answers with its report from that frame, so an onDestroyed handler reads how the entity's voice stood as it went.

Parameters

entityId string — The entity's stable id.

Returns AudioVoice? — An AudioVoice, or nil.

local v = audio.voice(e.id); print(v and v.mixerState)

globals/audio/voiceAccounting

audio.voiceAccounting() -> AudioVoiceAccounting

How many voices the mixer can hold, how many are in use, how many are free — read off the mixer's own tracks, so the free count is the one a play call is granted or refused against. The two pools are reported apart: capacity / inUse / free are the main track, which carries the NON-spatial voices, while a spatial voice plays through its own sub-track and is counted by spatialInUse instead. sourcesHolding counts both pools from the sources that own them, so it equals inUse + spatialInUse while every voice answers to a source.

Returns AudioVoiceAccounting — An AudioVoiceAccounting.

local v = audio.voiceAccounting(); print(v.inUse .. "/" .. v.capacity)
local v = audio.voiceAccounting(); print(v.sourcesHolding - (v.inUse + v.spatialInUse))

globals/audio/voices

audio.voices() -> { AudioVoice }

Every live audio source with the mixer's opinion of it.

Returns { AudioVoice } — An array of AudioVoice.

for _, v in audio.voices() do print(v.clip, v.gain.effective) end

globals/audio/whySilent

audio.whySilent(entityId: string) -> (string?, string?)

Why the source on an entity is making no sound. Returns nil when it IS sounding, and one of noBackend, noDevice, notResident, neverStarted, refused, paused, ended, gainZero, channelSilent, masterSilent, outOfRange when it is not — the nearest cause, so the answer names the thing to change. A second return carries the mixer's own words when it refused the source, and "no audio source on this entity" when nothing there plays at all.

Parameters

entityId string — The entity's stable id.

Returns (string?, string?) — (reason, detail).

local why = audio.whySilent(e.id); if why then print(why) end

globals/av/is_live

av.is_live() -> boolean

True if a live-stream session is currently active.

Returns boolean — Whether the live encoder is running.

if av.is_live() then av.stop_live() end

globals/av/is_recording

av.is_recording() -> boolean

True if a recording session is currently active.

Returns boolean — Whether a recording is in progress.

print("recording:", av.is_recording())

globals/av/live

av.live(opts: LiveOpts?) -> (string?, string?)

Start the live-stream encoder. The stream is served at /engine/live.stream and reverse-proxied at /stream/<instance>/live.stream as a binary length-prefixed protocol consumed by the multiviewer UI's WebCodecs decoder. A browser engine encodes it with the browser's WebCodecs encoder, and the proxy reads it over the ZeroMind bridge and serves it at the same /stream/<instance>/live.stream. When texture_handle is set, the encoder reads from that GPU texture's guid (a Camera pointed at it via setTargetTexture) instead of the scene's viewport: that's how spectator cameras work. A stream runs beside an av.record take. Returns the stream URL once the encoder is open and streaming, or nil and the reason it is not: no encoder on this engine that streams H.264, a stream already running, an odd dimension, a session the encoder refuses (a size it will not encode, a spectator texture with no size to stream at). The engine opens the encoder, so the call waits for it: run it where it can yield, wrapping it in task.spawn from a callback that cannot. The reason also stands on av.status().liveError, which is where a stream that ends on a failure later says why.

Parameters

opts LiveOpts (optional) — Encoder options.

Returns (string?, string?) — Stream URL, or nil. Why the stream was refused, when it was.

local url, why = av.live({ width = 1280, height = 720, fps = 60 })

globals/av/record

av.record(path: string, opts: RecordOpts?) -> (string?, string?, Durability?)

Start recording the engine output to a VFS path. Default dir is /zero/runtime/recordings/ when path is not absolute: that directory is writable in every mode and stays out of the world's sync, so a take named relative to it is kept for the session and read back with vfs.read. A destination under /zero/source/ is world content, which the world's sync uploads when the take is filed. The take runs until av.stop_recording() unless opts bounds it with max_duration_sec (seconds of the take's own clock) or frames (captured frames); max_duration_sec wins when both are given, and the bound in force reads back as av.status().recordingBound. With no chroma/range opts the format defaults to full-range 4:4:4 HEVC where the GPU supports it, else 4:2:0. In the browser the browser's WebCodecs encoder converts the picture itself, so the take is 4:2:0 in the code range that encoder chooses, which an H.264 or AV1 file states, and grain above 0, chroma = "444" and any stated range are refused by name, and the master mix is encoded by the browser's WebCodecs AudioEncoder. On the "software" backend the take is H.264 encoded on the CPU, which costs the run it records: read av.status().recordingAchievedFps against recordingRequestedFps to see the rate it reached. What the take did with the master mix reads back as av.status().recordingAudio. cadence picks the clock the take stamps its frames from. "realtime" (the default) stamps each frame with the wall-clock slot it was captured in, so a recorded session is watched back at the speed it happened and an engine ticking under fps leaves slots empty. "frame" stamps every rendered frame one fixed slot after the last, so a timeline whose own clock advances a step per rendered frame — a cutscene, a scripted demo, anything on a fixed timestep — is delivered at the length that timeline runs to, however slowly the engine drew it: a take of frames = n at fps is n / fps seconds of film, and a max_duration_sec bound counts that film's seconds. A "frame" take records silent, because the master mix plays in wall-clock seconds and cannot share a file with a fixed-step picture; recordingAudio says so, and an explicit audio = true beside it is refused. Record the sound as a second "realtime" take. camera names the camera the take draws its film from — an entity proxy, an entity id, or an entity name. That camera holds the viewport for as long as the take runs, above the priority contest and above camera.setEditorOverride, so the film is its view and the frames carry everything the presented frame carries. Only frames that camera drew go into the film, and a camera that never draws the viewport ends the take with the reason on av.status().recordingError — so a take is the view it named or it is no take. The camera belongs to the take: nothing is written to it, and the viewport is back under its own contest the moment the take ends. It reads back as av.status().recordingCamera while the take runs. Omitted, the take records whichever camera holds the viewport, which in an engine on the editor profile is the editor's own fly camera rather than the scene's. renderLayers is the render-layer include spec the viewport is drawn under for as long as the take runs — the same token string a capture takes: all seeds every layer, name adds one and !name drops one, so "all !EditorUI !debug" films the scene without the editor's chrome or the authoring overlays (gizmos, light and probe icons, frustums, collider wireframes) over it, and "all !ui !EditorUI !debug" drops the authored HUD as well. The viewport admits geometry and screens by that one spec, so it states the whole picture. It belongs to the take: nothing is written to the camera it is stated against, and the moment the take ends — its bound reached, stopped, or refused — the viewport is back under the camera's own spec. While a take states layers, the window shows what the film holds, and av.status().recordingLayers reads the spec back. Omitted, the take records the engine output as presented. width / height are the film's size: the whole of the presented frame (in an editor layout, the rect the scene is drawn into) is resampled to it, and a side left out takes the presented frame's. Returns the destination path of a session that is open and recording, or nil and the reason it is not: an adapter that cannot encode, a take already running, an option the encoder rejects, a resolution the device refuses, a destination the VFS cannot write (a read-only path, a directory, a host filesystem path). The engine opens the session, so the call waits for it: run it where it can yield, wrapping it in task.spawn from a callback that cannot. How a take finished reads back as av.status().recordingEnd; a refused request puts its reason there and on recordingError and leaves no take report behind, while a request refused because a take is already running leaves that take's report as it is.

Parameters

path string — VFS destination path, returned in its /zero form; a relative name records under /zero/runtime/recordings/.
opts RecordOpts (optional) — Encoder options (optional).

Returns (string?, string?, Durability?) — Destination VFS path of the open recording, or nil. Why the recording was refused, when it was. Where the film will land, as a THIRD return value: { durable = true } when the take is filed where this call named it and the next session reads it back, { durable = false, survivesRestart = …, warning = …, playShadow = …, shadowed = { … } } when the play shadow takes it: live in this session and discarded on a guarded play-exit unless kept, with survivesRestart saying whether a restart keeps it and playShadow naming the routes that keep it, and { durable = false, survivesRestart = false, warning = …, scratch = { … } } when it is filed in /source/tmp/ scratch, which is gone when the engine stops or restarts. The take is filed when it ENDS, so this is the verdict a write landing there gets on the terms standing at THIS call, answered here because here is where a caller can still act on it: a film recorded during play is kept with vfs.promotePlayShadow(path) and play keeps running. durable is present whatever the answer is, so its absence is never a reading. It is the same answer, in the same words, that asset.create returns and the MCP write_file / capture tools attach to their own results. A take opened in edit that runs on into play is filed on play's terms, so av.status().recordingDurability — which answers for the take still running — is where a take spanning a mode change reads.

local clip = av.record("intro.mp4", { fps = 60 })
local film = av.record("cut.mp4", { fps = 24, frames = 24 * 181, cadence = "frame" })
local clean = av.record("take.mp4", { fps = 24, renderLayers = "all !EditorUI !debug" })
local shot = av.record("film.mp4", { fps = 24, frames = 240, camera = "FilmCamera" })
local take, why, d = av.record("take.mp4", { fps = 24 })
if d and d.playShadow then log.warn(d.warning .. " " .. d.playShadow) end

globals/av/save

av.save(path: string, opts: SaveOpts?) -> (boolean, string, SaveReport?)

Save a finished take to the device of the person at the screen, so the recording outlives the session: /zero/runtime/recordings/ is gone when the engine stops (and, in the browser, when the tab closes). path names the take the way av.record did: a relative name is the take of that name under /zero/runtime/recordings/, and any VFS path works too. The person decides where it lands. In the browser the take downloads at once when the save rides a click or key press of the person at the screen, so call it from a button's click handler; a save with no such gesture behind it (an agent, a timer, the end of a take) makes the page show the person an offer they save with a click of their own, because a browser starts a download only for a person who just acted. A native engine opens the host's save dialog in their downloads folder, with the take's name filled in, and writes it where they choose; an engine with no window has no person at its screen and refuses. The call waits for the first answer, so run it where it can yield (task.spawn from a callback). Returns true, a sentence saying where the take went, and the report { state, detail, name, size, path } with state "saved" (on the device) or "offered" (in front of the person, as the page's offer or the save dialog); or false and why it was refused: a take still recording (it is filed when it ends), no file at the path, a host path, a name with a folder in it, no window, or a take too large for the browser to hold, which the page shows the person as well. The saved file is the take byte for byte. userfile.save saves any other file the engine holds.

Parameters

path string — The take: its name under /zero/runtime/recordings/, or a VFS path.
opts SaveOpts (optional) — { name }, the file name the device gets (the take's own when omitted).

Returns (boolean, string, SaveReport?) — Whether the device took the take. Where it went, or why it was refused. The save report, when the device took the take.

local ok, where = av.save("take.mp4")
local ok, where, report = av.save("take.mp4", { name = "boss-fight.mp4" })

globals/av/status

av.status() -> AvStatus

Report the encoder subsystem's state. Always available regardless of GPU support. backend is the encode backend in use: "vulkan" or "vaapi" on an adapter with a media engine, "software" where a native engine encodes on the CPU, and "webcodecs" in the browser, where the browser's own WebCodecs encoder takes the frames. hardware is true for the first two, and in the browser when the browser reports a hardware encoder for the default codec, so a caller that pays for the take in engine time knows which it is getting. codecs lists what the backend encodes with the recording default first (h264 in the browser, which every engine plays back with video.create). reason is "supported", "not_probed" before the engine has asked its encoder, "unsupported_adapter" for a native adapter with no encoder, or "unsupported_browser" for a browser whose WebCodecs encodes none of the codecs the engine offers. live is true while the av.live stream is running, and liveError is why the most recent stream was refused or ended on a failure: nil while one runs and after av.stop_live() ended one that had opened, and the reason that names av.stop_live() after it ended one still opening. A take of its own reads back on the recording fields: recording is the destination of the take in flight, recordingBound what will end it, recordingEnd how the most recent one ended, and recordingError why one produced no file. recordingAudio is the codec the take is writing the master mix with ("opus"), or the reason the file carries no audio track — read it to tell a film with a soundtrack from a silent one. recordingLayers is the render-layer include spec the armed take is drawing the viewport under, in the words its caller wrote, and nil for a take that stated none — the reading that answers what is in the picture rather than how much of it there is. recordingCamera is the entity id of the camera the take's most recent captured frame was drawn from, and stands as the source of the most recent take once that take has ended — it is read off the frame the engine drew, so it answers which view a film holds whether or not the take named a camera. recordingCadence is the clock the take stamps its frames from, "realtime" or "frame", and so what the tally below is a reading against. What the take produced reads off recordingFrames, recordingBytes (every byte the take has produced so far, climbing while it runs and ending equal to the size of the file), recordingSeconds (the timeline those frames cover), recordingAchievedFps (the rate they arrived at) and recordingRequestedFps (the rate asked for) — live while a take runs, and its final tally once it ends. On "realtime" the requested rate is a ceiling and an engine ticking under it reaches less; on "frame" every rendered frame is a slot of the recorded timeline, so recordingSeconds is that timeline's length and recordingAchievedFps is the rate it plays back at. recordingDurability is where the film in flight lands on the terms standing now — the report av.record returns as its third value, asked again for the take still running, so a caller polling status learns the play shadow is taking the film in time to keep it, a take that entered play after it opened included. It is present while a take names a destination.

Returns AvStatus — Encoder status table.

local s = av.status(); print(s.backend, s.hardware, s.recordingAudio)

globals/av/stop_live

av.stop_live() -> boolean

Stop any active live-stream session.

Returns boolean — True if a session was stopped, false if none was active.

av.stop_live()

globals/av/stop_recording

av.stop_recording(handle: string?) -> (boolean, string?)

Stop the active recording (or the one for the given promise handle). Returns true when a recording was armed at call time. A false return carries a second value naming how the most recent recording already ended (the bound it reached, or the failure that cut it short), and nil when no recording has run at all. The engine finalizes the take once the call returns. A take stopped before its first frame is captured runs on until that frame lands and ends holding it, and one whose first frame has not landed 120 frames after the stop ends with none. Wait for av.is_recording() to go false, then read what it produced off av.status().

Parameters

handle string (optional) — Promise handle of a specific recording (optional).

Returns (boolean, string?) — True if a recording was stopped. How the most recent recording ended, when nothing was armed.

local stopped, ended = av.stop_recording()

globals/await

await(promiseHandle: PromiseHandle | string): any

Yield the current coroutine until the given promise settles and return its value; a promise that has already settled returns at once, with no yield. A rejected promise raises its error at the call (a string at the caller's line, any other error value as itself), and handle:rejection() reads why it was rejected, { message, origin?, stack? }; a rejection is kept apart from any value. Pair with any FFI function that returns a promise handle (asset.load, http.get_json, scene.save, delay, ...). Every coroutine that awaits a promise receives its result: the ones waiting when it settles and any that await it later, for as long as the engine keeps the promise, which is while a PromiseHandle naming it is alive, a coroutine waits on it, or its handle has left the VM as data and it has not been delivered yet. A PromiseHandle reads as its id string, and the id names the same promise, so hold the handle, not only its id, to await it later. A PromiseHandle is a future, a value to store and await later: handle:await() is this same call, handle:state() reads pending, resolved or rejected without taking the value, handle:rejection() reads why it was rejected, handle:source() names the line that created it, and @builtin::modules.future makes futures of your own; Future<T> types one whose await delivers a T. An id the engine keeps no promise by, an extra argument, or a value that is neither a handle nor an id raises an error at the call naming the reason. Behaves exactly as task.await.

Parameters

promiseHandle PromiseHandle | string — promise handle returned by an async FFI call, or its id

Returns any — Whatever value the resolved promise carries (nil if the promise resolved without a value).

globals/base64/decode

base64.decode(text: string) -> (string?, string?)

Decode standard-alphabet base64 text back to the original binary string.

Parameters

text string — Base64 text to decode.

Returns (string?, string?) decoded bytes on success, or (nil, errmsg).

local bytes = base64.decode(encoded)

globals/base64/encode

base64.encode(bytes: buffer | string) -> string

Encode a binary string to standard-alphabet (padded) base64 text.

Parameters

bytes buffer | string — Binary bytes to encode.

Returns string — Base64 text.

local text = base64.encode(jpegBytes)

globals/batch

batch(fn)

Execute a function with batched mutations. Operations inside the function are collected and executed in order at the end, instead of immediately. Use for bulk operations where you want to control when mutations flush. Within the batch, operations still execute in order — spawn before component.add before impulse.

Parameters

fn function — Function containing batched operations

globals/blend/destroyLayout

blend.destroyLayout(handle: number) -> boolean

Drop the layout from the registry.

Parameters

handle number — Layout handle.

Returns boolean — True if the layout existed and was removed.

globals/blend/layout

blend.layout(slots: { BlendSlot }, totalStride: number?) -> number?

Register a record-stride layout. Each slot is { offset, stride, op } where op is "lerp" / "slerp" / "sum" / "step". Slerp slots must have stride 4. totalStride defaults to max(offset + stride) across slots; pass an explicit value when records contain padding past the last slot.

Parameters

slots { BlendSlot } — Array of slot tables.
totalStride number (optional) — Optional explicit record stride.

Returns number? — Layout handle, or nil.

local l = blend.layout({ { offset = 0, stride = 3, op = "lerp" } })

globals/blend/lerpInto

blend.lerpInto(outBuffer: Substrate.TypedBuffer, layout: number, aBuffer: Substrate.TypedBuffer, bBuffer: Substrate.TypedBuffer, t: number) -> boolean

Two-input crossfade shortcut. Equivalent to blend.weightedInto(out, layout, { {a, 1-t}, {b, t} }). Faster for the common A/B fade case because it skips the inputs-table walk.

Parameters

outBuffer Substrate.TypedBuffer — The buffer written into.
layout number — Layout handle.
aBuffer Substrate.TypedBuffer — The A side of the fade.
bBuffer Substrate.TypedBuffer — The B side of the fade.
t number — Crossfade weight on B (0..1).

Returns boolean — True on success.

globals/blend/weightedInto

blend.weightedInto(outBuffer: Substrate.TypedBuffer, layout: number, inputs: { BlendInput }) -> boolean

Combine N weighted input buffers into the output buffer using the layout's slot ops. The output buffer's length must be a whole multiple of layout.totalStride; every input buffer must be at least as long as the output. Returns false on any handle / size mismatch.

Parameters

outBuffer Substrate.TypedBuffer — The buffer written into.
layout number — Layout handle.
inputs { BlendInput } — Array of { buffer, weight }.

Returns boolean — True on success.

globals/bundle/update

bundle.update(entityId: string, bundleRef: BundleRef?) -> any

Re-compose a bundle from an entity's current hierarchy and write it back to the bundle's on-disk path. The VFS write triggers the engine's generic asset hot-reload pipeline, which fires onAssetReload(field) on every component subscribed to this bundle's guid via a declared asset field — those components reconcile per their own policy.

Parameters

entityId string — The entity whose hierarchy is captured into the bundle.
bundleRef BundleRef (optional) — Optional. When omitted, the ref is inferred from the entity's Asset.source field. When given, the explicit ref wins.

Returns any — True on success (forwarded from the bundle assetType's :update).

bundle.update(entityId)                     -- infer from Asset
bundle.update(entityId, { guid = "..." })   -- explicit ref

globals/camera/active

camera.active() -> string?

Entity id of the on-screen render camera this frame — whichever camera wins the viewport by priority (the editor fly-camera in edit mode, the gameplay camera in play). Render features, billboards, and input bases that must follow the human's on-screen view read this.

Returns string? — Entity id of the on-screen camera, or nil if none is active — including the frame after that camera's entity is despawned.

local camId = camera.active()

globals/camera/cut

camera.cut()

Declare that the camera on screen cuts: the next frame it draws stands somewhere it did not travel to. Motion vectors are the difference between where a surface projects now and where it projected on the camera's previous frame, and everything temporal reads that difference — the shutter reconstructs the frame by walking it, a temporal resolve reprojects its history along it. Across a cut that difference describes a displacement no surface made, so the frame is reconstructed from taps a whole screen away and belongs to neither shot. A declared cut leaves the camera with no previous frame for exactly one frame, which is the state its very first frame is already in, so every consumer reads zero motion across the cut. Declare it in the same step that places the camera at the new station; declaring it again before that frame draws still costs the one frame. Handing the viewport from one camera to another is already a cut without being declared one: the incoming camera stands where it always stood, and the engine performs the handover, so it is what states it.

camera.cut(); entity(camId).position = { 40, 6, -12 }

globals/camera/editor

camera.editor() -> string?

Entity id of the editor fly-camera (the EditorOnly authoring camera), or nil if the scene has none. This is the camera the editor viewport renders through, so it is the one a capture of the screen sees. Its pose is its entity transform: assign entity(id).position to move it and aim it with the camera toolbox's lookAt, which makes a screen capture repeatable instead of whatever pose the instance booted with.

Returns string? — Entity id of the editor camera, or nil.

local camId = camera.editor()
local id = camera.editor(); entity(id).position = { 12, 8, 12 }; tools.use("camera", "lookAt", id, { 0, 0, 0 })

globals/camera/editorOverride

camera.editorOverride() -> string?

Entity id currently overriding viewport selection, or nil when the viewport is decided by highest-priority-wins.

Returns string? — Entity id of the overriding camera, or nil.

local owner = camera.editorOverride()

globals/camera/get

camera.get(target: (string | EntityRef)) -> CameraReport?

One camera's report from the observation — the same record camera.list yields, for the camera the caller names. Takes an entity id, an entity name, or an entity proxy, the same way the camera tools do.

Parameters

target (string | EntityRef) — Entity id, entity name, or entity proxy of the camera to report on.

Returns CameraReport? — The report, or nil when nothing resolves or that camera has none.

local c = camera.get(camera.active()); print(c.frame.far, c.authored.far)
local c = camera.get("minimapCam"); print(c.rendering, c.reason)

globals/camera/list

camera.list() -> { CameraReport }

Every camera in the world as a compact row each, ordered the way the renderer resolves the on-screen camera: highest priority first. Reads the same observation camera.observe does, so a row can never disagree with the full report about whether a camera is enabled or which one drew.

Returns { CameraReport } — One row per camera entity, or an empty list before the first frame.

for _, c in ipairs(camera.list()) do print(c.name, c.enabled, c.rendering) end

globals/camera/main

camera.main() -> string?

Entity id of the main scene camera — the scene camera the viewport is drawn from, and while the editor fly-camera holds the screen, the scene camera that would take it. The gameplay/PlayerPrototype camera, an agent-placed scene camera, or a cutscene camera. Never the editor camera; nil if the scene has only the editor camera. It comes off the same selection the frame does, so writing a pose to it moves what is drawn whenever a scene camera is on screen. The scene camera with the highest authored priority takes it; cameras tied on priority settle on the order the frame visits them, so a scene that needs a specific camera — a prototype and the clone play makes of it both stand at 0 — states a distinct priority rather than resting on that order. For the camera drawn on screen whichever partition owns it, use camera.active().

Returns string? — Entity id of the main scene camera, or nil — including the frame after that camera's entity is despawned, before the scene elects another.

local camId = camera.main(); local cam = camId and entity(camId)

globals/camera/motionTally

camera.motionTally() -> { frames: number, withoutHistory: number }

Frames the camera on screen has drawn, and how many of them had no previous frame to difference their motion vectors against — its first frame, every declared cut, and every frame the viewport changes hands on. Both counts are monotonic across the session, so two readings either side of a run say what happened in between.

Returns { frames: number, withoutHistory: number } — { frames, withoutHistory }.

local before = camera.motionTally().withoutHistory

globals/camera/observe

camera.observe() -> CameraObservation?

Every camera in the world, for the frame that has just been drawn. lastRender retains the latest submitted frame, target and pending build count while the camera exists, including frames when it is disabled. engineFrame identifies the script frame (time.frameCount()) that drew. Answers "why is this camera not showing what I expect" in one call: rendering says whether each camera drew and reason names the single cause when it did not — "disabled", "entityInactive", "targetMissing", "noLayers", "outranked", "notDrawn", "paced" for a texture camera whose per-frame render waits for the next frame the frame pacer admits, or "awaitingPipelines" for a deferrable render queued behind pipeline builds. Explicit camera renders draw the requested frame and report its pending builds in lastRender. pipelinesPending counts the builds a camera is waiting on: what a held render is queued behind, and — for a camera that DID draw — what the frame it drew was drawn WITHOUT. It is absent once a camera's frame was drawn with everything the scene asked for. Each camera carries both projections: authored is what the Camera component holds and frame is what the renderer actually built, with mismatch naming every field the two disagree on — so a clip range or a lens the frame did not use is one field read. frame, viewProj, frustum and the cost numbers describe a camera that drew; every cost is for that one frame. One snapshot is published per drawn frame, from after the frame is drawn, so a read describes the last frame rather than the world at the instant of the call — a write and a read in one script step return the frame that ran before the write. Put a task.wait() between them to compare a camera either side of a change; frame counts the frames observed, so a poll can wait for it to advance.

Returns CameraObservation? — The observation, or nil before the first frame has been drawn.

local obs = camera.observe(); for _, c in ipairs(obs.cameras) do print(c.name, c.rendering, c.reason) end
local dark = {}; for _, c in ipairs(camera.observe().cameras) do if not c.rendering then table.insert(dark, c.name .. ": " .. tostring(c.reason)) end end

globals/camera/setEditorOverride

camera.setEditorOverride(entityId: string?)

Give one camera the viewport outright, or pass nil to clear it. While set, that camera IS the on-screen camera and priority is never consulted, so no authored priority can take the viewport from it — which is what makes an authoring camera safe to fly over a scene holding a camera at any priority. An override naming a camera that is despawned or disabled falls back to highest-priority-wins rather than blanking the screen. The override holds across a change of engine.mode: the flip rebuilds the scene and materialises its cameras again, and an override whose camera the rebuilt scene holds under the same entity id names it again once the rebuild has ended. One whose camera the rebuilt scene does not hold (a temporary camera, or one play replaces with a clone of its own) ends there: a warning notice and the engine log name the camera it held, and the wld.play / wld.edit result reports it under viewportOverride. An override on a camera that draws nothing in the new mode (an EditorOnly camera sits out play) stands, and the highest-priority camera holds the viewport until its camera draws again; viewportOverride.onScreen names the camera holding it.

Parameters

entityId string (optional) — Entity id of the camera to route the viewport to, or nil to clear.

camera.setEditorOverride(camera.editor())
camera.setEditorOverride(nil)

globals/camera/viewData

camera.viewData(target: ((string | EntityRef)?)?) -> CameraView?

Camera render data. Called with no argument it is the active viewport camera's data for this frame: world position, which projection it drew and the field describing that frame, viewport pixel size, the 6 world-space frustum planes (the same inward-pointing, normalized planes the renderer culls with), and the view-projection matrix. The camera state a render feature needs for camera-relative work — LOD selection, frustum culling, billboards. Render features also get it as ctx.camera. Called with an entity id it is that camera's data, read from the frame's camera observation, and carries the identity the bare form has no room for: which camera it describes, which frame it was built for, what it rendered into, and the render layers it resolved to.

Parameters

target ((string | EntityRef)?) (optional) — Entity id, name, or proxy of the camera to read, or nil for the viewport camera.

Returns CameraView? — The camera view data, or nil when that camera drew no frame.

local c = camera.viewData(); if c then print(c.position.x, c.fovY) end
local minimap = camera.viewData("minimapCam"); print(minimap.output.target, minimap.viewport.w)

globals/cameraInput/devices

cameraInput.devices() -> { CameraDevice }

Every capture device the platform offers: id is what cameraInput.open takes as device, name the label a person recognises, facing which way it points. Until camera access has been granted, id and name read as empty strings and facing as "unknown": the list is part of what the permission protects, so a first open takes the default camera or a facing, and the list names the devices after that.

Returns { CameraDevice } — Array of { id, name, facing, default }.

for _, d in ipairs(cameraInput.devices()) do print(d.name, d.facing) end

globals/cameraInput/feeds

cameraInput.feeds() -> { CameraFeedEntry }

Every open feed, with the owning context that opened it (scope, the same tag scopes.list() shows), whose end stops the feed.

Returns { CameraFeedEntry } — Array of { id, scope, state, device, texture }.

for _, f in ipairs(cameraInput.feeds()) do print(f.id, f.state, f.scope) end

globals/cameraInput/onDeviceChange

cameraInput.onDeviceChange(callback: (devices: { CameraDevice }) -> ()) -> () -> ()

Call callback with the new device list every time a camera is added or removed, or the platform names its devices after access is granted. Each listener is watched by a task of the context that registered it, so a listener a component registered ends when that component is removed, and every other listener goes on hearing changes.

Parameters

callback (devices: { CameraDevice }) -> () — Receives the device list, as cameraInput.devices() returns it.

Returns () -> () — A function that removes this listener.

local off = cameraInput.onDeviceChange(function(devices) print(#devices) end)

globals/cameraInput/open

cameraInput.open(opts: CameraOpenOpts?) -> (CameraFeed?, string?)

Open a camera as a feed. Returns the feed at once; it reads permissionPending while the platform asks for access, then running (or denied / failed with a reason). Wait with feed:awaitRunning().

feed.texture is a texture handle that shows the camera's current frame: put it on a surface as a session material's base_color_texture (renderer.material.create, worn with Model:applySessionMaterial) or in a slot of a material that exists (renderer.material.setTexture(key, "base_color_texture", feed.texture)), in UI as { type = "image", props = { src = feed.texture } }, and av.record films it like any other scene content.

device names a camera by the id or name cameraInput.devices() lists; facing asks for the "front" or "back" camera instead; with neither, the platform's default camera opens. width, height and fps are hints: the device delivers the nearest it can, and feed:status() reports what it delivers. A second feed on a camera already open shares it.

Returns nil and the reason for a request refused before any camera is touched: a hint outside its range, a facing other than front or back, a device the platform does not offer.

Parameters

opts CameraOpenOpts (optional) — { device, facing, width, height, fps }, every field optional.

Returns (CameraFeed?, string?) — The feed, or nil and the reason.

local feed = cameraInput.open({ facing = "front", width = 1280, height = 720 })
local screen = feed and renderer.material.create({ shader = "unlit", textures = { base_color_texture = feed.texture } }, "tv_screen")

globals/cameraInput/stopAll

cameraInput.stopAll() -> number

Stop every open feed and release every camera, whatever opened them.

Returns number — How many feeds stopped.

cameraInput.stopAll()

globals/channel/create

channel.create(opts: ChannelOpts) -> number?

Register a keyframe channel. times is the sorted keyframe time array; values is the packed value array (layout depends on interp); stride is the floats-per-sample width; interp is "step" | "linear" | "slerp" | "cubicHermite". Returns the channel handle, or nil on malformed input.

Parameters

opts ChannelOpts — { times, values, stride, interp }.

Returns number? — Channel handle, or nil.

local h = channel.create({ times = ts, values = vs, stride = 3, interp = "linear" })

globals/channel/destroy

channel.destroy(handle: number) -> boolean

Drop the channel from the registry.

Parameters

handle number — Channel handle.

Returns boolean — True if the channel existed and was removed.

globals/channel/sampleInto

channel.sampleInto(ch: number, time: number, buf: Substrate.TypedBuffer, offset: number) -> boolean

Sample the channel at time and write stride floats into the buffer starting at f32 index offset. Returns false on unknown handle, layout mismatch, or out-of-bounds; the buffer is unchanged on failure.

Parameters

ch number — Channel handle.
time number — Sample time in seconds.
buf Substrate.TypedBuffer — The buffer written into.
offset number — Starting f32 index in the buffer.

Returns boolean — True on success.

globals/channel/sampleManyInto

channel.sampleManyInto(ch: number, time: number, buf: Substrate.TypedBuffer, offsets: { number }) -> boolean

Sample once, blit the result into every position in offsets. Saves the per-offset binary search when one channel feeds many bones / particles / parameters.

Parameters

ch number — Channel handle.
time number — Sample time in seconds.
buf Substrate.TypedBuffer — The buffer written into.
offsets { number } — Array of f32 indices.

Returns boolean — True on success.

globals/channel/sampleQuat

channel.sampleQuat(ch: number, time: number) -> (number?, number?, number?, number?)

Convenience accessor for stride-4 quaternion channels.

Parameters

ch number — Channel handle.
time number — Sample time.

Returns (number?, number?, number?, number?) — (x, y, z, w) or nil.

globals/channel/sampleVec3

channel.sampleVec3(ch: number, time: number) -> (number?, number?, number?)

Convenience accessor for stride-3 channels. Returns the three components as multiret, or nil if the channel is unknown / has a different stride.

Parameters

ch number — Channel handle.
time number — Sample time.

Returns (number?, number?, number?) — (x, y, z) or nil.

local x, y, z = channel.sampleVec3(h, t)

globals/color/coerce

color.coerce(value: any?) -> Color?

Read a value written in any of the shapes a colour is authored in — a hex string, an {r=,g=,b=,a=} map, or an {r,g,b,a} array — as an sRGB color table. Returns nil when the value does not describe a colour, so a caller can name the value it was handed instead of substituting one. Channels absent from a map or array read as 0; alpha absent reads as 1.

Parameters

value any (optional) — Value to read as a colour.

Returns Color? — sRGB color table, or nil when value is not a colour.

local c = color.coerce("#5a5a62") or color.coerce({ 0.2, 0.7, 0.2 })

globals/color/complementary

color.complementary(c: Color) -> Color

Complementary color — rotate hue 180° in Oklch space.

Parameters

c Color — Input color.

Returns Color — Complementary sRGB color.

local accent = color.complementary(primary)

globals/color/darken

color.darken(c: Color, amount: number) -> Color

Decrease the lightness of a color in Oklch perceptual space.

Parameters

c Color — Input color.
amount number — Lightness decrease 0-1.

Returns Color — Darkened sRGB color.

local pressed = color.darken(base, 0.1)

globals/color/desaturate

color.desaturate(c: Color, amount: number) -> Color

Decrease the chroma (saturation) of a color in Oklch space.

Parameters

c Color — Input color.
amount number — Chroma decrease (typically 0-0.2).

Returns Color — Less saturated sRGB color.

local muted = color.desaturate(base, 0.05)

globals/color/hex

color.hex(hexString: string) -> Color?

Parse a hex color string into an sRGB color table. Accepts 3, 4, 6, or 8 hex digits with or without a leading # (e.g. "#f00", "f00f", "#ff0000", "ff000080"). Returns nil on parse failure.

Parameters

hexString string — Hex color string.

Returns Color? — sRGB color table or nil.

local fromCss = color.hex("#ff8800")

globals/color/hsl

color.hsl(h: number, s: number, l: number) -> Color

Build a color from HSL (h: 0-360, s: 0-1, l: 0-1). Returned as sRGB.

Parameters

h number — Hue (degrees, 0-360).
s number — Saturation (0-1).
l number — Lightness (0-1).

Returns Color — sRGB color table { r, g, b, a = 1 }.

local teal = color.hsl(180, 0.5, 0.5)

globals/color/hsla

color.hsla(h: number, s: number, l: number, a: number) -> Color

Build a color from HSLA, returned as sRGB.

Parameters

h number — Hue (0-360).
s number — Saturation (0-1).
l number — Lightness (0-1).
a number — Alpha (0-1).

Returns Color — sRGB color table { r, g, b, a }.

local fadedTeal = color.hsla(180, 0.5, 0.5, 0.3)

globals/color/hsv

color.hsv(h: number, s: number, v: number) -> Color

Build a color from HSV (h: 0-360, s: 0-1, v: 0-1).

Parameters

h number — Hue (0-360).
s number — Saturation (0-1).
v number — Value / brightness (0-1).

Returns Color — sRGB color table { r, g, b, a = 1 }.

local primary = color.hsv(220, 0.7, 0.9)

globals/color/lighten

color.lighten(c: Color, amount: number) -> Color

Increase the lightness of a color in Oklch perceptual space.

Parameters

c Color — Input color.
amount number — Lightness increase 0-1.

Returns Color — Lightened sRGB color.

local hover = color.lighten(base, 0.1)

globals/color/linear

color.linear(r: number, g: number, b: number, a: number?) -> Color

Build a color from linear RGB values (not gamma-corrected), output converted to sRGB. Useful for GPU-correct blending. Alpha defaults to 1.

Parameters

r number — Linear red (0-1).
g number — Linear green (0-1).
b number — Linear blue (0-1).
a number (optional) — Alpha (0-1, default 1).

Returns Color — sRGB color table { r, g, b, a }.

local gpuBlue = color.linear(0.0, 0.0, 1.0)

globals/color/mix

color.mix(c1: Color, c2: Color, t: number) -> Color

Perceptually blend two colors in Oklch space — better than RGB mixing for gradients.

Parameters

c1 Color — First color.
c2 Color — Second color.
t number — Blend factor 0-1 (0 = c1, 1 = c2).

Returns Color — Blended sRGB color.

local mid = color.mix(color.rgb(255, 0, 0), color.rgb(0, 0, 255), 0.5)

globals/color/mixRgb

color.mixRgb(c1: Color, c2: Color, t: number) -> Color

Linearly blend two colors in sRGB space — simple, but not perceptually uniform. Prefer color.mix for natural gradients.

Parameters

c1 Color — First color.
c2 Color — Second color.
t number — Blend factor 0-1.

Returns Color — Blended sRGB color.

local plain = color.mixRgb(a, b, 0.5)

globals/color/oklch

color.oklch(l: number, c: number, h: number) -> Color

Build a color from Oklch perceptual color space (l: 0-1, c: 0-0.4, h: 0-360). Ideal for perceptually uniform gradients and color manipulation.

Parameters

l number — Lightness (0-1).
c number — Chroma / saturation (0-0.4).
h number — Hue (0-360).

Returns Color — sRGB color table { r, g, b, a = 1 }.

local accent = color.oklch(0.7, 0.15, 30)

globals/color/rgb

color.rgb(r: number, g: number, b: number) -> Color

Build an sRGB color from CSS-style 0-255 RGB channels. Alpha defaults to 1. Channels are normalised to 0-1 on the way out so the result composes with every other color helper.

Parameters

r number — Red channel (0-255).
g number — Green channel (0-255).
b number — Blue channel (0-255).

Returns Color — sRGB color table { r, g, b, a = 1 }, normalised to 0-1.

local red = color.rgb(255, 0, 0)

globals/color/rgba

color.rgba(r: number, g: number, b: number, a: number) -> Color

Build an sRGB color from CSS-style 0-255 RGB channels with explicit alpha. RGB are normalised to 0-1; alpha is taken as-is in the 0-1 range.

Parameters

r number — Red channel (0-255).
g number — Green channel (0-255).
b number — Blue channel (0-255).
a number — Alpha (0-1).

Returns Color — sRGB color table { r, g, b, a }.

local halfRed = color.rgba(255, 0, 0, 0.5)

globals/color/rotateHue

color.rotateHue(c: Color, degrees: number) -> Color

Rotate the hue of a color by a given number of degrees in Oklch space.

Parameters

c Color — Input color.
degrees number — Hue rotation (positive or negative).

Returns Color — Hue-rotated sRGB color.

local triadic = color.rotateHue(base, 120)

globals/color/saturate

color.saturate(c: Color, amount: number) -> Color

Increase the chroma (saturation) of a color in Oklch space.

Parameters

c Color — Input color.
amount number — Chroma increase (typically 0-0.2).

Returns Color — More saturated sRGB color.

local pop = color.saturate(base, 0.05)

globals/color/toHex

color.toHex(c: Color) -> string

Convert a color to a hex string. Returns "#rrggbb" or "#rrggbbaa" if alpha is not 1.

Parameters

c Color — Input color.

Returns string — Hex color string.

print(color.toHex(color.rgb(255, 136, 0))) -- "#ff8800"

globals/color/toHsl

color.toHsl(c: Color) -> HslColor

Convert a color to HSL.

Parameters

c Color — Input color.

Returns HslColor — HSL color table { h, s, l, a } (h: 0-360, s/l: 0-1).

local hsl = color.toHsl(base)

globals/color/toLinear

color.toLinear(c: Color) -> Color

Convert a color from sRGB to linear RGB space — useful for GPU calculations that need linear-space values.

Parameters

c Color — Input sRGB color.

Returns Color — Linear RGB color table.

local gpu = color.toLinear(base)

globals/color/toOklch

color.toOklch(c: Color) -> OklchColor

Convert a color to Oklch perceptual color space.

Parameters

c Color — Input color.

Returns OklchColor — Oklch color table { l, c, h, a } (l: 0-1, c: 0-0.4, h: 0-360).

local okl = color.toOklch(base)

globals/color/withAlpha

color.withAlpha(c: Color, a: number) -> Color

Return a copy of a color with a different alpha value.

Parameters

c Color — Input color.
a number — New alpha (0-1).

Returns Color with modified alpha.

local ghost = color.withAlpha(base, 0.3)

globals/compute/absentReasons

compute.absentReasons() -> { string }

Every reason compute.diagnose reports, sorted. resident is the one that means the resource is there.

Returns { string } — The closed set, as strings.

for _, r in ipairs(compute.absentReasons()) do print(r) end

globals/compute/beginBvh

compute.beginBvh(instances: { any }, opts: { [string]: any }?) -> (number?, string?)

Start the build compute.buildBvh runs, without running any of it. Takes the same instances and options and reports the same non-resident guids, and returns an id compute.stepBvh advances a bounded slice at a time and compute.finishBvh collects. Each mesh the instances name is copied as this is called — once per guid however many instances share it — so the CPU mesh may be unloaded on the next line and the build still finishes on the copy it holds. compute.buildBvhSliced is the whole loop as one call.

Parameters

instances { any } — Array of { guid, transform, attributes? } mesh instances.
opts { [string]: any } (optional) — Optional { maxLeaf? } — max triangles per leaf.

Returns (number?, string?) — The build id, or (nil, err) naming any non-resident guid.

local id = compute.beginBvh(gather.instances)

globals/compute/buildBvh

compute.buildBvh(instances: { any }, opts: { [string]: any }?) -> (any, string?)

Build a bounding-volume hierarchy over the world-space triangles of a set of mesh instances and upload it as two named compute buffers — geometry never passes through the scripting heap. Each instance is { guid, transform, attributes? }: guid names a mesh resident in the meshcpu store (materialise with ref:load() / meshcpu.load), transform is 16 numbers, row-major, translation in slots 4/8/12, and attributes is up to 40 floats stamped onto every triangle of that instance (surface colors, material ids, physics tags — whatever the consuming shader wants per-surface). Triangles pack 18 vec4 each (v0/v1/v2, n0/n1/n2, uvs, then 10 attribute vec4s — float slots 32.. carry the instance attributes, zero when absent) in BVH leaf order; nodes 2 vec4 each (min + first-or-left, max + leaf-tagged count-or-right). Consumers: GI baking, ray-traced passes, GPU picking, navmesh and SDF generation.

Parameters

instances { any } — Array of { guid, transform, attributes? } mesh instances.
opts { [string]: any } (optional) — Optional { maxLeaf? } — max triangles per leaf.

Returns (any, string?) — { nodes, tris, nodeCount, triCount } — nodes and tris are buffer handles the caller owns, passed to a dispatch like any other and destroyed when the hierarchy is done with. Or (nil, err) naming any non-resident guid.

local bvh = compute.buildBvh(instances)
asset.resolve("raytrace", "computeShader"):dispatch({ buffers = { bvh.nodes, bvh.tris }, workgroups = { 64, 1, 1 } })

globals/compute/buildBvhSliced

compute.buildBvhSliced(instances: { any }, opts: { [string]: any }?) -> (any, string?)

The hierarchy compute.buildBvh builds, spread over as many frames as it takes: a slice of the build per frame, so a scene's triangle count costs the frame loop budgetMs at a time instead of the whole build at once. Yields, so it is called from a task. The result is the same pair of buffers and the same counts compute.buildBvh returns.

Parameters

instances { any } — Array of { guid, transform, attributes? } mesh instances.
opts { [string]: any } (optional) — Optional { maxLeaf?, budgetMs? } — max triangles per leaf, and the wall time one frame may spend on the build (default 4 ms).

Returns (any, string?) — { nodes, tris, nodeCount, triCount }, or (nil, err).

local built, err = compute.buildBvhSliced(gather.instances, { budgetMs = 4 })

globals/compute/bvhBuilds

compute.bvhBuilds() -> { any }

What the builds started by compute.beginBvh and not yet finished are costing, oldest id first. Each row is { id, phase, triangles, units, slices, cpuMs, uploadedBytes }: phase is "gather", "build", "serialize", "upload" or "ready", triangles how many have been gathered, units the work units run, slices the compute.stepBvh calls they ran in, cpuMs the wall time spent inside those calls, and uploadedBytes how much of the hierarchy has reached the GPU.

Returns { any } — Array of build rows.

print(#compute.bvhBuilds(), "hierarchies in flight")

globals/compute/cancelBvh

compute.cancelBvh(id: number) -> boolean

Drop a build along with the triangles it has gathered.

Parameters

id number — Build id from compute.beginBvh.

Returns boolean — True when the id named a build.

compute.cancelBvh(id)

globals/compute/compile

compute.compile(nameOrHandle: string | { [string]: any } | AssetRef, opts: { [string]: any }?) -> (boolean, string?)

Compile a compute shader from inline WGSL + a declarative binding schema — the same codegen a .computeShader asset uses. The engine generates the @group/@binding declarations from bindings/params, so the source writes only @compute fn main. Symmetric with registerShader, but with zero-scaffolding bindings (incl. textures, samplers, storage textures and a params uniform). For asset-backed shaders prefer authoring a .computeShader (compiled automatically); use this for dynamic/generated compute shaders.

Parameters

nameOrHandle string | { [string]: any } | AssetRef — Shader identity to register under (string or asset handle).
opts { [string]: any } (optional) — { source, entryPoint?, bindings, params? } — bindings is an ordered list of { name, kind, access?, element?, format?, array?, sample? }.

Returns (boolean, string?) — True when the program is queued to register; false with the reason when the schema or the module is one the device would refuse.

compute.compile("my_sim", { source = WGSL, bindings = { { name = "data", kind = "buffer", access = "read_write", element = "f32" } } })

globals/compute/compileByName

compute.compileByName(ref: string | { [string]: any } | AssetRef)

Optional explicit pre-warm for a .computeShader asset (idempotent — fingerprint-guarded). NORMALLY UNNECESSARY: compute.dispatch / dispatchEx auto-compile a .computeShader on first use. Reach for this only to avoid the one-frame first-dispatch warm-up in a latency-critical spot. Accepts an identity/guid string or a resolved asset handle (its .identity is used).

Parameters

ref string | { [string]: any } | AssetRef — A .computeShader identity/guid string, or a resolved asset handle.

compute.compileByName("@builtin::shaders.compute_double")

globals/compute/copyBufferToTexture

compute.copyBufferToTexture(bufferName: string, textureKey: string, width: number, height: number, format: string?, rowBytes: number?) -> boolean

Copy a compute buffer into a cached GPU texture under textureKey, staying on the GPU. The path for an image a compute pass produced: the buffer holds tightly-packed rows in the format's texel layout, and the result is an ordinary cached texture — sample it from a material, or pack it into the shared feature-texture array. The copy reads rows 256-byte aligned: rows packed tight must be a multiple of 256 bytes (at rgba16f, any width that is a multiple of 32), and any other width is laid out with its rows padded to that alignment and rowBytes saying how far apart they are.

Parameters

bufferName string — Source compute buffer.
textureKey string — Cache key to register the texture under.
width number — Texture width in texels.
height number — Texture height in texels.
format string (optional) — Texel format: "rgba16f" (default), "rgba32f", "rgba8".
rowBytes number (optional) — Bytes from one row's start to the next when rows are padded past width texels — a multiple of 256 at least a row's texels long.

Returns boolean — True when the copy was queued.

compute.copyBufferToTexture("gi_resolved", "lm_wall", 128, 128)
compute.copyBufferToTexture("heights_packed", "heights", 300, 300, "rgba16f", 2560)

globals/compute/createBuffer

compute.createBuffer(name: string, opts: { [string]: any }) -> boolean

Allocate a buffer under name, sized in bytes.

Parameters

name string — The name a dispatch binds it by.
opts { [string]: any } — { size, readback? } — size in bytes.

Returns boolean — True once allocated.

globals/compute/createSampler

compute.createSampler(name: string, opts: { [string]: any }?) -> boolean

Create a named GPU sampler. opts: filter/wrap settings.

Parameters

name string
opts { [string]: any } (optional)

Returns boolean

globals/compute/createStorageTexture2D

compute.createStorageTexture2D(name: string, opts: { [string]: any }) -> boolean

Create a 2D storage texture (compute-writable render target). opts: { width, height, format? }.

Parameters

name string
opts { [string]: any }

Returns boolean

globals/compute/createTexture3D

compute.createTexture3D(name: string, opts: { [string]: any }) -> boolean

Create a 3D texture volume. opts: { width, height, depth, format?, storage? }.

Parameters

name string — Unique volume name.
opts { [string]: any } — Dimensions + format (r8/r16f/r32f/rgba8/rgba16f/rgba32f).

Returns boolean — True on success (mutation queued).

globals/compute/createTextureHistory

compute.createTextureHistory(name: string, opts: { [string]: any }) -> boolean

Create a temporal history buffer (ping-pong textures) for a target. opts: { width, height, format? }.

Parameters

name string
opts { [string]: any }

Returns boolean

globals/compute/destroyBuffer

compute.destroyBuffer(name: string) -> boolean

Release the buffer allocated under name.

Parameters

name string — The name it was created under.

Returns boolean — True if a buffer under that name was released.

globals/compute/destroySampler

compute.destroySampler(name: string) -> boolean

Release a named sampler created by compute.createSampler and free it. The counterpart to that call, alongside destroyBuffer, destroyTexture, destroyTexture3D, destroyStorageTexture2D and destroyTextureHistory. The manager's own defaults (linear_clamp, linear_repeat, nearest_clamp) are kept for the session, since a compute pass binds them by name.

Parameters

name string — Sampler name.

Returns boolean — True when the release was queued.

compute.createSampler("soft", { filter = true }); compute.destroySampler("soft")

globals/compute/destroyShader

compute.destroyShader(name: string) -> boolean

Destroy a named compute shader pipeline.

Parameters

name string — Shader name.

Returns boolean — True on success.

globals/compute/destroyShaderEx

compute.destroyShaderEx(name: string) -> boolean

Destroy a shader registered via registerShaderEx.

Parameters

name string

Returns boolean

globals/compute/destroyStorageTexture2D

compute.destroyStorageTexture2D(name: string) -> boolean

Destroy a named 2D storage texture.

Parameters

name string

Returns boolean

globals/compute/destroyTexture

compute.destroyTexture(textureKey: string) -> boolean

Release the cached GPU texture copyBufferToTexture registered under textureKey, freeing its memory. Call it once the image is no longer sampled. Writing the same key again replaces the texture, so a key you keep re-using holds one allocation.

Parameters

textureKey string — Cache key the texture was registered under.

Returns boolean — True when the release was queued.

compute.destroyTexture("lm_wall")

globals/compute/destroyTexture3D

compute.destroyTexture3D(name: string) -> boolean

Destroy a named 3D volume and free its GPU memory.

Parameters

name string

Returns boolean

globals/compute/destroyTextureHistory

compute.destroyTextureHistory(name: string) -> boolean

Destroy a named texture-history buffer.

Parameters

name string

Returns boolean

globals/compute/diagnose

compute.diagnose(key: string) -> { [string]: any }

Whether a resource is filed under key right now, and when none is, which state the inventory says the key is in. A key out of a dispatch failure resolves here; a mistyped one reports why it does not.

Parameters

key string — The resource key, verbatim.

Returns { [string]: any } — { key, exists, reason, resource?, current? }. resource is the row when one is filed under the key. reason is one of compute.absentReasons(). current names the live key when the owner holds a resource under the same name at a different serial.

local d = compute.diagnose(key)
if not d.exists then print(d.reason, d.current) end

globals/compute/dispatch

compute.dispatch(shaderNameOrHandle: string | { [string]: any } | AssetRef, opts: DispatchOpts) -> boolean

Dispatch a compute shader with bound buffers. Accepts a shader name string or an asset handle from asset.load().

Parameters

shaderNameOrHandle string | { [string]: any } | AssetRef — Shader name or asset handle.
opts DispatchOpts — { buffers, workgroups }. A zero in any workgroup dimension is refused and recorded as a dispatch failure — clamp a computed count with math.max(1, math.ceil(n / 64)).

Returns boolean — True once the dispatch is queued. What the engine then did with it is in compute.failing().

compute.dispatch("blur", { buffers = { "src", "dst" }, workgroups = { 8, 8 } })

globals/compute/dispatchEx

compute.dispatchEx(shaderNameOrHandle: string | { [string]: any } | AssetRef, opts: { [string]: any }) -> boolean

Dispatch a compute shader with extended texture/storage/sampler bindings. Asset-backed .computeShaders resolve to their stable guid (collision-safe, lazily compiled on first dispatch); raw registerShaderEx names pass through. resources covers the bindings the shader DECLARES. A params: block's uniform is engine-owned — the compile creates and packs it, setParam writes it, and the dispatch binds it — so it takes no entry here.

Parameters

shaderNameOrHandle string | { [string]: any } | AssetRef — Shader name or asset handle.
opts { [string]: any } — { resources, workgroups } — each resource is { kind, name }. A zero in any workgroup dimension is refused and recorded as a dispatch failure — clamp a computed count with math.max(1, math.ceil(n / 64)).

Returns boolean — True once the dispatch is queued. What the engine then did with it is in compute.failing().

globals/compute/dispatchOnVertices

compute.dispatchOnVertices(shaderNameOrHandle: string | { [string]: any } | AssetRef, opts: DispatchOnVerticesOpts) -> boolean

Dispatch a compute shader with a model's vertex buffer bound at binding 0 (read_write). Use to mutate vertex positions directly. Asset-backed .computeShaders resolve to their stable guid (collision-safe, lazily compiled on first dispatch); raw registerShader names pass through.

Parameters

shaderNameOrHandle string | { [string]: any } | AssetRef — Shader name or asset handle.
opts DispatchOnVerticesOpts — { model, buffers?, workgroups }. A zero in any workgroup dimension is refused and recorded as a dispatch failure — clamp a computed count with math.max(1, math.ceil(n / 64)).

Returns boolean — True once the dispatch is queued. What the engine then did with it is in compute.failing().

compute.dispatchOnVertices("flatten", { model = "@user/mesh", workgroups = { 64 } })

globals/compute/failing

compute.failing() -> { { [string]: any } }

Every compute dispatch whose most recent run FAILED, one record per (shader, target) pair. A dispatch is recorded into a command encoder frames after the call that asked for it returned, so a pass that stops running reports here rather than through that call's return value: each record carries the shader key, the target it writes (a mesh guid for a dispatch over vertices, the buffers it bound for one that writes only those), how many dispatches and failures it has had, and lastError. An empty result means every dispatch the engine has been given is running.

Returns { { [string]: any } } — Array of { shader, target, dispatches, failures, ok, lastFrame, lastFailedFrame?, lastError? }.

for _, d in ipairs(compute.failing()) do print(d.shader, d.lastError) end

globals/compute/finishBvh

compute.finishBvh(id: number) -> (any, string?)

Hand over a finished build's hierarchy as the same two buffers compute.buildBvh returns, and release the build. The slices put every byte of it on the GPU as they ran, so this costs the frame it is called in the handover and nothing of the scene.

Parameters

id number — Build id from compute.beginBvh, stepped until "ready".

Returns (any, string?) — { nodes, tris, nodeCount, triCount }, or (nil, err) when the id names no build or the build still has work left.

local built = compute.finishBvh(id)

globals/compute/getReadbackResult

compute.getReadbackResult(resultKey: string) -> { number }?

Poll for a completed read-back and return its bytes as a 1-indexed array of f32 values, nil if pending. The f32 reinterpretation applies to whatever the buffer holds: bytes written as u32 1, 2, 3, 4 read back here as 1.4e-45, 2.8e-45, 4.2e-45, 5.6e-45 — use getReadbackResultU32() for those, or getReadbackResultBytes() for a buffer the rest of the buffer surface accepts. Result is consumed on retrieval, and polling a key that was never issued raises rather than reading as forever-pending.

Parameters

resultKey string — Key returned by readBuffer().

Returns { number }? — 1-indexed array of f32 values, or nil if not ready.

local floats = compute.getReadbackResult(key)

globals/compute/getReadbackResultBytes

compute.getReadbackResultBytes(resultKey: string) -> buffer?

Poll for a completed read-back and get its raw bytes as a buffer, copied once. The read counterpart of writeBufferBytes: read values out with buffer.readf32 / buffer.readu32, or hand the buffer straight to writeBuffer — a payload that stays packed never becomes a table. Result is consumed on retrieval.

Parameters

resultKey string — Key returned by readBuffer().

Returns buffer? — The read-back's bytes, or nil if not ready.

local bytes = compute.getReadbackResultBytes(key)
local firstTexel = buffer.readf32(bytes, 0)

globals/compute/getReadbackResultU32

compute.getReadbackResultU32(resultKey: string) -> { number }?

Poll for a completed read-back interpreting bytes as u32. Returns array of integer values if ready, nil if pending.

Parameters

resultKey string — Key returned by readBuffer().

Returns { number }? — Array of u32 values, or nil if not ready.

globals/compute/isReadbackReady

compute.isReadbackReady(resultKey: string) -> boolean

Check if a readback result is available without consuming it. Raises for a key this engine never issued, or whose result was already drained — nil/false already means "still in flight", so a mistyped key reports itself instead of polling forever. Use readbackState() to test that case without raising.

Parameters

resultKey string — Key returned by readBuffer().

Returns boolean — True if the result is ready.

globals/compute/markBaseline

compute.markBaseline() -> number

Record which compute resources are resident right now, and make that the point compute.resetToBaseline returns to.

A caller that runs work in someone else's engine — a sweep of test suites sharing one, an editor tool driving a session — allocates as it goes: a pass builds the grid its quality asks for and keeps it, a shader files the uniform it was compiled with. Kept is right for a frame that draws again and wrong for the next caller, which then reads an inventory it did not make. This is how a caller says where its own allocations start.

Returns number — The mark's generation, counting from 1 — what compute.resetToBaseline takes to restore to this mark.

local mark = compute.markBaseline() -- then run, then compute.resetToBaseline(mark)

globals/compute/observe

compute.observe() -> { [string]: any }

Every GPU resource the compute subsystem is holding right now — its storage and uniform buffers, its 3D textures, its 2D storage targets, its history pairs and its samplers — with what each one costs and which shader asked for it. This is the call to reach for when compute is holding memory and you do not know what, or when a key out of a dispatch failure needs matching against what exists.

Returns { [string]: any } — { published, generation, resources, totals }. Each row of resources carries key, kind (buffer / uniformBuffer / texture3d / storageTexture2d / textureHistory / sampler), owner ({ shader, name, serial }, read off the key), bytes, format, width, height, depth, usage (the bits it was created with — storage, copySrc, copyDst, vertex, index, indirect, uniform, sampled, sampler) and createdFrame. totals is { count, bytes, byKind }, what the rows sum to — and totals.bytes is the compute figure of renderer.gpuMemory(), read off the same registries. published is false when no renderer has published a reading yet, which is the engine saying it cannot answer rather than answering with nothing. The reading is the one the renderer published, republished on a frame where a registry gained or lost an entry: a resource created earlier in this same script is in the next reading, so wait a frame before asking about it, and generation moves when it arrives.

local r = compute.observe()
print(r.totals.count, r.totals.bytes)
for _, res in ipairs(r.resources) do print(res.key, res.kind, res.bytes) end

globals/compute/program/compile

compute.program.compile(key: string, spec: { [string]: any }) -> (boolean, string?)

Register a compiled program under key from WGSL plus a declared binding schema. The engine generates the @group/@binding declarations from the schema, expands #includes, naga-validates, and registers the result.

Parameters

key string — The key to register under.
spec { [string]: any } — { source, entryPoint?, bindings, params, ranges? } — the parsed schema. A buffer binding stating range = true (or every array buffer binding, with ranges = true) takes a range of a buffer: a dispatch binds it with zero_base_NAME / zero_len_NAME in its params, and the body's NAME[i] and arrayLength(&NAME) address that range. One that states no range binds the whole buffer.

Returns (boolean, string?) — True when the program is queued to register; false with the reason when the schema or the module is one the device would refuse.

globals/compute/program/destroy

compute.program.destroy(key: string) -> boolean

Release the program registered under key.

Parameters

key string — The program's key.

Returns boolean — True on success.

globals/compute/program/dispatch

compute.program.dispatch(key: string, opts: { [string]: any }) -> boolean

Dispatch the program under key with one buffer per declared storage binding, in declaration order.

Parameters

key string — The program's key.
opts { [string]: any } — { buffers, workgroups, params? }. params gives values of the program's params: uniform, by name, to this dispatch alone: two dispatches of one program in a frame each read their own, and a value left out reads what the program holds. A zero in any workgroup dimension is refused and recorded as a dispatch failure — clamp a computed count with math.max(1, math.ceil(n / 64)).

Returns boolean — True once the dispatch is queued. What the engine then did with it is in compute.program.status(key).

globals/compute/program/dispatchEx

compute.program.dispatchEx(key: string, opts: { [string]: any }) -> boolean

Dispatch the program under key with explicit resources — one { kind, name } per declared binding, in declaration order.

Parameters

key string — The program's key.
opts { [string]: any } — { resources, workgroups }. A zero in any workgroup dimension is refused and recorded as a dispatch failure — clamp a computed count with math.max(1, math.ceil(n / 64)).

Returns boolean — True once the dispatch is queued. What the engine then did with it is in compute.program.status(key).

globals/compute/program/dispatchOnVertices

compute.program.dispatchOnVertices(key: string, opts: { [string]: any }) -> boolean

Dispatch the program under key over a mesh's vertices. The mesh opts.model names fills the shader's vertices binding, and opts.buffers fills the remaining storage bindings.

Parameters

key string — The program's key.
opts { [string]: any } — { model, buffers?, workgroups, params? }. params gives values of the program's params: uniform to this dispatch alone. A zero in any workgroup dimension is refused and recorded as a dispatch failure — clamp a computed count with math.max(1, math.ceil(n / 64)).

Returns boolean — True once the dispatch is queued. What the engine then did with it is in compute.program.status(key).

globals/compute/program/dropList

compute.program.dropList(key: string) -> boolean

Forget the compute list under key.

Parameters

key string — The list's key.

Returns boolean — True once the removal is queued.

globals/compute/program/forPlugin

compute.program.forPlugin(key: string, bindings: number | { string }) -> any

The program under key as a handle a script passes a plugin in a call, for it to dispatch over the GPU buffers and textures it was handed. bindings is how many storage buffers the shader binds, or the kind of each binding in declaration order as bindings.yaml spells them (buffer, texture3d, storage3d, storage2d).

Parameters

key string — The program's key.
bindings number | { string } — A count of buffers, or each binding's kind.

Returns any — A handle for a plugin call.

globals/compute/program/list

compute.program.list(key: string, entries: { { [string]: any } }) -> boolean

Register the compute list under key, replacing one already there: an ordered run of dispatches and buffer writes that runList queues in one call, each value a run supplies named as a slot.

Parameters

key string — The key to register the list under.
entries { { [string]: any } } — The dispatches and writes, in the order they run. A dispatch is { program, buffers, workgroups, params?, slots?, when? }, slots mapping a param name to the slot that gives it its value. A write is { write = bufferName, offset?, values, unsigned?, slots?, when? }: offset in 32-bit words, values written as f32 (u32 when unsigned), slots mapping a value's place, from 1, to the slot that gives it. when names a slot: a run giving it 0 leaves the entry out.

Returns boolean — True once the registration is queued.

globals/compute/program/record

compute.program.record(fn: ({ { [string]: any } }) -> ()) -> { { [string]: any } }

Run fn and answer every compute.program.dispatch it made, in order, as the entries of a compute list: { program, buffers, workgroups, params } each. The dispatches are queued as they always are; the answer is what compute.program.list takes to replay them. fn is handed the entries as they are made, so it can add slots to one it just dispatched.

Parameters

fn ({ { [string]: any } }) -> () — The work to run, given the entries recorded so far.

Returns { { [string]: any } } — The dispatches fn made, as list entries.

globals/compute/program/runList

compute.program.runList(key: string, slots: { [string]: number }?) -> boolean

Queue the compute list under key: every entry in order, each value an entry names a slot for taken from slots, every other value as the list holds it.

Parameters

key string — The list's key.
slots { [string]: number } (optional) — The value of each slot for this run, by name.

Returns boolean — True once the run is queued.

globals/compute/program/setParam

compute.program.setParam(key: string, prop: string, value: number) -> boolean

Write one scalar of the program's params: uniform. A value set before the program's first compile is the value it starts with.

Parameters

key string — The program's key.
prop string — Parameter name as declared.
value number — New scalar value.

Returns boolean — True on success.

globals/compute/program/status

compute.program.status(key: string) -> { { [string]: any } }

What the engine did with the dispatches of the program under key, one record per target.

Parameters

key string — The program's key.

Returns { { [string]: any } } — Array of dispatch records, most recently dispatched first.

globals/compute/programState

compute.programState(ref: string | { [string]: any } | AssetRef) -> (string, string?)

Where a shader's compiled program stands. A registration is queued from script and the pipeline is built on the render side frames later, so the call that asked for the compile cannot say whether it produced a program: "absent" (the engine holds nothing under this key and nothing is in flight — never asked for, or released), "pending" (asked for, not on the device yet — a recompile of a resident program reads pending too, because what it produces is a different program from the one bound now), "ready" (compiled and resident, so a dispatch binds it), or "failed" (the most recent registration produced no program), returned with the reason as a second value. A .computeShader whose bindings.yaml its compile refuses reads "failed", with the reason naming the file and the line, because no program is compiled from the file as it stands. Wait for "ready" before a dispatch whose result is read back, rather than for a count of frames.

Parameters

ref string | { [string]: any } | AssetRef — A .computeShader identity/guid string, a resolved asset handle, or the name a raw registration chose.

Returns (string, string?) — "absent", "pending", "ready" or "failed", and the reason when "failed".

if compute.programState(carve) == "ready" then carve:dispatch(opts) end

globals/compute/readBuffer

compute.readBuffer(name: string) -> string

Start a GPU→CPU read of the buffer under name.

Parameters

name string — The name it was created under.

Returns string — The result key to poll with getReadbackResult*. A read that could not start answers with the empty key, which every drain reports as unknown — the same shape a caller already handles.

globals/compute/readTexture3D

compute.readTexture3D(name: string) -> string

Start a GPU→CPU readback of a named 3D volume. Returns a result key to poll.

Parameters

name string

Returns string

globals/compute/readbackPromise

compute.readbackPromise(resultKey: string) -> any

The promise that settles when a read-back reaches a terminal state, delivered or given up on. task.await it to wait for the read without polling: where the engine can wait on the device the coroutine resumes in the frame the device answered, so a task that reads, uses the answer and reads again does all of it in one frame; elsewhere it resumes on the next frame the answer reaches the engine. Answers nil when the read-back is not in flight (already delivered, failed, drained or never issued), and readbackState() then says which.

Parameters

resultKey string — Key returned by readBuffer().

Returns any — A promise to await, or nil when there is nothing to wait for.

local promise = compute.readbackPromise(key); if promise then task.await(promise) end

globals/compute/readbackState

compute.readbackState(resultKey: string) -> string

Where a readback key stands, without consuming it and without raising: "pending" (issued, GPU has not delivered), "ready" (delivered, waiting to be drained), "failed" (given up on, with the reason as a second return), or "unknown" (never issued by readBuffer(), or already drained — a result is delivered once).

Every read leaves "pending": one the GPU does not answer inside the bound the engine holds it to reaches "failed" naming the buffer. A wait on readbackState(key) ~= "pending" therefore ends on what the read did, which a count of engine frames cannot say.

Parameters

resultKey string — Key returned by readBuffer().

Returns string — "pending", "ready", "failed" (and the reason), or "unknown".

if compute.readbackState(key) == "ready" then ... end

globals/compute/registerShader

compute.registerShader(nameOrHandle: string | { [string]: any } | AssetRef, opts: ShaderOpts?) -> boolean

Parameters

nameOrHandle string | { [string]: any } | AssetRef — Shader name or asset handle.
opts ShaderOpts (optional) — Shader options. bindings comes from the source's own @group(0) @binding(n) declarations when omitted; supplying a count that disagrees with them raises. Every readOnlyBindings entry names one of those declared bindings, as a whole number from 0 to bindings - 1; an entry outside that run raises.

Returns boolean — True on success.

compute.registerShader("blur", { source = WGSL, entryPoint = "main" })

globals/compute/registerShaderEx

compute.registerShaderEx(nameOrHandle: string | { [string]: any } | AssetRef, opts: { [string]: any }?) -> boolean

Parameters

nameOrHandle string | { [string]: any } | AssetRef
opts { [string]: any } (optional)

Returns boolean

globals/compute/releaseBaseline

compute.releaseBaseline(generation: number) -> boolean

Drop the mark generation names, and every mark taken after it, once the caller that took it is done resetting to it. It touches the marks alone.

Parameters

generation number — The mark to drop, as compute.markBaseline returned it.

Returns boolean — True when the mark stood.

compute.releaseBaseline(mark) -- the run is over

globals/compute/resetToBaseline

compute.resetToBaseline(generation: number?) -> number

Release every compute resource allocated since compute.markBaseline, leaving the ones it recorded where they are.

Each kind this closes is one a caller allocated under a name it chose, so the next caller finds that name free. A storage buffer is such a resource and is not closed here: every one a caller can make is a substrate buffer, and substrate.resetToBaseline is what puts those back. A compiled shader's params uniform is not, and is left where it is: it is filed under its shader's guid, which no other caller can ask for, and it belongs to the .computeShader asset that compiled it, released when that is. mark is closed; nothing the mark recorded is touched.

Parameters

generation number (optional) — The mark to restore to, as compute.markBaseline returned it; the newest mark when omitted. The marks taken after it go with the reset.

Returns number — How many resources were released.

local released = compute.resetToBaseline(mark)

globals/compute/resources

compute.resources(owner: any?) -> { any }

The resource rows on their own, optionally narrowed to what one shader owns.

Parameters

owner any (optional) — A .computeShader ref, its guid, or its asset identity. Omit for every resource compute holds. A value carrying no shader raises, so a narrowing that cannot be done reads as an error rather than as the whole inventory. A guid stands for itself, so resources outlive the asset that made them and stay reachable by their owner.

Returns { any } — An array of rows in the shape compute.observe().resources carries. Empty when the owner holds nothing.

for _, r in ipairs(compute.resources(shaderRef)) do print(r.key, r.bytes) end

globals/compute/setParam

compute.setParam(nameOrHandle: string | { [string]: any } | AssetRef, prop: string, value: number) -> boolean

Set a named scalar parameter on a .computeShader (a params: entry in its bindings.yaml). Updates the shader's params uniform in place; the next dispatch sees the new value. No effect on raw registerShader shaders, which have no params block.

Parameters

nameOrHandle string | { [string]: any } | AssetRef — Shader identity (the .computeShader asset name), or the handle asset.load / asset.resolve returns — the same forms dispatch takes.
prop string — Parameter name as declared in bindings.yaml.
value number — New scalar value (numbers only).

Returns boolean — True on success.

compute.setParam("my_sim", "scale", 4.0)

globals/compute/stepBvh

compute.stepBvh(id: number, budgetMs: number?) -> (string?, string?)

Advance a build by as many work units as budgetMs buys, and report whether it has finished: "pending" means there is work left, "ready" means compute.finishBvh will hand over the buffers. The slices carry the hierarchy onto the GPU as well as building it, so a build that reads "ready" has already uploaded every byte of itself. A slice always runs at least one unit, so a budget of 0 advances the build by exactly one and the largest single unit sets the floor under a slice.

Parameters

id number — Build id from compute.beginBvh.
budgetMs number (optional) — Wall time this slice may spend, in milliseconds (default 4).

Returns (string?, string?) — "pending" or "ready", or (nil, err) when the id names no build.

while compute.stepBvh(id, 4) == "pending" do task.wait() end

globals/compute/textureForPlugin

compute.textureForPlugin(name: string) -> any

The texture this script created under name as a handle a script passes a plugin in a call. The handle holds the texture by an id of its own, never by the name: once the texture is destroyed the handle refuses, even when the name is made again. Only a texture made by compute.createTexture3D or compute.createStorageTexture2D can be handed over.

Parameters

name string — The texture's name.

Returns any — A handle for a plugin call.

globals/compute/textureFormatBytes

compute.textureFormatBytes(format: string) -> number

Bytes-per-voxel for a texture format string (rgba16f, r8, ...).

Parameters

format string

Returns number

globals/compute/writeBuffer

compute.writeBuffer(name: string, values: { number } | buffer | string, offset: number?) -> boolean

Write words into the buffer under name.

Parameters

name string — The name it was created under.
values { number } | buffer | string — The floats to write, or a buffer / binary string already holding them.
offset number (optional) — 32-bit word offset to write at.

Returns boolean — True on success.

globals/compute/writeBufferBytes

compute.writeBufferBytes(name: string, bytes: buffer | string, offsetBytes: number?) -> boolean

Write packed bytes into the buffer under name.

Parameters

name string — The name it was created under.
bytes buffer | string — The payload.
offsetBytes number (optional) — Byte offset to write at.

Returns boolean — True on success.

globals/compute/writeBufferU32

compute.writeBufferU32(name: string, values: { number } | buffer | string, offsetBytes: number?) -> boolean

Write 32-bit words into the buffer under name.

Parameters

name string — The name it was created under.
values { number } | buffer | string — The words to write.
offsetBytes number (optional) — Byte offset to write at.

Returns boolean — True on success.

globals/compute/writeFloatsTexture3D

compute.writeFloatsTexture3D(name: string, floats: { number }, formatOrOpts: (string | { [string]: any })?) -> boolean

Upload float values into a named 3D volume, packed via the given format (default rgba16f).

Parameters

name string
floats { number }
formatOrOpts (string | { [string]: any }) (optional)

Returns boolean

globals/compute/writeTexture3D

compute.writeTexture3D(name: string, data: buffer | string | { number }) -> boolean

Upload raw bytes (u8) into a named 3D volume. A buffer or a binary string holds the volume's byte layout verbatim and crosses in one copy — the shape a file's voxel payload arrives in; an array carries one byte value (0..255) per entry.

Parameters

name string — Volume name.
data buffer | string | { number } — Voxel bytes as a buffer, a binary string, or an array of bytes.

Returns boolean — True on success (mutation queued).

globals/computed

computed(fn: () -> any): any

Mark a public.X = computed(fn) field as a derived read-only property — auto-injected by the prelude's component_proxy.

globals/debugger/__diagnostics

debugger.__diagnostics() -> DebuggerDiagnostics

Internal diagnostic counters for debugging the debugger itself: { installs, debugbreakHits }.

Returns DebuggerDiagnostics — Diagnostic counters.

globals/debugger/addWatch

debugger.addWatch(expr: string) -> number

Parameters

expr string — Luau expression.

Returns number — Watch id.

globals/debugger/continue_

debugger.continue_() -> boolean

Resume the paused thread. The thread resumes on the next scheduler tick. A task's thread settles like any resumed task: a wait it reaches is armed, what it returns is its result, and an error it raises is the task's error, reported the way every task error is. A task cancelled while paused does not run on. A coroutine a task resumed itself runs on with nothing waiting for it, and an error it raises is logged.

Returns boolean — True if a thread was paused, false if nothing was paused.

globals/debugger/disableAll

debugger.disableAll()

Disable every registered breakpoint. Records persist; bytecode BREAK ops are cleared.

globals/debugger/disconnect

debugger.disconnect(handle: number) -> boolean

Disconnect an onBreak or onResume callback.

Parameters

handle number — Handle returned by onBreak/onResume.

Returns boolean — True if the handle existed.

globals/debugger/enableAll

debugger.enableAll()

Enable every registered breakpoint and re-install them in the VM bytecode.

globals/debugger/evaluate

debugger.evaluate(expr: string, frame: number?) -> (string?, string?)

Evaluate an expression against the paused frame's environment. Returns (value, error).

Parameters

expr string — Luau expression.
frame number (optional) — 1-based frame index (default 1).

Returns (string?, string?) — (value, error).

globals/debugger/getLocals

debugger.getLocals(frame: number?) -> { [string]: string }

Locals captured at the active pause for the given frame index (1 = top). Values are stringified for safe display.

Parameters

frame number (optional) — 1-based frame index (default 1).

Returns { [string]: string } — { [name] = string }.

globals/debugger/getPauseInfo

debugger.getPauseInfo() -> PauseInfo?

Info about the active pause, or nil if nothing is paused.

Returns PauseInfo? — { path, line, reason } or nil.

globals/debugger/getStack

debugger.getStack() -> { Frame }

Captured stack from the active pause, top frame first. Empty when nothing is paused.

Returns { Frame } — Array of Frame tables.

globals/debugger/getUpvalues

debugger.getUpvalues(frame: number?) -> { [string]: string }

Upvalues captured at the active pause for the given frame.

Parameters

frame number (optional) — 1-based frame index.

Returns { [string]: string } — { [name] = string }.

globals/debugger/getWatchValue

debugger.getWatchValue(id: number) -> (string?, string?)

Re-evaluate the watch expression against the paused frame's environment and return (value, error).

Parameters

id number — Watch id.

Returns (string?, string?) — (value, error).

globals/debugger/getWatches

debugger.getWatches() -> { Watch }

Snapshot of all watches with their last evaluated value and error, sorted by id.

Returns { Watch } — Array of Watch tables.

globals/debugger/isPauseOnError

debugger.isPauseOnError() -> boolean

Current pause-on-error toggle state for this VM.

Returns boolean — True if enabled.

globals/debugger/isPaused

debugger.isPaused() -> boolean

Whether the debugger currently has a paused thread.

Returns boolean — True if paused.

globals/debugger/listBreakpoints

debugger.listBreakpoints() -> { Breakpoint }

Snapshot of every registered breakpoint, sorted by id ascending. Each entry reports whether it is installed: chunkNames lists the loaded chunks carrying it, and pendingReason says why an empty list is empty.

Returns { Breakpoint } — Array of breakpoint tables.

globals/debugger/onBreak

debugger.onBreak(fn: (PauseInfo) -> ()) -> number

Register a callback invoked on every pause with { path, line, reason }. Returns a handle usable with debugger.disconnect.

Parameters

fn (PauseInfo) -> () — Callback.

Returns number — Handle.

globals/debugger/onResume

debugger.onResume(fn: () -> ()) -> number

Parameters

fn () -> () — Callback.

Returns number — Handle.

globals/debugger/removeBreakpoint

debugger.removeBreakpoint(id: number) -> boolean

Remove the breakpoint with the given id.

Parameters

id number — Breakpoint id returned by setBreakpoint.

Returns boolean — True if removed, false if the id was unknown.

globals/debugger/removeWatch

debugger.removeWatch(id: number) -> boolean

Remove the watch with the given id.

Parameters

id number — Watch id.

Returns boolean — True if removed.

globals/debugger/setBreakpoint

debugger.setBreakpoint(path: string, line: number, opts: BreakpointOpts?) -> Breakpoint

Set a breakpoint at line in the script path names — its VFS path, its require identity, or the chunk name it loaded under. An installed breakpoint carries resolvedLine and lists the loaded chunks holding it in chunkNames; one whose script is not loaded carries pendingReason, an empty chunkNames, and installs itself when that script loads.

Parameters

path string — VFS path, require identity, or chunk name.
line number — 1-based source line.
opts BreakpointOpts (optional) — { condition?, logMessage?, hitCount?, notify?, enabled? }. notify makes it a TRACEPOINT: the hit fires onBreak handlers against the frame it stopped in — getStack, getLocals, getUpvalues and evaluate all answer for that frame — and then runs on, instead of holding the thread until something resumes it. That is how code observes a moment without stopping it; halting is for a person at a keyboard. A breakpoint on any line drops the enclosing function out of native code until its chunk reloads, so one left armed on a per-frame function costs more than it reports.

Returns Breakpoint — The breakpoint table.

local bp = debugger.setBreakpoint("/zero/source/main.luau", 42)
print(bp.pendingReason or ("installed in " .. bp.chunkNames[1]))

globals/debugger/setPauseOnError

debugger.setPauseOnError(enabled: boolean)

When true, uncaught Luau errors fire the onBreak callback (observation only — the error still propagates).

Parameters

enabled boolean — Toggle state.

globals/debugger/stepInto

debugger.stepInto() -> boolean

Run until the next line, descending into any function call.

Returns boolean — True if a step was scheduled.

globals/debugger/stepOut

debugger.stepOut() -> boolean

Run until the current frame returns; pauses in the caller.

Returns boolean — True if a step was scheduled.

globals/debugger/stepOver

debugger.stepOver() -> boolean

Run until the next line in the current frame. Calls inside the current line are skipped.

Returns boolean — True if a step was scheduled.

globals/debugger/toggleBreakpoint

debugger.toggleBreakpoint(path: string, line: number) -> Breakpoint?

Toggle a breakpoint at the given line: removes if present, adds otherwise.

Parameters

path string — VFS path, require identity, or chunk name.
line number — 1-based line.

Returns Breakpoint? — Breakpoint table if added, nil if removed.

globals/declare

declare(spec)

Declare top-level component metadata: executionOrder (i32), role (string — the job this component fills when several types fill the same one, e.g. "collider"), bindings ({string}), syncedFunctions ({string}), initKeys ({string} — init-table keys the component normalises itself, beyond its public fields), nativeFields ({[string]: string | {string}} — fields of the SAME-NAMED native ECS component this component computes, each mapped to the field(s) of its own it computes them from, e.g. { lightType = "kind" }; a refusal naming one of those fields answers with the name that holds instead of a direct row write the next refresh replaces), translate ({[string]: string} — for procedural translation, the attribute that feeds each named field on its own, e.g. { intensity = "glow" }: an element carrying a glow attribute becomes an entity carrying this component with intensity from it. The component's own name is the attribute carrying it whole, one record per element, and needs no declaration). Per-field schema (type, default, Sync/NoSync) lives in public = {...} via Field.<kind>(default, mode) constructors — see man field. Must be called at top level, not inside any function.

Parameters

spec table — { executionOrder = N, role = "collider", bindings = {...}, syncedFunctions = {...}, initKeys = {...}, nativeFields = { nativeField = "ownField" }, translate = { field = "attribute" } }

globals/delay

delay(seconds) -> promise

Returns a promise that resolves after seconds of REAL time — the engine's own clock, which engine.timeScale does not reach, so await(delay(2)) pauses for two wall-clock seconds at every scale. task.delay(seconds, fn) schedules on the simulation clock the scale multiplies, and task.waitRealtime(seconds) is this clock spelled as a wait.

Parameters

seconds number — Delay duration in seconds

Returns PromiseHandle — Promise handle for use with await()

globals/editorTools

editorTools(scope: string) -> table

Bind a toolbox handle whose methods call each tool and return its value directly, raising on failure. Superseded by tools.use("", "", ...).

globals/egress/credentialNames

egress.credentialNames() -> { string }

List the names in the local credential store. Names only — secret values are never exposed to Luau.

Returns { string } — Array of configured credential names.

for _, n in ipairs(egress.credentialNames()) do print(n) end

globals/egress/fetch

egress.fetch(name: string, method: string, url: string, headers: Headers?, body: JsonBody?, response: EgressResponseType?) -> PromiseHandle?

Perform an HTTP request using a named credential. A bound world forwards the request to ZeroMind, which resolves its world secret and enforces the allowed destination. Await the returned handle; a refused forwarded request returns the HTTP error JSON with __http_error = true. With a local credential store, an unknown name or a URL outside its allowed prefix returns nil immediately. Credential values stay with the store or forwarding server and never enter world code.

Parameters

name string — Local credential name or bound-world secret handle.
method string — HTTP method, e.g. "GET" or "POST".
url string — Request URL allowed by the selected credential's destination rules.
headers Headers (optional) — Extra header key-value pairs.
body JsonBody (optional) — JSON body (encoded automatically).
response EgressResponseType (optional) — "json" (default) or "bytes".

Returns PromiseHandle? — Promise handle for task.await(), or nil on an immediate local refusal.

local h = egress.fetch("meshy", "POST", url, nil, { prompt = p })

globals/egress/hasCredential

egress.hasCredential(name: string) -> boolean

Whether the local credential store holds a name. Returns only a boolean. World secrets are resolved by ZeroMind when a forwarded request runs; this probe answers for the local store.

Parameters

name string — Credential name.

Returns boolean — True if configured.

if not egress.hasCredential("meshy") then error("set MESHY_API_KEY") end

globals/engine/discardPlayChanges

engine.discardPlayChanges() -> { paths: { string }, keep: string }

Arm the leave-play safeguard's deliberate discard for your next engine.mode = "edit" in the play session this is called from, so it proceeds and discards your unaccepted changes. While other callers stay in the session that is your own work: the /source edits you made in play, which revert, and the entities you tuned. The last caller to leave takes the engine to edit, and its discard throws away everything play made: the entities and tuning, and every /source file edit play holds on its shadow. Called by a caller outside the session, it arms the discard of engine.setMode("edit", { endSession = true }), which ends the session for every caller in it. A call that names no caller (a transport that attests no caller for it) cannot be told apart from any caller, so the exit that ends the session takes an arm made for the whole session from the other side: an exit that names no caller takes the arm a named caller made, and a named caller's exit takes the arm a call that named no caller made. Returns paths, the source files the flip would throw away if it ran now, and keep, the routes that keep any of them first; a notice names the same paths. The flip throws away what the shadow holds when it runs, so an edit made after the arm goes with the rest, and the flip names every path it took in a notice and the engine log (vfs.discardPlayShadow() states the same set). A discarded source edit whose pre-play copy is not on this engine yet keeps the play bytes until that copy lands and is put back; the flip names each such path, and vfs.playDiscardsOwed() lists them until then.

Returns { paths: { string }, keep: string }

globals/engine/editWaiters

engine.editWaiters() -> { EditWaiter }

Everyone waiting for this engine to leave play so their edit-mode work can run, oldest first. Each entry is { key, id, name, reason, heldSeconds }: id is the waiting caller, the same value vfs.currentAuthor().id reads, and heldSeconds is how long it has been waiting. A session holding play reads this to see whose work it is holding. Empty in edit mode and whenever nobody has said they are waiting.

Returns { EditWaiter } — The waiter records.

for _, who in ipairs(engine.editWaiters()) do
log.warn(("%s has waited %.0fs for edit mode"):format(who.name or who.key, who.heldSeconds))
end

globals/engine/engineRecord

engine.engineRecord(record: { [any]: any })

Declare record an engine record: a table a subsystem keeps as its own account of what the engine holds — the renderer's registry of live features, its capture views — and puts back through its own reset. engine.snapshotModuleState neither copies nor walks it, and engine.restoreModuleState leaves it as that reset left it, so a module restore never seats a record back around what the reset did: a feature a suite destroyed stays destroyed, and one it made again under its identity stays the one the renderer ticks.

Parameters

record { [any]: any } — The table.

engine.engineRecord(activeFeatures)

globals/engine/gameplayReady

engine.gameplayReady -> boolean

Whether gameplay simulation is running: not paused, and the play scene materialized. Read-only.

Returns boolean

globals/engine/gpuCompute

engine.gpuCompute -> boolean

Whether this process holds a live GPU device, so compute dispatch is available. Read-only.

Returns boolean

globals/engine/headless

engine.headless -> boolean

Whether this boot renders offscreen with no window a person can see. Content that only serves someone at a display stands down when it reads true. Read-only.

Returns boolean

globals/engine/holdTimeScale

engine.holdTimeScale(scale: number, options: TimeScaleHoldOptions?) -> () -> ()

Run the simulation at scale for as long as the hold stands, and return the release. engine.timeScale is one number for the whole process, so a rate one agent sets is the rate every other agent's entities and measurements run at; a hold records who set it and refuses a plain engine.timeScale write from anywhere else while it stands, so the rate cannot move under a measurement without the measurement hearing about it. Holds nest: the innermost names the rate, and the rate that stood before the outermost is back once the last release is called. Each release takes its own hold off the stack whatever order the releases come in. exclusive takes the rate for the owner key the call states: while that hold stands, a hold is admitted only when it states the same key, and every other one is refused with an error naming the key and the rate. That is what lets one caller pin the rate its run is measured at while another agent drives the same engine.

Parameters

scale number — The rate to run at: 1 is real time, 0.5 half speed, 0 frozen, 2 double. Finite and >= 0.
options TimeScaleHoldOptions (optional) — owner is the key this hold is taken under, a non-empty string, and an exclusive hold states one. A hold that states no key takes the key of the agent the call is attributed to, or a minted hold#n when the call carries no agent identity, so every hold answers to a key engine.releaseTimeScale accepts and engine.timeScaleOwner reports as id. exclusive takes the rate for the stated key until the hold is released.

Returns () -> () — A function that releases this hold. Calling it twice releases once.

local release = engine.holdTimeScale(0.08) ; task.wait(0.12) ; release()
local release = engine.holdTimeScale(1, { exclusive = true, owner = "duel-measurement" })

globals/engine/markBaseline

engine.markBaseline() -> number

Read this module's own memory — the time-scale holds standing, who wrote the rate last, the rate that stood before any hold, the onTimeScaleChange subscribers and the highest watcher id each registry has handed out — and answer with a generation engine.resetToBaseline puts exactly that back from. Take it once, before the work whose changes are to be undone.

Returns number — The generation of this reading.

local mark = engine.markBaseline() -- then run, then engine.resetToBaseline(mark)

globals/engine/markScriptingBaseline

engine.markScriptingBaseline() -> number

Record the scripting registries — world-event subscriptions, the four lifecycle-watcher lists, and the require cache — as they stand right now, as a mark engine.resetScriptingState() restores to. Marks stack: a reset names the generation it restores to, and the marks taken after that one go with it. Returns the new mark's generation, counting from 1. Mark once the engine is serving rather than while it boots: the registries keep growing as the prelude subscribes, the world entrypoint runs and the startup scene loads, so a mark taken partway through sits below the rest of that work and the first reset would remove it.

Returns number

engine.markScriptingBaseline()
world.on("player_join", function() end)
engine.resetScriptingState() -- the subscription above is gone

globals/engine/mode

engine.mode -> "edit" | "play"

The engine mode this process is in, edit or play. Assigning it takes the flip, side effects and all.

Returns "edit" | "play"

globals/engine/modeOwner

engine.modeOwner -> { id: string?, agent: string?, name: string?, heldSeconds: number }?

Who flipped the engine out of the mode it booted into, and how long ago, while that caller still holds it. In play, engine.playSession names every caller the session belongs to now. id is the caller key, the same value vfs.currentAuthor().id reads, and nil when that flip carried no agent identity; agent is the name the world's roster holds for that caller now, the one to ask by, and name the username of its account, which every agent of the account shares. Nil in the mode the engine booted into (a flip back to it ends the hold), once the caller that flipped it has left the play session it entered, and once that caller is no longer on the world's roster. Read-only.

Returns { id: string?, agent: string?, name: string?, heldSeconds: number }?

globals/engine/offDeviceRebuilt

engine.offDeviceRebuilt(id: number) -> boolean

Remove an engine.onDeviceRebuilt subscriber by its watcher id. Returns true when a live watcher carried that id, false when it named none — already removed, or never registered.

Parameters

id number — Watcher id returned by engine.onDeviceRebuilt.

Returns boolean

local id = engine.onDeviceRebuilt(function() end)
engine.offDeviceRebuilt(id)

globals/engine/offMemoryPressure

engine.offMemoryPressure(id: number) -> boolean

Remove an engine.onMemoryPressure subscriber by its watcher id. Returns true when a live watcher carried that id, false when it named none: already removed, or never registered.

Parameters

id number — Watcher id returned by engine.onMemoryPressure.

Returns boolean

local id = engine.onMemoryPressure(function() end)
engine.offMemoryPressure(id)

globals/engine/offModeChange

engine.offModeChange(id: number) -> boolean

Remove an engine.onModeChange subscriber by its watcher id. Returns true when a live watcher carried that id, false when it named none — already removed, or never registered.

Parameters

id number — Watcher id returned by engine.onModeChange.

Returns boolean

local id = engine.onModeChange(function() end)
engine.offModeChange(id)

globals/engine/offPauseChange

engine.offPauseChange(id: number) -> boolean

Remove an engine.onPauseChange subscriber by its watcher id. Returns true when a live watcher carried that id, false when it named none.

Parameters

id number — Watcher id returned by engine.onPauseChange.

Returns boolean

globals/engine/offTimeScaleChange

engine.offTimeScaleChange(id: number) -> boolean

Remove an engine.onTimeScaleChange subscriber by its watcher id. Returns true when a live watcher carried that id, false when it named none.

Parameters

id number — Watcher id returned by engine.onTimeScaleChange.

Returns boolean

local id = engine.onTimeScaleChange(function() end)
engine.offTimeScaleChange(id)

globals/engine/offWorldLoaded

engine.offWorldLoaded(id: number) -> boolean

Remove an onWorldLoaded subscriber by its watcher id.

Parameters

id number

Returns boolean

globals/engine/offWorldReady

engine.offWorldReady(id: number) -> boolean

Remove an engine.onWorldReady subscriber by its watcher id. Returns true when a live watcher carried that id, false when it named none.

Parameters

id number — Watcher id returned by engine.onWorldReady.

Returns boolean

globals/engine/offWorldUnloading

engine.offWorldUnloading(id: number) -> boolean

Remove an engine.onWorldUnloading subscriber by its watcher id. Returns true when a live watcher carried that id, false when it named none.

Parameters

id number — Watcher id returned by engine.onWorldUnloading.

Returns boolean

globals/engine/onDeviceRebuilt

engine.onDeviceRebuilt(callback: (number) -> ()) -> number

Register a callback that fires after the engine has answered a lost render device by building another one. The callback receives the new device generation — a number that counts the devices this session has run on, and moves exactly once per rebuild. Returns a watcher id.

A device is lost when the driver resets, when the GPU is taken away, or when a browser reclaims a WebGPU context. Everything the engine can re-derive by itself it does: meshes, materials, shaders, render passes and the UI are all back on the new device before this fires. What it cannot re-derive is what YOUR content made and only the GPU held — a texture uploaded from pixels a script computed, a compute buffer it filled, a render target it created. Make those again here.

Content that owns no GPU resource of its own needs no subscriber: asset handles re-materialise on their next use.

Parameters

callback (number) -> () — Function invoked as (generation: number).

Returns number

engine.onDeviceRebuilt(function(generation)
-- the noise field lived only on the GPU, so it is computed again
print("upload the noise field again for device generation", generation)
end)

globals/engine/onMemoryPressure

engine.onMemoryPressure(callback: (string, string) -> ()) -> number

Register a callback that fires when the system says it is short of memory, after the engine has released what that level of pressure releases. The callback receives the level ("normal", "moderate", "low" or "critical") and who reported it: onTrimMemory(<level>) for Android's report, script for renderer.signalMemoryPressure. Returns a watcher id.

The engine frees what it can rebuild by itself: mip levels the frame does not sample, render-target storage nothing is drawing into, and from "critical" half the texture budget. What it cannot free is what YOUR content keeps: a cache of generated meshes, a table of decoded audio, pixels kept for a later upload. Drop those here and build them again when they are next needed. A level stands for a minute after the last report, and renderer.memoryPressure() reads the level in force and what the engine released.

Parameters

callback (string, string) -> () — Function invoked as (level: string, source: string).

Returns number

engine.onMemoryPressure(function(level, source)
if level ~= "normal" then
decodedClips = {} -- decoded again on next play
end
end)

globals/engine/onModeChange

engine.onModeChange(callback: (string, string) -> ()) -> number

Parameters

callback (string, string) -> ()

Returns number

globals/engine/onPauseChange

engine.onPauseChange(callback: (boolean, boolean) -> ()) -> number

Register a callback that fires synchronously whenever the gameplay pause flag flips via an explicit engine.paused write. Callback receives (newPaused, oldPaused) as booleans. Returns a watcher id. Pause is independent of engine.mode: pausing play mode returns the editor authoring surface (free camera + EditorOnly entities) over the frozen play world, and resuming hides it again. Mode-driven pause resets (the edit=paused / play=running defaults applied on a mode flip) are delivered through onModeChange, not this hook.

Parameters

callback (boolean, boolean) -> () — Function invoked as (newPaused: boolean, oldPaused: boolean).

Returns number

local id = engine.onPauseChange(function(paused)
print(paused and "frozen" or "running")
end)

globals/engine/onTimeScaleChange

engine.onTimeScaleChange(callback: (number, number, TimeScaleOwner?) -> ()) -> number

Register a callback that fires on every move of the simulation rate, whoever moved it: an engine.timeScale write, a hold taken at a rate other than the one in force, a release that hands a different rate back. The callback receives (newScale, oldScale, owner), where owner is the record engine.timeScaleOwner answers with. It counts transitions, so a write or a hold landing on the rate already in force moves nothing and announces nothing; engine.timeScaleOwner is what answers who holds a rate that has stayed where it is. Several agents drive one engine at once, so this is how a run whose numbers depend on the rate learns that somebody else moved it instead of reading the number and finding out afterwards. Returns a watcher id.

Parameters

callback (number, number, TimeScaleOwner?) -> () — Function invoked as (newScale: number, oldScale: number, owner: TimeScaleOwner?).

Returns number

local id = engine.onTimeScaleChange(function(new, old, who)
log.warn(("rate %g -> %g by %s"):format(old, new, who and who.id or "nobody"))
end)

globals/engine/onWorldLoaded

engine.onWorldLoaded(callback: () -> ()) -> number

Register a callback fired (no args) when the world is fully LOADED — its .world_entrypoint.luau ran AND its onWorldLoad returned (the startup scene loaded, defaults seeded, editor UI mounted). This is strictly AFTER onWorldReady (content synced): ready = "bytes are in the VFS"; loaded = "the entrypoint has run". LATCHED — a callback registered after the world is already loaded fires immediately, so a late consumer never misses it and never has to poll. Read the same state synchronously via engine.worldLoaded.

Parameters

callback () -> () — Function invoked with no arguments.

Returns number

globals/engine/onWorldReady

engine.onWorldReady(callback: () -> ()) -> number

Register a callback fired (no args) when the bound world's content has been synced into the VFS and the world is ready to load. This is the race-free, user-space hook that drives the whole world-VM lifecycle: the builtin world-entrypoint loader subscribes to it and, when it fires, loadstring(vfs.read(...))s /source/.world_entrypoint.luau and runs its onWorldLoad — exactly the way a scene entrypoint loads. The trusted VM fires this (via world.markReady()) ONLY once the bytes are in the VFS, so a subscriber never sees a half-synced world. Returns a watcher id.

Parameters

callback () -> () — Function invoked with no arguments.

Returns number

globals/engine/onWorldUnloading

engine.onWorldUnloading(callback: () -> ()) -> number

Symmetric teardown of engine.onWorldReady: register a callback fired (no args) when the bound world is unbinding/swapping out. The builtin loader runs the world entrypoint's onWorldUnload here, so the world entrypoint has the same load/unload parity a scene entrypoint has. Returns a watcher id.

Parameters

callback () -> () — Function invoked with no arguments.

Returns number

globals/engine/pauseOwner

engine.pauseOwner -> { id: string?, agent: string?, name: string?, heldSeconds: number }?

Who asked for the standing gameplay pause, and for how long they have held it. One process runs one simulation, so a pause another agent took freezes update(dt), fixedUpdate(dt), physics and every gameplay clock for everybody: a component reading its own state while this is non-nil is reading the same instant over and over, and a sampling loop across task.wait returns the identical sample every time. id is the caller key, the same value vfs.currentAuthor().id reads, and nil when that call carried no agent identity; agent is the name the world's roster holds for that caller now, and name the username of its account, which every agent of the account shares. Nil while the engine is running, and nil when the standing pause is the mode's own rather than something somebody asked for. Read-only.

Returns { id: string?, agent: string?, name: string?, heldSeconds: number }?

globals/engine/paused

engine.paused -> boolean

Whether gameplay is paused: update(dt) component callbacks are gated off while editorUpdate(dt) keeps firing in edit mode. A component's gameplay task (one spawned from awake, start, onEnable, update, fixedUpdate or another gameplay task) is held for the whole pause: a task.wait it started keeps the time it had left, and a task.waitRealtime is held too, but its deadline runs through the pause, so it comes due on the first frame after it. A task spawned from editorUpdate and a task.spawnSystem task run while paused. The scripting-and-tasks guide's "When a task runs" states the whole rule.

Returns boolean

globals/engine/platform

engine.platform -> "native" | "browser" | "android" | "unknown"

Which build of the engine this process is: native for the desktop binary, browser for the wasm one, android for the Android app. unknown says the binary named no build, which a binary older than the field does. Read-only.

Returns "native" | "browser" | "android" | "unknown"

globals/engine/playSession

engine.playSession -> { agents: { { key: string, agent: string? } }, takesPart: boolean }?

The play session the engine is running, as you stand towards it. Play belongs to the callers that take part in it: the caller that entered play, and every caller that writes engine.mode = "play" while it runs. agents are those callers still connected, in the order they joined, each with its stable key (vfs.currentAuthor().stable) and the name the world's roster holds for it. takesPart is whether you are one of them: your /source writes are play work, and your engine.mode = "edit" takes you out of the session, which keeps running for the others and ends with the last of them. While it is false you author beside the session: your /source writes land on canonical source, and the session is theirs to end. Nil in edit mode, in the runtime profile, and while the engine plays for nobody in particular (a Play button press, a boot into play). Read-only.

Returns { agents: { { key: string, agent: string? } }, takesPart: boolean }?

globals/engine/profile

engine.profile -> "editor" | "runtime"

The boot profile this process started under, editor or runtime. Read-only.

Returns "editor" | "runtime"

globals/engine/releaseBaseline

engine.releaseBaseline(generation: number) -> boolean

Drop the mark generation names, and every mark taken after it, once the caller that took it is done resetting to it. It touches the marks alone.

Parameters

generation number — The mark to drop, as engine.markBaseline returned it.

Returns boolean — True when the mark stood.

engine.releaseBaseline(mark) -- the run is over

globals/engine/releaseScriptingBaseline

engine.releaseScriptingBaseline(generation: number) -> boolean

Drop the scripting mark generation names, and every mark taken after it, once the caller that took it is done resetting to it. The registries stay as they stand. A mark holds the require cache's keys as they stood, so a caller that marks once per run drops each mark when its run is over.

Parameters

generation number — The mark to drop, as engine.markScriptingBaseline returned it.

Returns boolean — True when the mark was in force.

engine.releaseScriptingBaseline(mark) -- the run is over

globals/engine/releaseTimeScale

engine.releaseTimeScale(owner: string) -> number

Parameters

owner string

Returns number

globals/engine/resetScriptingState

engine.resetScriptingState(generation: number?) -> { [string]: number }

Drop every world-event subscription, lifecycle watcher and cached module registered since the mark generation names — the newest engine.markScriptingBaseline() when omitted — leaving everything registered before it in place, including the builtin world-entrypoint loader, which subscribes at VM boot and so always sits below any mark. The marks taken after that one go with the reset. A module with a frame on the stack keeps its cache entry and its registrations. Raises when no such mark is in force. Returns per-registry counts of what was removed: worldEvents, modeWatchers, worldReadyWatchers, worldUnloadingWatchers, pauseWatchers, deviceRebuiltWatchers, memoryPressureWatchers, modules, and total.

Parameters

generation number (optional) — The generation to restore to, as markScriptingBaseline returned it; the newest mark when omitted.

Returns { [string]: number }

globals/engine/resetToBaseline

engine.resetToBaseline(generation: number?) -> number

Put this module's own memory back as engine.markBaseline read it: every time-scale hold taken since comes off, every onTimeScaleChange subscriber added since is removed, the last writer and the base rate read as they did, the rate in force follows, and the record of who registered each watcher forgets the ids handed out since. The marks taken after generation go with the reset.

Parameters

generation number (optional) — The mark to restore to; the newest when omitted.

Returns number — How many things were put back.

engine.resetToBaseline(mark)

globals/engine/restoreModuleState

engine.restoreModuleState(snapshot: { [string]: any }, options: { remint: { [string]: string }?, gone: ((any) -> boolean)?, successor: ((any) -> any?)?, only: { [string]: boolean }? }?) -> { [string]: any }

Put every module's state back to what engine.snapshotModuleState read. A table is refilled in place, so a closure holding it sees the restored contents, and the upvalue is pointed back at it, so a module that assigned its local a new table is reading the restored one again.

The same snapshot restores to the same state however many times it is used, so one reading taken before a run puts every later point in that run back to it.

A value whose reading names an object the engine no longer holds — a buffer whose storage was destroyed since, answering alive false — is left as it stands: what it remembered cannot be given back to it, and what it holds now is what it went on to make. The module holding it is left whole, because what it remembers beside that object — the count it sized the object for, the flag that says it made one — is the state of the object it now holds instead; it is counted under stale. A data table naming a gone object in one of its fields is a registry, and is refilled around that field. Called from a coroutine that can yield, the restore gives the frame back every few milliseconds. the engine re-minted since replaced by the id it now uses.

Parameters

snapshot { [string]: any }
options { remint: { [string]: string }?, gone: ((any) -> boolean)?, successor: ((any) -> any?)?, only: { [string]: boolean }? } (optional) — remint maps an id the reading holds to the id the engine now uses for the same thing — a render feature's guid after the feature was made again under its identity. gone answers whether a value the reading holds names something the engine no longer holds — a record carrying a render feature's guid when no feature carries it now; a handle answering alive false is known without it. successor answers, for an object that has gone, the live object standing in for it — the handle under a buffer's name after the buffer was made again under it — and the reading is put back holding that one, its own fields left as they are. A handle that already answered alive false when the reading was taken is the module's own record of something gone before it, and is put back as it is. same answers whether a value a module holds now already stands in for the gone object a successor was found for — the module's own live handle to the buffer made again under the name — and such a value is kept, the successor released. only names the modules to put back, by module name, and every other module's reading is left untouched — a second pass over the modules an earlier one left as they stood, once what stands in for the objects they named exists.

Returns { [string]: any } — { restored, refused, recached, stale, staleModules, staleObjects, failed, failedModules } — how many values were written, how many the engine would not write, how many require-cache entries were put back to the module they held, how many modules hold a value left as it stood because its reading names something the engine no longer holds, and those modules by name, each with the gone object the reading named; then how many modules hold a value whose restore raised, and those by name with the error. A value whose restore raises is counted and passed over, and the rest of the reading is put back.

engine.restoreModuleState(snap)
engine.restoreModuleState(snap, { remint = { [oldGuid] = newGuid } })

globals/engine/scriptingRegistryCounts

engine.scriptingRegistryCounts() -> { [string]: number }

How many subscriptions each scripting registry holds right now, plus the size of the require cache and the generation of the mark in force. Keys: worldEvents, modeWatchers, worldReadyWatchers, worldUnloadingWatchers, pauseWatchers, deviceRebuiltWatchers, memoryPressureWatchers, modules, and baselineGeneration (nil when no mark has been taken).

Returns { [string]: number }

globals/engine/setMode

engine.setMode(mode: string, options: { strict: boolean?, endSession: boolean? }?) -> { mode: string, bypassed: { any }, leftPlaySession: boolean }

Change the engine mode with per-call control over the play gate, and read back what the change went past. engine.mode = value is the same flip with the defaults.

options.strict = false lets THIS call enter play while your own content carries error-severity diagnostics. It settles with the call: the world's lsp.strict_mode is untouched, so no other session and no later session of the world sees a different gate. The returned bypassed array holds the diagnostics the call went past — each { path, line, col, code, message, severity } — and the engine log carries the same list. An error in content another session wrote never gates the flip, so it never appears here; a push still refuses to publish while any of them stands.

Parameters

mode string — "edit" or "play".
options { strict: boolean?, endSession: boolean? } (optional) — { strict: boolean?, endSession: boolean? }. strict = false waives the play gate for this call; true or omitted honours the world's lsp.strict_mode. endSession = true with "edit" ends the running play session for every caller in it, whether or not you take part, and answers for everything the session holds (engine.discardPlayChanges() arms its discard); without it, an "edit" from one of several callers in play takes only that caller out, and one from a caller outside the session is refused.

Returns { mode: string, bypassed: { any }, leftPlaySession: boolean } — { mode, bypassed, leftPlaySession }: the mode now in force, the diagnostics this call entered play past (empty when it went past none), and whether an "edit" took this caller out of a play session that keeps running for other callers, in which case mode still reads "play".

local report = engine.setMode("play", { strict = false })
for _, d in ipairs(report.bypassed) do
print(("entered play past %s:%d — %s"):format(d.path, d.line, d.message))
end

globals/engine/snapshotModuleState

engine.snapshotModuleState() -> { [string]: any }

Read the state every cached module holds — the upvalues of the closures it hands back — and answer with it, so engine.restoreModuleState can put exactly that back later.

Take it once, before the work whose changes are to be undone. Data is copied, so a table a module goes on mutating does not change what was read; an object — a table carrying a metatable or methods of its own — is kept by reference and its fields copied, so restoring returns the object every other holder is still attached to rather than a lookalike that answers none of its methods.

A module with a frame on the stack while the reading is taken — the caller, and whatever called it — is mid-operation, and what it holds is left out: neither copied here nor written back by the restore, however another module reaches it. Called from a coroutine that can yield, the walk gives the frame back every few milliseconds, so the engine keeps stepping through a reading that takes seconds.

Returns { [string]: any } — { entries, cache, modules, inUse, unreachable } — what was captured, the require cache as { key, value } pairs, how many modules it covers, how many were left out for executing, and the count of values that could not be read.

local snap = engine.snapshotModuleState()

globals/engine/timeScale

engine.timeScale -> number

The global time scale applied to the fixed-timestep accumulator, to update(dt), and to the seconds task.wait and task.delay count: 1.0 is real time, 0.0 frozen, 2.0 double speed. So a task.wait(1) returns after four wall-clock seconds at 0.25 and never comes due at 0. task.waitRealtime(1) is one wall-clock second at every scale. One number for the whole process, shared by every agent driving this engine: engine.timeScaleOwner names who it answers to, engine.onTimeScaleChange reports every move of it, and engine.holdTimeScale takes it for a named owner so a write from anywhere else is refused while the hold stands. A script computing a delta of its own from a wall clock multiplies it by this, and one reading time.deltaTime() or a callback's dt already has it applied.

Returns number

globals/engine/timeScaleHolds

engine.timeScaleHolds() -> { string }

Returns { string }

globals/engine/timeScaleOwner

engine.timeScaleOwner -> { id: string?, name: string?, scale: number, exclusive: boolean, held: boolean }?

Who engine.timeScale answers to: the hold naming the rate while one stands, and otherwise whoever wrote the rate outright last. scale is the rate in force, the same number engine.timeScale reads. While held is true, id is a key engine.releaseTimeScale hands the rate back by; for a write it is the agent the call was attributed to, and nil when the call carried no agent identity. name is that agent's username when its token carried one. exclusive is true while the hold refuses every other key. Nil when nothing has moved the rate this session. Read-only.

Returns { id: string?, name: string?, scale: number, exclusive: boolean, held: boolean }?

globals/engine/vertexStride

engine.vertexStride -> number

Byte stride of the engine's standard GPU Vertex layout, which a mesh built from a compute buffer sizes and strides its writes to. Read-only.

Returns number

globals/engine/waitForEdit

engine.waitForEdit(reason: string?) -> () -> ()

Say that this caller's edit-mode work is waiting on the play session, and get back the call that takes the wait off.

One process runs one simulation, so a play session holds the engine in play for every caller outside it, and an agent whose edit-mode work cannot proceed has no way to say so. While this wait stands it rides the run-state block of every response any session gets from this engine, so the callers in the play session learn somebody is waiting for it. Read engine.playSession for who they are, and engine.modeOwner for who entered play and how long ago.

A wait is keyed by the caller the call was attributed to, so a named caller that registers twice moves its own entry forward rather than queueing behind itself: one caller waits once. A call carrying no agent identity is named by nothing to key on, so each such ask stands as its own waiter and the release it handed back takes that one off. Reaching edit mode takes every wait off, so a caller whose release is lost with the task that held it leaves nothing standing.

Parameters

reason string (optional) — What this caller is waiting to do, shown beside its name.

Returns () -> () — A function that takes this caller's wait off.

local done = engine.waitForEdit("promote my scene edits")
-- ... once the engine reaches edit and the work is taken ...
done()

globals/engine/watcherSources

engine.watcherSources() -> { [string]: { [number]: string } }

Who registered each watcher still standing, by registry: modeWatchers, pauseWatchers, deviceRebuiltWatchers, memoryPressureWatchers, worldReadyWatchers and worldUnloadingWatchers, each a map from watcher id to the source and line that registered it. A watcher registered before this module loaded is absent. What a reading of the registry counts names when it finds a watcher more than it began with.

Returns { [string]: { [number]: string } } — The sources, by registry and watcher id.

for id, at in pairs(engine.watcherSources().modeWatchers) do print(id, at) end

globals/engine/worldLoaded

engine.worldLoaded -> boolean

Whether the world entrypoint's onWorldLoad has run to completion. Read-only.

Returns boolean

globals/entity/__batchReadIds

entity.__batchReadIds()

Read a component field across many entities by id, in one crossing.

globals/entity/__batchReadToBuffer

entity.__batchReadToBuffer()

Read a component field into a typed substrate buffer.

globals/entity/__batchWriteBound

entity.__batchWriteBound()

Write a component field across a binding handle, in one crossing.

globals/entity/__batchWriteFromBuffer

entity.__batchWriteFromBuffer()

Write a component field from a typed substrate buffer.

globals/entity/__batchWriteIds

entity.__batchWriteIds()

Write a component field across many entities by id, in one crossing.

globals/entity/batchAddComponent

entity.batchAddComponent(targets: { string | entityRef }, type_name: string, data: table?) -> number

Add the same component type to many entities in one call. Returns the count of entities the component was added to — an entity already carrying an unnamed instance of the same type is skipped rather than double-added.

Parameters

targets { string | entityRef } — Array of entity ids or entity proxies (e.g. the return of entity.batchSpawn or entity.findAll).
type_name string — Component type to add to every entity.
data table (optional) — Init data: one table applied identically to every entity — the same shape the second arg to entity(id).component.add(type, data) takes — or an array holding one such table per target, in the targets' order, so each entity is added with its own values.

Returns number — How many entities had the component added.

local n = entity.batchAddComponent(ids, "Debris", { lifetime = 5 })
entity.batchAddComponent({ a, b }, "Debris", { { lifetime = 2 }, { lifetime = 7 } })

globals/entity/batchDespawn

entity.batchDespawn(targets: { string | entityRef }) -> number

Despawn many entities in one call. Locked or unresolvable entities are skipped. Returns the count queued for despawn.

Parameters

targets { string | entityRef } — Array of entity ids, entity proxies, or display names (e.g. the return of entity.batchSpawn / entity.findAll).

Returns number — Count of entities queued for despawn.

local n = entity.batchDespawn(ids)

globals/entity/batchProxy

entity.batchProxy(targets: { string | entityRef }) -> { entityRef? }

Resolve an array of entity ids to proxies in one call. Each output slot is the standard entity(id) proxy; ids missing from the frame cache surface as nil at that index. Use when iterating over a snapshot of entities so per-id lookups don't dominate the hot path.

Parameters

targets { string | entityRef } — Array of entity ids or entity proxies.

Returns { entityRef? } — Array of proxies (nil for missing ids).

local proxies = entity.batchProxy(ids)

globals/entity/batchRead

entity.batchRead(target: { string | entityRef } | binding, component: string?, field: string?, sink: TypedBuffer?) -> { any? } | number

Read a component-field across many entities in one call. Polymorphic on the shape of target and sink:

entity.batchRead(ids) / (ids, comp) / (ids, comp, field) — returns one value per entity (a whole snapshot, one component table, or one field value). Missing entities/components/fields surface as nil at that slot.
entity.batchRead(binding, comp, field, buffer) — reads each entity's field directly into a typed CPU substrate buffer (substrate.createBuffer({type="vec3"}), etc.) with no per-entity Lua table allocation. Returns the count of successful reads. Transform.worldMatrix reads each entity's world matrix — its own transform folded with its ancestors', the one entity(id).worldMatrix() answers — as 16 column-major numbers: into a mat4 buffer, or as an array of 16 per entity. It is read, never written. target accepts an entity-id array or a ecs.bindEntities(ids) handle. Buffer sinks require a binding — the typed kernel is binding-only.

Parameters

target { string | entityRef } | binding — Array of entity ids or entity proxies, or a binding handle from ecs.bindEntities(ids).
component string (optional) — Component type name (e.g. "Transform").
field string (optional) — Field name (e.g. "position").
sink TypedBuffer (optional) — Typed CPU buffer from substrate.createBuffer({...}) to memcpy field values into. Required when target is a binding.

Returns { any? } | number — Per-entity values when reading into Lua tables; count of reads when reading into a buffer.

local snapshot = entity.batchRead(ids)
local positions = entity.batchRead(ids, "Transform", "position")
entity.batchRead(ecs.bindEntities(ids), "Transform", "worldMatrix", matrices)

globals/entity/batchReadToBuffer

entity.batchReadToBuffer(binding: number, component: string, field: string, buffer: number) -> number

FFI primitive backing entity.batchRead(binding, ..., buffer). Prefer the unified entity.batchRead, which auto-dispatches by argument shape. Reads each entity's component field directly into a typed CPU substrate buffer, with no per-entity Lua table allocation. After the call, read the buffer via buf:read(0, count*stride).

Parameters

binding number — Binding id from ecs.bindEntities(ids).id, live while its handle is held.
component string — Component type name.
field string — Field name to read.
buffer number — Destination buffer id (must be the matching type).

Returns number — Count of successful reads.

entity.batchReadToBuffer(binding.id, "Transform", "position", buf.id)

globals/entity/batchSpawn

entity.batchSpawn(count: number, name_prefix: string?, opts: { temporary: boolean?, parent: (string | entityRef)? }?) -> { string }

Spawn count entities in one call. Returns an array of the new entity ids in spawn order. Each entity is given a display name of <name_prefix><i> (or entity<i> if the prefix is omitted). Prefer this over looping entity.spawn when creating large entity counts.

Parameters

count number — How many entities to spawn (capped at 1,000,000).
name_prefix string (optional) — Display-name prefix appended with the 1-based index. Defaults to "entity".
opts { temporary: boolean?, parent: (string | entityRef)? } (optional) — { temporary?, parent? } — a temporary batch is one the scene does not save, the same thing entity.spawn({ temporary = true }) asks for. The component goes on in one crossing, so a batch of a hundred thousand costs what one does. A batch spawned inside an entrypoint that is itself temporary is temporary whatever this says. parent — an id or an entity — makes every entity in the batch its child as it is made, so the batch stands, moves and goes with it.

Returns { string } — Array of newly-spawned entity ids.

local ids = entity.batchSpawn(100, "grass_")
local particles = entity.batchSpawn(5000, "spark_", { temporary = true })
local parts = entity.batchSpawn(8, "part_", { temporary = true, parent = owner })

globals/entity/batchWrite

entity.batchWrite(target: { string | entityRef } | binding, component: string, field: string, source: { any? } | TypedBuffer) -> number

Write a single component-field across many entities in one call. Polymorphic on the shape of target and source:

entity.batchWrite(ids, comp, field, values) — per-call entity-id resolution; values is an array the same length as ids (nil slots are skipped). Use for one-shot writes.
entity.batchWrite(binding, comp, field, values) — binding handle from ecs.bindEntities(ids); skips per-call id resolution. Use for per-frame writes against a stable entity set.
entity.batchWrite(binding, comp, field, buffer) — typed CPU buffer source (substrate.createBuffer({type="vec3"}), etc.), with no per-entity table allocation. Returns the count of successful writes. Buffer sources require a binding — the typed kernel is binding-only.

Parameters

target { string | entityRef } | binding — Array of entity ids or entity proxies, or a binding handle from ecs.bindEntities(ids).
component string — Component type name.
field string — Field name to write.
source { any? } | TypedBuffer — Per-entity values array (nil entries are skipped), or a typed CPU buffer from substrate.createBuffer({...}). A buffer source requires a binding target.

Returns number — Count of successful writes.

entity.batchWrite(ids, "Transform", "position", positions)

globals/entity/batchWriteBound

entity.batchWriteBound(binding: number, component: string, field: string, values: { any? }) -> number

FFI primitive backing entity.batchWrite(binding, ...) with a per-entity values table. Prefer the unified entity.batchWrite, which auto-dispatches by argument shape; this entry stays for power users / debug code that wants to skip dispatch overhead.

Parameters

binding number — Binding id from ecs.bindEntities(ids).id, live while its handle is held.
component string — Component type name.
field string — Field name to write.
values { any? } — Per-entity source values (nil = skip). Length must match the binding's entity count.

Returns number — Count of successful writes.

entity.batchWriteBound(binding.id, "Transform", "position", values)

globals/entity/batchWriteFromBuffer

entity.batchWriteFromBuffer(binding: number, component: string, field: string, buffer: number) -> number

FFI primitive backing entity.batchWrite(binding, ..., buffer). Prefer the unified entity.batchWrite, which auto-dispatches by argument shape. Caller fills a typed substrate buffer (substrate.createBuffer({type="vec3"})) once via buf:write(...), then this memcpys 12 (vec3) or 16 (quat) bytes per entity into the component field. Buffer count and binding count should match — a mismatch processes the smaller of the two.

Parameters

binding number — Binding id from ecs.bindEntities(ids).id, live while its handle is held.
component string — Component type name.
field string — Field name to write.
buffer number — Buffer id from substrate.createBuffer({type="vec3", len=N}).id.

Returns number — Count of successful writes.

entity.batchWriteFromBuffer(binding.id, "Transform", "position", buf.id)

globals/entity/capture

entity.capture(builder: () -> ()) -> ({ string }, any?, { string }, { any })

Run builder inside an entity capture scope and return the entity ids it minted, in creation order, the error it raised (if any), and the ids among them that a component the builder attached minted in its own lifecycle. Every id minted while the builder runs is recorded — through entity.spawn, entity.spawnSynced, entity.batchSpawn, and entity.instantiate alike. Scopes nest: an id minted inside an inner capture is recorded by that capture AND every enclosing one — the innermost capture answers, so a nested build shapes its own entities, not the ones around it. A builder that raises still returns its ids, so the caller can despawn what a failed build left behind; the scope closes either way and never outlives this call. While the builder runs, an operation whose result cannot be composed into a record is refused rather than applied, and so is any operation aimed at an entity the builder did not mint — a builder that returned while something it did was refused comes back with an error naming every refusal.

Parameters

builder () -> () — Function run inside the scope; the entities it creates are what comes back.

Returns ({ string }, any?, { string }, { any }) — Entity ids minted while the builder ran, in creation order; the error it raised (or the refusals it hit), or nil; the ids a component the builder attached minted in its own lifecycle; and the refusals the scope issued, each with message, at and source.

local ids, err, reproduced, refusals = entity.capture(function() entity.spawn("chair") end)

globals/entity/capturing

entity.capturing() -> boolean

Whether an entity.capture builder (a scene build) is running on the calling thread. A component a build adds takes the world's Preset Manager rows in every mode, so a scene built in a play session holds what the same build holds in edit mode.

Returns boolean — true inside a capture scope.

if entity.capturing() then return end

globals/entity/despawn

entity.despawn(target: string | entityRef)

Despawn an entity and all its components. Pass an id string or an entity proxy to despawn that ONE entity. Pass a name to despawn EVERY entity with that name — names are not unique, so a name argument despawns all matches, not one arbitrary match. A despawned id becomes invalid after this call. Raises if no entity matches; for a bulk name despawn, locked entities are skipped with a logged summary and only raise if every match is locked.

Parameters

target string | entityRef — Entity id, name, or entity proxy. A name despawns all entities sharing that name.

entity.despawn(id)
entity.despawn("Enemy") -- despawns every entity named "Enemy"

globals/entity/duplicate

entity.duplicate(sourceId: string | entityRef, name: string?, opts: table?) -> string?

Duplicate an entity with all its components (transform, script components, attributes, visuals, material) and its descendants. Returns the new entity's id, or nil when sourceId names no live entity. Descendants marked temporary are left out of the copy: they are scaffolding whatever spawned them re-creates, so a component that regenerates its own children rebuilds them on the copy rather than the copy carrying a second set. includeTemporary copies them too, for the hierarchy that IS the temporary thing.

Parameters

sourceId string | entityRef — Entity id or entity proxy of the source entity to clone.
name string (optional) — Display name for the copy (defaults to source name + " (copy)").
opts table (optional) — { includeTemporary?: boolean, descendants?: boolean, components?: table, name?: string }: descendants = false copies the source entity alone, leaving its children to it; components lays fields over the copy's script components by type ({ SoftBody = { mass = 2 } }), merged over the state copied from the source; name is the same field the name argument sets, and wins when both are given.

Returns string? — The new entity's id, or nil when the source is not live.

local copyId = entity.duplicate(id); if copyId then entity(copyId).position = { 1, 0, 0 } end
local copyId = entity.duplicate(id, "Turret", { includeTemporary = true })
local lone = entity.duplicate(id, nil, { descendants = false, components = { SoftBody = { mass = 2 } } })

globals/entity/exists

entity.exists(idOrProxy: string | entityRef) -> boolean

Check whether an entity currently exists in the scene. Accepts an entity-id string or an entity proxy, matched by entity id — so it agrees exactly with entity(id). A name is a different kind of identifier: a string that misses as an id but names a live entity raises rather than answering false, since false there is indistinguishable from absence. Check by name with entity.find(name) ~= nil.

Parameters

idOrProxy string | entityRef — Entity id or an entity proxy.

Returns boolean — true if the entity exists.

if entity.exists(id) then ... end

globals/entity/find

entity.find(nameOrGlob: string) -> entityRef?

Find the first entity matching nameOrGlob. A plain string matches an exact id or Name component; a string containing * (any run of characters) or ? (any single character) matches Names as a glob, so entity.find("enemy_*") is the first entity whose name starts with enemy_. A glob addresses Names only, never ids. Same-frame pending spawns are searched too, and anything queued for despawn in the same frame is skipped. Names are NOT unique — use entity.findAll when every match matters; find is the first entity findAll lists. Inside a build (a scene's content(), a module's builder), a name the build's own entities carry resolves to those first, so entity.spawn("door") followed by entity.find("door") is the door the build just made even when the world around it holds another door.

Parameters

nameOrGlob string — Exact entity Name or id, or a * / ? glob over Names.

Returns entityRef? — First matching entity proxy, or nil.

local e = entity.find("enemy_*")

globals/entity/findAll

entity.findAll(nameOrGlob: string?) -> { entityRef }

Enumerate entity proxies. With a nameOrGlob argument, returns every entity whose Name component or id matches (names are not unique): a plain string matches exactly, while a * / ? glob matches Names. With no argument, returns every entity in the current snapshot — findAll("") is the exact-match filter for the empty name, which normally matches nothing. Same-frame pending spawns are included and same-frame despawns filtered out. Elements are entity proxies, not id strings — for ids, wrap the result: entity.ids(entity.findAll(...)). Inside a build (a scene's content(), a module's builder), the entities the build made come first and the world's same-named entities follow.

Parameters

nameOrGlob string (optional) — Exact entity Name or id to filter by, or a * / ? glob over Names. Omit to enumerate every entity.

Returns { entityRef } — Array of entity proxies, possibly empty.

for _, e in entity.findAll("enemy_*") do e:despawn() end

globals/entity/getChildren

entity.getChildren(id: string | entityRef) -> { entityRef }

Get an array of the direct children as entity proxies. Each element carries .name, .id, .position, .component, and the rest of the per-entity surface, the same shape entity.findAll returns. A string that names a live entity rather than identifying it raises, naming entity.find(name), the lookup that takes a name.

Parameters

id string | entityRef — Entity id or entity proxy.

Returns { entityRef } — Array of child entity proxies, possibly empty.

for _, c in entity.getChildren(id) do c.internal = true end

globals/entity/getDescendants

entity.getDescendants(id: string | entityRef) -> { entityRef }

Get every descendant (children, grandchildren, and deeper) of the given entity as entity proxies in breadth-first order, excluding the entity itself. Resolves the whole subtree in one linear pass over the entity set, so a large subtree costs proportionally to the entity count rather than to the subtree size times the entity count. A string that names a live entity rather than identifying it raises, naming entity.find(name), the lookup that takes a name.

Parameters

id string | entityRef — Entity id or entity proxy.

Returns { entityRef } — Array of descendant entity proxies, possibly empty.

local all = entity.getDescendants(id)

globals/entity/getParent

entity.getParent(id: string | entityRef) -> entityRef?

Get the parent entity proxy, or nil if the entity is a root entity. The returned proxy carries .name, .id, .position, .component, and the rest of the per-entity surface — the same shape entity.find returns. A string that names a live entity rather than identifying it raises, naming entity.find(name), the lookup that takes a name: answering it as an id would read the same as a root entity.

Parameters

id string | entityRef — Entity id or entity proxy.

Returns entityRef? — Parent entity proxy, or nil.

local p = entity.getParent(id)

globals/entity/hierarchy

entity.hierarchy()

Hierarchy helpers reached as entity.hierarchy.<member>.

globals/entity/hierarchy/swap

entity.hierarchy.swap(entityId: string, assetPath: string, opts: { [string]: any }?) -> (string?, string?)

Replace a blockout entity with a generated or imported asset, fitting the asset to the source's bounds. Answers the new entity id, or nil and an error string.

Parameters

entityId string
assetPath string
opts { [string]: any } (optional)

Returns (string?, string?)

globals/entity/ids

entity.ids(value: any) -> { string }

Coerce an entity, a list of entities, or nil into a list of entity id strings. Accepts id strings and entityRefs in any mix, so the entity-answering calls and a stored list of ids both feed it. A single entity becomes a one-element list and nil becomes an empty list. Order is preserved and duplicates are kept.

Parameters

value any — An entity, a list of entities, or nil.

Returns { string } — Array of entity id strings.

local ids = entity.ids(entity.getChildren(root))
local ids = entity.ids(entity.findAll())

globals/entity/instantiate

entity.instantiate(handle: EntityTemplate, count: number, fn: ((number) -> EntityInstantiateOverrides?)?) -> { string }

Spawn count instances of a template registered with entity.template. Each instance gets a fresh entity id; the optional fn(i) callback runs per instance (i in 1..=count) and may return an overrides table. Override keys: name, position, rotation, scale, parent, temporary / active / internal, attributes, components (script components, merged over the template body's data for that type — a type the template lacks is added fresh), and ecs (native components, merged the same way). Each override supersedes the template's shared config for that instance. The whole batch crosses in one call and lands as a single deferred mutation the engine expands into bulk work — per-instance cost drops from a full round trip to one callback plus one mutation. Inside queue() the batch is deferred onto the cross-frame ring; outside, it lands in the next frame's drain. Returns the array of newly-minted entity ids in spawn order.

Parameters

handle EntityTemplate — Template handle from entity.template.
count number — Number of instances to spawn (capped at 1,000,000).
fn ((number) -> EntityInstantiateOverrides?) (optional) — Per-instance override callback (i) -> table?.

Returns { string } — Array of newly-spawned entity ids in spawn order.

local ids = entity.instantiate(h, 50, function(i) return { position = { i, 0, 0 } } end)

globals/entity/proxies

entity.proxies(value: any) -> { entityRef }

Coerce an entity, a list of entities, or nil into a list of entityRefs.

Parameters

value any — An entity, a list of entities, or nil.

Returns { entityRef } — Array of entityRefs.

globals/entity/refs

entity.refs(value: any) -> { entityRef }

Coerce an entity, a list of entities, or nil into a list of entityRefs — the form carrying .position, .component, and the rest of the per-entity surface. Accepts id strings and entityRefs in any mix. A single entity becomes a one-element list and nil becomes an empty list. An id whose entity has despawned raises, naming its position.

Parameters

value any — An entity, a list of entities, or nil.

Returns { entityRef } — Array of entityRefs.

for _, e in entity.refs(savedIds) do e.position = { 0, 1, 0 } end

globals/entity/sceneWalks

entity.sceneWalks() -> number

How many whole-scene walks entity.findAll has made since the module loaded. A walk is priced by the number of entities the scene holds, so a caller that must not scale with the scene — a per-frame push, a write that reaches its target through a record — is read either side of its work against this count, which a walk raises by one.

Returns number — The count of walks.

local before = entity.sceneWalks(); entity(id):rename("door"); assert(entity.sceneWalks() == before)

globals/entity/spawn

entity.spawn(nameOrOpts: (string | SpawnOpts)?, opts: SpawnOpts?) -> entityRef

Spawn a new entity and return its PROXY (the same value entity(id) yields) — act on it immediately (entity.spawn(name).component.add(...), .localPosition = ...) with no second entity(id) round trip. The proxy still exposes .id for the rare site that needs the raw string. An entity with no components carries only a Transform and is invisible; pass components to give it the components that make it visible in the same call — entity.spawn { name = "crate", components = { Model = { model = "cube" } } } — or add them afterwards through the returned proxy. Mirrors entity.find / entity.findAll, which also return proxies. The options table can be passed on its own with the name inside it — entity.spawn { name = "turret", position = { 1, 2, 3 } } is the same call as entity.spawn("turret", { position = { 1, 2, 3 } }).

Parameters

nameOrOpts (string | SpawnOpts) (optional) — Display name for the entity, or the options table itself.
opts SpawnOpts (optional) — Options: components = component types to attach to the new entity, keyed by type name with each value the component's init table (attached in sorted type order; a failing add raises), internal = take the entity out of the default entity listings (it still renders — entity(id):hide() stops the draw), parent = parent entity id or proxy, temporary = skip this entity (and descendants) from scene/world saves, position / rotation / scale = place the entity's Transform at spawn, attributes = per-entity attribute values keyed by name ({ role = "guard" }, the same values entity(id).attribute.set(key, value) writes and entity.template / entity.instantiate take), synced = flag the entity multiplayer-synced (what entity.spawnSynced implies), active = false spawns it inactive, scene_owner = the scene layer guid the entity is attributed to (inherited from the spawning scene when unset), id = restore a previously-assigned entity id (scene_loader use; leave unset for a normal spawn). An unrecognised key is rejected loudly.

Returns entityRef — Proxy for the new entity (carries .id, .component, transform properties, etc.).

local e = entity.spawn("crate", { components = { Model = { model = "cube" } } })
local e = entity.spawn { name = "turret", position = { 1, 2, 3 } }
local guard = entity.spawn("guard", { attributes = { role = "guard", hp = 3 } })

globals/entity/spawnSynced

entity.spawnSynced(nameOrOpts: (string | SpawnOpts)?, opts: SpawnOpts?) -> entityRef

Spawn an entity already flagged multiplayer-synced at the root — the explicit form of entity.spawn for SHARED, host-authoritative content. The entity's existence broadcasts to every peer; joiners receive it from the relay snapshot instead of spawning their own copy. Use this (NOT entity.spawn) for anything that must be the SAME object on all clients: enemies, pickups, projectiles, dynamic world props. Call it ONLY where exactly one client runs the code — a scene's onHostLoad (host-only) phase, or behind multiplayer.isHost(). Calling it in all-client code makes every client spawn+sync its own copy — the double-spawn "explosion". Equivalent to entity.spawn(name, { synced = true }); identical in every other respect.

Parameters

nameOrOpts (string | SpawnOpts) (optional) — Display name for the entity, or the options table itself.
opts SpawnOpts (optional) — Same options as entity.spawn (synced is already implied).

Returns entityRef — Proxy for the new synced entity.

if multiplayer.isHost() then entity.spawnSynced("Goblin") end

globals/entity/template

entity.template(def: EntityTemplateDef) -> EntityTemplate

Construct a reusable spawn template. Captures a shared entity config ONCE and returns an EntityTemplate handle for entity.instantiate(handle, count, fn?) — one call per batch instead of one per entity. The template lives as long as a script holds the handle: once nothing reaches it, the collector releases the template, and the entities already instantiated from it stay as they are. Keep the handle (a module-level local, a field) for as long as more instances may follow. def keys: components (script components, { [type] = init-data }), ecs (array of native ecs.X{...} components), temporary (instances skip scene/world saves), active (spawn state), internal (instances are taken out of the default entity listings; they still render), attributes ({ key = value } applied to every instance). Every value is a shared default; a per-instance entity.instantiate override supersedes it. The template body is captured by value — later edits to the source table do not affect templates already created.

Parameters

def EntityTemplateDef — Template definition: components / ecs / temporary / active / internal / attributes. Per-instance name / position / rotation / scale / parent and any override go through the instantiate callback.

Returns EntityTemplate — The template's handle for entity.instantiate.

local h = entity.template({ components = { Model = { model = "cube" } } })

globals/entity/tree

entity.tree(opts: { [string]: any }?) -> { [string]: any }

A windowed, lean view over the scene's entity tree, in one crossing. Rows carry id, name, parentId, depth, childCount, active, sceneLayer and componentNames — names only, never component values — so the call costs the rows it answers with rather than the size of the scene. Entities group under scene layers, per-layer roots and children name-sorted; internal entities and their subtrees stay out. expanded names the ids whose children unfold, and a collapsed node still reports its childCount; filter keeps the rows whose name or id contains the needle plus every ancestor on a path to one, auto-unfolded, with the actual matches flagged matched. offset / limit window the flattened rows, layer scopes the window and its total to one layer while layers still reports every layer's row count, and revision echoes getEntitiesRevision("structure"), which moves only on structural change.

Parameters

opts { [string]: any } (optional) — { layer?, expanded?, filter?, offset?, limit? }.

Returns { [string]: any } — { rows, total, layers, revision }.

local view = entity.tree({ filter = "crate", limit = 50 })

globals/environment/capture

environment.capture(x: number, y: number, z: number) -> boolean

Bake the scene into the environment from (x, y, z) as the single global reflection (slot 0 + one full-coverage probe). Every PBR surface reflects it. Queued — takes effect on the next frame. For multiple proximity-blended probes use the reflectionProbe system instead.

Parameters

x number — World X of the capture position.
y number — World Y of the capture position.
z number — World Z of the capture position.

Returns boolean — True — the capture was queued.

environment.capture(0, 2, 0)

globals/environment/captureSky

environment.captureSky(x: number?, y: number?, z: number?) -> boolean

Render the SKY alone into the environment's sky slot from (x, y, z) and arm the sky fallback. A reflective surface no probe covers then reflects the sky rather than black, and a partially covered one blends the shortfall against it. The capture holds whatever the scene's sky draws — a gradient, a physical atmosphere, a skybox material — with no geometry in it, so it stays correct wherever the camera goes. Once captured, the slot follows the sky the scene draws: a sky that changes is recaptured from the same position. Queued — takes effect on the next frame.

Parameters

x number (optional) — World X of the capture position. Defaults to 0.
y number (optional) — World Y of the capture position — the altitude a height-dependent atmosphere is sampled at. Defaults to 0.
z number (optional) — World Z of the capture position. Defaults to 0.

Returns boolean — True — the sky capture was queued.

environment.captureSky()

globals/environment/captureSlot

environment.captureSlot(slot: number, x: number, y: number, z: number) -> boolean

Bake the scene into reflection-probe slot from (x, y, z). Renders the FULL scene (geometry + sky) six times from that point into that slot. Register the probe's position+radius via setProbes so surfaces blend it by proximity. Queued — takes effect next frame.

Parameters

slot number — Reflection-probe slot (0-based).
x number — World X of the capture position.
y number — World Y of the capture position.
z number — World Z of the capture position.

Returns boolean — True — the capture was queued.

environment.captureSlot(0, 0, 2, 0)

globals/environment/captureSlotToAsset

environment.captureSlotToAsset(name: string, slot: number, x: number, y: number, z: number, timeoutFrames: number?) -> (string?, string?)

Bake the scene into reflection-probe slot from (x, y, z) AND persist the 6 rendered faces into a faces6 .texture cubemap asset at /source/<name>.texture/ (px/nx/py/ny/pz/nz PNGs + a cube.yaml sidecar). Survives an engine restart and syncs like any other texture. Yields a few frames while the bake + GPU readback complete; must be called from a task/coroutine context (component hook, task.spawn, or execute).

Parameters

name string — Destination asset identity (writes /source/<name>.texture/).
slot number — Reflection-probe slot (0-based).
x number — World X of the capture position.
y number — World Y of the capture position.
z number — World Z of the capture position.
timeoutFrames number (optional) — Optional max frames to wait for the readback (default 900, past the frames the engine gives a readback before it gives up on one, so what ends the wait is the readback's own answer).

Returns (string?, string?) — The asset path on success, or (nil, errorMessage) on failure.

environment.captureSlotToAsset("probe_lobby", 0, 0, 2, 0)

globals/environment/captureToAsset

environment.captureToAsset(name: string, x: number, y: number, z: number) -> (string?, string?)

Bake the single global reflection AND persist it to a faces6 .texture asset (slot 0). Yields a few frames; call from a task/coroutine context.

Parameters

name string — Destination asset identity (writes /source/<name>.texture/).
x number — World X of the capture position.
y number — World Y of the capture position.
z number — World Z of the capture position.

Returns (string?, string?) — The asset path on success, or (nil, errorMessage) on failure.

environment.captureToAsset("env_main", 0, 2, 0)

globals/environment/ensureSkyFallback

environment.ensureSkyFallback() -> boolean

Ensure the scene's sky is in the environment's sky slot: a reflective surface no probe covers then reflects the sky rather than black, and a partially covered one blends the shortfall against it. Queues a capture when the sky slot holds none, and re-arms the fallback when a capture is there but switched off. The engine's own state answers both questions, so everything that stands a sky up can call this and one capture is shared between them. Once captured, the slot follows the sky the scene draws on its own.

Returns boolean — True if a capture was queued, false if the sky slot already holds one.

environment.ensureSkyFallback()

globals/environment/loadFromAsset

environment.loadFromAsset(name: string) -> (boolean, string?)

Load a persisted global reflection asset into slot 0 and make it the active single reflection (one full-coverage probe).

Parameters

name string — Source asset identity (reads /source/<name>.texture/).

Returns (boolean, string?) — True on success, or (false, errorMessage) on failure.

environment.loadFromAsset("env_main")

globals/environment/loadSlotFromAsset

environment.loadSlotFromAsset(name: string, slot: number) -> (boolean, string?)

Load a persisted faces6 .texture cubemap (written by captureSlotToAsset) into reflection-probe slot WITHOUT re-rendering the scene. Reads the 6 face PNGs from /source/<name>.texture/ and uploads them into the slot's cube layers. How a persisted probe restores its baked environment on reload.

Parameters

name string — Source asset identity (reads /source/<name>.texture/).
slot number — Reflection-probe slot (0-based).

Returns (boolean, string?) — True on success, or (false, errorMessage) on failure.

environment.loadSlotFromAsset("probe_lobby", 0)

globals/environment/setProbes

environment.setProbes(probes: { any }) -> boolean

Set the active reflection probes' blend data. probes is an array of { x, y, z, radius } (or { position = {x,y,z}, radius = r }); index i is probe slot i. Surfaces blend the probe slots by proximity to these positions, gathering the highest priority first — each rank takes the coverage the ranks above it left, so a small interior probe ranked above a large exterior one wins outright wherever it reaches full weight. Coverage left over reflects the sky once captureSky has run. Queued for next frame.

Parameters

probes { any } — Array of { x, y, z, radius, priority? }, one per active probe slot. priority defaults to 0.

Returns boolean — True — the probe data was queued.

environment.setProbes({ { x = 0, y = 2, z = 0, radius = 12 } })

globals/environment/setSkyFallback

environment.setSkyFallback(active: boolean) -> boolean

Arm or disarm the sky fallback against the sky already captured, with no recapture. Disarmed, reflections come from the probes alone. Arming is refused while the sky slot holds no capture (captureSky fills it), since an uncaptured slot reflects black; renderer.reflectionEnvironment() reports whether the fallback ended up armed.

Parameters

active boolean — Whether reflections fall back to the captured sky.

Returns boolean — True — the change was queued.

environment.setSkyFallback(false)

globals/environment/stateFromAsset

environment.stateFromAsset(identity: string) -> (boolean, string?)

Make a faces6 .texture cubemap the environment that cameras stating it reflect. A camera rendering an isolated environment reflects only what it states — never the world's sky or reflection probes — so this is what a mirror or a pane of glass in its frame shows. The cube is stated under identity, which is the name a camera's environmentMap gives it; stating one already stated, from the same bytes, reads nothing. Called from a task, it waits for faces whose bytes are still being fetched (the web build fetches builtin images); called outside one, a face not yet fetched is answered as missing.

Parameters

identity string — The cube texture, by the identity asset.resolve takes.

Returns (boolean, string?) — True once the environment is queued, or (false, errorMessage).

task.spawn(environment.stateFromAsset, "@builtin::textures.studio_environment")

globals/error

error(message, level?)

Raise an error. Halts execution.

globals/font/coverage

font.coverage(bytes: buffer | string) -> string?

The codepoints a font file maps to a glyph, as the coverage text font.declare reads: comma-separated inclusive hexadecimal ranges.

Parameters

bytes buffer | string — Raw font-file bytes (TTF / OTF).

Returns string? — Coverage text, or nil if the bytes don't parse as a font.

local cov = font.coverage(assert(vfs.read("/zero/source/fonts/MyCJK.otf")))

globals/font/declare

font.declare(name: string, coverage: string, opts: table) -> number

Declare an on-demand face: a face known by the codepoints it covers, whose bytes are read only when a text surface first meets one of those codepoints with no installed face that draws it. Until then the face costs its coverage and nothing else, so a session that never shows its script never reads its file. When a label, a text.* object, or font.prefetch reaches a covered character, the engine requests the face: its file is read and registered with font.register(name, bytes, opts), and every text surface draws the character from it from then on. A .font asset whose face.json lists onDemand chunks declares each of them through this call.

Parameters

name string — The name the face is declared and registered under.
coverage string — The codepoints it covers: comma-separated inclusive hexadecimal ranges ("3040-30ff,4e00"), the text font.coverage(bytes) writes.
opts table — { file: string, family: string?, role: string?, aliases: {string}? } naming the VFS path of the font file (TTF / OTF, or a baked ZFNT) and the font.register options it installs with.

Returns number — How many codepoints the coverage holds.

font.declare("MyCJK", assert(font.coverage(assert(vfs.read("/zero/source/fonts/MyCJK.otf")))), { file = "/zero/source/fonts/MyCJK.otf" })

globals/font/glyph

font.glyph(name: string, codepoint: number) -> any

Read one glyph's vectorized outline from a registered font, in font units (resolution-independent — scale by fontSize / unitsPerEm).

Parameters

name string — Registered family name.
codepoint number — Unicode codepoint (e.g. string.byte("A")).

Returns any — { advance, unitsPerEm, bbox = {xMin,yMin,xMax,yMax}, contours } where each contour is { start = {x,y}, segments = { {kind="line|quad|cubic", ...} } }, or nil if the font isn't registered.

local g = font.glyph("Inter", string.byte("A"))

globals/font/list

font.list() -> { string }

List every registered font family name.

Returns { string } — Array of family-name strings.

for _, fam in font.list() do print(fam) end

globals/font/observe

font.observe() -> { any }

What the text system is holding for fonts: one row per family the shaper can resolve, with its face count, the numeric weights those faces carry, whether any of them is slanted, and whether the family arrived through a registration rather than from the platform. weights is what a style's weight can name for that family. The same rows are fonts in text.observe().

Returns { any } — Array of { family, faces, weights, italic, loaded }.

for _, f in ipairs(font.observe()) do print(f.family, #f.weights) end

globals/font/onDemand

font.onDemand() -> { any }

Every on-demand face font.declare declared, in declaration order, and where each stands: declared (its file never read), requested (a character it covers reached a text surface that could not draw it, and its file is being read), installed (registered on every text surface; bytes is the size of the file it installed from) or failed (error says why). requestedBy is the character whose miss requested it, and family the family group it installs into (opts.family of its font.declare, else its own name); ui.listFonts() lists each face under that family's row.

Returns { any } — Array of { name, family, state, codepoints, requestedBy, bytes, error }.

for _, f in ipairs(font.onDemand()) do print(f.name, f.state, f.requestedBy) end

globals/font/parse

font.parse(bytes: buffer | string) -> string?

Parse a font file (TTF / OTF raw bytes) ONCE into the baked, vectorized glyph format (ZFNT): per-glyph vector outlines + metrics + character map, plus the original bytes. Heavy — run at import time (the .font assetType's onCreate / the font importer), then store the result as the asset payload. font.register loads it cheaply.

Parameters

bytes buffer | string — Raw font-file bytes (binary-safe) — TTF / OTF.

Returns string? — Baked ZFNT payload (binary-safe string), or nil if the bytes don't parse as a font.

local zfnt = font.parse(assert(vfs.read("/zero/source/Inter.ttf")))

globals/font/prefetch

font.prefetch(text: string) -> number

Request, ahead of drawing it, every on-demand face covering a character of text that no requested or installed face covers: a screen about to show a script starts reading its faces before its first frame.

Parameters

text string — Text whose characters the faces must cover.

Returns number — How many faces this call requested.

font.prefetch("日本語")

globals/font/ready

font.ready(text: string) -> boolean

Whether text draws without waiting on an on-demand face: no character of it is covered by a face still declared or requested. Call font.prefetch(text) first, then open the screen once this answers true, and its first frame draws every glyph. True as well for a character no declared face covers (ui.missingGlyphs() names it once drawn) and for a face that failed (font.onDemand() says why).

Parameters

text string — Text to be drawn.

Returns boolean — Whether every on-demand face the text needs is installed.

font.prefetch(title); task.spawn(function() while not font.ready(title) do task.wait() end; ui.showScreen("title") end)

globals/font/reconcile

font.reconcile() -> { any }

Every family the text shaper can resolve, held against what the shaper does with it. family is the name, faces how many faces of it the font database holds, weights the numeric weights those faces carry, loaded whether it arrived through a registration rather than from the platform, registered whether content registered the name, selectable whether some style naming the family reaches it, matched whether fontFamily = family on its own reaches it — the family name at the default weight over Latin text — weight the weight it needs when the default is not it, shapedWith the family the name selects, spelled the way the font file spells it, which for a registration name is the family declared inside the file rather than the name the asset carries, and reason why when the face that answered is not that family. A family is probed at its own weights and over content from several scripts, so a family reachable only at one weight or covering only one script is reported selectable, with matched false and weight naming what the style must carry. Every probe object is destroyed again, so the live text-object count is where it was.

Returns { any } — Array of { family, faces, weights, loaded, registered, selectable, matched, weight, shapedWith, reason }.

for _, f in ipairs(font.reconcile()) do if f.selectable and not f.matched then print(f.family, f.weight) end end

globals/font/register

font.register(name: string, zfnt: string, opts: table?) -> any

Register a baked font (ZFNT from font.parse) under name, making it usable on every text surface via fontFamily = "<name>". Loads the vectorized glyph data into the runtime store (for font.glyph / font.textMesh) and feeds the embedded face to the 2D text and egui UI systems. Raw font bytes (TTF / OTF) register too: their tables are read and the glyph outlines are baked the first time font.glyph / font.textMesh reads one. A name declared with font.declare is installed by this call. Registering a name again replaces what the earlier call installed under it: a name, group or slot the earlier call gave the face and this one does not stops selecting it, and text laid out under a name whose face changed is laid out again. opts groups several weight/style faces under one CSS family and maps web-font names onto it: opts.family is the shared group key, opts.role is "regular" | "bold" | "italic" | "bolditalic", and opts.aliases is a list of extra selectable names (web fonts + CSS generics like "Arial", "sans-serif") that resolve to this group, matched case-insensitively. With a group set, font-weight / font-style on a font-family pick the real metric-compatible face instead of a synthesized one.

Parameters

name string — Family name to register under.
zfnt string — Baked ZFNT payload from font.parse (binary-safe string).
opts table (optional) — { family: string?, role: string?, aliases: {string}? } — group key, weight/style role, and case-insensitive selectable aliases.

Returns any — { family, faces, glyphCount } on success, or nil on failure.

local info = font.register("Inter", assert(font.parse(assert(vfs.read("/zero/source/Inter.ttf")))))

globals/font/textMesh

font.textMesh(name: string, text: string, opts: table?) -> any

Tessellate a string into renderable mesh geometry from a registered font's glyph outlines — true 3D text, laid out left-to-right by advance (newlines drop a line). Hand the result to renderer.mesh.create() (GPU) or asset.create("mesh") (persistable).

Parameters

name string — Registered family name.
text string — String to lay out.
opts table (optional) — { size?=1, depth?=0 (extrude, EM units), tolerance?=0.0015, letterSpacing?=0, lineHeight?=0 }.

Returns any — { positions, indices, normals, uvs } as flat float / u32 arrays, or nil if the font isn't registered or the string is all whitespace.

local geom = font.textMesh("Inter", "Hello", { size = 1, depth = 0.1 })

globals/frameStream/attach

frameStream.attach(texture: string, stream: string, opts: AttachOpts?) -> (string?, string?)

Carry an image the GPU drew out to an open byte stream, frame after frame. texture is the guid of the render target it was drawn into — renderer.texture.create({ width = W, height = H }) makes one, and a Camera component draws into it as its textureHandle; the session reads that target back when a frame comes due, so what the camera drew last reaches the far end. stream is a handle from stream.open. What reaches the stream is one frame's pixels then the next frame's, with nothing between them: a frame is width * height * bytesPerPixel bytes of tight rows, written in a single call so a consumer reads a whole frame or none of it. Each frame is read back off the render thread, so the stream never holds the renderer up. fps caps how often a frame is taken and defaults to one per rendered frame; format accepts "rgb24" (3 bytes per pixel, the default) or "rgba8" (4) — a call with a format outside those two raises, naming both; flipY writes the last texture row first. Returns the session handle, or nil and the reason an empty texture, a handle naming no open stream, a stream another session already carries, or a non-positive fps was refused with.

Parameters

texture string — Guid of the render target the image was drawn into (a Camera's textureHandle).
stream string — Stream handle from stream.open.
opts AttachOpts (optional) — Rate, pixel layout and row order (optional).

Returns (string?, string?) — Session handle, or nil and the refusal reason.

local session = frameStream.attach(rt.guid, handle, { fps = 30 })

globals/frameStream/detach

frameStream.detach(handle: string) -> boolean

End the session and free the staging buffers it read frames back through. The stream stays open — whoever opened it closes it.

Parameters

handle string — Session handle from frameStream.attach.

Returns boolean — True if a session was ended, false if handle already named none.

frameStream.detach(session)

globals/frameStream/list

frameStream.list() -> { string }

Every live session handle, in a stable order.

Returns { string } — Array of session handles.

for _, h in frameStream.list() do frameStream.detach(h) end

globals/frameStream/status

frameStream.status(handle: string) -> FrameStreamStatus?

Report what the session has carried and lost. frames counts the frames the stream accepted and bytes the bytes they carried. dropped counts the frames it refused, of which droppedBackpressure is the part refused because the consumer was behind; stalledReadbacks counts the frames that came due while every staging buffer still held a copy on its way from the GPU. achievedFps is the rate the accepted frames arrived at, across the span from the first to the most recent, and reads 0 until two have been accepted — compare it against requestedFps to see a display running slower than it was asked to. lastCopyFrame is the engine frame, in time.frameCount() numbering, the copy of the most recent accepted frame was taken on, and is nil until one is accepted: every copy one engine frame takes of a texture holds the same image, so two sessions on one texture whose lastCopyFrame agree hold the same image as the newest frame of their streams. lastOutcome names what became of the most recent frame offered. nil when handle names no live session.

Parameters

handle string — Session handle from frameStream.attach.

Returns FrameStreamStatus? — Session status, or nil when handle names no live session.

local s = frameStream.status(session); print(s.frames, s.dropped, s.achievedFps)

globals/getLookHitTable

getLookHitTable() -> table | nil

Get the physics raycast hit at screen center (crosshair).

Returns table | nil — Hit result {entityId, point, normal, distance} or nil

globals/getPointerHitTable

getPointerHitTable() -> table | nil

Get the physics raycast hit under the mouse pointer.

Returns table | nil — Hit result {entityId, point, normal, distance} or nil

globals/getTime

getTime() -> number

Seconds since the engine started, stamped once per frame: so every script in a frame reads the same now and the frame's logic is consistent with itself. It reads the WALL clock: engine.timeScale does not scale it and engine.paused does not stop it, so across a paused span this advances by the whole span while update(dt) receives nothing. The clock gameplay and animation run on is the dt a lifecycle callback receives, and task.wait counts in that same scaled clock. Because it holds still for the whole frame, it cannot measure a duration INSIDE one: two reads with no yield between them return the same number, and os.clock() is the live monotonic clock for that. One epoch on every platform. Frames are a third quantity again: time.frameCount() counts the frames the engine ran and time.drawnFrameCount() the frames it drew: and how many of either a second buys is decided by the machine, so waiting for engine work counts frames rather than seconds. The engine keeps stepping between tool calls, so both this clock and the gameplay one advance across the real seconds between two of them; the core/engine guide's "How fast time passes" has the whole model.

Returns number — Seconds since the engine started, as of this frame

globals/getViewportSize

getViewportSize() -> { width: number, height: number }

Get the viewport dimensions in pixels as a single table { width, height }. Returns one table, not two numbers.

Returns table — { width = number, height = number } in pixels

globals/getmetatable

getmetatable(table) -> mt | nil

Get a table's metatable.

globals/home/folder

home.folder() -> (string?, string?)

The home folder mounted in this session, /zero/home/<username>, or nil and the reason no home folder is mounted.

Returns (string?, string?) the folder, or nil and why there is none.

local folder, why = home.folder()

globals/home/status

home.status() -> Status

Where the home folder stands in this session. state is "mounted" (with folder, the path to read and write, owner, the account it belongs to, world, the guid of the home world that holds it, files, how many files the home world held when it was mounted, and from, "launch" or "zeromind", saying where that world was named), "mounting" while a world load is mounting it, or "unmounted" with reason saying why no home folder is mounted.

Returns Status — The home folder's status.

local s = home.status(); if s.state == "mounted" then vfs.write(s.folder .. "/notes.md", "hi") end

globals/http/get_bytes

http.get_bytes(url: string, headers: Headers?, timeoutMs: TimeoutMs?, expectedStatuses: ExpectedStatuses?) -> PromiseId

Async HTTP GET returning raw bytes (binary-safe string). Suitable for piping into vfs.write to download a file.

Parameters

url string — Request URL.
headers Headers (optional) — Header key-value pairs (optional).
timeoutMs TimeoutMs (optional) — How long the request is given to be answered, in milliseconds; the engine's own bound when left out, none when 0 (optional).
expectedStatuses ExpectedStatuses (optional) — HTTP statuses this caller handles as an answer it expected, e.g. { 404 }: a response with one settles the future as any other non-2xx does and is logged as information rather than an error (optional).

Returns PromiseId — A future (PromiseHandle): f:await() or task.await(f) delivers the response, f:state() reads whether it has.

local bytes = task.await(http.get_bytes("https://example.com/sound.ogg"))

globals/http/get_json

http.get_json(url: string, headers: Headers?, timeoutMs: TimeoutMs?, expectedStatuses: ExpectedStatuses?) -> PromiseId

Async HTTP GET returning JSON. Returns a future (a PromiseHandle) at once, with the request in flight: keep it, f:await() (or task.await(f)) when the response is needed, and f:state() to see whether it has arrived without waiting.

Parameters

url string — Request URL.
headers Headers (optional) — Header key-value pairs (optional).
timeoutMs TimeoutMs (optional) — How long the request is given to be answered, in milliseconds; the engine's own bound when left out, none when 0 (optional).
expectedStatuses ExpectedStatuses (optional) — HTTP statuses this caller handles as an answer it expected, e.g. { 404 }: a response with one settles the future as any other non-2xx does and is logged as information rather than an error (optional).

Returns PromiseId — A future (PromiseHandle): f:await() or task.await(f) delivers the response, f:state() reads whether it has.

local data = task.await(http.get_json("https://api.example.com/info"))
local record = task.await(http.get_json(recordUrl, nil, nil, { 404 })) -- a missing record is an answer

globals/http/post_bytes

http.post_bytes(url: string, headers: Headers?, body: JsonBody?, timeoutMs: TimeoutMs?, expectedStatuses: ExpectedStatuses?) -> PromiseId

Async HTTP POST returning raw bytes — use for APIs that accept JSON input but return binary output (audio, images).

Parameters

url string — Request URL.
headers Headers (optional) — Header key-value pairs (optional).
body JsonBody (optional) — JSON body (optional).
timeoutMs TimeoutMs (optional) — How long the request is given to be answered, in milliseconds; the engine's own bound when left out, none when 0 (optional).
expectedStatuses ExpectedStatuses (optional) — HTTP statuses this caller handles as an answer it expected, e.g. { 404 }: a response with one settles the future as any other non-2xx does and is logged as information rather than an error (optional).

Returns PromiseId — A future (PromiseHandle): f:await() or task.await(f) delivers the response, f:state() reads whether it has.

local audio = task.await(http.post_bytes(ttsUrl, nil, { text = "hello" }))

globals/http/post_json

http.post_json(url: string, headers: Headers?, body: JsonBody?, timeoutMs: TimeoutMs?, expectedStatuses: ExpectedStatuses?) -> PromiseId

Async HTTP POST returning JSON. Body is a Luau table; the FFI layer JSON-encodes it before the request goes out.

Parameters

url string — Request URL.
headers Headers (optional) — Header key-value pairs (optional).
body JsonBody (optional) — JSON body (optional).
timeoutMs TimeoutMs (optional) — How long the request is given to be answered, in milliseconds; the engine's own bound when left out, none when 0 (optional).
expectedStatuses ExpectedStatuses (optional) — HTTP statuses this caller handles as an answer it expected, e.g. { 404 }: a response with one settles the future as any other non-2xx does and is logged as information rather than an error (optional).

Returns PromiseId — A future (PromiseHandle): f:await() or task.await(f) delivers the response, f:state() reads whether it has.

local r = task.await(http.post_json(url, nil, { name = "Alice" }))

globals/http/request

http.request(method: string, url: string, headers: Headers?, body: JsonBody?, timeoutMs: TimeoutMs?, expectedStatuses: ExpectedStatuses?) -> PromiseId

Async HTTP request with an arbitrary verb (GET/POST/PUT/PATCH/ DELETE/…) returning JSON. Body is a Luau table; an empty 2xx response resolves to an empty table.

Parameters

method string — HTTP verb (case-insensitive).
url string — Request URL.
headers Headers (optional) — Header key-value pairs (optional).
body JsonBody (optional) — JSON body (optional).
timeoutMs TimeoutMs (optional) — How long the request is given to be answered, in milliseconds; the engine's own bound when left out, none when 0 (optional).
expectedStatuses ExpectedStatuses (optional) — HTTP statuses this caller handles as an answer it expected, e.g. { 404 }: a response with one settles the future as any other non-2xx does and is logged as information rather than an error (optional).

Returns PromiseId — A future (PromiseHandle): f:await() or task.await(f) delivers the response, f:state() reads whether it has.

local w = task.await(http.request("PATCH", url, hdrs, { description = "hi" }))

globals/http/request_raw

http.request_raw(method: string, url: string, headers: Headers?, body: buffer | string | nil?, timeoutMs: TimeoutMs?, expectedStatuses: ExpectedStatuses?) -> PromiseId

Async HTTP request with an arbitrary verb and a RAW binary request body (a binary-safe string), for content-addressed blob uploads. The resolved value is the response body text.

Parameters

method string — HTTP verb (case-insensitive).
url string — Request URL.
headers Headers (optional) — Header key-value pairs (optional).
body buffer | string | nil (optional) — Raw binary request body (optional).
timeoutMs TimeoutMs (optional) — How long the request is given to be answered, in milliseconds; the engine's own bound when left out, none when 0 (optional).
expectedStatuses ExpectedStatuses (optional) — HTTP statuses this caller handles as an answer it expected, e.g. { 404 }: a response with one settles the future as any other non-2xx does and is logged as information rather than an error (optional).

Returns PromiseId — A future (PromiseHandle): f:await() or task.await(f) delivers the response, f:state() reads whether it has.

local r = task.await(http.request_raw("POST", blobsUrl, hdrs, pngBytes))

globals/httpServer/address

httpServer.address(path: string) -> (string?, string?)

The URL a path answers on — scheme, host, port and the /app mount, ready to be fetched or printed for someone to open. Takes the same path spelling route does, and reads the interface and port from the socket routes answer on: the address listen opened while one is open, and the engine's own server otherwise.

Parameters

path string — Path under the /app mount, e.g. "/status".

Returns (string?, string?) — The URL, or nil plus the reason there is none — this engine holds no address, or the path is not one a route can be registered at.

print(httpServer.address("/status")) --> http://127.0.0.1:7607/app/status

globals/httpServer/listen

httpServer.listen(target: string) -> (HttpListener?, string?)

Hold an interface and port of this world's own, and answer content routes on it.

The host in target is the interface bound, and the whole of what decides who can reach those routes: "127.0.0.1:8080" answers programs on this machine, "0.0.0.0:8080" answers any host that routes to this machine on that port — a phone on the same wifi, and whatever else the network lets through. Bind loopback unless you want that. A port of 0 asks the operating system for a free one, which the returned record reports, and http:// may be spelled out in front.

This address serves the routes registered under the /app mount. The engine's own /engine/* tree answers on the loopback server it booted with, whose interface stays what the boot bound.

The address belongs to the chunk that opened it and is released when that chunk runs again, so an edited module holds the address its current source names. Asking for the address already held is the same address back.

Parameters

target string — Interface and port to hold, e.g. "0.0.0.0:8080".

Returns (HttpListener?, string?) — The listener record, or nil plus the reason the target or the bind was refused.

local l = assert(httpServer.listen("0.0.0.0:8080"))

globals/httpServer/route

httpServer.route(method: string, path: string, handler: HttpHandler, options: HttpRouteOptions?) -> (number?, string?)

Serve one method and path from this engine, answering each matching request with handler.

The path is relative to the /app mount, and a trailing /* segment matches the rest of the path — "/files/*" answers /app/files/a/b, with "a/b" in request.wildcard. An exact path answers ahead of a wildcard, and among wildcards the longest one wins.

One method and path is served by one handler. Registering an address another chunk serves returns nil and a reason naming the handle and the chunk holding it; httpServer.routes() finds that handle and httpServer.unroute frees the address. Registering an address this same chunk already serves takes it back and releases the handler it replaces, so a chunk that runs twice serves the handler it just built.

The handler runs on the script thread; a route registered from a worker job keeps its handler in that worker's VM, which runs it once per request as the job that registered it while the request waits for its value and the engine thread waits for nothing. The route lasts until it is unregistered or the world changes or the scripting state resets, whatever became of the job that made it. Raising inside the handler answers 500 and writes the error to the engine log; returning something that is not a response table or a string answers 500 saying what arrived.

Parameters

method string — HTTP verb, e.g. "GET" or "POST".
path string — Path under the /app mount, e.g. "/status" or "/files/*".
handler HttpHandler — Called with the request table; returns a response table or a body string.
options HttpRouteOptions (optional) — { timeoutMs? } — how long a request waits for this handler.

Returns (number?, string?) — The route handle, or nil plus the reason it was not registered.

local h = httpServer.route("GET", "/status", function(req)

globals/httpServer/routes

httpServer.routes() -> { HttpRoute }

Every route this engine currently serves, in registration order — handle, method, registered path, the address it answers on, its full URL, the chunk that registered it, and how long a request for it waits.

Returns { HttpRoute } — An array of route records.

for _, r in ipairs(httpServer.routes()) do print(r.method, r.url, r.owner) end

globals/httpServer/status

httpServer.status() -> HttpServerStatus

Whether this engine serves content routes, on which interface, port and mount, who can reach them, and how many routes and waiting requests it holds. host, port, url and reach are read from the socket routes answer on — the one listen opened while one is open, and the engine's own server otherwise — and listeners carries every address, each with its own reach. When supported is false, reason says why: a browser tab answers HTTP requests and holds no address of its own.

Returns HttpServerStatus — { supported, reason?, host?, port?, url?, reach?, prefix, routeCount, pending, listeners }.

local s = httpServer.status(); print(s.url, s.reach)

globals/httpServer/unlisten

httpServer.unlisten() -> boolean

Release the address listen opened. Returns once the socket is free, so the same port binds again straight after.

Returns boolean — True when an address was held.

httpServer.unlisten()

globals/httpServer/unroute

httpServer.unroute(handle: number) -> boolean

Stop serving a route and release its handler. The address is free for another registration once this returns true.

Parameters

handle number — The handle httpServer.route returned.

Returns boolean — True when a route with this handle was registered.

httpServer.unroute(h)

globals/i18n/choice

i18n.choice() -> string?

The locale the player chose for themselves, or nil when they chose none and the chain decides.

Returns string? — The player's choice.

globals/i18n/choose

i18n.choose(tag: string?) -> ChooseResult

Choose tag as the player's locale, kept on this device for their ZeroMind account (or for the signed-out player), so it wins the chain from now on and on every later boot. nil clears the choice and the chain decides again. A tag no offered locale matches is refused, naming the offered tags.

Parameters

tag string (optional) — A tag from i18n.offered(), or nil to clear the choice.

Returns ChooseResult — { ok, locale?, decidedBy?, reason?, offered? }.

local r = i18n.choose("pt-BR")
if not r.ok then log.warn(r.reason) end

globals/i18n/emulate

i18n.emulate(tags: { string }?) -> { ok: boolean, locale: string?, platform: { string }?, reason: string? }

Stand tags in for the platform's preferred languages, the way a visitor whose device or browser reads them would arrive, and negotiate again: i18n.preferred(), i18n.pick, i18n.locale() and i18n.onChange all answer as they would for that visitor. nil reads the platform's own list again. A report the platform makes afterwards (a language change, a desktop window regaining focus) replaces it. The player's own choice (i18n.choose) still comes first in the chain, so clear it with i18n.choose(nil) to see the list decide.

Parameters

tags { string } (optional) — BCP 47 tags, most preferred first, or nil for the platform's own list.

Returns { ok: boolean, locale: string?, platform: { string }?, reason: string? } — { ok, locale?, platform?, reason? }: the locale and list now in force.

i18n.emulate({ "pt-BR", "en" })   -- try the world as a Brazilian visitor
i18n.emulate(nil)                  -- back to this machine's own languages

globals/i18n/fallback

i18n.fallback() -> { string }

The chosen locale's fallback chain: the locales a string missing from its catalog is looked up in, in order.

Returns { string } — The fallback locales, en last; empty for en itself.

globals/i18n/locale

i18n.locale() -> string

The locale the engine chose for the player for its own text, a BCP 47 tag such as en or pt-BR. It is always one of i18n.offered(); en, the engine's source language, until a negotiation states another. To choose among your own content's languages, use i18n.pick.

Returns string — The chosen locale.

if i18n.locale() == "pt-BR" then label.text = "Olá" end

globals/i18n/offChange

i18n.offChange(id: number) -> boolean

Remove an i18n.onChange subscriber by its watcher id. Returns true when a live watcher carried that id (false for one already removed, or ended with the component that made it).

Parameters

id number — Watcher id returned by i18n.onChange.

Returns boolean

globals/i18n/offered

i18n.offered() -> { Locale }

The locales the engine offers for its own text, sorted by tag. Only an offered locale can be chosen or negotiated. Content picks among its own languages with i18n.pick.

Returns { Locale } — One { tag, name, nativeName, fallback } per offered locale.

globals/i18n/onChange

i18n.onChange(callback: (string, { string }) -> ()) -> number

Run callback each time the player's language changes: the platform's preferred list changes (the player changed their OS, browser or device language while the engine runs, or i18n.emulate stood a list in), the player chooses or clears a choice, the ZeroMind language moves, or the offered locales change. Called as (locale, preferred), after the engine renegotiated, so i18n.pick inside it answers for the new state. A change i18n.emulate or i18n.choose makes reaches this VM's subscribers before that call returns; one the platform reports reaches them within a tenth of a second. A subscription made from a component ends with that component. Returns a watcher id for i18n.offChange.

Parameters

callback (string, { string }) -> () — Function invoked as (locale: string, preferred: {string}).

Returns number — The watcher id.

local id = i18n.onChange(function(locale, preferred)
log.info(("locale %s, platform %s"):format(locale, table.concat(preferred, ",")))
end)

globals/i18n/pick

i18n.pick(available: { string }) -> string

Which of available, the languages your content has text for, this player reads: the tag the engine's chain picks from the player's own choice, the language ZeroMind shows them and their platform's list, matched by the same rules (same tag, the tag truncated, the same language). The first of available when none of them matches, so list your fallback first. Call it again (or from i18n.onChange) to follow a change while they play.

Parameters

available { string } — BCP 47 tags your content has text for, fallback first.

Returns string — One of available, exactly as written there.

local lang = i18n.pick({ "en", "pt", "es" })
label.text = ({ en = "Welcome!", pt = "Bem-vindo!", es = "¡Bienvenido!" })[lang]

globals/i18n/preferred

i18n.preferred() -> { string }

The languages the player's platform says they read, as BCP 47 tags, most preferred first: the OS language list on Linux, Windows and macOS, navigator.languages in a browser, the device (or per-app) language list on Android. Empty when the platform states none.

Returns { string } — The platform's preferred languages.

local first = i18n.preferred()[1]

globals/i18n/sources

i18n.sources() -> Sources

Every link of the chain the locale was chosen from: the player's choice, the zeromind language, the platform list, and which link decided (decidedBy: choice, zeromind, platform or default) with the tag it matched.

Returns Sources — { locale, decidedBy, matched, choice?, zeromind?, platform, signedIn }.

globals/ipairs

ipairs(table) -> iterator

Iterate array portion of table (1, 2, 3...).

globals/jobs/context

jobs.context() -> JobContext?

What the running item runs as, on a worker: worker, the pool slot of its worker; key, the item's key (nil for an item dispatched without one, and outside an item, as in a watcher); fresh, true for the first item a key runs in this VM since the key bound, so a module tells a cold cache from a missing entry. On the engine thread it answers nil, which is also how code asks whether it runs on a worker.

Returns JobContext? — { worker, key, fresh } on a worker; nil on the engine thread.

local ctx = jobs.context(); if ctx and ctx.fresh then cache = {} end

globals/jobs/dispatch

jobs.dispatch(module: string, entry: string, args: any?, opts: DispatchOptions?) -> PromiseHandle

Run the function entry of the module module on a worker thread and answer a promise of its result, so a long computation holds no frame. module is a require path, read from this line as require reads it. The module and the modules it requires run in a worker VM that boots the prelude and requires modules as the engine thread's VM does, so it has the same globals and modules/api modules: the entry computes from its arguments and returns a value, which the code awaiting the promise applies to the world. A call to a function only the engine thread serves (one that reads or changes the ECS world, the component instances its VM holds, or a table the engine thread has not published yet) is forwarded: from a job it runs on the engine thread, with the same arguments, after every write the job recorded before it, as the principal the job runs as, and its answer or its error returns to the job. The task that made it waits as it waits on a promise, so the worker keeps serving its other tasks and the time parked never counts toward the deadline; each forwarded call takes at least one engine frame, and job:info().forwarded_calls and the system.schedule.frame.forwarded_calls profiler count report them. Arguments and answers cross as copies; a call whose argument holds a function, coroutine or userdata handle, whose answer holds an object of the engine thread's VM, or that is made where the script cannot wait (a metamethod, the top level of a module being required, a batch or queue body) raises an error naming the function and the reason. A watcher the module registers (engine.onModeChange, modules.watch, world.on, vfs.watch and the other engine events) is kept by the worker VM and fires when that VM next starts an item, for the events raised since it registered; save.onMarksLanded is refused there, because the engine reads the value its callback returns inside its own frame. A function the module registers with renderfeature.setTick, jobs.register or httpserver.route stays in the worker VM that registered it, and that VM stays in service for it after the item returns, until the function is unregistered or the world changes or the scripting state resets: the engine posts each call to the worker and never waits for it, and reads the latest completed output (a tick's passes are drawn one frame after the tick that made them, and a frame whose call is still running keeps the last completed output and counts a late call). The entry is a task of the worker VM's own scheduler: it waits on time (task.wait), on frames (task.waitFrames) and on promises (task.await), starts tasks and scopes, and returns its result. The promise settles once the engine thread has applied every write the entry and the tasks it started made before it returned, so the code awaiting it reads all of those writes, as it would after making the same calls on the engine thread; an entry whose last write the engine thread applies in the frame the entry returned settles as soon as one that writes nothing, and otherwise one frame later. It waits as long as the writes do (while no world is open, say) and then resolves with the entry's result. It rejects only when the engine resets the scripting state or changes its world first, and then none of those writes is applied. The tasks it started keep running until they end or the item's deadline ends them; a write one of them makes after the entry returned is not part of what the promise waits for. A promise the worker makes settles on that worker (delay, http.* and jobs.dispatch serve one), and a call whose promise an engine-thread system settles (queue, shell.run of a command it would queue) is refused, naming it. An item runs on whichever worker takes it. With opts.key (a string of 1 to 256 bytes) it runs in the worker VM the key is bound to, after every item of the key dispatched before it has settled and its writes have applied, one at a time and in order, so module state an item of the key leaves (a cache, a plan, a parsed asset) is there for the next; jobs.context().fresh tells the first item of a key in a VM since the key bound, jobs.release ends the binding, and jobs.warm runs a module's one-time setup in every worker VM before items need it. A key's item waits for its VM even while other workers are idle, and /runtime/workers reports the wait per key. An item awaiting an item of its own key is refused at the await, naming the key: the awaiting item holds the key's turn, and the awaited item starts only after it settles; dispatching to its own key without awaiting runs that item after it. Items of a key queued before the world changes or the scripting state resets are refused unrun, naming the reset. opts.supersede = true with a key rejects every item of the key that has not started, naming this one; the running item is untouched. The pool holds at most 65,536 keys at once, and a dispatch with a key past that is refused at the call, naming the bound. args and the result cross as copies: nil, booleans, numbers, strings, buffers, vectors and tables of those, and any other value is refused at this call naming its key path. The promise rejects with the entry's error and stack, or names the deadline when one run of the entry, from a resume to the next yield, takes longer than opts.deadline seconds (default and at most the longest hold on one run, 60 seconds on every platform, native and browser alike; any number above 0 up to it, which a caller may lower and may not raise: a larger one is refused at the call naming the cap). The clock starts at each resume and stops at each yield, so time parked on a promise, task.wait or a frame never counts, every yield resets it, and no sum across runs is kept: an entry that keeps yielding runs until it returns or is cancelled. The tasks it resumed are held to the same deadline. Cancelling the promise (p:cancel()) halts a running entry and keeps a queued one from running. An edit of the module reaches the next dispatch.

Parameters

module string — Require path of the module, as require takes it.
entry string — Name of the function in the module's export table.
args any (optional) — The value the entry receives, copied.
opts DispatchOptions (optional) — Optional { deadline: number?, name: string?, key: string?, supersede: boolean? }: the deadline in seconds, what a rejection calls the item, the key it runs as, and whether it supersedes the key's items that have not started.

Returns PromiseHandle — A PromiseHandle settling with a copy of the entry's result.

local heights = jobs.dispatch("@root::terrain.noise", "heights", { size = 256, seed = 7 }):await()
local p = jobs.dispatch("@root::sim.solver", "solve", state, { deadline = 2, name = "solver" }); p:cancel()
local route = jobs.dispatch("@root::paths.graph", "route", { map = "level3", from = a, to = b }, { key = "level3" }):await()
local preview = jobs.dispatch("@root::graph.cook", "cook", edit, { key = graphId, supersede = true })

globals/jobs/find

jobs.find(name: string) -> JobHandle?

Look up a registered job by name. Returns a JobHandle or nil for anonymous / unknown names. Single FFI crossing — returns the id directly, no registry snapshot.

Parameters

name string — Job name supplied to jobs.register.

Returns JobHandle? — JobHandle or nil.

local job = jobs.find("animation_blend_main")

globals/jobs/inspect

jobs.inspect(target: JobHandle | string) -> JobInfo?

Return a snapshot row by job handle or by name without retrieving a full handle. Single FFI crossing — pulls only the matching row.

Parameters

target JobHandle | string — JobHandle, or job name string.

Returns JobInfo? — Snapshot row or nil.

local info = jobs.inspect("animation_blend_main")

globals/jobs/list

jobs.list(phase: string?) -> { JobInfo }

List registered job summaries. Pass a phase name to filter to a single phase. Single FFI crossing — only the requested rows cross the bridge.

Parameters

phase string (optional) — Optional phase filter — nil returns every job.

Returns { JobInfo } — Array of JobInfo rows.

local rows = jobs.list("main")

globals/jobs/onRelease

jobs.onRelease(watcher: (key: string) -> ()) -> () -> ()

Hold watcher in the VM of the module that calls it: each time a key bound to that VM is released (jobs.release), watcher(key) runs there as a task of the VM's scheduler, once per release, so the module drops what it keeps for the key. A worker VM keeps its own watchers, as it keeps engine.on* watchers; a module registers at its top level, which every worker VM that requires it runs. The engine thread's VM holds no key, so a watcher there is held and never called. Answers the function that removes the watcher.

Parameters

watcher (key: string) -> () — Called with the released key.

Returns () -> () — A function that removes the watcher.

jobs.onRelease(function(key) plans[key] = nil end)

globals/jobs/parallelFor

jobs.parallelFor(module: string, entry: string, over: number | { first: number, last: number } | buffer, args: any?, opts: { deadline: number?, name: string?, chunks: number?, stride: number? }?) -> PromiseHandle

Split a loop across the worker pool and answer a promise of the chunks' results in chunk order, so a loop over many indices or many elements of a buffer holds no frame and uses every worker. over is a count of indices (1..over), a table { first = a, last = b } of whole numbers, or a buffer whose elements are opts.stride bytes each (default 1). The range is cut into opts.chunks runs (default four per worker, at most one per element), as even as it divides, and each run is an item the pool's workers take: entry, a function of the module module, is called with one table chunk = { index, count, first, last, args, data }, where first and last are the run's indices (for a buffer, its element numbers from 1), args is a copy of args, and data is the run's slice of a buffer. The promise settles once every chunk has returned and the engine thread has applied every write the chunks made before they returned, with { chunk 1 result, chunk 2 result, ... } in chunk order whichever chunk finished first. A chunk that raises, or runs past its deadline (opts.deadline seconds for one run of each chunk between a resume and the next yield, default and at most the hold on one run, 60 seconds on every platform), rejects the promise naming the chunk (chunk 3 of 8) and cancels the chunks still queued or running; cancelling the promise (p:cancel()) cancels every chunk. It runs from the engine thread and from a worker: a worker that awaits the call runs queued chunks itself, so a call made inside a chunk completes however many workers are busy. The module, its worker VMs, args and result copies follow the rules of jobs.dispatch.

Parameters

module string — Require path of the module, as require takes it.
entry string — Name of the function in the module's export table; it receives the chunk table.
over number | { first: number, last: number } | buffer — A count of indices, a { first, last } range, or a buffer.
args any (optional) — A value every chunk receives as chunk.args, copied.
opts { deadline: number?, name: string?, chunks: number?, stride: number? } (optional) — Optional { deadline: number?, name: string?, chunks: number?, stride: number? }: the deadline in seconds of one run of each chunk, what a rejection calls the call, the number of chunks, and the bytes in one element of a buffer.

Returns PromiseHandle — A PromiseHandle settling with a table of the chunks' results in chunk order.

local sums = jobs.parallelFor("@root::sim.reduce", "sumChunk", 100000, { scale = 2 }):await()
local out = jobs.parallelFor("@root::pixels.tint", "tint", pixels, nil, { stride = 4, chunks = 16 }):await()

globals/jobs/register

jobs.register(descriptor: table) -> JobHandle?

Register a substrate job. Returns a JobHandle on success, nil on validation failure. Dispatches by executor.kind:\n - "kernel" / "stub" → standard __jobs.register (JSON-only descriptor).\n - "luau" → __jobs.register_luau(descriptor, executor.run) so the Luau function survives the JSON crossing as a stable registry ref. The dispatcher invokes the run closure once per frame; the closure captures any bindings/buffers it needs.\n - "compute" → the shader reference resolves to its registration key, and the job queues one dispatch per frame.\n - "worker" → __jobs.register_worker(descriptor, executor.input, executor.apply): each frame input() runs on the engine thread, its value is copied to a worker where the module's entry runs, and the result reaches apply(result) on the engine thread the frame it lands, one item in flight at a time. An input that returns nil queues no item that frame. A job that errors logs its name and message once, and job:info().last_error holds it. Without a key, its items can run on different workers, each with module state of its own, so what carries from one item to the next travels through the result and input. key (a string of 1 to 256 bytes) runs every item of the job as that key, in the worker VM the key is bound to, so module state one item leaves there is there for the next (see jobs.dispatch); job:info().waited_for_vm_ms is the time its items waited for that VM. A job registered from a worker VM keeps its function (run, or input and apply) in that VM: the dispatcher posts each frame's call to the worker and never waits for it, the writes it records apply one frame later, a raise marks the job errored one frame later, and job:info().worker reports the calls, the late calls and the failed calls. A worker job whose input or apply is not a function, whose deadline is not a number of seconds above 0 and at most the hold on one run (60 seconds on every platform), or whose key is not a string of 1 to 256 bytes, is raised at the call, naming why. deadline (default and at most that hold: it may be lowered and never raised) is the longest one run of an item may take between a resume and the next yield, counted on monotonic time per run and never summed across runs.\n\norigin is auto-filled with the VFS path of the calling script unless the descriptor already supplies one — surfaced under /zero/runtime/jobs/<phase>/<key>/origin.txt for agent traceability. Single FFI crossing — auto-origin runs Rust-side via lua_getinfo, no separate stack-inspection trip.

Parameters

descriptor table — Job declaration with the JobDescriptor shape — name?, phase, reads?, writes?, executor, ordering?, pure?, origin?, metadata?. Param is typed as table rather than JobDescriptor because the LSP doesn't yet narrow string literals to their literal types in record fields, so a JobDescriptor annotation rejects the tagged-union executor discriminator on every call site (literal kind = "kernel" infers as kind: string, doesn't subtype KernelExecutor.kind: "kernel"). Runtime validation in __jobs.register enforces the actual structure; see the JobDescriptor type alias above for the canonical shape.

Returns JobHandle? — JobHandle or nil.

local job = jobs.register({ phase = "main", executor = { kind = "kernel", kernel = "copy_buffer" }, reads = {{resource={kind="buffer",id=src.id},mode="r"}}, writes = {{resource={kind="buffer",id=dst.id},mode="w"}} })
local job = jobs.register({ phase = "main", executor = { kind = "luau", run = function() print("tick") end } })
local job = jobs.register({ phase = "main", executor = { kind = "compute", shader = asset.resolve("carve", "computeShader"), buffers = { "heights" }, workgroups = { 64 } } })
local job = jobs.register({ name = "crowd", phase = "main", executor = { kind = "worker", module = "@root::crowd.sim", entry = "step", input = function() return { count = 100 } end, apply = function(result) print(result) end } })

globals/jobs/release

jobs.release(key: string) -> boolean

End the binding of key to its worker VM once the items of the key dispatched before this call have settled. That VM then runs its jobs.onRelease watchers with the key, each as a task of its scheduler, and the key's next item binds anew and runs with jobs.context().fresh true. Without a release, a VM keeps each key's state for as long as it lives (until the world changes or the scripting state resets); /runtime/workers lists the keys each worker holds. Runs from the engine thread and from a worker.

Parameters

key string — The key, as jobs.dispatch was given it in opts.key.

Returns boolean — True when the pool held the key: a binding, or items queued or running, which the release runs after; false, doing nothing, for a key with none of those, a key whose only entry is a release under way among them.

jobs.release(graphId)

globals/jobs/shared/atomic

jobs.shared.atomic(t: SharedTable<T>, key: string | number, fn: (current: T?) -> T?) -> T?

Read-modify-write of one key of a shared table without a race. fn receives the key's current value (nil when absent; a table arrives as a plain copy) and returns the new one (nil removes the key); the new value is stored only if no other write reached the table meanwhile, otherwise fn runs again on the newer value, so fn must be quick, pure and must not yield. A call that lost four attempts in a row takes its turn at the table, and the other atomic calls on a worker wait for it, so atomic calls never starve one another; after 256 attempts that all lost to plain writes the call raises naming the table and the key. Counters, flags and compare-and-set are this one call; a change across several keys is one atomic on the parent key that returns the whole new table.

Parameters

t SharedTable<T> — A shared table.
key string | number — The key to update.
fn (current: T?) -> T (optional) — (current) -> new.

Returns T? — The value now stored under the key.

jobs.shared.atomic(counts, "done", function(n) return (n or 0) + 1 end)
jobs.shared.atomic(flags, "claimed", function(was) if was then return was end return true end)

globals/jobs/shared/queue

jobs.shared.queue(name: string, opts: SharedQueueOptions?) -> SharedQueue<T>

The shared queue named name, made on first use: a bounded first-in-first-out queue between any VMs, for handing work or results from one job to another, or from a job to the game. push(v) puts a copy of v at the back from anywhere; pop() takes the front value and waits for one when there is none (a worker parks, and on the engine thread only the calling task suspends, so a frame is never held); tryPop() takes it without waiting and answers nil when empty. A value is any non-nil data that a shared table holds. A full queue raises naming the queue and its capacity: it never drops. The capacity is opts.capacity (default 1024, at most 65,536); asking for the name again with another capacity raises. The queue is process memory with the lifetime of a shared table: when the world changes or the scripting state resets, every task parked on pop is rejected with the reason.

Parameters

name string — The queue's name, at most 256 bytes.
opts SharedQueueOptions (optional) — Optional { capacity: number? }.

Returns SharedQueue<T> — The queue.

local work = jobs.shared.queue("terrain.chunks", { capacity = 256 }); work:push({ x = 3, z = 4 })
-- in a job: while true do local chunk = work:pop(); build(chunk) end

globals/jobs/shared/revision

jobs.shared.revision(t: SharedTable<T>) -> number

A number that every write to the shared table, or to any table nested in it, advances. One atomic read: game code that polls a table each frame compares it with the number it saw and reads the table only when it changed.

Parameters

t SharedTable<T> — A shared table.

Returns number — The revision.

local seen = 0; -- each frame: local now = jobs.shared.revision(progress); if now ~= seen then seen = now; redraw(progress) end

globals/jobs/shared/snapshot

jobs.shared.snapshot(t: SharedTable<T>) -> { [string | number]: T }

An ordinary table holding a deep copy of a shared table, for the code that wants a stable view (JSON, comparison, the table library on a frozen picture). Each table of the copy is read at one version; tables read at different moments may be of different versions.

Parameters

t SharedTable<T> — A shared table.

Returns { [string | number]: T } — A plain table.

local picture = jobs.shared.snapshot(progress)

globals/jobs/shared/table

jobs.shared.table(name: string) -> SharedTable<T>

The shared table named name, made on first use. The engine's VM and every worker VM read and write the same table with the syntax of an ordinary table: t.x = 1, t.player.hp -= 5, #t, for k, v in t, pairs, next, table.insert, table.remove, table.sort, table.concat, table.find, table.unpack, table.clone, table.clear, rawget, rawset. A job streams results to the game through it while it runs, and the game reads them each frame. It holds data: booleans, numbers, integers, strings, vectors, buffers and tables of those; assigning a table copies it in (later edits of the local table are not seen), and a function, coroutine, userdata or a table with a metatable raises naming the key path. A value reads back as the type it was written with. t.a where t.a is a table is that table, so t.a.b = 1 writes through; a table replaced or removed raises on any use, naming where it was. Each access is atomic on its table; a loop walks the table as it stood when the loop started. The store is process memory: not persisted and not replicated (runtime_data and synced component fields are those). It lasts until the world changes or the scripting state resets (play entry keeps it); after that a held table raises, naming the reason. A table holds at most 1,048,576 entries and 64 MiB (32 MiB in the browser), nested at most 64 deep; the process holds at most 4,096 names and 256 MiB (128 MiB in the browser); a write past a bound raises and changes nothing.

Parameters

name string — The table's name, at most 256 bytes.

Returns SharedTable<T> — The shared table.

local progress = jobs.shared.table("terrain.progress"); progress.done = 0
-- in a job: progress.done += 1 -- in the game each frame: print(progress.done)

globals/jobs/warm

jobs.warm(module: string, entry: string) -> PromiseHandle

Run the function entry of the module module once in every worker VM, to build the state the module keeps in its own VM (a registry, compiled rules, a table it loads) before an item needs it. The pool starts every worker it may start; each VM up runs the entry as its next item, and a VM made later (after a panic, or one the pool starts later) runs it before its first item, whichever boot made it. The entry is called with nil and builds state in its own VM: it runs under the hold on one run (60 seconds from a resume to the next yield), and its whole run, waits included, under the warm bound of 60 seconds of wall time from its start to its end, read between runs, so a run in progress at the bound goes on to its next yield: an entry that does not yield runs at most the bound plus one hold, about 120 seconds. It may wait on time and frames, and it does not wait on the pool: an await of a jobs.dispatch, a jobs.parallelFor or a jobs.warm promise by the entry or a task it started, while the entry runs, raises, naming what it awaited, whether or not the promise has settled, and so does making a future combinator over one, at the call (dispatching without awaiting is fine: the item runs once the VM is warm). No item starts on the VM until its warm pass is over; tasks the VM already holds go on between the entry's waits. The promise settles once every VM up at the call has run it: it resolves when every run ended well, and rejects naming each worker whose run raised, panicked or ran past the warm bound, and its error. It is kept for the life of the scripting state: a second call for the same module and entry runs nothing more and answers the same. An entry that raises is logged, naming its module, entry and worker, and the VM serves on. An entry that panics the engine or runs past the warm bound runs on no later VM until the scripting state resets: it is logged once and ended, and the VM made after a panic runs the rest of the pass before any item. A panic in code outside the entry while it runs (the VM's boot, a watcher, a task another entry left behind) ends that VM, and the VM made next runs it once more; these endings are counted per entry, across every worker of the pool, for the epoch, and the second one fails the entry for the epoch, naming the panic. The entry runs on workers only; the calling thread pays the dispatch. Call it from the engine thread or from an item, awaited or not, whatever principal the item runs as: the item's worker runs the warm job for its own VM before the item resumes. Call it when the module is first used in a world, so a world that never uses it pays nothing; /runtime/workers lists each worker's warm runs with their milliseconds, and the failed ones with why.

Parameters

module string — Require path of the module, as require takes it.
entry string — Name of the function in the module's export table.

Returns PromiseHandle — A PromiseHandle settling once every worker VM up at the call has run the entry: nil when every run ended well, otherwise rejected naming each failed run.

jobs.warm("@root::graph.nodes", "buildRegistry")
jobs.warm("@root::paths.graph", "loadTables"):await()

globals/layers/active

layers.active -> any

The root scene's proxy, re-resolved on every read.

Returns any

globals/layers/active/clearDirty

layers.active:clearDirty(opts?) -> boolean

Discard this layer's unsaved edits and respawn the live scene from canonical. Must run in a coroutine.

Parameters

opts table (optional) — Options

Returns boolean — True if overlay state was removed.

globals/layers/active/hasDirty

layers.active:hasDirty(opts?) -> boolean

True iff any unsaved dirty state exists for this layer — the manifest or at least one per-entity overlay file.

Parameters

opts table (optional) — Options

Returns boolean — True when unsaved dirty state exists.

globals/layers/active/load_additive

layers.active:load_additive(ref) -> SceneProxy

Load a scene as an additive overlay on top of this one, routed through the canonical load path (v6 player/camera config, lighting, entrypoint discovery).

Parameters

ref string — Scene asset reference to overlay

Returns SceneProxy — The new overlay layer's proxy.

globals/layers/active/offReady

layers.active:offReady(handle) -> boolean

Remove an onReady subscription by its handle.

Parameters

handle number — Handle returned by onReady

Returns boolean — True if the subscription was removed.

globals/layers/active/off_edit_load

layers.active:off_edit_load(handle) -> boolean

Remove an on_edit_load subscription by its handle.

Parameters

handle number — Handle returned by on_edit_load

Returns boolean — True if the subscription was removed.

globals/layers/active/off_load

layers.active:off_load(handle) -> boolean

Remove an on_load subscription by its handle.

Parameters

handle number — Handle returned by on_load

Returns boolean — True if the subscription was removed.

globals/layers/active/off_load_play

layers.active:off_load_play(handle) -> boolean

Remove an on_load_play subscription by its handle.

Parameters

handle number — Handle returned by on_load_play

Returns boolean — True if the subscription was removed.

globals/layers/active/off_unload

layers.active:off_unload(handle) -> boolean

Remove an on_unload subscription by its handle.

Parameters

handle number — Handle returned by on_unload

Returns boolean — True if the subscription was removed.

globals/layers/active/onReady

layers.active:onReady(cb) -> number

Subscribe to this scene reaching the 'ready' state. LATCHED — a callback registered after the scene is already ready fires immediately.

Parameters

cb function — Called with the SceneProxy once the scene is ready

Returns number — Subscription handle for offReady.

globals/layers/active/on_edit_load

layers.active:on_edit_load(cb) -> number

Subscribe to THIS scene's load, firing only when it loads with engine.mode == "edit".

Parameters

cb function — Called when this scene loads in edit mode

Returns number — Subscription handle for off_edit_load.

globals/layers/active/on_load

layers.active:on_load(cb) -> number

Subscribe to THIS scene's load event (fires only for this layer, unlike the global layers.onLoad).

Parameters

cb function — Called on this scene's load

Returns number — Subscription handle for off_load.

globals/layers/active/on_load_play

layers.active:on_load_play(cb) -> number

Subscribe to THIS scene's load, firing only when it loads with engine.mode == "play".

Parameters

cb function — Called when this scene loads in play mode

Returns number — Subscription handle for off_load_play.

globals/layers/active/on_unload

layers.active:on_unload(cb) -> number

Subscribe to THIS scene's unload event (fires only for this layer).

Parameters

cb function — Called on this scene's unload

Returns number — Subscription handle for off_unload.

globals/layers/active/promoteDirty

layers.active:promoteDirty(opts?) -> string?

Promote the dirty overlay to canonical — compose canonical + overlay, write canonical, then clear the dirty state, all but the records waiting for a parent that no record and no entity provides, which stay in the overlay. Raises when the canonical write is refused, leaving the overlay holding this session's edits.

Parameters

opts table — { to?

Returns string — Canonical path on success; nil when nothing to promote.

globals/layers/active/reload

layers.active:reload(opts?) -> SceneBuildReport?

Round-trip reload: unload then reload the same scene ref, firing the full unload / before-load / load lifecycle, and running the scene's build.luau against what it resolves now. The rebuild saves the scene, as a write to build.luau does, so scene.json holds what it built and the next play loads it; a library scene is rebuilt in the live scene alone. Must run in a coroutine.

Parameters

opts table — { rebuild?

Returns table — What the scene's build.luau did: { built = true, content, editorOnly, saved }, saved saying whether scene.json now holds the rebuild (false for a library scene), carrying { refused, message } when an operation the build ran was refused, or { built = false, reason, message }.

globals/layers/active/root

layers.active:root() -> SceneProxy

Return the root SceneProxy — walks up the additive-overlay parent chain to the non-additive root layer.

Returns SceneProxy — The root layer's proxy.

globals/layers/active/save

layers.active:save(opts?) -> string

Publish the layer's current state to canonical scene.json — the ONLY path that writes canonical (edit/play cycles never save). A record waiting for a parent that no record and no entity provides stays in the dirty overlay, where it is built once its parent arrives. Additive overlays cannot be saved.

Parameters

opts table — { to?

Returns string — Full VFS path of the canonical scene.json written.

globals/layers/active/set_visible

layers.active:set_visible(v)

Show or hide this layer's entities.

Parameters

v boolean — Whether the layer's entities are shown

globals/layers/active/unload

layers.active:unload()

Unload this layer: despawn every entity attributed to it and remove its slot from SceneLayers.

globals/layers/active/writeDirty

layers.active:writeDirty(opts?) -> string

Force-drain every pending edit into the dirty overlay directory (deltas only — changed-entity files + manifest); never touches canonical.

Parameters

opts table (optional) — Write options

Returns string — Absolute VFS path of the dirty manifest.

globals/layers/camera

layers.camera -> any

The active root scene's camera handle, the same value layers.active.camera answers.

Returns any

globals/layers/cost

layers.cost() -> { SceneLayerCost }

What each loaded scene's per-frame tick costs, attributed to the layer that owns it — the update / editorUpdate its entrypoint declares, timed where it runs. totalMs is a SUM across the window layers.observe().window reports, so divide by calls (or read avgMs) for the per-tick figure; a tick that runs every frame makes that the per-frame figure. Call layers.resetCostWindow() first to time a particular stretch. A layer whose entrypoint declares no tick is absent.

Returns { SceneLayerCost } — An array of SceneLayerCost.

layers.resetCostWindow(); task.wait(1); for _, c in layers.cost() do print(c.name, c.avgMs) end

globals/layers/find

layers.find(ref: AssetRef<scene> | string) -> any?

The loaded layer for a scene, matched on guid — the canonical identity, since display names can collide and paths drift when assets move. A layer torn down but not yet pumped out of the engine's loaded list reads as gone.

Parameters

ref AssetRef<scene> | string — A scene AssetRef, or an identity string resolved through asset.ref.

Returns any? — The scene proxy, or nil when that scene has no loaded layer.

local layer = layers.find("scenes.arena")

globals/layers/fireBeforeLoad

layers.fireBeforeLoad(proxy: any?) -> nil

Announce that a scene layer is about to load: clears any pending unload for that layer slot, marks the proxy loading, and fans out to every layers.onBeforeLoad subscriber. The scene-load pipeline calls this.

Parameters

proxy any (optional) — The scene proxy about to load.

Returns nil

layers.fireBeforeLoad(sceneProxy)

globals/layers/fireLoad

layers.fireLoad(proxy: any?) -> nil

Announce that a scene layer has loaded, fanning out to every layers.onLoad subscriber. The layer is pinned as the active one for the duration of the fan-out, so entities a subscriber spawns are attributed to it rather than landing orphaned. The scene-load pipeline calls this.

Parameters

proxy any (optional) — The loaded scene proxy.

Returns nil

layers.fireLoad(sceneProxy)

globals/layers/fireUnload

layers.fireUnload(proxy: any?) -> nil

Announce that a scene layer is unloading: fans out to every layers.onUnload subscriber, then drops the layer's cached proxy and per-layer state so the next load of that scene rebuilds from disk. The unload path calls this.

Parameters

proxy any (optional) — The scene proxy being unloaded.

Returns nil

layers.fireUnload(sceneProxy)

globals/layers/install

layers.install() -> nil

Install the layers global. layers.active is exposed as a property whose every read resolves the current root scene, so it tracks scene changes without manual invalidation; other keys resolve against this module. The prelude calls this once at boot.

Returns nil

layers.install()

globals/layers/inventory

layers.inventory() -> { SceneLayerInventory }

What each loaded layer holds: the entities the engine attributes to it, whether it came up whole, and how many failures it carries. unattributed in layers.observe().totals counts what exists in the world that no layer claims.

Returns { SceneLayerInventory } — An array of SceneLayerInventory.

for _, l in layers.inventory() do print(l.name, l.entities, l.ok) end

globals/layers/is_loaded

layers.is_loaded(ref: AssetRef<scene> | string) -> boolean

Whether a scene currently has a loaded layer — the boolean form of layers.find. A scene counts as loaded from the frame the engine holds a layer slot for it — the same slot its entities are attributed to — until an unload is issued against that slot. So a gate like if layers.is_loaded(ref) then layers.unload(ref) end sees the layer on the frame its entities exist.

Parameters

ref AssetRef<scene> | string — A scene AssetRef, or an identity string.

Returns boolean — True when the scene is loaded as a layer.

if not layers.is_loaded("scenes.hud") then layers.load("scenes.hud", { additive = true }) end

globals/layers/lastLoad

layers.lastLoad() -> SceneLoadReport?

The most recent load's report: what it loaded, what root it replaced and which overlays went with it, the entity counts on each side, how long each phase took, and every failure it produced. Nil on an engine that has loaded nothing — which is how "nothing has loaded" reads differently from a load that changed nothing.

Returns SceneLoadReport? — A SceneLoadReport, or nil.

local r = layers.lastLoad(); print(r.name, r.outcome, r.entities.added)

globals/layers/lastUnload

layers.lastUnload() -> SceneUnloadReport?

The most recent unload's report: the layer it took down under the name it was loaded with, the overlays it cascaded, and the entities that went with them. A guid no longer resolves to a name once its layer is gone, so this is where that name survives.

Returns SceneUnloadReport? — A SceneUnloadReport, or nil.

local u = layers.lastUnload(); print(u.name, u.entities.removed)

globals/layers/list

layers.list() -> { any }

Every loaded scene layer as a proxy, root and additive alike, in the order the engine reports them. A proxy carries name, sceneIdentity (the scene asset it was loaded from, e.g. scenes.main), guid, path, additive, persistent and visible. The root one — the open scene — is also layers.active.

Returns { any } — Array of scene proxies — empty before any scene is loaded.

for _, layer in ipairs(layers.list()) do print(layer.name, layer.sceneIdentity, layer.additive) end
print("open scene:", layers.active.sceneIdentity)

globals/layers/load

layers.load(ref: AssetRef<scene> | string, opts: LoadOpts?) -> any

Parameters

ref AssetRef<scene> | string
opts LoadOpts (optional)

Returns any

globals/layers/loadHistory

layers.loadHistory() -> { SceneLoadReport }

Every load report the engine still holds, oldest first. Bounded — old reports fall off the front, so a long session's memory does not grow with how many times a scene was swapped.

Returns { SceneLoadReport } — An array of SceneLoadReport.

for _, r in layers.loadHistory() do print(r.name, r.durationMs) end

globals/layers/loadInFlight

layers.loadInFlight() -> number

Returns the number of scene loads currently in flight (queued but not yet visible via onLoad dispatch). Returns 0 when the engine is in a stable load state. Used by engine.mode = ... to block flips while a load is mid-air; agents can read this to wait for a load to finish before driving the next operation.

Returns number

globals/layers/localPlayer

layers.localPlayer -> any

The active root scene's local player handle, the same value layers.active.players.localPlayer answers.

Returns any

globals/layers/observe

layers.observe() -> SceneObservation

What every scene load did, and what each loaded scene costs. One read covering the last load's report (what it produced, what it replaced, what it failed to produce and why, and how long each phase took), the load and unload history, a per-layer inventory of what the engine attributes to each layer, and the per-frame cost of each layer's entrypoint tick. Answers in edit mode as well as play.

Returns SceneObservation — A SceneObservation.

local o = layers.observe(); print(o.lastLoad.outcome, o.lastLoad.durationMs)
for _, c in layers.observe().cost do print(c.name, c.avgMs) end

globals/layers/offBeforeLoad

layers.offBeforeLoad(h: number) -> boolean

Cancel a layers.onBeforeLoad subscription.

Parameters

h number — The handle layers.onBeforeLoad returned.

Returns boolean — True when a subscription was removed.

layers.offBeforeLoad(h)

globals/layers/offEntityChanged

layers.offEntityChanged(h: number) -> boolean

Remove a subscription made with layers.onEntityChanged.

Parameters

h number — The handle returned by layers.onEntityChanged.

Returns boolean — True when the subscription existed and was removed.

layers.offEntityChanged(handle)

globals/layers/offLoad

layers.offLoad(h: number) -> boolean

Cancel a layers.onLoad subscription.

Parameters

h number — The handle layers.onLoad returned.

Returns boolean — True when a subscription was removed.

layers.offLoad(h)

globals/layers/offUnload

layers.offUnload(h: number) -> boolean

Cancel a layers.onUnload subscription.

Parameters

h number — The handle layers.onUnload returned.

Returns boolean — True when a subscription was removed.

layers.offUnload(h)

globals/layers/onBeforeLoad

layers.onBeforeLoad(cb: (any) -> ()) -> number

Run a callback just before a scene layer loads, while the previous layer's entities are still present.

Parameters

cb (any) -> () — Receives the scene proxy about to load.

Returns number — A handle to pass to layers.offBeforeLoad.

local h = layers.onBeforeLoad(function(scene) print("loading", scene.name) end)

globals/layers/onEntityChanged

layers.onEntityChanged(cb: (any) -> ()) -> number

Subscribe to authored entity changes. The callback runs once per frame with every entity edited since the previous frame, batched by layer as { { scene = string, entities = { string } } } — a moved transform, an edited component field, a spawn, or a despawn (the id of a despawned entity arrives with entity.exists already false). Any number of subscribers can watch the same edits.

Scope: authored edits in edit mode — what lands in the scene's dirty overlay. Mutations a component makes from its own update are runtime behavior and do not appear, so a subscriber that rebuilds derived data cannot re-trigger itself.

Parameters

cb (any) -> () — Called with the change batch.

Returns number — A handle for layers.offEntityChanged.

layers.onEntityChanged(function(batch)
for _, row in ipairs(batch) do
for _, id in ipairs(row.entities) do print("changed", id) end
end
end)

globals/layers/onLoad

layers.onLoad(cb: (any) -> ()) -> number

Run a callback once a scene layer has loaded — the point where its entities exist and player / camera spawners can attach to them.

Parameters

cb (any) -> () — Receives the loaded scene proxy.

Returns number — A handle to pass to layers.offLoad.

local h = layers.onLoad(function(scene) print("loaded", scene.name) end)

globals/layers/onUnload

layers.onUnload(cb: (any) -> ()) -> number

Run a callback as a scene layer unloads, while its entities are still addressable — the place to release anything keyed to them.

Parameters

cb (any) -> () — Receives the scene proxy being unloaded.

Returns number — A handle to pass to layers.offUnload.

local h = layers.onUnload(function(scene) print("unloaded", scene.name) end)

globals/layers/problems

layers.problems(ref: (AssetRef<scene> | string | any)?) -> { SceneLoadFailure }

What a layer failed to produce, and why. Each entry names the phase it happened in, one reason from the closed set, and the engine's own words — plus the entity, component or lifecycle hook it is about when it is about one.

Parameters

ref (AssetRef<scene> | string | any) (optional) — A scene AssetRef, an identity string, or a scene proxy. Omit for the active root layer.

Returns { SceneLoadFailure } — An array of SceneLoadFailure — empty for a layer that came up whole.

for _, f in layers.problems() do print(f.reason, f.entity, f.message) end

globals/layers/rebuildInFlight

layers.rebuildInFlight() -> boolean

Whether the engine is rebuilding the live scene right now: a scene load is carrying entities in, an edit↔play flip's transition is materialising the layer set, or a scene's build.luau build is landing what it produced. A flip unloads the root layer and loads it again for the new mode across many frames, and each mode materialises a different set of entities; a build lands its records and saves the scene over several frames. Either way the live entities are a stage of a scene being built while this reads true. A caller whose answer belongs to the settled scene, a test taking a root or a validator judging the live tree, polls it down to false first.

Returns boolean — true while a load, a mode-flip transition or a scene build is converging.

if not layers.rebuildInFlight() then print(#entity.findAll(), "entities settled") end

globals/layers/reload

layers.reload(ref: (AssetRef<scene> | string)?) -> any?

Unload and re-load a scene layer in place, so an edited scene asset takes effect without rebuilding the surrounding layer stack. The scene's build.luau runs against what it resolves right now, so a build script whose inputs moved (a component that now exists, an asset that now resolves) produces the scene it describes today. The rebuild saves the scene, as a write to build.luau does, so scene.json holds what the reload built and the next play loads it; unsaved edits in the scene are saved with it. A library scene (under /libs/) is rebuilt in the live scene alone. A reload asked for while a build of the scene is landing (a write to build.luau rebuilds over several frames) waits for that build to land and save first.

Parameters

ref (AssetRef<scene> | string) (optional) — A scene AssetRef, or an identity string. Omit to reload the active root scene.

Returns any? — What the scene's build.luau did: { built = true, content, editorOnly, saved } with the entity counts each half placed and whether scene.json now holds them (false for a library scene), carrying refused and a message reading them out when an operation the build ran was refused, or { built = false, reason, message } naming what stood in the way. Nil when no layer matched, which is a no-op.

layers.reload("scenes.arena")

globals/layers/resetCostWindow

layers.resetCostWindow() -> nil

Open a new cost window, discarding what the previous one measured. Call this before timing a stretch of frames; the load history is untouched.

Returns nil

layers.resetCostWindow()

globals/layers/unload

layers.unload(refOrProxy: (AssetRef<scene> | string | any)?) -> nil

Unload a scene layer. Unloading the root cascades through its additive overlays first, most-recently-loaded first, so none is left as a layer the engine still lists after its entities are gone; persistent additive layers survive the cascade. A scene with no loaded layer is a no-op. An unload made while a load is part way through a scene's records waits for that load to apply the rest.

Parameters

refOrProxy (AssetRef<scene> | string | any) (optional) — A scene AssetRef, an identity string, or a scene proxy. Omit to unload the active root scene.

Returns nil

layers.unload("scenes.hud")
layers.unload() -- the active root, plus its non-persistent overlays

globals/layers/whyPartial

layers.whyPartial(ref: (AssetRef<scene> | string | any)?) -> (string?, string?)

Why a layer is not whole. Returns nil when it IS — everything the scene declared was produced — and otherwise the nearest cause from the closed set loaderRaised, entrypointCompileFailed, entrypointBodyRaised, entrypointRaised, buildRaised, entityFailed, parentMissing, parentRefused, parentAbandoned, componentUnresolved, componentRefused, subscriberRaised, updateRaised. A second return carries the engine's own words for that cause.

Parameters

ref (AssetRef<scene> | string | any) (optional) — A scene AssetRef, an identity string, or a scene proxy. Omit for the active root layer.

Returns (string?, string?) — (reason, detail).

local why, detail = layers.whyPartial(); if why then print(why, detail) end

globals/library/has

library.has(path: string) -> boolean

Check if a library asset exists at the given path. Answers from the content the engine holds: an imported library's namespace ("@mylib") and its assets answer true once its pinned commit is mounted, and false while only its marker is there.

Parameters

path string — Library asset path (e.g. "@builtin/models/Sample/DamagedHelmet").

Returns boolean — True if the asset exists in the library.

assert(library.has("@builtin/models/Cube"))

globals/library/import

library.import(namespace: string, worldRef: string) -> LibraryImport

Import another world as a library under the given namespace. Resolves the world, pins its current commit, and writes the library marker at /source/libs/@<namespace>, then waits while the engine fetches the pinned commit from ZeroMind and mounts it. When it returns, the imported tree answers to require("@<namespace>::path") and is listed by library.list(). A library that does not mount raises naming why, and running the import again retries it.

Parameters

namespace string — Library namespace, with or without the leading @ (e.g. "@mylib" or "mylib").
worldRef string — The upstream world's guid, or its name as it appears in world.list().

Returns LibraryImport — Record describing the import: the local name, the marker path, the upstream world_guid, and the pinned commit as version.

library.import("@mylib", "my-shared-world")

globals/library/list

library.list(assetType: string?) -> { LibraryAsset }

List all available library assets. Optionally filter by asset type — call asset.categories() for the live set.

Parameters

assetType string (optional) — Asset type filter (optional).

Returns { LibraryAsset } — Array of asset descriptors. path is the @builtin::* identity, type is one of the valid type strings, format is the original file extension (e.g. "glb", "luau", "wgsl").

for _, a in ipairs(library.list("model")) do print(a.path) end

globals/loadstring

loadstring(code, chunkname?) -> function | (nil, error)

Compile a Luau source string into a callable function. Returns the compiled function on success, or nil + error message on failure. The returned function can be called to execute the code.

Parameters

code string — Luau source code to compile
chunkname string (optional) — Name for the chunk (shown in errors). Default: '=(loadstring)'

Returns function | (nil, string) — Compiled function, or nil + error string

globals/logs/clear

logs.clear() -> boolean

Drop all buffered log entries. Lifetime per-level counts (logs.count) are preserved.

Returns boolean — True on success.

logs.clear()

globals/logs/count

logs.count(opts: LogQueryOpts?) -> LogCounts

Aggregate counters for the log ring. Lifetime counts survive eviction, so errors reflects the total seen even if the lines have scrolled out of the buffer. opts takes the same filter table as logs.query, and matched is how many held entries it selects, counted without materialising them — limit and newest_first bound and order what a query RETURNS, so they leave matched alone. mcp is how many held entries record your own tool traffic; a query leaves those out, so with no opts, matched + mcp is everything held. last_seq is the cursor for incremental polling: read it before an action, then pass it as logs.query({ since = <that> }) afterwards to see only what the action logged.

Parameters

opts LogQueryOpts (optional) — Filter options, as logs.query takes.

Returns LogCounts — Counts summary table.

print("errors:", logs.count().errors)
local before = logs.count().last_seq

globals/logs/errors

logs.errors(limit: number?) -> { LogEntry }

Most-recent ERROR-level entries (newest first). limit defaults to 100.

Parameters

limit number (optional) — Maximum entries to return.

Returns { LogEntry } — Array of ERROR log-entry tables.

for _, e in ipairs(logs.errors(20)) do print(e.message) end

globals/logs/find

logs.find(text: string, limit: number?) -> { LogEntry }

Case-insensitive substring search over log messages. limit defaults to 200 (keeps the most recent matches). Searches what the engine logged, so looking for a marker cannot return the call that looked for it; logs.query({ contains = ..., include_mcp = true }) searches your own tool traffic too.

Parameters

text string — Substring to search for.
limit number (optional) — Maximum entries to return.

Returns { LogEntry } — Array of matching log-entry tables in chronological order.

local hits = logs.find("MY_MARKER")

globals/logs/query

logs.query(opts: LogQueryOpts?) -> { LogEntry }

Query the engine's in-memory log ring — the filtered view of what also reads as plain text at /zero/runtime/logs/engine. Answers about what the engine logged: the MCP record of your own tool traffic is left out, because the call carrying the query is one of those records and an unqualified search would match itself. type = "MCP" selects them; include_mcp = true mixes them in with everything else. On a world several sessions share, origin = "local" narrows the answer to the lines this session's own authoring caused.

Parameters

opts LogQueryOpts (optional) — Filter options.

Returns { LogEntry } — Array of matching log-entry tables.

logs.query({ entity = "guard-1", limit = 20 })
logs.query({ level = "error", context = 3 })
for _, e in ipairs(logs.query({ level = "warn", limit = 50 })) do print(e.message) end

globals/logs/tail

logs.tail(limit: number?) -> { LogEntry }

Most-recent entries of any level in chronological order. limit defaults to 100.

Parameters

limit number (optional) — Maximum entries to return.

Returns { LogEntry } — Array of the most recent log-entry tables.

for _, e in ipairs(logs.tail(20)) do print(e.level, e.message) end

globals/logs/template

logs.template(message: string) -> string

Normalize a message to its template — the same line with the parts that vary between occurrences (numbers, hashes, entity ids) masked out. Two messages that differ only in those parts share a template, which is what turns "this error repeated 400 times" into one row instead of 400. The engine keys its own error retention by the same normalization, so grouping built on this agrees with what survives ring eviction.

Parameters

message string — Log message to normalize.

Returns string — The message template.

local key = logs.template(entry.message)

globals/logs/warnings

logs.warnings(limit: number?) -> { LogEntry }

Most-recent WARN+ entries (newest first). limit defaults to 100.

Parameters

limit number (optional) — Maximum entries to return.

Returns { LogEntry } — Array of WARN+ log-entry tables.

print(#logs.warnings(), "warnings")

globals/lsp/analysisFlags

lsp.analysisFlags() -> AnalysisFlags

The feature-flag configuration upstream Luau's analysis library (the linter and .luaurc parser behind these checks) runs under, read back from Luau's flag registry. It is upstream's own: every Luau* flag the analysis sources declare is on unless upstream's experimental list names it, and every other flag they declare keeps the value Luau registers it at. The flags the analysis library reads from the parser and the VM are shared: they run at the value the engine's runtime gives them, and one that differs from upstream's carries its deviation. flags lists every flag those sources declare or read, sorted by name; atUpstream counts the linked ones whose value is upstream's. available is false, with reason, on a build that does not link the analysis library.

Returns AnalysisFlags — { available, reason?, declared, shared, linked, atUpstream, flags }.

globals/lsp/check

lsp.check(path: string, opts: CheckFileOpts?) -> DiagnosticsResult

Validate a single .luau file in the VFS and return its diagnostics. A path the check could not read comes back as one lsp-check-* error naming the path and the reason, and a file past the 2 MiB analysis bound as one source-too-large warning with checked = false, so checked with errors == 0 means a code body was analysed and is clean. opts.upstream (true, or the options of lsp.upstreamCheck) also checks the file with upstream Luau's type checker and sets upstream to what lsp.upstreamCheck returns for it, whose modules say how the check of each module it reached ended and how long it took; the diagnostics and counts stay the engine checker's.

Parameters

path string — VFS path.
opts CheckFileOpts (optional) — { severity?, limit?, context?, upstream? }.

Returns DiagnosticsResult — Array of diagnostic tables.

local diags = lsp.check("/zero/source/main.luau")

globals/lsp/checkAll

lsp.checkAll(opts: CheckAllOpts?) -> CheckAllResult

Validate the user's Luau scripts and return an aggregate summary plus diagnostic list. opts.scope = "user" (default) skips library mounts; "all" includes them. The sweep is time-budgeted (ZERO_EXECUTE_INLINE_BUDGET_MS, ~20s by default): if the budget elapses it returns the partial result gathered so far with budgetExceeded = true rather than blocking the engine. opts.budgetMs sets the budget for the call and opts.maxFiles caps the files it checks. opts.upstream (true, or the options of lsp.upstreamCheck) also checks each file with upstream Luau's type checker once the engine's own pass is done, within the same budget, and sets upstream to what that pass found: every module it reached, once, with how its check ended ("completed", "timedOut", "internal", or "notAnalysable" with the reason) and how long it took, reused marking a module checked before the pass reached it and unchanged since. The pass starts no file once the budget is spent, so it ends within the budget plus the check of the one file in progress.

Parameters

opts CheckAllOpts (optional) — { scope?, severity?, limit?, budgetMs?, maxFiles?, upstream? }.

Returns CheckAllResult — { filesChecked, errors, warnings, info, hints, budgetExceeded, diagnostics, upstream? }.

globals/lsp/checkCode

lsp.checkCode(source: string, opts: CheckOpts?) -> DiagnosticsResult

Validate inline Luau source without a backing file. Useful for checking code before writing it to disk. A source past the 2 MiB analysis bound comes back as one source-too-large warning with checked = false.

Parameters

source string — Luau source.
opts CheckOpts (optional) — { severity?, limit?, context? }.

Returns DiagnosticsResult — Array of diagnostic tables.

globals/lsp/checkDirty

lsp.checkDirty() -> DiagnosticsResult

Drain the dirty-file set populated by the hot-reload hook, validate each, and return the combined diagnostic list.

Returns DiagnosticsResult — Array of diagnostic tables.

globals/lsp/definitions

lsp.definitions(opts: DefinitionsOpts?) -> Definitions

The engine's API as a Luau definitions source, upstream Luau's .d.luau form (declare statements and export type aliases), generated from the live API registry: the documented members, the --!global-types aliases and the asset-category shapes lsp.check types against, bound under the same names. It declares each global the calling script's globals hold, so a user script, a trusted script and a worker each get the globals they have; a name upstream's builtin environment declares keeps upstream's type and is left out. lsp.upstreamCheck loads the same source into every module it checks, and an external editor running luau-lsp reads it as a definitions file (lsp.exportDefinitions writes it out). It is generated again whenever the registry changes, a library hot reload included. package (@zero) keys each global's documentation symbol (@zero/global/<name>, @zero/global/<name>.<member>). Each type string the registry holds is declared as written, or listed in unrepresented with the reason and declared any; a declaration upstream refuses to load (an alias that reaches itself through aliases alone) is listed with upstream's reason, so the rest of the source loads. opts.documentation = true also returns documentation, the JSON documentation file luau-lsp reads beside the definitions: each symbol's signature and description. available is false, with reason, on a build that does not link the analysis library.

Parameters

opts DefinitionsOpts (optional) — { documentation? }.

Returns Definitions — { available, reason?, package, source, bytes, globals, types, unrepresented, documentation? }.

local d = lsp.definitions()
print(d.bytes, #d.globals, #d.unrepresented)

globals/lsp/describe

lsp.describe(path: string, opts: DescribeOpts?) -> DocEntry?

Inspect a single documented entry. Returns the full doc table (signature, args, returns, examples, level), or nil. The path is resolved independently of which root the doc is registered under and of separator style, so the spelling that reads off the API surface (renderer.texture.create) finds the entry registered as globals/renderer/texture/create. A path naming a binding the engine registered internally answers with the entry a Luau module publishes over it where there is one, so the signature is the call content makes; opts.includeInternal answers with the internally registered entry itself. When a path does not resolve, lsp.describePaths says what the registry holds near it.

Parameters

path string — Doc path (e.g. "asset/resolve", "renderer.texture.create").
opts DescribeOpts (optional) — Optional { includeInternal? } — default prefers the published entry.

Returns DocEntry? — Full doc table or nil.

local doc = lsp.describe("renderer.texture.create")

globals/lsp/describePaths

lsp.describePaths(path: string) -> { string }

List the registered doc paths related to path. A path that names an entry returns every root it is registered under (the first is what lsp.describe resolves to); a path that names a namespace returns the entries registered under it. Empty when the registry holds nothing near the path — so a lookup that returns nil can always be turned into the list of what does exist.

Parameters

path string — Doc path in any spelling ("renderer.texture", "ecs/query").

Returns { string } — Array of registered doc paths, most canonical first.

for _, p in ipairs(lsp.describePaths("renderer.texture")) do print(p) end

globals/lsp/describeTool

lsp.describeTool(path: string) -> string?

Return the full documentation text for a code-mode tool.

Parameters

path string — Tool path (e.g. "scene/spawnLight").

Returns string? — Full tool docs or nil.

globals/lsp/docsByKind

lsp.docsByKind(kind: string) -> { MethodSummary }

List every doc whose registration kind matches kind. Valid: "binding", "runtime_tool", "module", "component", "library", "lua_export".

Parameters

kind string — Registration kind.

Returns { MethodSummary } — Array of doc summary tables.

globals/lsp/exportDefinitions

lsp.exportDefinitions(dir: string) -> ExportDefinitionsResult

Write the engine's API out for an external editor (VS Code, Neovim, or any editor running luau-lsp), so it types the engine's globals: <dir>/zero.d.luau (the definitions lsp.definitions returns), <dir>/zero.docs.json (the documentation file naming each global's and member's signature and description) and <dir>/.luaurc (nonstrict mode). Point luau-lsp at the first two (its types.definitionFiles and types.documentationFiles settings) and it types the engine's globals: entity.spawn("x") checks against its signature and entity.spwan("x") is an unknown member. The files hold the globals the calling script's globals hold; call it from the tier whose globals the editor should know. The files land in the VFS: userfile.save(path) (the Files panel's "Save to this device") puts each one on the machine of the person at the screen, and over HTTP the engine's GET /engine/vfs/<path> reads it. The engine's own checker reads a *.d.luau file the way upstream does, by loading it, so the files can stay in a world's content. Writing again replaces them. globals, types and unrepresented are the lists lsp.definitions returns; the guide topics/external-editor walks through the editor setup.

Parameters

dir string — VFS directory to write into, such as /zero/source/editor.

Returns ExportDefinitionsResult — { available, reason?, files, bytes, globals, types, unrepresented }.

local r = lsp.exportDefinitions("/zero/source/editor")
print(r.files[1], r.bytes, #r.globals)

globals/lsp/getStrictMode

lsp.getStrictMode() -> StrictMode

Return the current strict mode.

Returns StrictMode — "off" | "soft" | "strict".

globals/lsp/isStrict

lsp.isStrict() -> boolean

Is the pre-execute LSP gate fully strict? False when off or in soft mode.

Returns boolean — True when fully strict.

globals/lsp/lastCheckGen

lsp.lastCheckGen() -> number

Generation counter — bumped each time the cache is rebuilt. UI polls this to know when to redraw.

Returns number — Generation number.

globals/lsp/methods

lsp.methods(namespace: string, opts: MethodsOpts?) -> { MethodSummary } | { string }

List every documented method / entry under a namespace. A broad namespace (ui, renderer) returns a large dump by default, so two options narrow it: opts.filter keeps only methods whose name (or doc path) contains the substring, case-insensitively; opts.namesOnly returns a plain list of method-name strings instead of the full per-method summary tables, which is much smaller and leaves nothing to unwrap. A name, in either form (a summary's name field or a namesOnly string), is the member's path below the namespace, dot-joined (texture.create under renderer, proxy.component.get under entity), so each one names a single member and lsp.describe(namespace .. "." .. name) answers for it. The listing answers with the surface content calls: an entry registered internally is left out where its signature spells the __ binding or a Luau module publishes the same member, and opts.includeInternal lists every registered entry instead.

Parameters

namespace string — Namespace name (e.g. "entity", "modules/Transform").
opts MethodsOpts (optional) — Optional { filter?, namesOnly?, includeInternal? }.

Returns { MethodSummary } | { string } — Array of method summary tables, or plain name strings when namesOnly is set (empty when the namespace is unknown or nothing matches the filter).

for _, name in ipairs(lsp.methods("ui", { namesOnly = true })) do print(name) end
lsp.methods("renderer", { filter = "shadow" })

globals/lsp/modules

lsp.modules() -> { ModuleEntry }

List every Luau library module the engine currently knows about — discovered via --!module headers, library scans, and manually-recorded docs.

Returns { ModuleEntry } — Array of module summary tables.

globals/lsp/namespaces

lsp.namespaces(opts: NamespacesOpts?) -> { NamespaceEntry }

List the documentation namespaces reachable from Luau. By default only namespaces exposing at least one PUBLIC method are returned, so the list matches what you can actually call — internal FFI plumbing (e.g. pause, native_entity), whose public surface lives elsewhere (engine.paused, the entity proxy, …), is left out. Pass { includeInternal = true } to list every namespace, internal ones included.

Parameters

opts NamespacesOpts (optional) — Optional { includeInternal? } — default lists public only.

Returns { NamespaceEntry } — Array of namespace summary tables.

for _, ns in ipairs(lsp.namespaces()) do print(ns.name) end

globals/lsp/readDirectives

lsp.readDirectives(source: string) -> DirectiveBlock

Parse the leading --! directive block of a Luau source string. Used by UIs that audit which files have skip directives and what they suppress.

Parameters

source string — Luau source text.

Returns DirectiveBlock — { mode, codes? }.

globals/lsp/search

lsp.search(query: string, opts: SearchOpts?) -> { MethodSummary }

Case-insensitive substring search across every registered doc's path, signature, and description. Hits answer with the surface content calls: an entry registered internally is left out where its signature spells the __ binding or a Luau module publishes the same member, and opts.includeInternal searches every registered entry.

Parameters

query string — Substring to search for.
opts SearchOpts (optional) — { limit? = 50, includeInternal? }.

Returns { MethodSummary } — Array of method summary tables.

globals/lsp/setStrict

lsp.setStrict(enabled: boolean) -> boolean

Toggle the pre-execute LSP gate. Returns true when the change was persisted to .world_settings, false when the play-mode write lock blocked the write.

Parameters

enabled boolean — True = strict, false = off.

Returns boolean — Persistence signal.

globals/lsp/setStrictMode

lsp.setStrictMode(mode: StrictMode) -> boolean

Set the pre-execute strict gate's mode. Returns true when the change was persisted to .world_settings, false when the play-mode write lock blocked the write.

Parameters

mode StrictMode — "off" | "soft" | "strict".

Returns boolean — Persistence signal.

globals/lsp/summary

lsp.summary() -> Summary

Counts only — does not re-run validation.

Returns Summary — Counts of cached diagnostics by severity.

globals/lsp/symbolAt

lsp.symbolAt(path: string, line: number, col: number, name: string, chain: string?) -> SymbolHit?

The identifier name where it stands in path at line:col (1-based, the file's own columns), with the static type it resolves to THERE — a local inside a function body, a parameter, a field, a global — and, for a name a statement in the file introduced, declaredAt = {line, col}. The answer a hover shows and a go-to-definition jumps to. nil when nothing at that position is a reference to name: a word in a comment or a string, a keyword.

Parameters

path string — VFS path of the file.
line number — 1-based line.
col number — 1-based column in the file's own text.
name string — The identifier at that position.
chain string (optional) — The access path the identifier ends — entity.component.add for a pointer on add — which is where its API page is filed.

Returns SymbolHit? — The type descriptor of the reference (kind, display, and for a function its params and returns), with name, symbolKind (local, parameter, function, field, global, namespace, module), declaredAt when the file introduced the name, members and memberCount for a table, modulePath and moduleFile for a required module, and doc — the comment block above the declaration (--!desc, --!arg, --!return, --!example), the required module's header, or the API page under chain. nil when nothing there is a reference to name.

local hit = lsp.symbolAt("/zero/source/main.luau", 12, 9, "add", "entity.component.add")

globals/lsp/tools

lsp.tools() -> { ToolEntry }

List every code-mode tool registered in the VFS under /zero/docs/tools/<category>/<tool>.

Returns { ToolEntry } — Array of tool summary tables.

globals/lsp/typeOf

lsp.typeOf(expr_source: string, context_path: string?) -> TypeDescriptor

Infer the static type of a Luau expression. When context_path is given, the file is loaded and walked so the inference env contains every local + alias in scope at its end. An expression past the 2 MiB analysis bound answers kind unknown, and a context file past it adds nothing to the env.

Parameters

expr_source string — Luau expression source (no surrounding chunk).
context_path string (optional) — VFS path whose scope should be visible.

Returns TypeDescriptor — Type descriptor table.

local td = lsp.typeOf("Transform.lookAt", "/zero/source/main.luau")

globals/lsp/upstreamCheck

lsp.upstreamCheck(path: string, opts: UpstreamCheckOpts?) -> UpstreamCheck

Type-check a .luau file with upstream Luau's own type checker (its Frontend), and every module the file requires, and return what upstream reports as data. What lsp.check, the execute gate and the play gate report is unchanged by it. Each require literal resolves to the source file lsp.check's require-unresolved pass holds it to, which is the file the runtime loads wherever the runtime loads one, in whichever form it is written (@builtin::a.b, @ns::, @root::, a bare dotted a.b, .x, ..x, ./x, ../x, ~.x, ~pkg.x, an absolute path, a folder asset), and requires lists each literal of the file with the file it reached (module) and the file the runtime's require loads for it from this file (runtimeModule), nil where either reaches none. findings carry upstream's error kind or lint rule, the module they are in and 1-based positions in that module as written; they cover the file, and with opts.includeRequired the type errors of every module it reaches come back as well. Upstream reads every module through its analysis view: the engine's typed declarations read as the plain Luau functions they declare, their generic, parameter and return annotations kept, a return type written after -> read after :, and every line and column as in the source; view lists where the view of the file differs from it as { line, col, written, reads }. A module whose findings come back that upstream cannot parse through its view or past the 2 MiB analysis bound is listed in notAnalysable with the reason, and when that is the file itself analysable is false. The checked modules are kept between calls: checkedModules names the ones this call type-checked, empty when nothing the file reaches changed, and a write to a source re-checks that module and the modules depending on it. The phases of a module's check whose work grows with the types they meet stop at opts.timeLimitMs, and modules says of each module the call reached how its check ended ("completed", "timedOut" or "internal") and how long it took, reused marking one an earlier call checked that is unchanged since. The modules held between calls are bounded: once a call returns their sources sum to at most held.limitBytes, the modules the least recent calls reached going first, and held and evicted report it. A failure inside the analysis library comes back as an "internal" finding with completed false, and the internal errors upstream reports for a module come back as one "internal" finding whose count says how many. opts.typeCheck = false reads the file and every module it reaches and resolves their requires without checking types. available is false, with reason, on a build that does not link the analysis library.

Parameters

path string — VFS path of a .luau file, a module identity, or a @scope::path.
opts UpstreamCheckOpts (optional) — { solver?, mode?, timeLimitMs?, includeRequired?, typeCheck? }.

Returns UpstreamCheck — { available, path, analysable, reason?, completed, solver, mode, elapsedMs, findings, notAnalysable, requires, view, checkedModules, timedOut, modules, held, evicted, errors, warnings }.

local r = lsp.upstreamCheck("@builtin::modules.boids", { mode = "strict" })

globals/luau_profile/begin

luau_profile.begin(name: string) -> number

Open a named manual region. Returns an opaque integer id; pass it back to end_region(id) to close and record elapsed wall-clock under name.

Parameters

name string — Region name; aggregated across opens.

Returns number — Region id.

local id = luau_profile.begin("walk"); ...; luau_profile.end_region(id)

globals/luau_profile/end_region

luau_profile.end_region(id: number)

Close a region previously opened by begin(name). Records elapsed wall-clock under the region's name. Silently no-ops on unknown id (typically a double-close or swapped-out VM).

Parameters

id number — Region id returned by begin().

luau_profile.end_region(id)

globals/luau_profile/folded

luau_profile.folded() -> string

The folded-stack dump as a string, one line per stack as <ticks> <stack_csv> — the format upstream Luau emits and tools/perfgraph.py consumes unchanged. The same bytes dump writes, handed back instead of written, so the profile can go wherever the caller keeps it: vfs.write puts it in the engine filesystem, where bash and vfs.read reach it.

Returns string — Folded-stack text; empty when nothing was sampled.

vfs.write("/source/tmp/sample.folded", luau_profile.folded())

globals/luau_profile/is_running

luau_profile.is_running() -> boolean

True iff the background sampler is currently running.

Returns boolean — Sampler running state.

if luau_profile.is_running() then luau_profile.stop() end

globals/luau_profile/reset

luau_profile.reset()

Clear every accumulated sample and region stat. The sampler keeps running if it was already on; only the data is wiped.

luau_profile.reset()

globals/luau_profile/sampling_available

luau_profile.sampling_available() -> boolean

True on platforms where the background sampler can run (native targets), false on WASM. Manual regions work everywhere — only the sampler is platform-gated.

Returns boolean — Whether start() would actually spawn a sampler.

if luau_profile.sampling_available() then luau_profile.start() end

globals/luau_profile/snapshot

luau_profile.snapshot(top_n: number?) -> Snapshot

Snapshot the current accumulator without touching the filesystem — cheap enough for per-frame UI polling. top_n truncates stacks to the N hottest entries; omitting it returns all stacks sorted descending by self_us. regions is always returned in full (sorted by total_us).

Parameters

top_n number (optional) — Truncate stacks to this many entries; omit for all.

Returns Snapshot — Profile snapshot.

local snap = luau_profile.snapshot(10)

globals/luau_profile/start

luau_profile.start(hz: number?) -> StartResult

Start the background Luau sampling profiler at hz samples per second (default 1000, clamped to [1, 100000]). Idempotent — calling while already running is a no-op. Returns { available, hz } — available = false on WASM (no std::thread). Manual regions work regardless.

Parameters

hz number (optional) — Sampling rate in Hz.

Returns StartResult — Effective sampler state.

luau_profile.start(500)

globals/luau_profile/stop

luau_profile.stop()

Stop the background sampler. Blocks until the sampler thread joins (typically <1ms). Safe when not running. Does not clear the accumulator — call reset() to drop samples.

luau_profile.stop()

globals/mathx/addScaledVec3

mathx.addScaledVec3(dstBuffer: Substrate.TypedBuffer, srcBuffer: Substrate.TypedBuffer, count: number, scale: number) -> boolean

dst[i] += src[i] * scale for count vec3 elements. Both buffers must hold at least count * 3 floats. Useful for particle integration (position += velocity * dt) and accumulator passes.

Parameters

dstBuffer Substrate.TypedBuffer — The buffer written into.
srcBuffer Substrate.TypedBuffer — The buffer read from.
count number — Number of vec3 elements.
scale number — Multiplier applied to every src element.

Returns boolean — True on success.

mathx.addScaledVec3(positions, velocities, n, dt)

globals/mathx/dampScalar

mathx.dampScalar(buffer: Substrate.TypedBuffer, offset: number, count: number, target: number, smoothTime: number, dt: number) -> boolean

Critically-damped exponential approach toward target for count scalars at buffer[offset .. offset+count]. smoothTime is the time constant (~ 0.16 ⇒ ~63% per frame at 60 Hz). Pass smoothTime <= 0 to snap to the target.

Parameters

buffer Substrate.TypedBuffer — The buffer to operate on.
offset number — Starting f32 index.
count number — Number of scalars.
target number — Target value all scalars approach.
smoothTime number — Time constant (≤ 0 snaps to target).
dt number — Frame time in seconds.

Returns boolean — True on success, false on a bad handle or out-of-range slice.

mathx.dampScalar(buf, 0, 16, 0.0, 0.16, dt)

globals/mathx/lerpVec3

mathx.lerpVec3(buffer: Substrate.TypedBuffer, offset: number, count: number, tx: number, ty: number, tz: number, t: number) -> boolean

Element-wise linear blend of count vec3s in buffer[offset .. offset+count*3] toward (tx, ty, tz) by t.

Parameters

buffer Substrate.TypedBuffer — The buffer to operate on.
offset number — Starting f32 index.
count number — Number of vec3 elements.
tx number — Target X.
ty number — Target Y.
tz number — Target Z.
t number — Blend amount (0..1).

Returns boolean — True on success.

mathx.lerpVec3(buf, 0, n, 0, 1, 0, 0.5)

globals/mathx/normalizeQuat

mathx.normalizeQuat(buffer: Substrate.TypedBuffer, offset: number, count: number) -> boolean

Re-normalise count quaternions in place. Zero-length quats become identity (0, 0, 0, 1) so downstream code never sees NaN.

Parameters

buffer Substrate.TypedBuffer — The buffer to operate on.
offset number — Starting f32 index.
count number — Number of quaternions.

Returns boolean — True on success.

mathx.normalizeQuat(buf, 0, n)

globals/mathx/slerpQuat

mathx.slerpQuat(buffer: Substrate.TypedBuffer, offset: number, count: number, tx: number, ty: number, tz: number, tw: number, t: number) -> boolean

Slerp count quaternions (xyzw) at buffer[offset..] toward (tx, ty, tz, tw) by t. Falls back to nlerp+normalize for very-close quats. Always picks the shortest-arc path.

Parameters

buffer Substrate.TypedBuffer — The buffer to operate on.
offset number — Starting f32 index.
count number — Number of quaternions.
tx number — Target quat X.
ty number — Target quat Y.
tz number — Target quat Z.
tw number — Target quat W.
t number — Slerp amount (0..1).

Returns boolean — True on success.

mathx.slerpQuat(buf, 0, n, 0, 0, 0, 1, 0.25)

globals/mathx/transformVec3

mathx.transformVec3(buffer: Substrate.TypedBuffer, offset: number, count: number, mat16: { number }) -> boolean

Treat each vec3 in buffer[offset..] as a position (w = 1), multiply by the 4x4 column-major matrix mat16 (16-element array), write .xyz of the result back. Layout matches glam, wgpu, and GLSL conventions.

Parameters

buffer Substrate.TypedBuffer — The buffer to operate on.
offset number — Starting f32 index.
count number — Number of vec3 elements.
mat16 { number } — Column-major 4x4 matrix as a 16-element array.

Returns boolean — True on success.

mathx.transformVec3(positions, 0, n, worldMatrix)

globals/mcpLog/clear

mcpLog.clear() -> boolean

Clear all entries from the engine's MCP log ring buffer.

Returns boolean — True on success.

mcpLog.clear()

globals/mcpLog/query

mcpLog.query(limit: number?) -> { McpLogEntry }

Return the most-recent MCP tool-call entries from the engine's MCP log ring buffer (newest last). Pass limit to cap how many entries are returned — omit for the full ring (up to 500 entries).

Parameters

limit number (optional) — Maximum number of entries to return.

Returns { McpLogEntry } — Array of tool-call entry tables.

for _, e in ipairs(mcpLog.query(50)) do print(e.tool_name, e.status) end

globals/microphone/awaitRunning

microphone.awaitRunning(timeout: number?) -> (MicState, string?)

Wait until the capture settles out of starting and permissionPending, and report where it landed. Returns as soon as the state settles, or when timeout seconds have passed, whichever comes first — a browser permission prompt nobody answers never settles, so the wait is always bounded.

Parameters

timeout number (optional) — Seconds to wait at most. Defaults to 10.

Returns (MicState, string?) — The state reached, and its reason where it has one.

microphone.start(); local state, why = microphone.awaitRunning()

globals/microphone/devices

microphone.devices() -> { MicDevice }

Every input device the platform offers. id is what microphone.start takes to select one and is stable across reboots where the platform provides a stable identifier; name is the label a person recognises.

An empty list is a legitimate answer, not a failure: a machine with no input hardware offers none, and a browser names none until microphone access has been granted at least once — the labels are part of what the permission protects.

Returns { MicDevice } — Array of { id, name, default }.

for _, d in ipairs(microphone.devices()) do print(d.name, d.default) end

globals/microphone/frequencies

microphone.frequencies() -> { number }

The frequency each spectrum bin is centred on, in Hz, as an array parallel to microphone.spectrum(). Derived from the capture's rate and transform size, so it changes only when a capture is started with different ones. Empty while no capture is running.

Returns { number } — Array of centre frequencies, one per bin.

local hz = microphone.frequencies(); print(hz[#hz]) -- the Nyquist frequency

globals/microphone/level

microphone.level() -> number

Loudness of the most recent analysis window, as an RMS amplitude in 0..1. A full-scale sine reads about 0.707 and silence reads 0.

Measured over only the samples that have arrived, so a capture that has just started reports the loudness of what it holds rather than a level diluted by a window it has not filled yet. 0 while no capture is running.

Returns number — RMS amplitude, 0..1.

if microphone.level() > 0.05 then print("someone is talking") end

globals/microphone/peak

microphone.peak() -> { [string]: number }?

The bin carrying the most energy and what it says: the frequency it is centred on, its amplitude, and the loudness of the whole window. A capture reading silence answers with amplitude 0 at bin 1.

Returns { [string]: number }? — { bin, hz, amplitude, level }, or nil while no capture is running.

local p = microphone.peak(); if p and p.amplitude > 0.05 then print(p.hz) end

globals/microphone/samples

microphone.samples(max: number?) -> buffer?

Captured mono PCM no caller has taken yet, oldest sample first, as a buffer of little-endian f32 read with buffer.readf32. The samples are removed, so successive calls walk forward through the capture and a caller doing its own analysis sees every frame once.

nil while no capture is running, and a zero-length buffer when the capture is running and nothing new has arrived. Samples nobody takes are discarded once the queue fills, and status().overruns counts every one.

Parameters

max number (optional) — How many samples to take at most. Omitted, everything held comes back.

Returns buffer? — Buffer of f32 samples, or nil when no capture is running.

local pcm = microphone.samples(); if pcm then print(buffer.len(pcm) // 4) end

globals/microphone/spectrum

microphone.spectrum() -> { number }

Amplitude per frequency bin over the most recent analysis window: fftSize / 2 + 1 numbers, DC at index 1 through the Nyquist frequency at the last. Bin i covers (i - 1) * status().binHz Hz.

Each value is an amplitude estimate rather than a raw transform magnitude, so a full-scale tone sitting on a bin centre reads about 1.0 and the numbers stay comparable across transform sizes.

The window is multiplied by a Hann taper before the transform. An untapered window ends abruptly at both edges and the transform reads that as energy spread across every bin, smearing one tone into a skirt that buries quieter tones beside it. Hann trades a slightly wider main lobe — a tone occupies about three bins rather than one — for sidelobes that fall away steeply, which is what lets neighbouring tones be told apart. Read a peak as "a tone near here", not "a tone exactly here".

Reading this takes no samples away from microphone.samples(). Empty while no capture is running.

Returns { number } — Array of amplitudes, one per bin.

local bins = microphone.spectrum(); print(#bins, bins[1])

globals/microphone/start

microphone.start(opts: MicOpts?) -> (MicState?, string?)

Open an input device and begin capturing. Returns the state the capture reached — "running" once a device is delivering, or "permissionPending" where the platform must ask for access first, which is the browser's normal path. Poll microphone.status() from there, or use microphone.awaitRunning(). Where access is refused without asking (a native engine's fake microphone presented as deny), it returns "denied".

A request that cannot be made at all returns nil and the reason: an fftSize that is not a whole power of two between 64 and 16384, a device no machine here offers, a rate the device does not capture at, or a capture that is already running.

Omitting device opens the platform default. Omitting sampleRate takes the device's own rate, which is what avoids a resample. fftSize is how many samples one analysis window covers and defaults to 1024 — at 48 kHz that spans ~21 ms and resolves ~47 Hz per bin.

Parameters

opts MicOpts (optional) — { device, sampleRate, fftSize }.

Returns (MicState?, string?) — The state reached, or nil and the reason the request was refused.

local state, why = microphone.start({ fftSize = 2048 })

globals/microphone/status

microphone.status() -> MicStatus

Where the capture stands.

reason carries the platform's own message: the refusal for denied, the device's message for failed, what is being waited on for permissionPending. binHz is the width of one spectrum bin and bins how many microphone.spectrum() returns.

framesCaptured counts every mono frame the device delivered whether or not anything drained it, so a silent room reads differently from a stalled device. overruns counts samples discarded because a consumer did not keep up — it standing still is what says the readings are continuous, and it climbing is why a caller sees gaps.

Returns MicStatus — { state, reason, device, sampleRate, fftSize, binHz, bins, framesCaptured, overruns }.

local s = microphone.status(); print(s.state, s.framesCaptured, s.overruns)

globals/microphone/stop

microphone.stop() -> boolean

Stop the capture and release the device. True when a capture was open or being opened at call time. The device is let go before this returns, so a stop followed by a start opens it again rather than finding it held.

Returns boolean — Whether a capture was active.

microphone.stop()

globals/modelImport/decompose

modelImport.decompose(bytes: buffer | string, format: string) -> PromiseHandle

Parse raw model bytes on a background thread. format is the real source extension ("fbx", "obj", "dae", "gltf", "glb", "stl", "ply", "3ds", …), forwarded to assimp as the format hint. Returns a promise handle: task.await it, then read the data with result(handle).

Parameters

bytes buffer | string — Raw model file bytes (from vfs.readAsync).
format string — The source file extension (lowercase, no dot).

Returns PromiseHandle — Promise handle for task.await.

local h = modelImport.decompose(bytes, "obj"); task.await(h)

globals/modelImport/decomposeFiles

modelImport.decomposeFiles(files: { ModelFile }, mainName: string) -> PromiseHandle

Parse a model plus its companion files on a background thread, so assimp resolves the model's external references (a .gltf's external .bin and image files, an .obj's .mtl colors/textures, MD5's .md5anim, …). files is an array of { name = basename, bytes = <bytes> } that MUST include the model file itself; mainName is that file's basename. Returns a promise handle: task.await it, then read the data with result(handle) — the same shape decompose produces.

Parameters

files { ModelFile } — Array of { name, bytes }: the model file plus its companions.
mainName string — Basename of the model file to import (one of files' names).

Returns PromiseHandle — Promise handle for task.await.

local h = modelImport.decomposeFiles(files, "CesiumMilkTruck.gltf"); task.await(h)

globals/modelImport/extractAnimation

modelImport.extractAnimation(sourcePath: string, clipName: string) -> PromiseHandle

Read a model source file and extract one animation clip to its .zanim payload, stashed for retrieval. The source extension decides the parser, so this is format-agnostic. Returns a promise handle: task.await it, then extractAnimationResult(handle) returns the bytes.

Parameters

sourcePath string — VFS path to the source model file.
clipName string — Clip name as returned by result(handle).animations[i].name.

Returns PromiseHandle — Promise handle for task.await.

globals/modelImport/extractAnimationResult

modelImport.extractAnimationResult(handle: PromiseHandle | string) -> string?

After awaiting an extractAnimation handle, return the extracted .zanim bytes (binary-safe), consuming them. The bytes to hand to asset.create("animation", name, { bytes }). Returns nil on failure or if already taken.

Parameters

handle PromiseHandle | string — Promise handle from extractAnimation.

Returns string? — The clip's .zanim payload bytes, or nil.

globals/modelImport/result

modelImport.result(handle: PromiseHandle | string) -> any?

Read the decomposed model after decompose's handle has been awaited. Every format decomposes into the same shape, so this is format-agnostic. Runs on the main thread; consumes the stored result.

Parameters

handle PromiseHandle | string — Promise handle from decompose.

Returns any? — { nodes, meshes, materials, textures, animations, hasSkeleton, skeletonRootNode?, skeleton? }, or nil.

globals/modelImport/retryHandle

modelImport.retryHandle(makeHandle: () -> any, retries: number?, yield: (() -> ())?) -> (PromiseHandle | string)?

Parameters

makeHandle () -> any
retries number (optional)
yield (() -> ()) (optional)

Returns (PromiseHandle | string)?

globals/modelImport/rigFromMeshSkin

modelImport.rigFromMeshSkin(meshBytes: buffer | string) -> string?

Lift the skeleton out of a skinned .mesh (ZMSH) payload and return it as a .rig JSON document: bones (hierarchy, rest pose, inverse-bind). The source rig a skinned mesh's clips retarget through. Returns nil when the bytes are not a mesh or carry no skin.

Parameters

meshBytes buffer | string — Raw ZMSH mesh bytes carrying a skin.

Returns string? — .rig JSON document, or nil.

globals/modelImport/rigFromSkeleton

modelImport.rigFromSkeleton(skeleton: AnimationSkeleton) -> string?

Build a .rig JSON document from the skeleton an animation-only file was authored on — result(handle).skeleton, the bones its clips drive with their local rest transforms. Forward kinematics over the locals resolves globals + inverse-bind. The source rig a standalone clip retargets through. Returns nil on malformed input.

Parameters

skeleton AnimationSkeleton — { names, parents, locals } (a decompose result's skeleton).

Returns string? — .rig JSON document, or nil.

local rigJson = modelImport.rigFromSkeleton(data.skeleton)

globals/multiplayer/beginOperation

multiplayer.beginOperation(description: string)

Begin recording an undoable operation. All mutations until commitOperation() are grouped into one undo entry.

Parameters

description string — Human-readable label.

multiplayer.beginOperation("move cube")

globals/multiplayer/canRedo

multiplayer.canRedo() -> boolean

Check if this client has any redoable operations.

Returns boolean — True if redo is available.

globals/multiplayer/canUndo

multiplayer.canUndo() -> boolean

Check if this client has any undoable operations.

Returns boolean — True if undo is available.

globals/multiplayer/cancelOperation

multiplayer.cancelOperation()

Cancel the current operation and restore all properties to their values at begin time.

multiplayer.cancelOperation()

globals/multiplayer/claimOwnership

multiplayer.claimOwnership(entityId: (string | entityRef)?) -> boolean

Request ownership of an entity. The relay grants a claim on an entity no peer owns (never claimed, or released by its owner) and tells every peer; an entity another peer owns stays theirs, and the refusal is counted in getDiagnostics().refusals and named by lastRefusal. The owner hands an entity on by releasing it.

Parameters

entityId (string | entityRef) (optional) — Entity id or proxy to claim.

Returns boolean — True once the claim is on its way to the relay.

globals/multiplayer/clearHistory

multiplayer.clearHistory()

Drop this client's whole undo/redo history — for boundaries where old edits stop being meaningful (a scene load, a test rig reset).

multiplayer.clearHistory()

globals/multiplayer/commitOperation

multiplayer.commitOperation()

Finalize the current operation and push it onto the undo stack. Only changes that actually differ from the start state are recorded.

multiplayer.commitOperation()

globals/multiplayer/connect

multiplayer.connect(relayUrl: string)

Connect to a multiplayer relay server for the current world. Uses the loaded world's world_id as the room prefix for scene isolation. A world must be loaded before connecting.

Parameters

relayUrl string — Relay server URL.

multiplayer.connect("https://relay.example.com")

globals/multiplayer/disconnect

multiplayer.disconnect()

Disconnect from the multiplayer relay server.

multiplayer.disconnect()

globals/multiplayer/explain

multiplayer.explain(entityId: (string | entityRef), componentType: string, property: string) -> DeliveryVerdict

Why a synced property is not reaching the peers this client shares its entity's room with. Answers from the engine's own registry, so a name the component never registered is reported as such instead of inferred from a second client's silence.

Parameters

entityId (string | entityRef) — Entity id or proxy.
componentType string — Component type name, e.g. "Health".
property string — Property name as written in the component's sync {} block.

Returns DeliveryVerdict — arriving true when the property is on its way; otherwise reason names the cause and property carries the registry's record of it when one exists.

local v = multiplayer.explain(player, "Health", "hp")
if not v.arriving then print(v.reason) end

globals/multiplayer/getDiagnostics

multiplayer.getDiagnostics() -> SyncDiagnostics

Get sync diagnostics — traffic counts, bandwidth, link quality, peer count. The counts — bytesSent/Received, datagramsSent/Received, rpcsSent/Received, ownershipChanges, refusals — are running totals for the session, so a sparse event stays readable long after it happened; subtract two samples for the rate over the interval between them. refusals counts the messages of this session's the relay refused (a claim on an entity another peer owns, a change to one, a room outside the world the connection was admitted to), and lastRefusal names the latest in the relay's words once there is one. bytesSentPerSec / bytesReceivedPerSec are averages over the last completed ~1 second window. rttMs is the smoothed round-trip time to the relay and packetLoss the fraction (0..1) of packets lost over the last 5 seconds; both read 0 until the transport has sampled a live connection. messagesAwaitingEntity counts the sync messages this peer is holding for an entity it has not received yet — each waits for the spawn that names it, applies the moment it arrives, and is released once its wait runs out.

Returns SyncDiagnostics — Diagnostics table.

local d = multiplayer.getDiagnostics()
if d.packetLoss > 0.05 then log.warn(("unstable connection, %d ms round trip"):format(d.rttMs)) end
print(d.messagesAwaitingEntity .. " message(s) waiting for their entity")

globals/multiplayer/getPeerId

multiplayer.getPeerId() -> number?

Get the peer ID the relay issued this client's connection.

Returns number? — Local peer ID, or nil exactly when isConnected() is false: while dialing, before the relay names the connection, and after it drops.

globals/multiplayer/getPeers

multiplayer.getPeers() -> { PeerInfo }

Get a list of all connected peers in the current session.

Returns { PeerInfo } — Array of peer info tables.

for _, p in ipairs(multiplayer.getPeers()) do print(p.name) end

globals/multiplayer/getRoomPeers

multiplayer.getRoomPeers(roomKey: string) -> { PeerInfo }

Get the peers this client shares the given room with, ordered by peer id. getPeers answers for the whole session — the union of every room this client is in — while this answers for one room, so a peer that leaves this room while staying in another disappears from here and remains in getPeers.

Parameters

roomKey string — Fully-qualified room key ({worldGuid}/{profile}/{mode}/{sceneGuid}).

Returns { PeerInfo } — Array of peer info tables for that room.

for _, p in ipairs(multiplayer.getRoomPeers(key)) do print(p.id) end

globals/multiplayer/getRooms

multiplayer.getRooms() -> { string }

The relay rooms this client has joined, sorted. A broadcast reaches only the peers that share one of these. They are the rooms of observe() as of the last frame, read from the rooms the session publishes every frame, so a component can ask every frame.

Returns { string } — Room keys.

for _, key in ipairs(multiplayer.getRooms()) do print(key) end

globals/multiplayer/getTickRate

multiplayer.getTickRate() -> number

Get the current sync tick rate (network updates per second).

Returns number — Sync ticks per second (default 20).

globals/multiplayer/heldMessages

multiplayer.heldMessages() -> { HeldMessage }

The sync messages this peer is holding for entities it has not received — what getDiagnostics().messagesAwaitingEntity counts, one entry each, with the entity sync id it names, its age and the grace it is held against.

Returns { HeldMessage } — Held messages.

for _, m in ipairs(multiplayer.heldMessages()) do print(m.kind, m.ageMs) end

globals/multiplayer/isConnected

multiplayer.isConnected() -> boolean

Check if a multiplayer session is connected to a relay and the relay has named this client. True exactly when getPeerId() answers an id.

Returns boolean — True if connected under the peer id the relay issued.

globals/multiplayer/isHost

multiplayer.isHost() -> boolean

Whether THIS client is the host (authoritative owner) of the current scene's play room — the relay room CREATOR, or offline / single-player. Host code spawns the shared synced world (via entity.spawnSynced or a scene's onHostLoad) and runs authoritative simulation; a non-host (JOINER) receives that content from the relay snapshot and must NOT re-create it. Gate ANY code that spawns synced entities or owns shared state with this so it runs on exactly one client — running it on every peer is the double-spawn 'explosion'.

Returns boolean — True on the host / offline / single-player; false on a confirmed joiner. Defaults to true when the role isn't known yet (degrade to host so single-player and pre-join code still run) — pair with a scene's onHostLoad hook when exact one-shot timing matters.

if multiplayer.isHost() then enemy = entity.spawnSynced("enemy") end

globals/multiplayer/isOperationOpen

multiplayer.isOperationOpen() -> boolean

Whether an operation is open: true between beginOperation and its commitOperation or cancelOperation. A caller that records file writes reads the file's prior contents only while one is open.

Returns boolean — True while an operation is open.

if multiplayer.isOperationOpen() then before = vfs.read(p) end

globals/multiplayer/isOwner

multiplayer.isOwner(entityId: (string | entityRef)?) -> boolean

Check if the local client owns the given entity (or the current entity if called from a component). Only the owner can modify synced properties directly.

Parameters

entityId (string | entityRef) (optional) — Entity id or proxy to check (defaults to self.entityId in component context).

Returns boolean — True if the local client is the owner.

globals/multiplayer/isRoomCreator

multiplayer.isRoomCreator(roomKey: string) -> boolean?

Whether this client created the given room — it was the FIRST peer to join it (race-free; the relay assigns it on join). In play mode the creator instantiates the scene's entities (synced) and every other joiner receives them from the relay snapshot, so the scene is never double-instantiated.

Parameters

roomKey string — Fully-qualified room key ({worldGuid}/{profile}/{mode}/{sceneGuid}).

Returns boolean? — True if this client created the room, false if it joined an existing one, nil if the relay hasn't reported a role yet (offline / not joined).

if multiplayer.isRoomCreator(key) ~= false then layers.load(scene) end

globals/multiplayer/joinRoom

multiplayer.joinRoom(roomKey: string)

Join a relay room. Room keys are built as {worldGuid}/{profile}/{mode}/{sceneGuid} — four segments, the {profile} one keeping a runtime peer (published content) and an editor peer (live content) in separate rooms even when both are in play mode. Rooms partition the relay's fan-out: only peers in the same room receive each other's broadcasts. getRooms() reports the keys this client is already in and roomFor(entity) the one an entity broadcasts into, so a key can be read rather than rebuilt. A relay that verifies its peers holds each connection to the world it was admitted to: a key whose first segment names another world is refused, and getDiagnostics().refusals counts the refusal and lastRefusal names it. No-op when not connected or already joined.

Parameters

roomKey string — Fully-qualified room key.

multiplayer.joinRoom(worldGuid .. "/" .. engine.profile .. "/play/" .. sceneGuid)

globals/multiplayer/leaveRoom

multiplayer.leaveRoom(roomKey: string)

Leave a relay room. The key is reported under observe().withdrawnRooms until joinRoom names it again. No-op when not connected or not joined.

Parameters

roomKey string — Fully-qualified room key.

globals/multiplayer/loopback

multiplayer.loopback() -> { [string]: any }

Loopback testing harness. Returns a table with enable(), disable(), flush(), receive() methods for testing sync without a relay server.

Returns { [string]: any } — Loopback API table.

globals/multiplayer/observe

multiplayer.observe() -> ReplicationObservation

Report what this peer is replicating and why a property is not arriving. Carries the rooms this client joined, one record per entity with a sync id — its owner, the room it broadcasts into, how many other peers share that room, and every REGISTERED synced component with its declared property names, wire indices, public/private table and dirty bits — the messages held for entities that have not arrived, and the registry's totals. Every property carries notArriving: one name from reasons, or nil when it is on its way. Answers in edit mode as well as play mode, for what the relay carries in each: in edit mode scene content is left out of the sync-id pass, so its changes travel to the other clients with the source they are written into and it reads entityNotSynced here. The document describes the session as of the last frame: when no frame has assembled one since the surface was last read, the call waits for the next frame to assemble it, and a caller that cannot yield receives the one published last.

Returns ReplicationObservation — The engine's current replication observation.

local o = multiplayer.observe()
print(#o.entities .. " entities, " .. o.totals.properties .. " synced properties")

globals/multiplayer/observeComponent

multiplayer.observeComponent(entityId: (string | entityRef), componentType: string) -> SyncedComponent?

The registered synced component of the named type on an entity's record. Matches a fully-qualified type (@builtin::components.Model) and the leaf name it ends in (Model) alike.

Parameters

entityId (string | entityRef) — Entity id or proxy.
componentType string — Component type name or its leaf.

Returns SyncedComponent? — The registered component instance, or nil when none of that type is registered on the entity.

local c = multiplayer.observeComponent(player, "Health")
print(#c.properties .. " declared properties")

globals/multiplayer/observeEntity

multiplayer.observeEntity(entityId: (string | entityRef)) -> EntityReplication?

The replication record for one entity — its sync id, owner, room, and the synced components registered on it.

Parameters

entityId (string | entityRef) — Entity id or proxy.

Returns EntityReplication? — The entity's record, or nil when the engine holds no sync record for it.

local r = multiplayer.observeEntity(e)
for _, c in ipairs(r.components) do print(c.componentType) end

globals/multiplayer/on

multiplayer.on(channel: string, callback: (number, ...any) -> ())

Subscribe to a custom message channel. The callback runs as callback(fromPeerId, ...args) whenever another peer calls multiplayer.send(channel, ...). Multiple callbacks per channel fire in registration order.

Parameters

channel string — Channel name to listen on.
callback (number, ...any) -> () — function(fromPeerId: number, ...) — the sender's peer id then the sent args.

globals/multiplayer/recordFileWrite

multiplayer.recordFileWrite(path: string, before: string?, after: string?) -> boolean

Record a file write into the open operation. Undoing the operation writes before back through the VFS (removing the file when before is nil), which runs the owning asset type's onChange as the original write did; redoing writes after. The path is resolved and held to the rules vfs.write applies, so a sidecar (.meta, .refs, .pull) or an asset root no script may write raises. A step's files are put back all together or not at all: each file is checked (its asset exists, and it holds what the step being reversed left in it) before any is written and again immediately before it is written; when a check or a write fails partway, the files the step already wrote go back to what they held, and a notice names each file of the step and what happened to it. Several writes of one path in an operation undo to the first before. Records nothing and returns false when no operation is open.

Parameters

path string — VFS path of the written file.
before string (optional) — The file's contents before the write, nil when it did not exist.
after string (optional) — The file's contents after the write, nil when the write removed it.

Returns boolean — True when the write was recorded.

local before = vfs.read(p); vfs.write(p, body); multiplayer.recordFileWrite(p, before, body)

globals/multiplayer/recordSpawn

multiplayer.recordSpawn(entityId: string)

Adopt an existing entity into the open operation as its spawn — for flows that create an entity before the operation opens (a drag preview adopted on drop). Undoing the operation despawns it.

Parameters

entityId string — Entity id to record as spawned by this operation.

multiplayer.recordSpawn(id)

globals/multiplayer/redo

multiplayer.redo() -> boolean

Redo this client's last undone operation.

Returns boolean — True if an operation was redone.

globals/multiplayer/releaseOwnership

multiplayer.releaseOwnership(entityId: (string | entityRef)?) -> boolean

Release ownership of an entity, leaving it unowned so another peer's claim takes it. Only the owner releases an entity.

Parameters

entityId (string | entityRef) (optional) — Entity id or proxy to release.

Returns boolean — True if ownership was released.

globals/multiplayer/roomFor

multiplayer.roomFor(entityId: (string | entityRef)) -> string?

The room key an entity's spawns and property deltas broadcast into: the room of its observeEntity record as of the last frame, read from the rooms the session publishes every frame, so a component can ask every frame.

Parameters

entityId (string | entityRef) — Entity id or proxy.

Returns string? — The room key, or nil when the engine has established no scene context for the entity.

local key = multiplayer.roomFor(player)
if key then multiplayer.joinRoom(key) end

globals/multiplayer/send

multiplayer.send(channel: string, ...: any?)

Broadcast a message on a named channel to every OTHER peer in the room. The relay forwards it transparently; peers receive it via multiplayer.on. The arguments are delivered to listeners in order, each one arriving as what it is, at the top level and at every depth of a table:

nil, a boolean, a number and a string arrive as themselves.
A vector arrives as a vector.
The proxy of a synced entity arrives as the receiving peer's proxy for that entity; the proxy of an entity no peer shares, as its id string.
A synced component (its proxy, or the componentRef envelope naming it) arrives as that component on the receiving peer; the proxy of a component that holds no sync id, as the envelope { __ref, entity, componentType } naming it.
A transform vector an entity hands out (e.position, e.rotation, e.localScale, ...) arrives as the table of its components, { x, y, z } or { x, y, z, w }.
Every other table is data and arrives as a new table, each value carried by these same rules: { x = 5, kind = "coin" } arrives as { x = 5, kind = "coin" }, and { who = e, n = 1 } with t.who the receiver's proxy for e. A table whose length #t is above 0 arrives as the list of its values at 1..#t alone; any other arrives keyed by its string keys and by each number key as the string of its whole part ({ [5] = "a" } arrives as { ["5"] = "a" }).
A value no peer can hold (a function, a thread, a buffer) arrives as nil. The same rules carry the arguments of a declare.syncedFunctions method, the payload of a Sync event and a synced table-valued field. No-op when not connected.

Parameters

channel string — Channel name listeners subscribe to via multiplayer.on.
... any (optional) — Zero or more values delivered to each listener after the sender's peer id.

globals/multiplayer/syncTotals

multiplayer.syncTotals() -> SyncTotals

What the sync registry holds across every entity: entities with a registered synced component, component instances, declared properties, declared functions, and the component instances holding a dirty property this tick.

Returns SyncTotals — The registry totals.

print(multiplayer.syncTotals().properties .. " synced properties registered")

globals/multiplayer/undo

multiplayer.undo() -> boolean

Undo this client's last edit-mode operation.

Returns boolean — True if an operation was undone.

globals/next

next(table, key?) -> key, value

Raw table iteration.

globals/notices/post

notices.post(template: string, params: { [string]: any }?, opts: NoticeOpts?)

Post a notice. template is a fixed sentence used to collapse repeats; put varying values in params. A param holding a picture — an image value from mcp.image or a capture tool — is delivered as a picture beside the notice's line rather than as text. opts.severity defaults to "info"; opts.includeLocation attaches the emitting call site. opts.actionable says whether the callers a notice is addressed to are expected to act on it, rather than take it in as background; an agent's event receiver starts an idle agent's turn for an actionable notice and hands the rest to its context.

Parameters

template string — Fixed sentence identifying the notice.
params { [string]: any } (optional) — Optional table of named values rendered alongside the template; a picture in it arrives as a picture.
opts NoticeOpts (optional) — Optional table: severity ("info" | "warn" | "error"), includeLocation (boolean), audience ("" | { string }?). Who sees the notice: "" addresses every caller; a list addresses those callers only, each entry an agent's name (as the agent toolbox's roster lists it) or a caller key <kind>:<id> such as "agent:". Defaults to the caller whose code is running, or every caller when none is. A notice reaches the agents on this engine: an entry naming no agent on this world's roster, an entry naming an agent the roster lists on another engine of the world, or an empty list, raises an error. actionable (boolean) states whether the callers it is addressed to are expected to act on it; omitted, a notice addressed to particular callers (the default audience included) is actionable, and one for every caller is background.

notices.post("wave complete", { wave = 3 })
notices.post("save slot corrupted, using defaults", { slot = id }, { severity = "warn" })
notices.post("bake finished", { shot = mcp.image("/source/tmp/bake.png") })
notices.post("boss defeated", nil, { audience = "*" })
notices.post("build ready for review", nil, { audience = { "terrain" }, actionable = true })

globals/notices/withdraw

notices.withdraw(template: string) -> number

Withdraw the notices this code posted under template that are still stored, whoever they are addressed to, because what they state has stopped being true: a caller that has not read one yet would be told something that no longer holds. A notice is keyed by its template and by the code that posted it, so a call reaches only notices posted from the same place. A later notices.post under the template queues again.

Parameters

template string — The template the notices were posted under.

Returns number — How many emissions the withdrawn notices held.

notices.withdraw("save slot corrupted, using defaults")

globals/nx/add

nx.add(dst: NxBuffer, x: NxScalarOrBuffer) -> boolean

b[i] += x (x scalar) or b[i] += x[i] (x buffer). Dispatches on type(x). For interleaved-stride writes use nx.addStrided.

Parameters

dst NxBuffer — Target buffer (mutated).
x NxScalarOrBuffer — Either a scalar or a same-shaped buffer.

Returns boolean — true on success, false on type / handle errors.

nx.add(b, 1.5)
nx.add(dst, src)

globals/nx/addStrided

nx.addStrided(b: NxBuffer, scalar: number, stride: number, offset: number?) -> boolean

Strided add: buf[k * stride + offset] += scalar for every valid k.

Parameters

b NxBuffer — Target buffer (mutated).
scalar number — Per-element addend.
stride number — Element stride.
offset number (optional) — Optional element offset (default 0).

Returns boolean — true on success, false on unknown handle.

nx.addStrided(buf, 1.0, 3, 1)

globals/nx/addStridedFrom

nx.addStridedFrom(dst: NxBuffer, src: NxBuffer, scale: number?, dst_stride: number, dst_off: number?, src_stride: number, src_off: number?) -> boolean

Strided BLAS-axpy from src into dst: dst[k * dst_stride + dst_off] += scale * src[k * src_stride + src_off].

Parameters

dst NxBuffer — Target buffer (mutated).
src NxBuffer — Source buffer.
scale number (optional) — Optional src scale factor (default 1.0).
dst_stride number — Destination element stride.
dst_off number (optional) — Optional destination offset (default 0).
src_stride number — Source element stride.
src_off number (optional) — Optional source offset (default 0).

Returns boolean — true on success, false on shape mismatch / unknown handle.

nx.addStridedFrom(dst, src, 1, 3, 0, 3, 0)

globals/nx/applyAbs

nx.applyAbs(b: NxBuffer) -> boolean

In-place b[i] = abs(b[i]).

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applyAbs(b)

globals/nx/applyCeil

nx.applyCeil(b: NxBuffer) -> boolean

In-place b[i] = ceil(b[i]).

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applyCeil(b)

globals/nx/applyCos

nx.applyCos(b: NxBuffer) -> boolean

In-place b[i] = cos(b[i]).

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applyCos(b)

globals/nx/applyExp

nx.applyExp(b: NxBuffer) -> boolean

In-place b[i] = exp(b[i]).

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applyExp(b)

globals/nx/applyFloor

nx.applyFloor(b: NxBuffer) -> boolean

In-place b[i] = floor(b[i]).

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applyFloor(b)

globals/nx/applyFract

nx.applyFract(b: NxBuffer) -> boolean

In-place b[i] = fract(b[i]) (fractional part).

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applyFract(b)

globals/nx/applyLog

nx.applyLog(b: NxBuffer) -> boolean

In-place b[i] = ln(b[i]).

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applyLog(b)

globals/nx/applyLog2

nx.applyLog2(b: NxBuffer) -> boolean

In-place b[i] = log2(b[i]).

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applyLog2(b)

globals/nx/applyNeg

nx.applyNeg(b: NxBuffer) -> boolean

In-place b[i] = -b[i].

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applyNeg(b)

globals/nx/applyRecip

nx.applyRecip(b: NxBuffer) -> boolean

In-place b[i] = 1 / b[i].

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applyRecip(b)

globals/nx/applyRecipSqrt

nx.applyRecipSqrt(b: NxBuffer) -> boolean

In-place b[i] = 1 / sqrt(b[i]).

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applyRecipSqrt(b)

globals/nx/applyRound

nx.applyRound(b: NxBuffer) -> boolean

In-place b[i] = round(b[i]).

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applyRound(b)

globals/nx/applySign

nx.applySign(b: NxBuffer) -> boolean

In-place b[i] = sign(b[i]) (returns -1, 0, or +1).

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applySign(b)

globals/nx/applySin

nx.applySin(b: NxBuffer) -> boolean

In-place b[i] = sin(b[i]).

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applySin(b)

globals/nx/applySqrt

nx.applySqrt(b: NxBuffer) -> boolean

In-place b[i] = sqrt(b[i]).

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applySqrt(b)

globals/nx/applySquare

nx.applySquare(b: NxBuffer) -> boolean

In-place b[i] = b[i] * b[i].

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applySquare(b)

globals/nx/applyTan

nx.applyTan(b: NxBuffer) -> boolean

In-place b[i] = tan(b[i]).

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applyTan(b)

globals/nx/applyTrunc

nx.applyTrunc(b: NxBuffer) -> boolean

In-place b[i] = trunc(b[i]).

Parameters

b NxBuffer — Target buffer.

Returns boolean — true on success, false on unknown handle.

nx.applyTrunc(b)

globals/nx/applyWindow

nx.applyWindow(signal: NxBuffer, window: NxBuffer) -> boolean

Element-wise signal[i] *= window[i] in place. Operates over the shorter of the two — passing a longer window to window a shorter clip is intentional, not an error.

Parameters

signal NxBuffer — Signal buffer (mutated).
window NxBuffer — Window buffer.

Returns boolean — true on success, false on unknown handle.

nx.applyWindow(signal, hann)

globals/nx/axpby

nx.axpby(dst: NxBuffer, a: number, src: NxBuffer, b: number) -> boolean

BLAS axpby: dst[i] = a*dst[i] + b*src[i].

Parameters

dst NxBuffer — Target buffer (mutated).
a number — Scale applied to dst.
src NxBuffer — Source buffer.
b number — Scale applied to src.

Returns boolean — true on success, false on stride mismatch / unknown handle.

nx.axpby(y, 0.5, x, 2.0)

globals/nx/clamp

nx.clamp(b: NxBuffer, min_v: number, max_v: number) -> boolean

In-place clamp: b[i] = clamp(b[i], min_v, max_v).

Parameters

b NxBuffer — Target buffer (mutated).
min_v number — Lower bound.
max_v number — Upper bound.

Returns boolean — true on success, false on unknown handle.

nx.clamp(b, 0.0, 1.0)

globals/nx/copy

nx.copy(dst: NxBuffer, src: NxBuffer) -> boolean

Copy every record from src into dst (memcpy fast path).

Parameters

dst NxBuffer — Target buffer (mutated).
src NxBuffer — Source buffer.

Returns boolean — true on success, false on shape mismatch or unknown handle.

nx.copy(dst, src)

globals/nx/copyStridedFrom

nx.copyStridedFrom(dst: NxBuffer, src: NxBuffer, scale: number?, dst_stride: number, dst_off: number?, src_stride: number, src_off: number?) -> boolean

Strided copy from src into dst with optional scaling: dst[k * dst_stride + dst_off] = scale * src[k * src_stride + src_off].

Parameters

dst NxBuffer — Target buffer (mutated).
src NxBuffer — Source buffer.
scale number (optional) — Optional src scale factor (default 1.0).
dst_stride number — Destination element stride.
dst_off number (optional) — Optional destination offset (default 0).
src_stride number — Source element stride.
src_off number (optional) — Optional source offset (default 0).

Returns boolean — true on success, false on shape mismatch / unknown handle.

nx.copyStridedFrom(dst, src, 1, 3, 0, 3, 0)

globals/nx/create

nx.create(type_: NxType, n: number) -> NxBuffer?

Allocate a CPU buffer of type × len records. Thin alias for substrate.createBuffer({type=type_, len=n, kind="cpu"}) — kept here so the public nx library is the canonical entry point and users never need to import buffer separately.

Parameters

type_ NxType — Element layout ("f32", "vec3", "vec4", "quat", "mat4").
n number — Record count.

Returns NxBuffer? — Freshly allocated buffer handle, or nil on failure.

local b = nx.create("vec3", 1024)

globals/nx/div

nx.div(dst: NxBuffer, x: NxScalarOrBuffer) -> boolean

b[i] /= x (x scalar) or b[i] /= x[i] (x buffer).

Parameters

dst NxBuffer — Target buffer (mutated).
x NxScalarOrBuffer — Either a scalar or a same-shaped buffer.

Returns boolean — true on success, false on type / handle errors.

nx.div(b, 2)
nx.div(dst, src)

globals/nx/dot

nx.dot(a: NxBuffer, b: NxBuffer) -> number?

Reduction: dot product of two same-shaped buffers.

Parameters

a NxBuffer — First buffer.
b NxBuffer — Second buffer.

Returns number? — Scalar dot product, or nil on shape mismatch / unknown handle.

local d = nx.dot(a, b)

globals/nx/fft1d

nx.fft1d(re: NxBuffer, im: NxBuffer, inverse: boolean?) -> boolean

In-place 1D FFT over parallel re / im CPU buffers. inverse=true runs the inverse transform scaled by 1/N (so ifft(fft(x)) ≈ x).

Parameters

re NxBuffer — Real-component buffer (mutated).
im NxBuffer — Imaginary-component buffer (mutated).
inverse boolean (optional) — When true runs the inverse transform.

Returns boolean — true on success, false on length mismatch / invalid handle. Raises an error naming the bytes when the transform's plan or working memory cannot be held.

nx.fft1d(re, im)
nx.fft1d(re, im, true)

globals/nx/fft2d

nx.fft2d(re: NxBuffer, im: NxBuffer, width: number, height: number, inverse: boolean?) -> boolean

In-place 2D FFT over row-major parallel re / im buffers of length width*height. inverse=true is scaled by 1 / (width * height).

Parameters

re NxBuffer — Real-component buffer (mutated).
im NxBuffer — Imaginary-component buffer (mutated).
width number — 2D width in samples.
height number — 2D height in samples.
inverse boolean (optional) — When true runs the inverse transform.

Returns boolean — true on success, false on length mismatch / invalid handle. Raises an error naming the bytes when the transform's plan or working memory cannot be held.

nx.fft2d(re, im, w, h)

globals/nx/fill

nx.fill(b: NxBuffer, value: number?) -> boolean

Fill the buffer with value (default 0.0). Equivalent to the scalar form of nx.add against a zeroed buffer, but skips the type-dispatch.

Parameters

b NxBuffer — Target buffer (mutated).
value number (optional) — Fill value (default 0.0).

Returns boolean — true on success, false on unknown handle.

nx.fill(b, 3.5)

globals/nx/fillRandomNormal

nx.fillRandomNormal(b: NxBuffer, mean: number?, stddev: number?, seed: NxSeed) -> boolean

Fill the buffer with Gaussian samples (Box-Muller), with the given mean and standard deviation, using a splitmix-keyed PRNG.

Parameters

b NxBuffer — Target buffer (mutated).
mean number (optional) — Optional mean (default 0.0).
stddev number (optional) — Optional standard deviation (default 1.0).
seed NxSeed — Optional seed — number, "frame", or nil (0).

Returns boolean — true on success, false on unknown handle.

nx.fillRandomNormal(b, 0, 1, 42)

globals/nx/fillRandomUniform

nx.fillRandomUniform(b: NxBuffer, min_v: number?, max_v: number?, seed: NxSeed) -> boolean

Fill the buffer with uniform-random samples in [min, max), using a splitmix-keyed deterministic PRNG.

Parameters

b NxBuffer — Target buffer (mutated).
min_v number (optional) — Optional lower bound (default 0.0).
max_v number (optional) — Optional upper bound (default 1.0).
seed NxSeed — Optional seed — number, "frame" (per-frame value), or nil (0).

Returns boolean — true on success, false on unknown handle.

nx.fillRandomUniform(b, -1, 1, "frame")

globals/nx/fillStrided

nx.fillStrided(b: NxBuffer, value: number, stride: number, offset: number?) -> boolean

Strided fill: buf[k * stride + offset] = value for every valid k.

Parameters

b NxBuffer — Target buffer (mutated).
value number — Per-element value.
stride number — Element stride.
offset number (optional) — Optional element offset (default 0).

Returns boolean — true on success, false on unknown handle.

nx.fillStrided(buf, 0, 3, 2)

globals/nx/fromTable

nx.fromTable(arr: { number }, type_: NxType?) -> NxBuffer?

Build a CPU buffer from a Lua table of numbers. The table is written through buf:write in a single FFI crossing — no per-element Lua loop. For large arrays, prefer one of the nx.* constructors + a kernel pass over building a Lua array first.

Parameters

arr { number } — Lua array of numbers.
type_ NxType (optional) — Optional element layout (default "f32").

Returns NxBuffer? — Newly allocated buffer holding arr, or nil on failure.

local b = nx.fromTable({ 0.1, 0.2, 0.3 })

globals/nx/full

nx.full(n: number, type_: NxType?, value: number?) -> NxBuffer?

Allocate a buffer of type × len records and fill with value.

Parameters

n number — Record count.
type_ NxType (optional) — Optional element layout (default "f32").
value number (optional) — Fill value (default 0.0).

Returns NxBuffer? — Buffer initialised to value.

local b = nx.full(1024, "f32", -1.0)

globals/nx/ifft1d

nx.ifft1d(re: NxBuffer, im: NxBuffer) -> boolean

Convenience: nx.fft1d(re, im, true).

Parameters

re NxBuffer — Real-component buffer (mutated).
im NxBuffer — Imaginary-component buffer (mutated).

Returns boolean — true on success, false on length mismatch / invalid handle. Raises an error naming the bytes when the transform's plan or working memory cannot be held.

nx.ifft1d(re, im)

globals/nx/ifft2d

nx.ifft2d(re: NxBuffer, im: NxBuffer, width: number, height: number) -> boolean

Convenience: nx.fft2d(re, im, w, h, true).

Parameters

re NxBuffer — Real-component buffer (mutated).
im NxBuffer — Imaginary-component buffer (mutated).
width number — 2D width.
height number — 2D height.

Returns boolean — true on success, false on length mismatch / invalid handle. Raises an error naming the bytes when the transform's plan or working memory cannot be held.

nx.ifft2d(re, im, w, h)

globals/nx/integratePosition

nx.integratePosition(pos: NxBuffer, vel: NxBuffer, dt: number) -> boolean

Per-vec3: pos[i] += vel[i] * dt — the position half of an Euler step. Both buffers must be vec3 (stride 3). The velocity step is the caller's: update vel BEFORE this call for semi-implicit Euler; updating it after gives forward Euler, which gains energy on stiff systems.

Parameters

pos NxBuffer — Position buffer (mutated).
vel NxBuffer — Velocity buffer.
dt number — Time step.

Returns boolean — true on success, false on shape mismatch / unknown handle.

nx.integratePosition(pos, vel, dt)

globals/nx/irfft1d

nx.irfft1d(re: NxBuffer, im: NxBuffer) -> NxBuffer?

Real-output inverse 1D FFT. Input re / im are length N/2 + 1. Returns a fresh CPU f32 buffer of length 2 * (N/2 + 1 - 1) = N real samples.

Parameters

re NxBuffer — Real-component buffer.
im NxBuffer — Imaginary-component buffer.

Returns NxBuffer? — Real-valued output buffer on success, nil on failure. Raises an error naming the bytes when the transform's plan or working memory cannot be held. A call that answers nil or raises leaves none of the buffers it created allocated.

local out = nx.irfft1d(re, im)

globals/nx/irfft2d

nx.irfft2d(re: NxBuffer, im: NxBuffer, width: number, height: number) -> NxBuffer?

Real-output inverse 2D FFT. Input re / im are (width/2 + 1) * height row-major. Returns a fresh f32 buffer of length width * height.

Parameters

re NxBuffer — Real-component buffer.
im NxBuffer — Imaginary-component buffer.
width number — 2D width.
height number — 2D height.

local out = nx.irfft2d(re, im, w, h)

globals/nx/lerpTo

nx.lerpTo(dst: NxBuffer, src: NxBuffer, t: number) -> boolean

dst[i] += t * (src[i] - dst[i]) — element-wise lerp toward src by t.

Parameters

dst NxBuffer — Target buffer (mutated).
src NxBuffer — Source buffer.
t number — Interpolation factor.

Returns boolean — true on success, false on stride mismatch / unknown handle.

nx.lerpTo(current, target, 0.1)

globals/nx/max

nx.max(b: NxBuffer) -> number?

Reduction: maximum of all elements.

Parameters

b NxBuffer — Source buffer.

Returns number? — Scalar max, or nil on unknown handle.

local m = nx.max(b)

globals/nx/maxOp

nx.maxOp(dst: NxBuffer, x: NxScalarOrBuffer) -> boolean

Element-wise b[i] = max(b[i], x) (scalar) or b[i] = max(b[i], x[i]) (buffer).

Parameters

dst NxBuffer — Target buffer (mutated).
x NxScalarOrBuffer — Scalar or same-shaped buffer.

Returns boolean — true on success, false on type / handle errors.

nx.maxOp(b, 0.0)

globals/nx/mean

nx.mean(b: NxBuffer) -> number?

Reduction: arithmetic mean of all elements.

Parameters

b NxBuffer — Source buffer.

Returns number? — Scalar mean, or nil on unknown handle.

local m = nx.mean(b)

globals/nx/min

nx.min(b: NxBuffer) -> number?

Reduction: minimum of all elements.

Parameters

b NxBuffer — Source buffer.

Returns number? — Scalar min, or nil on unknown handle.

local m = nx.min(b)

globals/nx/minOp

nx.minOp(dst: NxBuffer, x: NxScalarOrBuffer) -> boolean

Element-wise b[i] = min(b[i], x) (scalar) or b[i] = min(b[i], x[i]) (buffer).

Parameters

dst NxBuffer — Target buffer (mutated).
x NxScalarOrBuffer — Scalar or same-shaped buffer.

Returns boolean — true on success, false on type / handle errors.

nx.minOp(b, 1.0)

globals/nx/mul

nx.mul(dst: NxBuffer, x: NxScalarOrBuffer) -> boolean

b[i] *= x (x scalar) or b[i] *= x[i] (x buffer).

Parameters

dst NxBuffer — Target buffer (mutated).
x NxScalarOrBuffer — Either a scalar or a same-shaped buffer.

Returns boolean — true on success, false on type / handle errors.

nx.mul(b, 2)
nx.mul(dst, src)

globals/nx/mulStrided

nx.mulStrided(b: NxBuffer, scalar: number, stride: number, offset: number?) -> boolean

Strided multiply: buf[k * stride + offset] *= scalar for every valid k.

Parameters

b NxBuffer — Target buffer (mutated).
scalar number — Per-element multiplier.
stride number — Element stride.
offset number (optional) — Optional element offset (default 0).

Returns boolean — true on success, false on unknown handle.

nx.mulStrided(buf, 2, 4, 0)

globals/nx/normL1

nx.normL1(b: NxBuffer) -> number?

Reduction: L1 norm — sum(|b[i]|).

Parameters

b NxBuffer — Source buffer.

Returns number? — Scalar L1 norm, or nil on unknown handle.

local n = nx.normL1(b)

globals/nx/normL2

nx.normL2(b: NxBuffer) -> number?

Reduction: L2 norm — sqrt(sum(b[i]^2)).

Parameters

b NxBuffer — Source buffer.

Returns number? — Scalar L2 norm, or nil on unknown handle.

local n = nx.normL2(b)

globals/nx/normalizeVec3

nx.normalizeVec3(b: NxBuffer) -> boolean

Normalise each vec3 in-place. Vectors below 1e-8 are left untouched.

Parameters

b NxBuffer — Vec3 buffer (mutated).

Returns boolean — true on success, false on stride mismatch / unknown handle.

nx.normalizeVec3(directions)

globals/nx/ones

nx.ones(n: number, type_: NxType?) -> NxBuffer?

Allocate a buffer of type × len records and fill with 1.0.

Parameters

n number — Record count.
type_ NxType (optional) — Optional element layout (default "f32").

Returns NxBuffer? — Buffer initialised to one.

local b = nx.ones(1024)

globals/nx/pow

nx.pow(b: NxBuffer, p: number) -> boolean

In-place b[i] = b[i] ^ p (scalar exponent).

Parameters

b NxBuffer — Target buffer (mutated).
p number — Scalar exponent.

Returns boolean — true on success, false on unknown handle.

nx.pow(b, 2.2)

globals/nx/quatFromYaw

nx.quatFromYaw(dst: NxBuffer, yaw: NxBuffer) -> boolean

Per-quat: dst[i] = (0, sin(yaw[i]/2), 0, cos(yaw[i]/2)) — the pure-Y axis-angle quaternion for each yaw value. dst must be stride-4 (quat); yaw must be stride-1 (f32).

Parameters

dst NxBuffer — Quaternion buffer (mutated).
yaw NxBuffer — Source yaw scalars buffer.

Returns boolean — true on success, false on shape mismatch / unknown handle.

nx.quatFromYaw(quats, yaws)

globals/nx/rfft1d

nx.rfft1d(signal: NxBuffer?) -> (NxBuffer?, NxBuffer?)

Real-input forward 1D FFT. Allocates two new CPU f32 buffers of length N/2 + 1 holding the (re, im) parts of the Hermitian-symmetric spectrum (same convention as NumPy np.fft.rfft).

Parameters

signal NxBuffer (optional) — Real-valued input buffer.

Returns (NxBuffer?, NxBuffer?) — (re_buf, im_buf) on success, nil otherwise. Raises an error naming the bytes when the transform's plan or working memory cannot be held. A call that answers nil or raises leaves none of the buffers it created allocated.

local re, im = nx.rfft1d(signal)

globals/nx/rfft2d

nx.rfft2d(signal: NxBuffer, width: number, height: number) -> (NxBuffer?, NxBuffer?)

Real-input forward 2D FFT. Input signal is row-major width*height. Returns (re_buf, im_buf) of length (width/2 + 1) * height each (matches np.fft.rfft2 layout).

Parameters

signal NxBuffer — Real-valued input buffer (row-major).
width number — 2D width.
height number — 2D height.

Returns (NxBuffer?, NxBuffer?) — (re_buf, im_buf) on success, nil on failure. Raises an error naming the bytes when the transform's plan or working memory cannot be held. A call that answers nil or raises leaves none of the buffers it created allocated.

local re, im = nx.rfft2d(image, w, h)

globals/nx/scale

nx.scale(b: NxBuffer, s: number) -> boolean

In-place b[i] *= s.

Parameters

b NxBuffer — Target buffer (mutated).
s number — Scalar multiplier.

Returns boolean — true on success, false on unknown handle.

nx.scale(b, 0.5)

globals/nx/sinCosTo

nx.sinCosTo(src: NxBuffer, sin_dst: NxBuffer, cos_dst: NxBuffer) -> boolean

Compute sin_dst[i] = sin(src[i]) and cos_dst[i] = cos(src[i]) in one pass using cheaper paired-trig argument reduction.

Parameters

src NxBuffer — Source angles buffer.
sin_dst NxBuffer — Destination buffer for the sine values.
cos_dst NxBuffer — Destination buffer for the cosine values.

Returns boolean — true on success, false on stride mismatch / unknown handle.

nx.sinCosTo(angles, s, c)

globals/nx/sub

nx.sub(dst: NxBuffer, x: NxScalarOrBuffer) -> boolean

b[i] -= x (x scalar) or b[i] -= x[i] (x buffer).

Parameters

dst NxBuffer — Target buffer (mutated).
x NxScalarOrBuffer — Either a scalar or a same-shaped buffer.

Returns boolean — true on success, false on type / handle errors.

nx.sub(b, 0.5)
nx.sub(dst, src)

globals/nx/sum

nx.sum(b: NxBuffer) -> number?

Reduction: sum of all elements.

Parameters

b NxBuffer — Source buffer.

Returns number? — Scalar sum, or nil on unknown handle.

local s = nx.sum(b)

globals/nx/wanderYaw

nx.wanderYaw(args: NxWanderArgs) -> boolean

Fused per-entity wander step. Each entity's yaw[i] walks by a uniform random delta in [-yawDelta, +yawDelta], then pos[i] advances forward in the (sin yaw, cos yaw) direction by step. Optional rot quat output writes a pure-Y axis-angle rotation. Replaces the per-entity Luau loop pattern (~12 ms / 5000 in interpreter) with a single Rust pass (~0.2 ms / 5000). Buffer requirements: pos vec3 (stride 3), yaw f32 (stride 1), rot optional quat (stride 4). All counts should match (kernel walks min(count_i)).

Parameters

args NxWanderArgs — Table with { pos, yaw, rot?, step?, yawDelta?, seed }.

Returns boolean — true on success, false on stride / shape failures.

nx.wanderYaw({ pos = pos, yaw = yaw, step = 0.5, yawDelta = 0.2, seed = "frame" })

globals/nx/window/blackman

nx.window.blackman(n: number) -> NxBuffer?

Allocate a fresh CPU f32 buffer of length n filled with a symmetric Blackman window (NumPy np.blackman convention).

Parameters

n number — Number of samples in the window.

Returns NxBuffer? — Buffer holding the window samples, or nil on failure.

local w = nx.window.blackman(1024)

globals/nx/window/hamming

nx.window.hamming(n: number) -> NxBuffer?

Allocate a fresh CPU f32 buffer of length n filled with a symmetric Hamming window (NumPy np.hamming convention).

Parameters

n number — Number of samples in the window.

Returns NxBuffer? — Buffer holding the window samples, or nil on failure.

local w = nx.window.hamming(1024)

globals/nx/window/hann

nx.window.hann(n: number) -> NxBuffer?

Allocate a fresh CPU f32 buffer of length n filled with a symmetric Hann window (NumPy np.hanning convention).

Parameters

n number — Number of samples in the window.

Returns NxBuffer? — Buffer holding the window samples, or nil on failure.

local w = nx.window.hann(1024)

globals/nx/zeros

nx.zeros(n: number, type_: NxType?) -> NxBuffer?

Allocate a buffer of type × len records and fill with 0.0.

Parameters

n number — Record count.
type_ NxType (optional) — Optional element layout (default "f32").

Returns NxBuffer? — Buffer initialised to zero.

local b = nx.zeros(2048)

globals/packages/list

packages.list() -> { PackageEntry }

List every registered package across scopes.

Returns { PackageEntry } — Array of package entries.

for _, p in ipairs(packages.list()) do print(p.name, p.scope) end

globals/packages/lookup

packages.lookup(name_or_scope: string, name: string?) -> PackageEntry?

Look up a single package by name (any scope) or by exact (scope, name).

Parameters

name_or_scope string — Package name, or scope if a second arg is given.
name string (optional) — Package name when the first arg is a scope.

Returns PackageEntry? — Package entry or nil.

local p = packages.lookup("@builtin", "audio")

globals/pairs

pairs(table) -> iterator

Iterate all key-value pairs.

globals/pcall

pcall(fn, ...) -> ok, result

Protected call. Returns false + error on failure.

globals/physics

physics: any

Lowercase alias of Physics. Same identity (#565). Prefer this in new code so the convention matches entity, scene, cam, ...

globals/physics/COLLIDER_COMPONENTS

physics.COLLIDER_COMPONENTS()

Every collider component, in the order a lookup walks them. Shape is the component's identity, so code that works on "whatever collider this entity has" walks this list rather than guessing a shape.

for _, name in ipairs(Physics.COLLIDER_COMPONENTS) do ... end

globals/physics/addCollider

physics.addCollider(entityId: string | entityRef, component: string, config: table?)

Add a collider component to an entity, naming the shape you want.

Parameters

entityId string | entityRef — Target entity id.
component string — One of Physics.COLLIDER_COMPONENTS.
config table (optional) — The component's own fields, e.g. { radius = 0.5 } for a sphere.

Physics.addCollider(id, "SphereCollider", { radius = 0.5 })

globals/physics/addConstraint

physics.addConstraint(entityId: string | entityRef, opts: table?)

Add a transform constraint to an entity.

Parameters

entityId string | entityRef — Target entity id.
opts table (optional) — Optional constraint description (targetEntityId, position, rotation, scale, lookAt, targetPosition, axes, weight).

Physics.addConstraint(id, { targetEntityId = parent, position = true })

globals/physics/addJoint

physics.addJoint(entityIdA: string | entityRef, entityIdB: string | entityRef, opts: table?)

Add a Joint component connecting two entities. Accepts either vec3-style anchor inputs (localAnchor = {x,y,z}) or pre-split scalar keys (localAnchorX/Y/Z).

Parameters

entityIdA string | entityRef — Entity that hosts the Joint component.
entityIdB string | entityRef — Connected entity.
opts table (optional) — Optional joint description (kind, anchors, axis, stiffness, damping, restLength, maxDistance, breakForce, breakTorque, region, remoteRegion). Either end may be a soft body (an entity with a SoftBody component): region anchors this end on a region of entity A's soft body and remoteRegion the other end on a region of B's, each as Physics.softRegion takes it ({ box = { min, max } } or { nearest = point }); an end with no region is anchored on the soft body as a whole. The anchor of a region end is measured from the centre of its particles.

Physics.addJoint(a, b, { kind = "fixed" })
Physics.addJoint(a, b, { kind = "hinge", axis = {x=0,y=1,z=0} })
Physics.addJoint(a, b, { kind = "rope", maxDistance = 8 })
Physics.addJoint(a, b, { kind = "fixed", breakForce = 1200, breakTorque = 800 })
Physics.addJoint(bag, hook, { kind = "ball", region = { nearest = {0, 0.5, 0} } })
Physics.addJoint(jellyA, jellyB, { kind = "fixed", region = { box = { min = {0.4, -0.5, -0.5}, max = {0.5, 0.5, 0.5} } }, remoteRegion = { box = { min = {-0.5, -0.5, -0.5}, max = {-0.4, 0.5, 0.5} } } })

globals/physics/addVelocity

physics.addVelocity(a: string | entityRef | number | vec3, b: (number | vec3)?, c: number?, d: number?)

Add to the linear velocity of an entity. Same call shapes as setVelocity.

Parameters

a string | entityRef | number | vec3 — dx, a {x, y, z} delta vector, or an entity id (explicit target).
b (number | vec3) (optional) — dy, dx, or the delta vector depending on call form.
c number (optional) — dz or dy depending on call form.
d number (optional) — Optional dz when targeting an explicit entity.

Physics.addVelocity(0, 5, 0)
Physics.addVelocity(entityId, 0, 5, 0)
Physics.addVelocity(entityId, {x=0, y=5, z=0})

globals/physics/addWheelCollider

physics.addWheelCollider(entityId: string | entityRef, config: table?)

Parameters

entityId string | entityRef — Target entity id.
config table (optional) — Optional wheel configuration (radius?, suspensionDistance?, springRate?, damperRate?, targetPosition?, maxSuspensionForce?, motorTorque?, brakeTorque?, steerAngle?, forwardFriction?, sidewaysFriction?, frictionSlip?, sideFrictionStiffness?, rollingResistance?, width?, is2D?, visual?).

Physics.addWheelCollider(id, { radius = 0.35, motorTorque = 500 })

globals/physics/applyForce

physics.applyForce(entityIdOrForce: string | entityRef | vec3, force: vec3?, region: PhysicsSoftRegion?)

Parameters

entityIdOrForce string | entityRef | vec3 — Entity id (when paired with force) OR a force vector for the script-context entity.
force vec3 (optional) — Optional force vector when targeting an explicit entity.
region PhysicsSoftRegion (optional) — Part of a soft body to push, as Physics.softRegion takes it; omit for the whole body.

Physics.applyForce({x=0, y=10, z=0})
Physics.applyForce(entityId, {x=0, y=10, z=0})
Physics.applyForce(cloth, {x=0, y=0, z=4}, { box = { min = {-0.5, -1, 0.4}, max = {0.5, 1, 0.5} } })

globals/physics/applyForceAtPoint

physics.applyForceAtPoint(entityId: string | entityRef, force: vec3, point: vec3, opts: PhysicsPushOptions?)

Parameters

entityId string | entityRef — Target entity id.
force vec3 — Force vector, in newtons.
point vec3 — World-space application point.
opts PhysicsPushOptions (optional) — { radius: number? }: how far into a soft body the force reaches, in metres.

Physics.applyForceAtPoint(id, {x=0,y=10,z=0}, {x=1,y=0,z=0})
Physics.applyForceAtPoint(jelly, {x=0,y=-40,z=0}, {x=0,y=1.5,z=0}, { radius = 0.3 })

globals/physics/applyImpulse

physics.applyImpulse(entityIdOrImpulse: string | entityRef | vec3, impulse: vec3?, region: PhysicsSoftRegion?)

Parameters

entityIdOrImpulse string | entityRef | vec3 — Entity id (with impulse) OR an impulse vector for the script-context entity.
impulse vec3 (optional) — Optional impulse vector when targeting an explicit entity.
region PhysicsSoftRegion (optional) — Part of a soft body to push, as Physics.softRegion takes it; omit for the whole body.

Physics.applyImpulse({x=0, y=5, z=0})
Physics.applyImpulse(entityId, {x=0, y=5, z=0})
Physics.applyImpulse(jelly, {x=0, y=-2, z=0}, { box = { min = {-0.5, 0.4, -0.5}, max = {0.5, 0.5, 0.5} } })

globals/physics/applyImpulseAtPoint

physics.applyImpulseAtPoint(entityId: string | entityRef, impulse: vec3, point: vec3, opts: PhysicsPushOptions?)

Parameters

entityId string | entityRef — The body's entity id or proxy.
impulse vec3 — The impulse, in newton-seconds.
point vec3 — Where it lands, in world space: a contact point.
opts PhysicsPushOptions (optional) — { radius: number? }: how far into a soft body the impulse reaches, in metres.

Physics.applyImpulseAtPoint(crate, {x=0, y=0, z=-30}, hit.point)
local contact = Physics.contacts(shell)[1]; if contact then Physics.applyImpulseAtPoint(shell, {x=-60, y=0, z=0}, contact.points[1].point, { radius = 0.4 }) end

globals/physics/applyRadialImpulse

physics.applyRadialImpulse(entityId: string | entityRef, center: vec3, magnitude: number, opts: PhysicsPushOptions?)

Parameters

entityId string | entityRef — The body's entity id or proxy.
center vec3 — The blast's world position.
magnitude number — The impulse at the centre, in newton-seconds.
opts PhysicsPushOptions (optional) — { radius: number? }: how far the push reaches, in metres.

Physics.applyRadialImpulse(balloon, {x=0, y=1, z=-1}, 8, { radius = 3 })
for _, id in pile do Physics.applyRadialImpulse(id, blast, 20, { radius = 4 }) end

globals/physics/applyTorque

physics.applyTorque(entityIdOrTorque: string | entityRef | vec3, torque: vec3?)

Parameters

entityIdOrTorque string | entityRef | vec3 — Entity id (with torque) OR a torque vector for the script-context entity.
torque vec3 (optional) — Optional torque vector when targeting an explicit entity.

Physics.applyTorque({x=0, y=1, z=0})
Physics.applyTorque(entityId, {x=0, y=1, z=0})

globals/physics/attachSoftRegion

physics.attachSoftRegion(softId: string | entityRef, targetId: string | entityRef, region: PhysicsSoftRegion?) -> number

Parameters

softId string | entityRef — The soft body's entity (it carries a SoftBody component).
targetId string | entityRef — The rigid body's entity (it carries Physics), by id, name or proxy.
region PhysicsSoftRegion (optional) — Which particles to hold, as Physics.softRegion takes it: { box = { min, max } }, { nearest = point }, with space = "world" for world coordinates; nil holds the whole body.

Returns number — The attachment's index in the component's attachments.

Physics.attachSoftRegion(flag, pole, { box = { min = {-0.5, -1, -1}, max = {-0.45, 1, 1} } })
Physics.attachSoftRegion(shell, chassis, { box = { min = {-0.5, -0.5, -0.5}, max = {0.5, -0.4, 0.5} } })

globals/physics/bodyState

physics.bodyState(entityId: string | entityRef) -> PhysicsBodyState?

Parameters

entityId string | entityRef — Entity id or proxy.

local b = Physics.bodyState(id); print(b.bodyType, b.mass, b.stillness)
local b = Physics.bodyState(id); if not b.exists then print(b.waitingFor or "no body was built") end

globals/physics/boxCast

physics.boxCast(origin: vec3, halfExtents: vec3, direction: vec3, maxDistance: number?, exclude: (string | {string})?) -> table?

Cast a box along a direction and return the first hit. A trigger (isTrigger) holds nothing up, so the sweep passes through it.

Parameters

origin vec3 — Box center at the start of the cast.
halfExtents vec3 — Half the size of the box on each axis.
direction vec3 — Cast direction, of any length: the engine normalises it, so maxDistance and distance are world units.
maxDistance number (optional) — Optional distance limit.
exclude (string | {string}) (optional) — Optional entity id, or array of entity ids, to exclude from hits. Applied while sweeping rather than to the answer, so a cast that starts inside an excluded collider reports what is behind it.

local hit = Physics.boxCast(o, {x=0.5,y=0.5,z=0.5}, dir, 10)
local hit = Physics.boxCast(o, {x=0.5,y=0.5,z=0.5}, dir, 10, { selfId, carriedId })

globals/physics/capsuleCast

physics.capsuleCast(origin: vec3, radius: number, halfHeight: number, direction: vec3, maxDistance: number?, exclude: (string | {string})?) -> table?

Parameters

origin vec3 — Capsule centre at the start of the cast.
radius number — Capsule radius.
halfHeight number — Distance from the centre to either cap centre. The capsule stands halfHeight + radius tall in each direction.
direction vec3 — Cast direction, of any length: the engine normalises it, so maxDistance and distance are world units.
maxDistance number (optional) — Optional distance limit.
exclude (string | {string}) (optional) — Optional entity id, or array of entity ids, to exclude from hits. Applied while sweeping rather than to the answer, so a cast that starts inside an excluded collider reports what is behind it.

local hit = Physics.capsuleCast(o, 0.3, 0.6, dir, 0.5)
local hit = Physics.capsuleCast(o, 0.3, 0.6, dir, 0.5, selfId)

globals/physics/colliderCount

physics.colliderCount() -> number

Returns number — colliders across the whole physics world.

if Physics.colliderCount() == 0 then print("nothing here is solid") end

globals/physics/colliderGeometry

physics.colliderGeometry(options: table?) -> table?

Read the physics world as drawable triangles: every collider triangulated in world space into one indexed mesh, in GPU buffers ready to draw.

The returned buffers are yours: destroy them when you replace them.

Parameters

options table (optional) — { tessellation = "low"|"medium"|"high", colors = { {r,g,b,a}, ... }, defaultColor = {r,g,b,a} }.

Returns table? — { vertices, indices, vertexCount, indexCount, colliders } where each entry of colliders is { entity, colliderName?, shapeType, role, exact, firstIndex, indexCount }.

local geo = Physics.colliderGeometry({ tessellation = "high" })

globals/physics/colliderManifest

physics.colliderManifest() -> table

List every physics collider in the world with what it is and what it takes part in — no geometry, so it is the cheap read to make before deciding what to do with each one.

Returns table — Array of { entity, colliderName?, shapeType, role, exact }.

for _, c in ipairs(Physics.colliderManifest()) do print(c.entity, c.role) end

globals/physics/colliderOn

physics.colliderOn(entityId: string | entityRef) -> string?

Which collider component an entity carries, or nil when it carries none.

Parameters

entityId string | entityRef — Target entity id.

Returns string? The component name, e.g. "SphereCollider".

local which = Physics.colliderOn(id)

globals/physics/colliderShapes

physics.colliderShapes(entityId: string | entityRef) -> table

Read an entity's resolved physics collider shape(s) as the physics engine sees them, including auto-sized colliders.

shapeType is one of box, sphere, capsule, convex, mesh, heightfield, compound, other — the shape the simulation is running, so a mesh collider reads mesh.

Parameters

entityId string | entityRef — Target entity id.

Returns table — Array of resolved collider shapes (empty if none): { shapeType, position, rotation, params, linePoints, name? }.

local shapes = Physics.colliderShapes(id)

globals/physics/contacts

physics.contacts(entityId: string | entityRef, otherId: (string | entityRef)?) -> { PhysicsContact }

Parameters

entityId string | entityRef — Entity id or proxy.
otherId (string | entityRef) (optional) — Optional entity id or proxy: report only the contacts with it.

Returns { PhysicsContact } — An array of PhysicsContact — empty when the body touches nothing, or when the entity carries neither a rigid body nor a soft body.

for _, c in Physics.contacts(id) do print(c.other, c.deepestPenetration) end
for _, c in Physics.contacts(truck, pole) do if c.totalImpulse > 5 then print("dented at", #c.particles, "particles") end end

globals/physics/cutSoftBody

physics.cutSoftBody(softId: string | entityRef, blade: { vec3 } | { point: vec3, normal: vec3 })

Parameters

softId string | entityRef — The soft body's entity.
blade { vec3 } | { point: vec3, normal: vec3 } — { point = vec3, normal = vec3 } for a plane, or the triangle's three world-space corners { a, b, c }.

Physics.cutSoftBody(jelly, { point = vector.create(0, 1, 0), normal = vector.create(1, 0, 0) })   -- in two along x = 0
Physics.cutSoftBody(jelly, { vector.create(0, -5, -5), vector.create(0, 5, -5), vector.create(0, 0, 10) })
-- slice off the part of a body 0.2 m above its centre
local c = Physics.softBodyState(jelly).centerOfMass
Physics.cutSoftBody(jelly, { point = vector.create(c.x, c.y + 0.2, c.z), normal = vector.create(0, 1, 0) })

globals/physics/debugLines

physics.debugLines(opts: table?) -> (PhysicsDebugLines?, string?)

Parameters

opts table (optional) — { layers: {string}?, entities: {(string | entityRef)}? }: layers names the layers to draw (every layer when omitted; an unknown name is an error naming the layers), entities keeps to those entities' objects (a wheel is kept by its own entity or its chassis').

Returns (PhysicsDebugLines?, string?) — A PhysicsDebugLines, or nil and the reason on a VM that holds no physics world (a worker).

local d = Physics.debugLines(); print(d.lineCount, #d.objects)
for _, o in Physics.debugLines({ layers = { "particles" } }).objects do print(o.entity, o.pinned, o.attached) end
local w = Physics.debugLines({ layers = { "wheels" }, entities = { truck } }); for _, o in w.objects do print(o.index, o.grounded, o.compression) end

globals/physics/detachSoftRegion

physics.detachSoftRegion(softId: string | entityRef, targetId: (string | entityRef)?) -> number

Release regions of a soft body held with attachSoftRegion (or authored in its SoftBody component's attachments).

Parameters

softId string | entityRef — The soft body's entity.
targetId (string | entityRef) (optional) — Release only the regions held to this rigid body; omit to release every one.

Returns number — How many attachments were released.

Physics.detachSoftRegion(flag, pole)
Physics.detachSoftRegion(shell)

globals/physics/drawnPose

physics.drawnPose(entityId: string | entityRef) -> PhysicsDrawnPose?

Parameters

entityId string | entityRef — Entity id or proxy. Any entity answers; one the simulation does not blend is drawn at its simulated pose.

Returns PhysicsDrawnPose? — A PhysicsDrawnPose, or nil when nothing in the scene answers to that id.

local d = Physics.drawnPose(car); print(d.position.z, d.simulatedPosition.z, d.interpolated)
local d = Physics.drawnPose(tyre); print(d.interpolated, d.interpolatedBy)

globals/physics/getAngularVelocity

physics.getAngularVelocity(entityId: (string | entityRef)?) -> vec3?

Read the angular velocity of an entity's rigid body.

Parameters

entityId (string | entityRef) (optional) — Target entity id or proxy; resolves from script context when omitted.

Returns vec3? — Angular velocity in rad/s, or nil if the entity has no rigid body.

local w = Physics.getAngularVelocity(id)

globals/physics/getGravity

physics.getGravity() -> vec3

Read the current world gravity vector.

Returns vec3 — Gravity vector in m/s² (negative y is "down" in the default world).

local g = Physics.getGravity()

globals/physics/getVelocity

physics.getVelocity(entityId: (string | entityRef)?) -> vec3?

Read the linear velocity of an entity's rigid body, or of its soft body as a whole (its particles' velocities averaged by mass).

Parameters

entityId (string | entityRef) (optional) — Target entity id or proxy; resolves from script context when omitted.

Returns vec3? — Velocity in m/s, or nil if the entity has no rigid body.

local v = Physics.getVelocity(id)

globals/physics/getWheelState

physics.getWheelState(entityId: string | entityRef) -> table?

Read a wheel collider's runtime state. Reads the native component the wheel system writes after each physics step.

Parameters

entityId string | entityRef — Target entity id (must carry a WheelCollider component).

local state = Physics.getWheelState(id)

globals/physics/hasLineOfSight

physics.hasLineOfSight(fromId: string, toId: string) -> boolean

Check whether two entities have line-of-sight between their origins.

Parameters

fromId string — Viewer entity id.
toId string — Target entity id.

Returns boolean — true when no collider sits between them (including coincident origins), false otherwise.

if Physics.hasLineOfSight(a, b) then ... end

globals/physics/ignoreCollision

physics.ignoreCollision(entityIdA: string | entityRef, entityIdB: string | entityRef, ignore: boolean?)

Parameters

entityIdA string | entityRef — First entity id.
entityIdB string | entityRef — Second entity id.
ignore boolean (optional) — When true (default) the pair is kept apart; false lets it collide again.

Physics.ignoreCollision(a, b, true)

globals/physics/isSleeping

physics.isSleeping(entityId: (string | entityRef)?) -> boolean?

Whether an entity's rigid or soft body is currently asleep (at rest and not simulating). A body sleeps once it stops moving, to save simulation cost.

Parameters

entityId (string | entityRef) (optional) — Target entity id or proxy; resolves from script context when omitted.

Returns boolean? — true if asleep, false if awake, or nil if the entity has neither a rigid nor a soft body.

if Physics.isSleeping(id) then Physics.wakeUp(id) end

globals/physics/jointBreaks

physics.jointBreaks() -> table

Returns table — Array of { entityId, connectedEntityId, kind, impulse, angularImpulse, force, torque, position }, oldest first.

for _, e in ipairs(Physics.jointBreaks()) do print(e.entityId, e.force) end

globals/physics/jointReaction

physics.jointReaction(entityId: string | entityRef) -> table?

Parameters

entityId string | entityRef — Entity carrying the Joint component.

Returns table? — { impulse, angularImpulse, force, torque, position }, or nil when the entity owns no joint.

local r = Physics.jointReaction(id); print(r and r.force)

globals/physics/observe

physics.observe(entityId: (string | entityRef)?, opts: table?) -> PhysicsObservation?

Parameters

entityId (string | entityRef) (optional) — Report on this one entity. Omit for every body in the world.
opts table (optional) — { bodies: boolean?, contactPoints: boolean? } — bodies = false builds the world accounting alone, and contactPoints = false keeps each contact pair's normal, depth, impulse and point count while leaving out the individual points. Both default to true.

local o = Physics.observe(); for _, b in o.bodies do print(b.entity, b.stillness) end
local o = Physics.observe(id); print(o.bodies[1].stillness, o.bodies[1].stillnessDetail)

globals/physics/onJointBreak

physics.onJointBreak(fn: (table) -> ()) -> () -> ()

Call fn for every joint that breaks from now on, with the same record jointBreaks returns.

Parameters

fn (table) -> () — Receives one break record per broken joint.

Returns () -> () — A function that removes this listener.

local off = Physics.onJointBreak(function(e) print(e.kind, e.force, e.position) end)

globals/physics/onSoftBodyTear

physics.onSoftBodyTear(fn: (PhysicsSoftBodyTear) -> ()) -> () -> ()

Parameters

fn (PhysicsSoftBodyTear) -> () — Receives one PhysicsSoftBodyTear per tear: { entity, pieces, position, kind, particles }.

Returns () -> () — A function that removes this listener.

local off = Physics.onSoftBodyTear(function(e) print(e.entity, #e.pieces, e.position) end)
Physics.onSoftBodyTear(function(e) for _, id in e.pieces do Physics.applyImpulse(id, { x = 0, y = 2, z = 0 }) end end)

globals/physics/overlapSphere

physics.overlapSphere(center: vec3, radius: number) -> table

Find every entity id whose colliders overlap a sphere. A trigger (isTrigger) holds nothing up, so it is left out.

Parameters

center vec3 — Sphere center in world space.
radius number — Sphere radius.

Returns table — Array of overlapping entity ids.

local ids = Physics.overlapSphere({x=0,y=0,z=0}, 5)

globals/physics/pinSoftRegion

physics.pinSoftRegion(softId: string | entityRef, region: PhysicsSoftRegion?, opts: { name: string?, target: vec3? }?) -> string

Parameters

softId string | entityRef — The soft body's entity (it carries a SoftBody component).
region PhysicsSoftRegion (optional) — Which particles to pin, as Physics.softRegion takes it: { box = { min, max } } or { nearest = point } in the entity's unit box (-0.5 to 0.5 on each axis), space = "world" for world coordinates; nil pins the whole body.
opts { name: string?, target: vec3? } (optional) — { name: string?, target: vec3? }: the name the pin is addressed by (by default pin1, pin2, ...), and a world position to move the pinned particles' centre to.

Returns string — The pin's name.

local corner = Physics.pinSoftRegion(cloth, { nearest = {-0.5, 0, 0.5} })
Physics.pinSoftRegion(cloth, { box = { min = {-0.5, -1, 0.45}, max = {0.5, 1, 0.5} } }, { name = "top" })
local grab = Physics.pinSoftRegion(jelly, { nearest = hit.point, space = "world" }, { target = hit.point })

globals/physics/pumpJointBreaks

physics.pumpJointBreaks()

Physics.pumpJointBreaks()

globals/physics/pumpSoftBodyTears

physics.pumpSoftBodyTears()

Physics.pumpSoftBodyTears()

globals/physics/raycast

physics.raycast(origin: vec3, direction: vec3, maxDistance: number?, exclude: (string | {string})?) -> table?

Parameters

origin vec3 — Ray origin in world space.
direction vec3 — Ray direction, of any length: the engine normalises it, so maxDistance and distance are world units.
maxDistance number (optional) — Maximum distance along the ray (defaults to 1000).
exclude (string | {string}) (optional) — Optional entity id, or array of entity ids, to exclude from hits.

local hit = Physics.raycast({x=0,y=2,z=0}, {x=0,y=-1,z=0})
local hit = Physics.raycast(origin, dir, 50, { selfId, carriedId })
local hit = Physics.raycast(muzzle, aim, 100); if hit and hit.softBody then Physics.applyImpulseAtPoint(hit.entityId, {x=0, y=0, z=-20}, hit.point, { radius = 0.3 }) end
if Physics.colliderCount() == 0 then hit = renderer.raycast(eye, down, 200) end

globals/physics/raycastAll

physics.raycastAll(origin: vec3, direction: vec3, maxDistance: number?, maxHits: number?, exclude: (string | {string})?) -> table

Parameters

origin vec3 — Ray origin in world space.
direction vec3 — Ray direction, of any length: the engine normalises it, so maxDistance and distance are world units.
maxDistance number (optional) — Optional distance limit along the ray.
maxHits number (optional) — Optional cap on the number of hits returned.
exclude (string | {string}) (optional) — Optional entity id, or array of entity ids, to exclude from hits.

local hits = Physics.raycastAll(origin, dir, 50, 4)

globals/physics/raycastBetween

physics.raycastBetween(fromId: string, toId: string, maxDistance: number?) -> table?

Cast a ray from one entity toward another and return the first hit.

Parameters

fromId string — Origin entity id.
toId string — Target entity id.
maxDistance number (optional) — Optional distance cap (default 1000).

Returns table? — Hit table, or nil if the entities are coincident or nothing was hit.

local hit = Physics.raycastBetween(a, b)

globals/physics/raycastScreen

physics.raycastScreen(sx: number, sy: number, maxDistance: number?, exclude: (string | {string})?) -> table?

Cast a ray from a screen pixel into the scene and return the first hit. Unprojects the pixel with screenToRay, then casts with raycast.

Parameters

sx number — Screen X in viewport-local pixels (the space of input.mouse_position and screenToRay).
sy number — Screen Y in viewport-local pixels.
maxDistance number (optional) — Maximum distance along the ray (defaults to 1000, matching raycast).
exclude (string | {string}) (optional) — Optional entity id, or array of entity ids, to exclude from hits.

local Zin = require("@builtin::modules.zinput")
local x, y = Zin.state.mousePosition(); local hit = Physics.raycastScreen(x, y)

globals/physics/releasePin

physics.releasePin(softId: string | entityRef, name: string?) -> number

Release a pin of a soft body, or every pin: its particles take their mass back, keep the velocity they had, and move with the body again.

Parameters

softId string | entityRef — The soft body's entity.
name string (optional) — The pin to release; omit to release every one.

Returns number — How many pins were released.

Physics.releasePin(cloth, corner)
Physics.releasePin(cloth)

globals/physics/removeCollider

physics.removeCollider(entityId: string | entityRef) -> string?

Remove whichever collider component an entity carries.

Parameters

entityId string | entityRef — Target entity id.

Returns string? The component that was removed, or nil when there was none.

Physics.removeCollider(id)

globals/physics/removeConstraint

physics.removeConstraint(entityId: string | entityRef, index: number?)

Remove transform constraints from an entity (if any are present).

Parameters

entityId string | entityRef — Target entity id.
index number (optional) — Optional constraint index (currently ignored — the whole component is removed).

Physics.removeConstraint(id)

globals/physics/removeJoint

physics.removeJoint(entityId: string | entityRef)

Remove the Joint component from an entity (if present).

Parameters

entityId string | entityRef — Target entity id.

Physics.removeJoint(id)

globals/physics/removeWheelCollider

physics.removeWheelCollider(entityId: string | entityRef)

Remove the WheelCollider component from an entity (if present).

Parameters

entityId string | entityRef — Target entity id.

Physics.removeWheelCollider(id)

globals/physics/repairSoftBody

physics.repairSoftBody(softId: string | entityRef) -> number

Parameters

softId string | entityRef — The soft body's entity (it carries a SoftBody component).

Returns number — The component's new repairs count.

Physics.repairSoftBody(shell)

globals/physics/setAngularDamping

physics.setAngularDamping(entityIdOrDamping: string | entityRef | number, damping: number?)

Set angular damping on an entity's rigid body. One-arg form targets the script-context entity.

Parameters

entityIdOrDamping string | entityRef | number — Entity id (with damping) OR damping value (script-context entity).
damping number (optional) — Optional explicit damping when targeting another entity.

Physics.setAngularDamping(0.1)
Physics.setAngularDamping(entityId, 0.1)

globals/physics/setAngularVelocity

physics.setAngularVelocity(a: string | entityRef | number | vec3, b: (number | vec3)?, c: number?, d: number?)

Set the angular velocity of an entity (radians/sec). Same call shapes as setVelocity.

Parameters

a string | entityRef | number | vec3 — x-component, a {x, y, z} vector, or an entity id (explicit target).
b (number | vec3) (optional) — y-component, x-component, or the vector depending on call form.
c number (optional) — z-component or y-component depending on call form.
d number (optional) — Optional z-component when targeting an explicit entity.

Physics.setAngularVelocity(0, 0, 1)
Physics.setAngularVelocity(entityId, 0, 0, 1)
Physics.setAngularVelocity(entityId, {x=0, y=0, z=1})

globals/physics/setBodyType

physics.setBodyType(entityId: string | entityRef, bodyType: string)

Change a rigid body's type at runtime. Mass, colliders, and joints are preserved — only the body's response to forces and position writes changes.

Parameters

entityId string | entityRef — Target entity id.
bodyType string — One of "dynamic", "kinematic", "static".

Physics.setBodyType(entityId, "kinematic")

globals/physics/setCcdEnabled

physics.setCcdEnabled(entityIdOrEnabled: string | entityRef | boolean, enabled: boolean?)

Enable or disable continuous collision detection on an entity's rigid body. One-arg form targets the script-context entity.

Parameters

entityIdOrEnabled string | entityRef | boolean — Entity id (with enabled) OR boolean (script-context entity).
enabled boolean (optional) — Optional explicit boolean when targeting another entity.

Physics.setCcdEnabled(true)
Physics.setCcdEnabled(entityId, true)

globals/physics/setCollisionGroups

physics.setCollisionGroups(entityId: string | entityRef, membership: number, filter: number)

Parameters

entityId string | entityRef — Target entity id.
membership number — Bitmask: which groups this collider belongs to.
filter number — Bitmask: which groups this collider can collide with.

Physics.setCollisionGroups(id, 0x0001, 0xFFFF)

globals/physics/setGravity

physics.setGravity(gravity: vec3)

Replace the world gravity vector.

Parameters

gravity vec3 — New gravity vector in m/s².

Physics.setGravity({x=0, y=-9.81, z=0})

globals/physics/setGravityScale

physics.setGravityScale(entityIdOrScale: string | entityRef | number, scale: number?)

Set the per-entity gravity scale (1.0 = normal, 0.0 = no gravity). One-arg form targets the script-context entity.

Parameters

entityIdOrScale string | entityRef | number — Entity id (with scale) OR scale value (script-context entity).
scale number (optional) — Optional explicit scale when targeting another entity.

Physics.setGravityScale(0.5)
Physics.setGravityScale(entityId, 0.5)

globals/physics/setJointMotor

physics.setJointMotor(entityId: string | entityRef, targetVelocity: number, maxForce: number)

Set a motor on an entity's joint.

Parameters

entityId string | entityRef — Target entity id (must carry a Joint component).
targetVelocity number — Desired joint velocity.
maxForce number — Maximum force the motor can apply.

Physics.setJointMotor(id, 5.0, 1000)

globals/physics/setLinearDamping

physics.setLinearDamping(entityIdOrDamping: string | entityRef | number, damping: number?)

Set linear damping on an entity's rigid body (0 = no damping). One-arg form targets the script-context entity.

Parameters

entityIdOrDamping string | entityRef | number — Entity id (with damping) OR damping value (script-context entity).
damping number (optional) — Optional explicit damping when targeting another entity.

Physics.setLinearDamping(0.05)
Physics.setLinearDamping(entityId, 0.05)

globals/physics/setMass

physics.setMass(entityIdOrMass: string | entityRef | number, mass: number?)

Set the mass of an entity's rigid body (kg). One-arg form targets the script-context entity.

Parameters

entityIdOrMass string | entityRef | number — Entity id (with mass) OR mass value (script-context entity).
mass number (optional) — Optional explicit mass when targeting another entity.

Physics.setMass(10)
Physics.setMass(entityId, 10)

globals/physics/setPinTarget

physics.setPinTarget(softId: string | entityRef, name: string, position: vec3?)

Parameters

softId string | entityRef — The soft body's entity.
name string — The pin's name, as pinSoftRegion returned it.
position vec3 (optional) — The world position for the pinned particles' centre; nil holds them where they stand.

Physics.setPinTarget(cloth, corner, {x=2, y=3, z=0})

globals/physics/setRotationLocks

physics.setRotationLocks(entityId: string | entityRef, x: boolean, y: boolean, z: boolean)

Lock or unlock rotation on specific axes.

Parameters

entityId string | entityRef — Target entity id.
x boolean — Lock rotation about the world X axis.
y boolean — Lock rotation about the world Y axis.
z boolean — Lock rotation about the world Z axis.

Physics.setRotationLocks(id, false, true, false)

globals/physics/setTranslationLocks

physics.setTranslationLocks(entityId: string | entityRef, x: boolean, y: boolean, z: boolean)

Lock or unlock translation on specific axes.

Parameters

entityId string | entityRef — Target entity id.
x boolean — Lock translation along the world X axis.
y boolean — Lock translation along the world Y axis.
z boolean — Lock translation along the world Z axis.

Physics.setTranslationLocks(id, false, false, true)

globals/physics/setVelocity

physics.setVelocity(a: string | entityRef | number | vec3, b: (number | vec3)?, c: number?, d: number?)

Set the linear velocity of an entity. Accepts (x, y, z) or a {x, y, z} vector for the script-context entity, or the same prefixed with an explicit entityId.

Parameters

a string | entityRef | number | vec3 — x-component, a {x, y, z} vector, or an entity id (explicit target).
b (number | vec3) (optional) — y-component, x-component, or the vector depending on call form.
c number (optional) — z-component or y-component depending on call form.
d number (optional) — Optional z-component when targeting an explicit entity.

Physics.setVelocity(0, 10, 0)
Physics.setVelocity(entityId, 0, 10, 0)
Physics.setVelocity(entityId, {x=0, y=10, z=0})

globals/physics/softBodySettings

physics.softBodySettings(changes: table?) -> PhysicsSoftBodySettings

Parameters

changes table (optional) — Optional PhysicsSoftBodySettings fields to change; the fields left out keep their values. A field the settings do not have, or a value of the wrong type or out of range, is refused by name.

Returns PhysicsSoftBodySettings — The settings with changes applied: what the next tick runs under.

print(Physics.softBodySettings().maxExtraSubsteps)
Physics.softBodySettings({ maxExtraSubsteps = 2 })   -- cheaper hard hits
Physics.softBodySettings({ tangleDetection = false, overlapConstraints = false })
Physics.softBodySettings({ femMaxIterations = 60 })   -- FEM bodies converge further each substep

globals/physics/softBodyState

physics.softBodyState(entityId: string | entityRef, opts: table?) -> PhysicsSoftBodyState?

Parameters

entityId string | entityRef — Entity id or proxy.
opts table (optional) — { positions: boolean?, contactPoints: boolean?, contactsWith: (string | entityRef)? }: positions = true adds every particle's world position; contactPoints = true adds each contact's individual points (each contact names its count, normal, depth, impulse and particles either way); contactsWith reports only the contacts with that one entity, so a read about one pair costs that pair alone.

Returns PhysicsSoftBodyState? — A PhysicsSoftBodyState, or nil when nothing in the scene answers to that id.

local s = Physics.softBodyState(jelly); print(s.particleCount, s.volumeRatio, s.sleeping)
print(Physics.softBodyState(shell).plastic)   -- metres of dent the shell kept
local s = Physics.softBodyState(shell); if s.diverged then print(s.divergence) end
for _, p in Physics.softBodyState(jelly, { positions = true }).positions do print(p.y) end
local r = Physics.softBodyState(jelly).render; print(r.drawn, r.vertexCount, r.uploadedThisFrame)
local s = Physics.softBodyState(jelly); if s.status == "pending" then print("waiting for", s.waitingFor) end

globals/physics/softBodyTears

physics.softBodyTears() -> { PhysicsSoftBodyTear }

Every soft body tear since the last call to this function, oldest first, with the entities of the pieces each split off: the same records onSoftBodyTear delivers. The 256 most recent are kept.

Returns { PhysicsSoftBodyTear } — Array of PhysicsSoftBodyTear.

for _, t in Physics.softBodyTears() do print(t.kind, t.entity, #t.pieces) end

globals/physics/softParticles

physics.softParticles(softId: string | entityRef, region: PhysicsSoftRegion?) -> PhysicsSoftParticles?

Parameters

softId string | entityRef — The soft body's entity.
region PhysicsSoftRegion (optional) — Which particles to read, as Physics.softRegion takes it; nil reads the whole body.

Returns PhysicsSoftParticles? — A PhysicsSoftParticles report, or nil when nothing in the scene answers to that id.

local top = Physics.softParticles(jelly, { box = { min = {-0.5, 0.4, -0.5}, max = {0.5, 0.5, 0.5} } }); print(top.count, top.center.y)
local near = Physics.softParticles(shell, { box = { min = {1.6, 0, -0.3}, max = {2.2, 1, 0.3} }, space = "world" })

globals/physics/softRegion

physics.softRegion(region: PhysicsSoftRegion?) -> { [string]: any }

Parameters

region PhysicsSoftRegion (optional) — { box = { min = v, max = v } | { v, v }, nearest = v, space = "local" | "world" }, or nil for the whole body.

Returns { [string]: any } — { kind = "Whole" | "Box" | "Nearest", min, max, point, world }.

Physics.softRegion({ box = { min = {-0.5, 0.45, -0.5}, max = {0.5, 0.5, 0.5} } })
Physics.softRegion({ nearest = {0, 0.5, 0} })

globals/physics/sphereCast

physics.sphereCast(origin: vec3, radius: number, direction: vec3, maxDistance: number?, exclude: (string | {string})?) -> table?

Cast a sphere along a direction and return the first hit. A trigger (isTrigger) holds nothing up, so the sweep passes through it.

Parameters

origin vec3 — Sphere center at the start of the cast.
radius number — Sphere radius.
direction vec3 — Cast direction, of any length: the engine normalises it, so maxDistance and distance are world units.
maxDistance number (optional) — Optional distance limit.
exclude (string | {string}) (optional) — Optional entity id, or array of entity ids, to exclude from hits. Applied while sweeping rather than to the answer, so a cast that starts inside an excluded collider reports what is behind it.

local hit = Physics.sphereCast(o, 0.5, dir, 10)
local hit = Physics.sphereCast(o, 0.5, dir, 10, selfId)

globals/physics/stepCost

physics.stepCost() -> PhysicsStepCost?

Returns PhysicsStepCost? — A PhysicsStepCost, or nil on a frame where the pipeline did not step — a paused simulation, or a world still bootstrapping.

local c = Physics.stepCost(); if c then print(c.stepMs, c.narrowPhaseMs) end
local c = Physics.stepCost(); if c then for _, b in c.softBodies do print(b.entity, b.ms, b.particles) end end
local seen, total = 0, 0
-- once a frame: every step since the last read, however many a frame ran
local c = Physics.stepCost(); if c then for _, s in c.recent do if s.step > seen then total += s.softBodyMs; seen = s.step end end end

globals/physics/stillnessReasons

physics.stillnessReasons() -> { string }

Every reason whyStill can answer with, in the order the engine considers them. Read from the engine, so the list is the one the answers come from.

Returns { string } — An array of reason names.

for _, reason in Physics.stillnessReasons() do print(reason) end

globals/physics/tearSoftBody

physics.tearSoftBody(softId: string | entityRef, region: PhysicsSoftRegion)

Parameters

softId string | entityRef — The soft body's entity.
region PhysicsSoftRegion — Which particles to tear through, as Physics.softRegion takes it: { box = { min, max } } or { nearest = point } in the entity's unit box (-0.5 to 0.5 on each axis), space = "world" for world coordinates.

Physics.tearSoftBody(cloth, { nearest = hit.point, space = "world" })
Physics.tearSoftBody(sheet, { box = { min = {-0.02, -1, -0.5}, max = {0.02, 1, 0.5} } })

globals/physics/touching

physics.touching(entityId: string | entityRef, otherId: string | entityRef) -> (boolean, number, { PhysicsContactPoint })

Whether two entities are touching, and how deeply.

Parameters

entityId string | entityRef — Entity id or proxy.
otherId string | entityRef — The other entity id or proxy.

Returns (boolean, number, { PhysicsContactPoint }) — (touching, deepestPenetration, points) — deepestPenetration is in metres and 0 for surfaces that meet without overlapping.

local hit, depth = Physics.touching(a, b); print(hit, depth)

globals/physics/vehicleState

physics.vehicleState(chassisId: string | entityRef) -> table?

Read a vehicle: the chassis body's forward speed and the state of every wheel under it (the wheels whose nearest rigid-body ancestor is this chassis).

Parameters

chassisId string | entityRef — The chassis entity (the rigid body the wheels sit under).

local car = Physics.vehicleState(carId); print(car.speed, car.groundedWheels)

globals/physics/wakeUp

physics.wakeUp(entityId: (string | entityRef)?)

Parameters

entityId (string | entityRef) (optional) — Target entity id or proxy; resolves from script context when omitted.

Physics.wakeUp(id)

globals/physics/whyStill

physics.whyStill(entityId: string | entityRef) -> (string?, string?)

Parameters

entityId string | entityRef — Entity id or proxy.

Returns (string?, string?) — (reason, detail).

local why, detail = Physics.whyStill(id); if why then print(why, detail) end

globals/physics/worldState

physics.worldState() -> PhysicsWorldState

Returns PhysicsWorldState — A PhysicsWorldState.

local w = Physics.worldState(); print(w.bodies.awake .. "/" .. w.bodies.total .. " awake")
print(Physics.worldState().contacts.touchingPairs .. " pairs touching")

globals/plugins/awaitingPerson

plugins.awaitingPerson(plugin: string) -> boolean

Whether a call to this plugin is waiting on a person to answer the engine's consent dialog.

Parameters

plugin string — The plugin asset's library identity, written as a LITERAL.

Returns boolean — True while a dialog for this plugin is unanswered.

if plugins.awaitingPerson(PLUGIN) then print("waiting for you") end

globals/plugins/call

plugins.call(plugin: string, method: string, ...: any?) -> any

Call a method the plugin at plugin exposes.

Parameters

plugin string — The plugin asset's identity, written as a LITERAL: the same string asset.identity answers for it (motion for a plugin in this world, @publisher::path.name for one in a library), or its path. A computed identity is refused: runtime code has no dynamic requires and a plugin is no exception, because a name the reference deriver cannot read at authoring time is a dependency that never enters the graph.
method string — A method the plugin's plugin.yaml exposes. One it does not is not callable, whatever the component carries.
... any (optional) — The call's arguments: numbers, strings, booleans, nil and tables of them. A whole number arrives as an integer.

Returns any — A promise of the plugin's answer, as a value: a table as a table, a number as a number. A refusal raises where it is awaited.

local agents = await(plugins.call("@builtin/plugins/dotbot.plugin", "agents"))

globals/plugins/isLoaded

plugins.isLoaded(plugin: string) -> boolean

Whether the plugin is loaded and answering calls.

Parameters

plugin string — The plugin asset's identity, written as a LITERAL.

Returns boolean — True while it is loaded. Never raises: a caller asking is usually waiting for a plugin that is still loading.

while not plugins.isLoaded("motion") do task.wait(0.1) end

globals/plugins/off

plugins.off(plugin: string, id: number) -> boolean

Stop a listener plugins.on started.

Parameters

plugin string — The same identity the listener was started with.
id number — The id plugins.on answered.

Returns boolean — Whether a listener was removed.

globals/plugins/on

plugins.on(plugin: string, listener: (string, any) -> ()) -> number

Listen to the events a plugin raises.

Parameters

plugin string — The plugin asset's library identity, written as a LITERAL.
listener (string, any) -> () — function(eventType, payload) — payload is the event's decoded JSON.

Returns number — The subscription id, for plugins.off.

local id = plugins.on("@builtin/plugins/dotbot.plugin", function(kind, body) end)

globals/plugins/openUrl

plugins.openUrl(plugin: string, url: string) -> any

Ask the person to open one of the plugin's own pages.

Parameters

plugin string — The plugin asset's library identity, written as a LITERAL.
url string — An https URL whose host the plugin declared under network.hosts. Anywhere else is refused — otherwise this would be an unbounded way to put an address in front of somebody.

Returns any — A promise: true when it opened, or the refusal.

await(plugins.openUrl("@builtin/plugins/dotbot.plugin", "https://claude.ai/login"))

globals/plugins/status

plugins.status() -> { [string]: any }

What the plugin host loaded, and what refused to load.

Returns { [string]: any } — started is whether the host has run; plugins lists each one with loaded, and the error that refused it where there is one. On the web build it also carries what the browser offers of JavaScript Promise Integration, which every plugin's host calls are lowered onto.

local report = plugins.status()
print(report.started, #report.plugins)

globals/postprocess/add

postprocess.add(name: string, shader: string | AssetRef, opts: PostprocessOpts?) -> (boolean, string?)

Register a fullscreen post-process effect. This call is what puts a pass into the frame — a post-process .shader asset defines an effect, and renders only once registered here. The chain applies the registration on the caller's own stack and the returned boolean is its answer, so a setProperty or setTexture naming the effect in the same call finds it. shader is a .shader asset reference whose shader.wgsl provides fn fragment(in: PostInput) -> vec4<f32> and whose properties.yaml declares the effect's properties; the engine generates the group(0) framework + schema-driven group(1) from that schema. Editing that shader afterwards recompiles this effect in place, keeping its enabled state, priority, layer and tuned property values. WGSL text is accepted in place of a shader reference, and then opts.properties is the whole schema — this form is deprecated and will be removed, author the effect as a .shader asset instead. Effects run in priority order (lower first, default 100). A registered effect runs over the live viewport's frame AND over every offscreen one — a capture from a world-space station, one orbiting an entity, one of a named camera, a render-to-texture camera. In each of those the effect's engine.view_proj / engine.prev_view_proj / engine.inv_view_proj are the camera THAT render was drawn from and engine.resolution is that target's own size, so a pass reconstructing world space from zero_scene_depth(uv) reconstructs against the station and lens the capture asked for. An offscreen capture is therefore an oracle for an authored grade: it photographs a chosen station without taking the on-screen camera from whoever else is driving the scene, and a capture's postProcessing = false is the one control that takes the chain off the frame it returns. An offscreen render keeps no view history of its own, so engine.prev_view_proj there holds that same matrix rather than the frame before it, and a pass taking camera motion from the two reads none.

Parameters

name string — Unique effect name.
shader string | AssetRef — A resolved shader asset reference, or author WGSL (fn fragment(in: PostInput) only).
opts PostprocessOpts (optional) — { priority = 100, enabled = true, layer = "scene"|"all" (omitted: follows space), space = "light"|"display", properties = {{name, type, default?, min?, max?, textureDefault?}}, passes = {{name, entry, scale?, inputs?}} } — with a shader asset, properties layers over the asset's own schema. layer picks which composited image the effect grades: "scene" runs it before the UI is drawn, over the scene's own light in a float target, so it grades the rendered picture and leaves every widget on screen as authored, and "all" runs it after the UI has landed, over the picture the display transform fitted to the screen, so the interface is graded along with the picture. A call that names no layer takes the one its space belongs to: a "light" effect grades "scene" and a "display" effect grades "all", so a bloom reads the scene's highlights where they still stand above 1.0 and a lookup table reads code values. Naming "all" is what an effect that must grade the interface says — and on a world several authors share, it grades everyone else's screens too. space names what fragment() reads and returns: "light" (the default) for a threshold/blur/grade on the scene, "display" for a lookup table, dither ladder or authored overlay working on code values. With a shader asset, defaults to what that asset's getSpace() declares. passes names the passes the effect draws, each an entry in the shader, a fraction of the view it draws at, and what it reads. With a shader asset, defaults to what that asset's getPasses() declares; a single-pass effect (the common case) declares none. textureDefault is what a type = "texture" property samples while nothing is bound to it: "white" (1,1,1,1 — the default), "black" (0,0,0,1), "normal" (0.5,0.5,1,1) or "transparent" (0,0,0,0). An effect that lays its texture over the scene wants "transparent", so the frame is untouched until setTexture binds a texture that exists.

Returns (boolean, string?) — True when the chain took the effect; false and the compiler's message when it refused it. A shader that does not compile draws nothing at any property value, so it is not registered and postprocess.list() never names it. __display__ and __blit__ are the renderer's own effects, and a registration under either name is refused with a message saying so. A call made from inside queue() or batch(), where the engine has not run the registration by the time the call returns, answers true for the queued request.

postprocess.add("vignette", asset.resolve("@builtin::shaders.post.vignette", "shader"))
postprocess.add("vignette", VIGNETTE_WGSL, { properties = {{ name = "intensity", type = "float", default = 0.5 }} })

globals/postprocess/describe

postprocess.describe(name: string) -> PostprocessDescription?

One effect by name, read in full: the chain state postprocess.status() lists for it, and on top of that properties — the schema the effect declared, each entry { name, type, default?, min?, max?, textureDefault? } in the shape add takes — and values, what each of those properties currently holds. A property's value is the one the last setProperty wrote, or the schema's own default where nothing has written one, and it comes back as a number for a scalar and as the array for a wider value, which is what setProperty takes, so a property read here is written straight back.

This is the read-back for a property write. setProperty answers whether the uniform took the value; this answers what the effect holds now, which is the reading a pass that writes its properties every frame needs and the one that tells a mistyped property name from an effect that is not grading. The schema and the values are the engine's own record of the effect — the schema it was registered with and every write the chain accepted into its uniform, the same record /runtime/fx/<name>/meta.json is serialized from. A write the chain refused is not in it, and neither is one made against a property the schema does not declare.

Before an effect is registered its schema lives on the .shader asset it will render: asset.resolve("@builtin::shaders.post.bloom", "shader"):getProperties() names what that shader declares.

Parameters

name string — Effect name.

Returns PostprocessDescription? — The effect's state, schema and live values, or nil when nothing is registered under the name. An effect the renderer registers itself declares no properties of its own, and its properties and values are empty.

local e = postprocess.describe("vignette")
print(if e then e.values.intensity else "not registered")
for _, p in ipairs(postprocess.describe("bloom").properties) do
print(p.name, p.type, p.min, p.max)
end

globals/postprocess/list

postprocess.list() -> { string }

List all registered post-process effect names in renderer priority order (lower priority runs first).

Returns { string } — Array of effect names.

for _, n in ipairs(postprocess.list()) do print(n) end

globals/postprocess/remove

postprocess.remove(name: string) -> boolean

Queue removal of a post-process effect. Takes effect on the next frame. Removing a name that isn't registered is a silent no-op.

Parameters

name string — Effect name to remove.

Returns boolean — True when the mutation was queued. False, named in a WARN, for the two effects the renderer registers itself: __display__, the chain's display transform, which stays in every frame and is tuned through its curve and exposure properties, and __blit__, the copy that carries every render into its target, every capture and the viewport included.

postprocess.remove("vignette")

globals/postprocess/setEnabled

postprocess.setEnabled(name: string, enabled: boolean) -> boolean

Queue an enable/disable toggle on a registered post-process effect. Targeting an unknown name is a silent no-op.

Parameters

name string — Effect name.
enabled boolean — True to enable, false to disable.

Returns boolean — True when the mutation was queued. False, named in a WARN, for a __display__ disable (the chain's display transform stays in every frame, and curve = 0 is its linear curve) and for a __blit__ enable (the copy that carries every render into its target joins no chain).

postprocess.setEnabled("bloom", false)

globals/postprocess/setProperty

postprocess.setProperty(name: string, prop: string, value: (number | { number })) -> boolean

Set a named material property on a registered post-process effect. The property must be declared in the effect's properties schema; read in WGSL as material.<prop>. value is a number or a number array (vec/color).

Parameters

name string — Effect name.
prop string — Declared property name.
value (number | { number }) — Number or array of numbers.

Returns boolean — True when the effect's uniform took the value; false when it did not — an effect that is not registered, or one that declares no property by that name, is named in a WARN in the engine log. A call made from inside queue() or batch(), where the engine has not run the write by the time the call returns, answers true for the queued request.

postprocess.setProperty("vignette", "intensity", 0.6)

globals/postprocess/setSampler

postprocess.setSampler(name: string, opts: { [string]: any }) -> boolean

Configure the per-effect user sampler shared by the effect's declared texture properties. opts.filter = "linear" (default) or "nearest". opts.wrap (alias .address) = "clamp" (default), "repeat", or "mirror" — applied to all axes.

Parameters

name string — Effect name.
opts { [string]: any } — { filter = "linear"|"nearest", wrap = "clamp"|"repeat"|"mirror" }.

Returns boolean — True — the mutation was queued.

postprocess.setSampler("blur", { filter = "linear", wrap = "clamp" })

globals/postprocess/setTexture

postprocess.setTexture(name: string, prop: string, path: string) -> boolean

Bind a texture to one of an effect's declared texture properties. Declare it in properties ({ name = "noise", type = "texture" }) and sample in WGSL as textureSample(noise, noise_sampler, in.uv). path is any TextureCache-resolvable spec (@builtin::textures.foo, color:1,0,0, default:white, a render-target name, ...). A path whose texture has not reached the GPU yet — one this same script created — is held and bound as soon as it does; postprocess.status() reports it under pendingTextures until then.

Parameters

name string — Effect name.
prop string — Declared texture-property name.
path string — Texture path / spec.

Returns boolean — True when the slot took the binding, including one held until its texture reaches the GPU; false when it did not — an effect that is not registered, or one that declares no texture property by that name, is named in a WARN in the engine log. A call made from inside queue() or batch(), where the engine has not run the binding by the time the call returns, answers true for the queued request.

postprocess.setTexture("__vsky_composite", "cloud", "cloud_target")

globals/postprocess/status

postprocess.status() -> { PostprocessStatus }

Every registered effect in chain order with the state that decides whether it reaches the frame — enabled flag, priority, layer, the shader's compile error when it has one, the .shader asset it renders when it was registered from one, the texture each declared slot is bound to (textures) and the bindings still waiting for their texture (pendingTextures). This is what the renderer draws with, so a survey of the chain answers "is this one affecting the picture right now?" without capturing a frame and reading pixels.

An effect this script has just registered is listed with pending = true until the renderer publishes it, since a registration is queued for the next frame.

Returns { PostprocessStatus } — Array of per-effect state, in the order the chain runs them.

for _, e in ipairs(postprocess.status()) do
print(e.name, e.enabled, e.layer, e.error)
end

globals/preset

preset: any

Preset namespace — apply named tuning bundles to components. Auto-injected by the prelude from @builtin::modules.preset.

globals/preset/addDefault

preset.addDefault(source: AssetRef<preset>, opts: { filter: string?, at: number? }?) -> number

Add one Preset Manager row for source, governing the declarer the preset targets. opts.filter limits it to new things whose path matches; opts.at places it (1 = first); it goes last otherwise.

Parameters

source AssetRef<preset> — The preset.
opts { filter: string?, at: number? } (optional) — { filter?, at? }.

Returns number — The new row's index.

preset.addDefault("pixel_art")

globals/preset/apply

preset.apply(source: AssetRef<preset>, into: AssetRef | ComponentTarget) -> ApplyResult

Write a preset's values into something its target governs: an asset whose settings the target declares, or a component on an entity, named { entity = <id>, component = <name> }, whose component the target is. Struct groups merge field by field. Only valid fields are written; the rest are returned in rejected.

Parameters

source AssetRef<preset> — The preset.
into AssetRef | ComponentTarget — The asset, or the component on an entity, to write into.

Returns ApplyResult — { applied, rejected, targetChanged }.

preset.apply("pixel_art", "hero_sprite")
preset.apply("red_tint", { entity = id, component = "Model" })

globals/preset/capture

preset.capture(from: AssetRef | ComponentTarget, name: string, opts: CaptureOpts?) -> AssetRef<preset>

Save the settings something has now as a new preset targeting the declarer that governs them: an asset's settings, or the live values of a component on an entity, named { entity = <id>, component = <name> }.

Parameters

from AssetRef | ComponentTarget — The asset, or the component on an entity, whose values are captured.
name string — A bare name (lands in /source/presets/) or an absolute .preset path.
opts CaptureOpts (optional) — { overwrite = true } re-authors an existing preset of that name, keeping its guid.

Returns AssetRef<preset> — The new preset.

local p = preset.capture("hero_sprite", "pixel_art")
local tint = preset.capture({ entity = id, component = "Model" }, "red_tint")

globals/preset/create

preset.create(name: string, entityId: string, componentType: string, opts: PresetCreateOpts?) -> PresetCreateResult

Capture a component's live values on an entity as a preset targeting that component; forwards to preset.capture, overwriting a preset of the same name.

Parameters

name string — Either a bare preset name (lands under /source/presets/) or an absolute path to a .preset asset or its /preset.yaml.
entityId string — Source entity id.
componentType string — Component to capture (e.g. "CharacterController").
opts PresetCreateOpts (optional) — { name = <display name> } sets the preset's .metadata title.

Returns PresetCreateResult — { path, entityId, component, properties }: the written preset.yaml, and the values it holds.

globals/preset/defaultFor

preset.defaultFor(declarer: AssetRef | string, path: string) -> ({ [string]: any }?, any?)

The values a new thing declarer governs would take from the Preset Manager at path, and the row that gives them — the layer itself, before the creating importer's and the caller's settings. A path is a /source/... path for an asset or an import, and an entity's hierarchy path for a component.

Parameters

declarer AssetRef | string — The declarer: an assetType, an importer or a component.
path string — The new thing's path.

Returns ({ [string]: any }?, any?) — (values, match), or nil when no row matches.

local values, match = preset.defaultFor("@builtin::importers.texture", "/source/sprites/hero.png")

globals/preset/defaults

preset.defaults() -> { any }

The world's Preset Manager rows, in order. Each row names the declarer it governs (target), the preset a new thing takes (preset) and a glob over the new thing's path (filter), resolved, with every problem that keeps it from applying.

Returns { any } — { { index, target, preset, filter, targetGuid, presetGuid, targetName, presetName, problems } }.

for _, row in ipairs(preset.defaults()) do print(row.index, row.targetName, row.presetName, row.filter) end

globals/preset/load

preset.load(source: AssetRef<preset>, overrides: table?) -> { [string]: any }

Read a preset as a plain table: a typed preset's valid values, or a component-only preset's properties as authored, with overrides laid over the top. Forwards to preset.values.

Parameters

source AssetRef<preset> — The preset.
overrides table (optional) — Optional table merged onto the values.

Returns { [string]: any } — Plain table.

globals/preset/removeDefault

preset.removeDefault(index: number)

Remove the Preset Manager row at index. The other rows are kept exactly as written, so removing succeeds while another row is broken.

Parameters

index number — The row's index, as preset.defaults() reports it.

preset.removeDefault(1)

globals/preset/setDefaults

preset.setDefaults(rows: { DefaultRowInput })

Replace the world's Preset Manager rows. Each row names a preset, optionally its declarer (the preset's own target when omitted) and a filter (empty matches everything). Every row is checked first; any problem refuses the whole write, naming every one. Rows change only what is created afterwards.

Parameters

rows { DefaultRowInput } — The complete ordered rows.

preset.setDefaults({ { preset = "pixel_art", filter = "sprites/**" } })

globals/preset/setValues

preset.setValues(source: AssetRef<preset>, values: { [string]: any })

Replace a preset's values. The whole table is checked against the target's schema first; any violation refuses the write, naming every one. A preset in the component-only form is written in the typed form, its target the component it names and its legacy name moved to the .metadata title, as a heal moves it.

Parameters

source AssetRef<preset> — The preset.
values { [string]: any } — The complete values table the preset holds from now on.

preset.setValues("pixel_art", { filter = "nearest", generateMipmaps = false })

globals/preset/values

preset.values(source: AssetRef<preset>) -> ({ [string]: any }, Report)

A preset's values and a report on them. Every value is checked against the target declarer's current schema: the fields that pass are returned, every field that fails is listed by path in report.rejected, report.pinned says the preset's .refs pins the target by guid, and report.targetChanged says the target differs from the version pinned. A preset still in the component-only form reads as authored, and report.rejected lists what its component's schema would refuse.

Parameters

source AssetRef<preset> — The preset.

Returns ({ [string]: any }, Report) — (values, report), where report is { format, target, targetGuid, pinned, targetChanged, rejected }.

local values, report = preset.values("pixel_art")

globals/print

print(...)

Print values to the engine console. Concatenates all arguments with tabs.

Returns nil

globals/profiler/begin

profiler.begin(name: string)

Start a named profiling block. Call profiler.finish(name) to record the duration. Blocks appear in profiler.stats() under "script.<name>" and inside captures.

Parameters

name string — Block name (e.g. "MyComponent.update").

profiler.begin("MyComponent.update"); ...; profiler.finish()

globals/profiler/counters

profiler.counters(pattern: string?) -> { ProfilerCount }

Get the profiler's counts: unitless per-frame tallies (for example the waits the engine thread made on the process-wide lock, under system.schedule.frame.process_local_waits), kept apart from the millisecond timings of stats. frame is the tally of the frame in progress, prev_frame the last complete frame's, max_frame the highest complete-frame tally and total the tally since the profiler started. Optional glob pattern filters by name (supports * and ?).

Parameters

pattern string (optional) — Filter pattern (e.g. "system.schedule.frame.*").

Returns { ProfilerCount } — Array of profiler counts.

for _, c in ipairs(profiler.counters("system.schedule.frame.*")) do print(c.name, c.prev_frame) end

globals/profiler/disableRing

profiler.disableRing()

Disable the ring buffer and clear its history.

profiler.disableRing()

globals/profiler/enableRing

profiler.enableRing(seconds: number?) -> boolean

Enable the always-recording ring buffer, retaining the last seconds of per-frame data (default 20). Query it AFTER the fact with profiler.retro() — latency-immune, since the data is historical. Editor profile only: returns false in the runtime profile. The enable is the gate; the ring costs nothing until on.

Parameters

seconds number (optional) — Seconds of history to retain (default 20).

Returns boolean — True if enabled, false if refused (runtime profile).

if profiler.enableRing(30) then ... end

globals/profiler/finish

profiler.finish(name: string?) -> number?

Finish a profiling block and record the elapsed duration as "script.<name>". Without an argument, closes the most-recently- begun block (LIFO stack). With a name, closes the most recent block whose name matches — useful when blocks of different names are nested.

Parameters

name string (optional) — Block name to finish. Omit to pop the top of the stack.

Returns number? — Elapsed milliseconds, or nil if no matching block was active.

local ms = profiler.finish("MyComponent.update")

globals/profiler/gpuFrame

profiler.gpuFrame() -> GpuFrameReport

Label-aggregated GPU pass timings over the last window_frames resolved frames, measured with GPU timestamp queries. supported is false when the device lacks timestamp queries — spans stays empty. Each span covers every render/compute pass recorded under one label — compute.<shader> per compute dispatch, scene.* for the scene passes, post.<effect> per post-process effect, feature.* for render-feature passes: ms is the median of its per-frame totals, min_ms/max_ms the range that median sits in, count the passes per frame and frames how much of the window carried it. at_floor marks a label whose every sample landed within a few ticks of the device's timestamp counter (tick_ms) — those passes ran and the device resolved no duration for them, which is not the same as a measured zero. ran is whether the label recorded a measured pass in the newest resolved frame, and last_frame the newest frame that did. The window outlives the work it describes, so a pass that stops being recorded leaves a row standing for up to window_frames frames carrying the median of the frames it did run in: read ran to answer whether a pass is running, frame - last_frame for how many resolved frames ago it last did, and ms as the cost of the frames it ran in. frame_span_ms (first pass begin to last pass end) and total_ms are medians too, so rows do not sum to total_ms, and the GPU may overlap passes so total_ms can exceed frame_span_ms. The readback is asynchronous: the window lags the live frame by a few frames.

Returns GpuFrameReport — GPU timing window, spans ranked by median ms descending.

local g = profiler.gpuFrame(); print(g.frame_span_ms, g.spans[1] and g.spans[1].label)

globals/profiler/hits

profiler.hits(label: string?) -> string?

Drain the watchdog's recorded hit frames into a capture stored under label (default "watch_hits") and clear the buffer. Returns the capture JSON (same shape as stopCapture), or nil if there were no hits.

Parameters

label string (optional) — Capture label to store under (default "watch_hits").

Returns string? — Capture JSON of the hit frames, or nil if none.

local json = profiler.hits()

globals/profiler/isCapturing

profiler.isCapturing() -> boolean

Check if a profiler capture is currently active.

Returns boolean — True if a capture is in progress.

if profiler.isCapturing() then ... end

globals/profiler/lastCapture

profiler.lastCapture() -> string?

Get the most recent completed capture result as a JSON string. Same shape as profiler.stopCapture(). Returns nil if no capture has been completed yet.

Returns string? — JSON string of the last capture, or nil.

local last = profiler.lastCapture()

globals/profiler/perfLine

profiler.perfLine() -> string?

The engine's own one-line performance summary, exactly as it rides a tool response: the headline frame time, and when the frame is slow or CPU moved since the last edit, the severity marker, the delta and the systems the time went to. A DIFFERENT measurement from everything below: this is the published smoothed frame, while stats and the profiler toolbox report the effective frame time (wall clock minus the cost of the call doing the asking), so a line composed from those disagrees with what other callers are told about the same engine. Reading it does not move the edit-to-edit baseline the delta is measured against.

Returns string? — The summary line, or nil before the first frame is published.

local line = profiler.perfLine()

globals/profiler/recordingHold

profiler.recordingHold() -> string?

Who holds the one windowed recording, as a JSON string { label, owner, started_at, held_seconds }, or nil when nothing is recording. owner is the caller key of whoever called startCapture, the same value vfs.currentAuthor().id reads for the calling caller, so a caller refused the slot tells its own earlier recording from a sibling session's; it is absent when that call attested no caller. Several agent sessions share one engine and one recording, which is why a refusal is answered with this.

Returns string? — JSON hold description, or nil when no recording is active.

local held = profiler.recordingHold()

globals/profiler/retro

profiler.retro(seconds: number?, label: string?) -> { [string]: any }?

Retroactively aggregate the last seconds of the ring (default: the whole ring). The full per-frame capture is retained under label (default "retro") for in-engine drill-down (the frame / hotspots tools); this RETURNS a compact structured aggregate table (frame-time distribution

per-system summary), never the raw per-frame array — bounded, so it is safe over the ZeroMind bridge. Code-facing primitive; the retro tool renders the agent-facing report. Latency-immune: the data is historical.

Parameters

seconds number (optional) — How many seconds back to include (default: whole ring).
label string (optional) — Capture label to store under (default "retro").

Returns { [string]: any }? — A compact aggregate { label, source, frames, seconds, exclude_agent, agent_frames, dt = { avg, p50, p90, p99, max, min }, summary = {...} }, or nil if the ring holds nothing.

local agg = profiler.retro(8, "collapse")

globals/profiler/ringStatus

profiler.ringStatus() -> string

Ring buffer status as a JSON string: { enabled, frames, capacity, span_seconds }.

Returns string — JSON status string.

local s = profiler.ringStatus()

globals/profiler/startCapture

profiler.startCapture(label: string?) -> boolean

Start recording per-frame profiler data, held by the calling caller. Each frame's system timings are captured until stopCapture() is called. Results are accessible via profiler.lastCapture() and VFS at /zero/runtime/profiler/<label>.json. The recording is one engine-wide slot, so a start while another recording holds it returns false and profiler.recordingHold() names the label and the caller holding it.

Parameters

label string (optional) — Capture label (default "capture").

Returns boolean — True if capture started, false when another recording holds the slot.

if profiler.startCapture("frame-spike") then ... end

globals/profiler/stats

profiler.stats(pattern: string?) -> { ProfilerStat }

Get current EMA profiling statistics from the SystemProfiler. Optional glob pattern filters by metric name (supports * and ? wildcards). Each entry carries two averages: avg_ms averages one RUN of the block and is folded when the block runs, so a block that has stopped running keeps the last value it saw; avg_frame_ms averages one FRAME and is folded every frame, including the frames the block did not run in, so it is the block's share of the current frame and falls back to zero once the block stops running.

Parameters

pattern string (optional) — Filter pattern (e.g. "schedule.", "system.schedule.render.").

Returns { ProfilerStat } — Array of profiler block stats.

for _, s in ipairs(profiler.stats("schedule.*")) do print(s.name, s.avg_frame_ms) end

globals/profiler/stopCapture

profiler.stopCapture() -> string?

Stop the active profiler capture and return its result as a JSON string. The capture is also saved to VFS at /zero/runtime/profiler/<label>.json. Top-level fields: label, frame_count, started_at, ended_at, frames, summary, callers. Compute duration as ended_at - started_at. callers holds { caller, total_ms, max_ms, frames } per caller whose queued Luau ran in the window, heaviest first, and each frame carries its own split as agent_by_caller.

Returns string? — JSON capture result, or nil if no capture was active.

local json = profiler.stopCapture()

globals/profiler/unwatch

profiler.unwatch()

Disarm the watchdog. Recorded hits are kept for a final profiler.hits().

profiler.unwatch()

globals/profiler/watch

profiler.watch(ceilingMs: number, mode: string?, excludeAgent: boolean?, maxHits: number?) -> boolean

Arm the frame-time watchdog. When a frame's EFFECTIVE time (total minus agent-injected execute cost) crosses ceilingMs, mode "record" logs every offending frame (read with profiler.hits()), and mode "pause" pauses gameplay ONCE to freeze the bad state, then disarms. Editor profile only: returns false in the runtime profile. excludeAgent (default true) keeps the agent's own calls from tripping it.

Parameters

ceilingMs number — Effective frame-time ceiling in ms.
mode string (optional) — "record" (default) or "pause".
excludeAgent boolean (optional) — Subtract agent cost before comparing (default true).
maxHits number (optional) — Max frames retained in record mode (default 240).

Returns boolean — True if armed, false if refused (runtime profile).

if profiler.watch(50, "pause") then ... end

globals/profiler/watchStatus

profiler.watchStatus() -> string

Watchdog status as a JSON string: { armed, ceiling_ms, mode, exclude_agent, hits, dropped_hits, tripped }.

Returns string — JSON status string.

local s = profiler.watchStatus()

globals/queue

queue(fn, opts?): (boolean, string?)

Defer FFI write paths inside fn to be applied across frames by the engine drainer. Returns (true, nil) on success or (false, err) if the body throws — partial batch is cleared so it never lands. Nested queue() is re-entrant (depth counter). Reads inside the body are not yet fenced (Stage 2). See modules.queue and docs/plans/2026-05-06-luau-queue-deferred-mutations.md.

Parameters

fn function — Body whose FFI writes get queued across frames
opts table (optional) — Reserved for future opts (currently { onError = "halt" })

Returns (boolean, string?) — (true, nil) on success; (false, err) if the body errored.

globals/rawequal

rawequal(a, b) -> boolean

Compare without metamethods.

globals/rawget

rawget(table, key) -> value

Get without metamethods.

globals/rawlen

rawlen(table) -> number

Length without metamethods.

globals/rawset

rawset(table, key, value)

Set without metamethods.

globals/reflectionProbe/add

reflectionProbe.add(x: number, y: number, z: number, opts: { [string]: any }?) -> string

Add a reflection probe at (x, y, z) in one call: spawns a probe entity carrying a ReflectionProbe component, which registers it and loads its persisted bake when one exists. A probe with no bake reflects nothing until it is baked (bakeAll(), or bake(id)). The probe is an editor gizmo, invisible in play mode. Returns the probe entity id.

Parameters

x number — World X.
y number — World Y.
z number — World Z.
opts { [string]: any } (optional) — Optional { radius = 12, probeId = "...", name = "..." }. probeId is the STABLE asset identity (so a re-created probe reloads the same baked cube); defaults to the entity id. The probe does NOT bake on add — call bakeAll() once the scene is built (baking is an authoring step).

Returns string — The probe entity id.

reflectionProbe.add(0, 3, 0, { radius = 15, probeId = "lobby" })

globals/reflectionProbe/apply

reflectionProbe.apply() -> number

Push the current active-probe blend data (live positions + radii) to the renderer. Builds a dense slot array so each probe's data lands at its cube slot; freed/missing slots, and the slot of a probe that holds no cube yet, become inert placeholders that reach no surface. Called automatically by add / bake / remove; call it directly after moving a probe entity.

Returns number — The number of probes applied that reflect (registered, live and baked).

globals/reflectionProbe/bake

reflectionProbe.bake(id: string) -> (string?, string?)

Bake the scene into probe id's cube slot from its current position AND persist it to a faces6 .texture asset (so it survives reload + syncs), then re-apply the probe set. Yields a few frames; call from a task/coroutine context (component hook via task.spawn, bakeAll, or execute).

Parameters

id string — Probe entity id.

Returns (string?, string?) — The asset path on success, or (nil, errorMessage) on failure.

globals/reflectionProbe/bakeAll

reflectionProbe.bakeAll() -> { baked: number, failed: number, errors: { string } }

Bake EVERY registered probe in the active layers, in one call. Captures the sky into the fallback slot, then each probe's scene from its position into its slot, persists it, and applies the full probe set. The agent/editor one-liner. Yields; call from a task/coroutine context (execute, a tool, or task.spawn).

Returns { baked: number, failed: number, errors: { string } } — { baked = N, failed = M, errors = { ... } }.

reflectionProbe.bakeAll()

globals/reflectionProbe/baking

reflectionProbe.baking(id: string?) -> boolean

Whether a bake under way will reach probe id: its own bake, or a bakeAll that has it still to reach. With no id, whether any bake is under way.

Parameters

id string (optional) — Optional probe entity id.

Returns boolean — True while a bake that reaches the probe (or any bake) runs.

reflectionProbe.baking(id)

globals/reflectionProbe/count

reflectionProbe.count() -> number

Number of registered probes.

Returns number

globals/reflectionProbe/ensureSkyFallback

reflectionProbe.ensureSkyFallback() -> boolean

Ensure the scene's sky is in the environment's sky fallback: a reflective surface no probe covers then reflects the sky rather than black, and a partially covered one blends the shortfall against it. Queues a capture when the sky slot holds none, and re-arms the fallback when a capture is there but switched off. The engine's own state answers both questions, so calling this on every probe that comes up costs one capture between them, and a scene that lost its fallback gets it back. Once captured, the fallback follows the sky the scene draws on its own.

Returns boolean — True if a capture was queued, false if the sky slot already holds one.

reflectionProbe.ensureSkyFallback()

globals/reflectionProbe/isBaked

reflectionProbe.isBaked(id: string) -> boolean

Whether probe id holds a cube it reflects: baked by bake / bakeAll, or loaded from its persisted bake by loadBaked. An unbaked probe adds nothing to the frame.

Parameters

id string — Probe entity id.

Returns boolean — True once the probe holds a cube; false for an unbaked or unregistered probe.

reflectionProbe.isBaked(id)

globals/reflectionProbe/list

reflectionProbe.list() -> { any }

List every registered probe: { { id, slot, radius, priority, baked, asset, position }, ... }. baked is whether the probe holds a cube it reflects; an unbaked probe adds nothing to the frame. asset names the persisted bake once there is one.

Returns { any } — The probe list.

globals/reflectionProbe/loadBaked

reflectionProbe.loadBaked(id: string) -> boolean

Load probe id's PERSISTED baked cube (probe_<key>.texture) into its slot WITHOUT re-rendering the scene — the runtime path. A probe bakes once at authoring time and loads the asset on every subsequent scene load. Returns false (not an error) when no baked asset exists yet.

Parameters

id string — Probe entity id.

Returns boolean — True if a baked asset was loaded, false if none exists / load failed.

globals/reflectionProbe/register

reflectionProbe.register(id: string, radius: number, key: string?) -> number?

Register a reflection probe for entity id with influence radius. Assigns a free cube slot and applies the updated probe set. Idempotent — a re-register keeps the same slot and just updates the radius. Called by the ReflectionProbe component's awake; rarely called directly.

Parameters

id string — Probe entity id.
radius number — Influence radius (world units) — surfaces within blend it.
key string (optional) — Optional STABLE asset identity (the probe's probeId). Defaults to id. The baked cube persists at probe_<key>.texture so an authored probe keeps the same asset across reloads even though its runtime entity id changes.

Returns number? — The assigned cube slot, or nil if all MAX_PROBES slots are taken.

globals/reflectionProbe/setPriority

reflectionProbe.setPriority(id: string, priority: number)

Set a probe's blend rank against the probes it overlaps, and re-apply. Probes are gathered highest rank first and each rank takes the coverage the ranks above it left, so a small interior probe ranked above the large exterior one it sits inside wins outright wherever it reaches full weight, while probes of equal rank crossfade by proximity as before.

Parameters

id string — Probe entity id.
priority number — Blend rank. Defaults to 0 on every probe.

reflectionProbe.setPriority(interiorId, 1)

globals/reflectionProbe/setProxy

reflectionProbe.setProxy(id: string, kind: string, x: number, y: number, z: number)

Anchor a probe's reflections to a proxy volume and re-apply. A cube records the environment from one point, so sampling it along the raw reflection vector puts everything it recorded at infinity and the reflection slides across a surface as the camera moves. Sizing a proxy to the geometry the probe recorded — a room's walls, say — keeps the reflection anchored to what it depicts.

Parameters

id string — Probe entity id.
kind string — "box" (sized by all three half-extents), "sphere" (sized by x), or "none" to sample along the raw reflection vector.
x number — Half-extent along X, in world units — the sphere radius for "sphere".
y number — Half-extent along Y.
z number — Half-extent along Z.

reflectionProbe.setProxy(id, "box", 5, 3, 4)  -- a 10x6x8 room

globals/reflectionProbe/setRadius

reflectionProbe.setRadius(id: string, radius: number)

Update a probe's influence radius and re-apply.

Parameters

id string — Probe entity id.
radius number — New influence radius.

globals/reflectionProbe/unregister

reflectionProbe.unregister(id: string)

Unregister entity id's probe, freeing its cube slot, and re-apply.

Parameters

id string — Probe entity id.

globals/renderer/anisotropy

renderer.anisotropy() -> number

The maximum anisotropy material textures are sampled with right now — the requested level clamped to what this device honours.

Returns number — The effective level, 1 through 16.

if renderer.anisotropy() < 4 then ... end

globals/renderer/atmospherics/held

renderer.atmospherics.held() -> boolean

Whether a hold is standing on the air right now.

Returns boolean — True while at least one renderer.atmospherics.hold stands.

if renderer.atmospherics.held() then print("clear air") end

globals/renderer/atmospherics/hold

renderer.atmospherics.hold(share: number?, options: AtmosphericsHoldOptions?) -> () -> ()

Hold the air between the camera and every surface at a stated share of what the scene authored, and return the release. At the default 0 the media contribute nothing and a surface renders in its own colour, which is what lets a reader judge an albedo, a tint or a material while another slice of a shared world drives the weather. The share reaches aerial perspective, height fog and volumetric light scattering; the sky, the sun and the light they put on a surface are untouched, because those are what the surface's colour is made of. Holds nest: the innermost names the share, and the authored air is back once the last release is called. Each release ends its own hold whatever order the releases come in, so two callers holding at once each end their own. exclusive takes the air for the owner key the call states: while that hold stands, a hold is admitted only when it states the same key, and every other one is refused with an error naming the key and the share holding it. That is what lets one caller clear the air to read a surface and keep it clear while another agent drives the same engine. The key is what an owner presents to take a nested hold of its own, and what renderer.atmospherics.release hands the air back by.

Parameters

share number (optional) — How much of the authored air reaches the image, in [0, 1]. Defaults to 0 -- no air at all.
options AtmosphericsHoldOptions (optional) — owner is the key this hold is taken under, and an exclusive hold states one. A hold that states no key is labelled with the caller the call is attributed to, which names the calling session where the transport attested one and the ZeroMind account otherwise. exclusive takes the air for the stated key until the hold is released.

Returns () -> () — A function that releases this hold. Calling it twice releases once.

local release = renderer.atmospherics.hold() ; local picture = tools.use("capture", "fromPosition", { position = { 0, 2, 8 } }) ; release()
local release = renderer.atmospherics.hold(0, { exclusive = true, owner = "albedo-read" })

globals/renderer/atmospherics/onChange

renderer.atmospherics.onChange(listener: (number) -> ()) -> () -> ()

Register a listener called with the share now in force whenever it changes — a hold taken, a hold released — and return the unsubscribe. A system that packs a medium into a GPU buffer registers here and re-packs what it has already pushed, so the buffer carries the share before the frame the hold was taken on is drawn rather than a frame later.

Parameters

listener (number) -> () — Called with the share now in force, in [0, 1].

Returns () -> () — A function that removes this listener.

local stop = renderer.atmospherics.onChange(function(share) print("atmosphere share", share) end)

globals/renderer/atmospherics/owner

renderer.atmospherics.owner() -> { id: string?, name: string?, share: number, exclusive: boolean }?

The hold naming the share of the authored air that reaches the image right now: who took it, what share it asked for, and whether it took the air exclusively. Several agents drive one engine at once and a hold any of them takes moves the air every surface in every other agent's frame is rendered through, so this is how a caller sees that another agent holds it before its own share is quietly replaced -- and, when exclusive is true, id is the key a hold of its own states to be admitted, and the key renderer.atmospherics.release hands the air back by. id and name are nil for a hold that stated no key and that the engine attributes to no agent.

Returns { id: string?, name: string?, share: number, exclusive: boolean }? — { id, name, share, exclusive } for the standing hold, or nil when the authored air is what reaches the image.

local who = renderer.atmospherics.owner()
if who ~= nil and who.exclusive then print(who.name, "holds the air at", who.share) end

globals/renderer/atmospherics/release

renderer.atmospherics.release(owner: string) -> number

Hand the air back by the key its holds were taken under, and report how many came off. A hold stands until its release is called, and the release is a closure the call that took the hold holds: a caller that takes a hold in one call and comes back in another, and a task that ends between the two, both leave the air pinned with nobody holding a release for it. Naming the key is how the authored air comes back, and how a caller refused by an exclusive hold takes one over.

Parameters

owner string — The key the holds to release were taken under -- what owner stated when they were taken, which renderer.atmospherics.owner reports.

Returns number — How many holds came off the stack.

renderer.atmospherics.release("albedo-read")

renderer.atmospherics.share() -> number

The share of the authored air that reaches the image: the innermost hold's share while one stands, and 1 otherwise. A system that packs a medium multiplies its extinction — aerial, a fog density — by this, and a hold then reaches that medium however it is being driven.

Returns number — A number in [0, 1]. 1 when nothing holds.

local density = state.density * renderer.atmospherics.share()

globals/renderer/blendedBatching

renderer.blendedBatching() -> boolean

Whether blended neighbours sharing a draw key draw together.

Returns boolean

globals/renderer/bounds/clear

renderer.bounds.clear(id: string) -> boolean

Withdraw the box an entity published, so it stops contributing to the entity's reported extent.

Parameters

id string — Entity id.

Returns boolean — True when there was a published box to withdraw.

renderer.bounds.clear(id)

globals/renderer/bounds/set

renderer.bounds.set(id: string, min: any?, max: any?) -> boolean

Publish the local-space box an entity's content-drawn geometry occupies. entity:bounds() and entity:hierarchyBounds() union it with whatever mesh geometry the entity has, each carried out of its own local space, so framing a camera on the entity frames what a feature actually draws.

Parameters

id string — Entity id.
min any (optional) — Local-space minimum corner — a Vec3 table or a 3-element array.
max any (optional) — Local-space maximum corner, same shape as min.

Returns boolean — True when the box was stored; false for a non-finite or inverted box.

renderer.bounds.set(id, cloud.boundsMin, cloud.boundsMax)

globals/renderer/captureView/channelId

renderer.captureView.channelId(name: string) -> number?

The debug channel a registered view draws on — what a feature passes as its pass debugChannel. Nil when no view is registered under name.

Parameters

name string — The view name.

Returns number? — The channel number or nil.

local debugChannel = renderer.captureView.channelId("lightmap") -- a render feature passes it as an enqueued pass's `debugChannel`

globals/renderer/captureView/list

renderer.captureView.list() -> { any }

Every registered capture view as { name, channel, description } records — what backs the discoverability of capture pass=<name> and the unknown-view error's suggestion list.

Returns { any } — An array of view records.

for _, v in ipairs(renderer.captureView.list()) do ... end

globals/renderer/captureView/ready

renderer.captureView.ready(name: string) -> boolean

Whether a registered view can draw yet. A view's passes are enqueued from the moment its render feature first runs, but they are skipped while the materials they name have no pipeline — their shader is still compiling — so for the first frames of a session a camera bound to the view renders the ORDINARY view into its target, and the image gives no sign of it. This reports the difference, and reports it before any camera is on the view, so it is answerable for the first camera bound to one. A view that registered a ready of its own is also held to it, so a view whose feature is not running, or whose draw has not landed, reads false. False for an unregistered name.

Parameters

name string — The view name.

Returns boolean — Whether this view's passes have resolved everything drawing needs.

repeat task.wait() until renderer.captureView.ready("zfighting")

globals/renderer/captureView/register

renderer.captureView.register(name: string, config: any?) -> number

Register (or update) a content capture view under name and return the debug CHANNEL number assigned to it. A render feature gates its pass to this channel (debugChannel = channel) so the pass draws only when a capture selects the view. Idempotent: re-registering the same name keeps its channel.

Parameters

name string — The view name, selected via capture pass=<name>.
config any (optional) — { description?, ensure?, ready?, warmup?, renderLayers? }. ensure is called before a capture of this view so the feature that draws it is live (e.g. create it on demand). ready is the view's own statement of whether what it draws is in place: a function(): (boolean, string?) answering true, or false and the reason. A capture of the view waits on it after ensure and before it renders, so a view whose draw lands over the frames after ensure (an entity it spawns, a feature that reads the state ensure built on its next frame, a material still compiling) is drawn rather than captured empty, and a view that stays unready is refused with that reason. warmup is how many present frames a capture lets the view accumulate, once it is ready, before it reads: set it when the feature retains prior-frame state (a temporal diff) so the first capture reads a warm result. renderLayers is the layer spec a capture of this view uses when the caller named none. A view that draws its own geometry and wants the scene's kept out of the frame (and out of the depth buffer it tests against) names only its own layer.

Returns number — The channel number assigned to the view.

local ch = renderer.captureView.register("lightmap", { description = "...", ensure = fn })

globals/renderer/captureView/resolve

renderer.captureView.resolve(name: string) -> any

Resolve a capture view by name to its { channel, ensure, ready, description, warmup, renderLayers } record, or nil when no view is registered under name.

Parameters

name string — The view name.

Returns any — The view record or nil.

local v = renderer.captureView.resolve("lightmap")

globals/renderer/captureView/unregister

renderer.captureView.unregister(name: string) -> boolean

Withdraw a capture view. A subsequent capture pass=<name> no longer resolves to it (falls through to the unknown-view error).

Parameters

name string — The view name.

Returns boolean — True when a view was registered under name.

renderer.captureView.unregister("lightmap")

globals/renderer/clearShadowHero

renderer.clearShadowHero() -> boolean

Release the hero caster, so every receiver samples the level its footprint selects again.

Returns boolean — Whether a caster was registered.

renderer.clearShadowHero()

globals/renderer/clearShadowProxy

renderer.clearShadowProxy(mesh: string?) -> number

Stop proxying mesh, so it rasterizes its own geometry into shadow pages again. Called with no argument, drops every registration.

Parameters

mesh string (optional) — The mesh to stop proxying. Omit to clear all of them.

Returns number — How many registrations were removed.

renderer.clearShadowProxy(statueMesh)
print(renderer.clearShadowProxy(), "proxies dropped")

globals/renderer/clusterErrorPx

renderer.clusterErrorPx(px: number?) -> number

The cluster cut's error threshold, in pixels. Every static mesh draws the coarsest clusters whose simplification error projects to at most this many pixels in the view drawing it — measured against that view's own pixel height and projection, so a small render target draws coarser geometry than the full-screen view does from the same place. Lower draws finer geometry at more cost; higher draws coarser geometry. A client graphics setting. Default 1.

Parameters

px number (optional) — A finite number above 0. Omit to read the threshold.

Returns number — The threshold in effect.

renderer.clusterErrorPx(0.5) -- finer geometry on this machine

globals/renderer/collect

renderer.collect() -> RuntimeCollection

Release every runtime texture, material, mesh and render feature nothing holds: no handle a script still reaches, no live owner, no reference from live engine state, no asset backing it, no hold. A root scene load runs this once the new scene stands, so what the previous scene's content created and nothing still wears goes with that scene; calling it directly collects at any other moment. A session material's handle counts as reached while the entity it was keyed for stands, and stops counting once that entity is gone. It reaches the GPU textures the device holds beside the registry's own: a texture the cache loaded for an asset goes once nothing live names it and is read back from that asset the next time something asks for it, while one no asset answers for stays, there being nothing to read it back from — a render pass's own target, a colour swatch, an atlas the engine built. A texture the ASSET path uploaded and whose asset has since been removed has nothing to come back from either, and the collection decides about it from its holders the way it does about every other resource: a handle a script still reaches, a live owner, a reference from live engine state, a hold. Features go first, then materials, then meshes, then textures, so a texture only a released material named goes with the material. Runs a full garbage collection first, so a handle nothing reaches counts as let go, and yields for the frame the census runs on. A handle the calling function still has in a variable — or in a temporary it has not overwritten — is one a script reaches, so a resource created in the function that collects is let go by the next collection rather than this one. A runtime mesh needs no call: once the garbage collector has taken its last handle, the engine judges it against the same holders, again whenever what displays it changes or its hold is released, and every few hundred frames while something holds it, and releases it once nothing does. Returns once what it released has left the device: the GPU objects of the textures and meshes it let go are freed over the frames after the call, and it waits for those frames (up to 60 drawn frames), so renderer.texture.list, renderer.mesh.list, renderer.material.list, renderer.textureMemory() and renderer.gpuMemory() read after it in the same script no longer count anything it reports released. landed says whether every release left the device within that wait; false names a texture or mesh something put back on the device before it could be seen gone, which renderer.references explains. Code that cannot yield (a metamethod, onDestroy, a module's top level while require runs it) is answered at once, without the wait: there landed is false while a release is still on its way, and the listings and totals stop counting it over the drawn frames that follow.

Returns RuntimeCollection — { released = { texture, material, mesh, feature }, kept, entries, landed }: the counts released per kind, how many stayed, every resource's status with action = "released" | "kept", and whether every release had left the device when the call returned.

local c = renderer.collect() print(c.released.texture, c.kept)
local c = renderer.collect() print(c.landed, renderer.gpuMemory().meshes)

globals/renderer/compiledShaders

renderer.compiledShaders() -> { string }

Every key renderer.compiledSource answers for — one per shader key (guid@checksum) or variant program a compile has run under, whether it succeeded or failed, for as long as the shader that compiled under it stands: a removed .shader takes its keys with it. What makes the composed-source surface enumerable rather than something to guess a key for.

Returns { string } — An array of keys, sorted.

for _, name in renderer.compiledShaders() do print(name) end

globals/renderer/compiledSource

renderer.compiledSource(shader: string) -> string?

The WGSL the shader compiler received under one name, exactly as it received it — the composed module, which is what a compile error's line numbers and handle indices are positions in. Answers for a shader key (guid@checksum), a program from renderer.shaderVariants(), or any reference to a shader asset (identity, path, guid), which reaches the key of the version that reference resolves to now; for a shader that declares no features, and for a shader whose compile FAILED, which is the case it exists for: a message about a function body carries a position and nothing else, and the text that position is in is this. The failed text stands for as long as shaderRef:compileStatus() reports that failure.

Parameters

shader string — A shader key, a shaderVariants() program name, or a reference to a shader asset.

Returns string? — The composed WGSL, or nil when no compile has run for it or its shader was removed.

local status = asset.resolve("myShader", "shader"):compileStatus()
if status.status == "failed" then
print(status.error)
print(renderer.compiledSource("myShader"))
end

globals/renderer/compositeSize

renderer.compositeSize() -> { width: number, height: number }

The size of the image the post-scene phases worked on in the last presented frame — the target the UI composites onto, which every pass after the scene reads as @scene.color and writes into, and which a screenSpace = "composite" render target follows. While the renderer presents the viewport itself that is the display's own size, whatever fraction of it the scene rasterized at; while a UI viewport panel owns the presentation it is the size the scene rasterized at, since the panel draws the scene target at its own rect and nothing upscales before the composite. Both read 0 before a frame has drawn.

Returns { width: number, height: number } in pixels.

local c = renderer.compositeSize()

globals/renderer/cullStats

renderer.cullStats() -> {

What the last completed frame decided to draw. total renderables went into the frustum test, culled fell outside it and visible survived. Of those, occlusion culling measured occlusionTested against the depth pyramid and proved occlusionCulled were entirely behind other geometry — both 0 while renderer.occlusionCulling() is false. A renderable the pyramid has no say over — one that laid no depth in the visibility raster, one whose bounds were never recorded, one straddling the near plane — is measured against nothing and counted in neither, so the gap between visible and occlusionTested reads how much of the frame the test could speak for.

This answers for the main camera. What the shadow page cull did with the frame's casters for each light is on that light's row in renderer.shadowViews().

views is the engine cull's own count, one row per camera view that rendered: view is "main" for the main camera and the camera's entity id for an offscreen one, kind says which, and frame is the frame the counts describe. The main view's row is first. drawn is whether a render drew the view in that frame: a frame the renderer declined still ran the cull's first test, so its row carries that test's counts and no second one, for a picture nothing rasterized. The counts are read back from the GPU and handed out oldest first, one frame per render, so a script reading once a frame reads the frames in the order they ran. A frame's counts wait at most two renders to be handed out: a row describes a frame as many frames back as the read took and at most two more, and when more frames' counts come back at once than that lets wait, the oldest of them are the frames a script does not read. instancesTested renderables were tested; instancesCulled breaks the rejections down by reason (hidden, layer, range, frustum, screenSize, and occludedPass1/occludedPass2 for occlusion against the depth of the frame that last drew the view and then this frame's); instancesRecovered came back in the second test. Then the clusters of the survivors: clustersTested, rejected by the level-of-detail cut (clustersLodRejected), the frustum (clustersFrustumCulled), facing away (clustersConeCulled) or occlusion (clustersOccludedPass1, of which clustersRecovered came back), and clustersDrawn / trianglesDrawn. binsPending counts material pipelines still compiling, whose clusters wait a frame; overflow counts clusters that did not fit the view's lists. total/culled/visible and occlusionTested/occlusionCulled above are the main view's count for the renderables drawn outside the cluster lists.

Returns { total: number, culled: number, visible: number, occlusionTested: number, occlusionCulled: number, views: { { [string]: any } } }

local s = renderer.cullStats(); print(s.visible - s.occlusionCulled, "drawn")
for _, v in renderer.cullStats().views do if v.view == "main" then print(v.clustersDrawn, "clusters,", v.instancesCulled.frustum, "outside the frustum") end end

globals/renderer/debugPass/builtins

renderer.debugPass.builtins() -> { string }

The built-in debug-pass names, one per channel in channel order — the engine's built-in pass vocabulary (final, albedo, normal, depth, …).

Returns { string } — An array of built-in pass names.

for _, n in ipairs(renderer.debugPass.builtins()) do ... end

globals/renderer/debugPass/channel

renderer.debugPass.channel(name: string) -> number?

The channel a debug-pass NAME renders on: a built-in pass, else a content capture view registered via renderer.captureView. Nil when the name is neither — the signal a selector uses to reject an unknown pass.

Parameters

name string — A debug-pass name (e.g. "normal", "depth", "lightmap").

Returns number? — The channel number, or nil for an unknown name.

local ch = renderer.debugPass.channel("normal")   -- 7

globals/renderer/debugPass/list

renderer.debugPass.list() -> { string }

Every selectable debug-pass name: the built-in passes plus every registered content capture view. What a debug-pass selector offers.

Returns { string } — An array of pass names.

local passes = renderer.debugPass.list()

globals/renderer/debugPass/name

renderer.debugPass.name(channel: number) -> string?

The canonical NAME for a debug channel: a built-in pass name for a built-in channel, else a registered capture view's name. Channel 0 is "final" (the lit image). Nil when no pass owns the channel.

Parameters

channel number — The channel number.

Returns string? — The pass name, or nil.

local name = renderer.debugPass.name(7)   -- "normal"

globals/renderer/destroy

renderer.destroy(handleOrKind: any?, id: string?) -> boolean

Free the GPU resource a renderer resource holds (the GPU-destroy verb). Takes any of the forms that name it: the handle a create returned, routed by its category so one call releases a mixed set of handles; the id a listing hands out, whose kind is read back off what the renderer holds under it — the runtime registry, the material definitions, the live features, and the device itself for an asset's own texture or mesh; or the kind with the id beside it, the shape renderer.hold and renderer.references take, which is what names the kind for an id two of them answer to. An id nothing holds anything under releases nothing and answers false. The on-disk asset, if any, is untouched. A CPU handle's :unload() frees the CPU copy separately.

Parameters

handleOrKind any (optional) — A MeshHandle, TextureHandle, MaterialHandle or feature handle; the id itself; or the kind ("texture", "material", "mesh", "feature") with the id as the second argument.
id string (optional) — The guid or registry key, when the first argument is a kind.

Returns boolean true if a GPU resource was known under the id.

renderer.destroy(meshHandle); renderer.destroy(materialHandle)
renderer.destroy(renderer.texture.list()[1].guid)
renderer.destroy("mesh", guid)

globals/renderer/deviceGeneration

renderer.deviceGeneration() -> number

Which render device this process is on, counted from the first.

A render device is lost when a driver resets, when the GPU is taken away, or when a browser reclaims a WebGPU context. The engine answers by building another device and re-deriving this session's resources onto it, and this number moves by one each time it does. Anything held across frames that was built from a GPU resource records this beside it and remakes it when the two differ; engine.onDeviceRebuilt is the hook that fires when it moves.

Returns number — The current device generation, counting from 1.

local generation = renderer.deviceGeneration()
engine.onDeviceRebuilt(function(g) print("device " .. g) end)

globals/renderer/deviceState

renderer.deviceState() -> string

Whether the render device this process draws through is the one it is using, one it is replacing, or one the platform has refused every quick attempt to replace.

"ready" is a live device. "rebuilding" is the window between a device reporting itself lost and another being in place: every GPU resource built from the old one is invalid, the frames in that window draw nothing, and anything reaching the GPU refuses. "abandoned" is after every quick attempt at another device was refused (a browser whose GPU process crashed twice answers no adapter for minutes): nothing is drawn, the engine log names the platform's last answer, and the engine keeps asking at a slower pace, so the state returns to "ready" on its own once the platform hands a device over. Restarting the engine (reloading the page on the web) builds one from scratch.

Work that spans the device — build a render target, draw into it, read it back — reads this to tell an operation that failed because the device went out from under it, which is worth doing again once renderer.deviceGeneration() moves, from one that failed on its own terms. The loss is reported before the next device exists, so the two readings answer different halves: this one says a replacement is coming, the generation says it arrived.

Returns string — "ready" | "rebuilding" | "abandoned".

if renderer.deviceState() == "rebuilding" then return end

globals/renderer/drawDiagnostics

renderer.drawDiagnostics() -> { DrawDiagnostic }

Every renderable that is NOT drawing what its material says — the one call for "why does this surface look wrong". Three states land here: a surface rendering as the magenta placeholder (substituted), one the renderer could bind nothing for at all (outcome = "skipped"), and one drawing a program whose most recent compile FAILED (stale), which is what a shader edited into brokenness looks like — the pipeline its last good compile built keeps drawing, so the picture is intact and answers to none of the edits since. Each row names the entity, the program asked for, the program bound, programStatus — the compile gate's word about the program the material NAMED — and the one cause from shaderCompileFailed / shaderNotRegistered / shaderNotCompiledYet / noGbufferEntry / renderStateKeyNotBuilt / noPipelineForTarget / unshaded, with the compiler's own message in detail or programError. Covers every renderable the renderer holds, whether or not a camera reached it: a row with observed = false and outcome = "notDrawn" carries the renderer's own resolution for one this frame drew nowhere, so a broken surface off-screen is reported the same as one in frame. An empty result means every renderable the renderer holds is drawing the program its material named and that program compiles. Answers on the deferred path as well as forward, and in edit mode as well as play.

Returns { DrawDiagnostic }

for _, d in renderer.drawDiagnostics() do print(d.entity, d.reason, d.detail) end
if #renderer.drawDiagnostics() == 0 then print("every surface is drawing what it says") end

globals/renderer/drawStats

renderer.drawStats() -> {

What the last completed frame actually submitted. draws counts every geometry draw call the frame issued — the camera's passes, each shadow view a shadow-casting light adds, and whatever a render feature draws — and instances counts the instances those draws covered. The pair is what separates one draw carrying five hundred instances from five hundred draws carrying one each, so it reads how well the scene batches rather than how many objects are in it.

compacted is how many of those instances the frame planned through draws whose instance count the GPU decides: the culler's own per-object answers packed into a dense run, so an object it rejects is absent from the draw instead of collapsing to nothing in the vertex stage. compactedDrawn is how many of them survived, counted on the GPU as it packed them — a pass that then skips a whole draw over its own layer or visibility answer leaves that draw's instances in both numbers.

The plan is made over the populations the frame draws, and the tests answer which of their instances the packing keeps. That packing runs before any pass has resolved the depth occlusion culling is tested against, so on its own it reads the frustum and screen-size answers alone. With setOcclusionCulling armed the frame packs the same plan a second time once the test has answered, and compactedDrawn then counts what came through occlusion as well.

compactedDrawn comes back from the buffer the GPU wrote, so it describes a frame that has finished while compacted describes the most recent plan, and it holds the last count the GPU wrote until another arrives — a frame that compacts nothing reads compacted 0 beside the count from the last frame that did. In a scene standing still the gap between the two is the front-end work culling removed.

materialBinds is how many times the frame's geometry passes set a material group's textures (group 1), and materialBindsElided how many times a pass reached that decision and found the group already bound. On the bindless backend every material group binds the one texture group of the frame, so a pass binds it once; on buckets a pass binds once per group it draws (see renderer.materialTable().groups). Their sum is how many times the decision was reached — once per unit of geometry submitted, which sits at or below draws, since a mesh of several primitives draws once per primitive under one set of binds. materialExtraBinds and materialExtraBindsElided are the same pair for group 2, the storage buffers a shader declares, bound per distinct buffer set, which only the shaders that declare them ever bind.

pipelineBinds and pipelineBindsElided are the same pair for the pipeline itself: how many times the frame's geometry passes set one, and how many times a pass reached that decision and found the pipeline it wanted already bound. Which pipeline a unit needs follows its shader, its material's render state and its mesh's vertex layout together, so a scene whose units share all three costs one set for the run of them, while units differing in any one of the three each pay their own. Their sum is how many units reached the pipeline decision, which sits at or above what the material pair reports: a unit the pass settles a pipeline for and then abandons — one whose material group resolved to nothing — counts here and never reaches the material decision.

Every figure here is the whole frame's. renderer.shadowViews() carries the main camera's own share of compacted and compactedDrawn.

clusters and clusterTriangles are the clusters and triangles the camera views drew from the cluster lists, from the GPU's own count.

drawsByPass splits draws by the part of the frame that submitted them — visibility (the id raster: one draw per raster bin per Hi-Z phase in each camera view, submitted whether or not the GPU cull left anything in the bin), resolve (one draw per resolve draw in each camera view: one per shader, and on the array-bucket backend one per shader and material group), legacy (geometry drawn outside cluster storage: skinned, morphed and compute-written meshes, line topology, shaders that write their own vertex stage), blended, exception (opaque materials whose depth state the visibility buffer cannot express), featureGeometry, featureDraw, shadow and splat (one draw per splat buffer a render feature draws) — and the nine sum to draws.

instances counts instances named by draws whose size the CPU decided; the visibility raster's work is clusters and clusterTriangles, which the GPU decided. pixelsResolved is how many pixels the resolve shaded, summed over camera views, and pixelsByShader the same split by shader. clusters, clusterTriangles, pixelsResolved and pixelsByShader come back from the GPU and describe the frame numbered frame, which is earlier than the frame draws describes. frame is the renderer's own frame number, counted from 0 when the renderer was built (again after a device loss), a count of its own that compares only with another drawStats().frame.

Returns { draws: number, instances: number, compacted: number, compactedDrawn: number, materialBinds: number, materialBindsElided: number, materialExtraBinds: number, materialExtraBindsElided: number, pipelineBinds: number, pipelineBindsElided: number, clusters: number, clusterTriangles: number, drawsByPass: { visibility: number, resolve: number, legacy: number, blended: number, exception: number, featureGeometry: number, featureDraw: number, shadow: number, splat: number }, pixelsResolved: number, pixelsByShader: { [string]: number }, frame: number }

local d = renderer.drawStats(); print(d.instances / math.max(d.draws, 1), "instances per draw")
print(renderer.drawStats().compactedDrawn, "of", renderer.drawStats().compacted, "survived the cull")
local d = renderer.drawStats(); print(d.materialBinds, "material group binds over", d.materialBinds + d.materialBindsElided, "units")
local d = renderer.drawStats(); print(d.pipelineBinds, "pipeline sets over", d.pipelineBinds + d.pipelineBindsElided, "units")
local d = renderer.drawStats(); print(d.clusters, "clusters,", d.clusterTriangles, "triangles from the cluster lists")
local d = renderer.drawStats(); print(d.drawsByPass.visibility, "id-raster draws,", d.drawsByPass.resolve, "resolve draws")
local d = renderer.drawStats(); for shader, px in d.pixelsByShader do print(shader, px) end

globals/renderer/feature/configure

renderer.feature.configure(handleOrRef: any?, settings: { [string]: any }) -> { [string]: any }

Lay settings over a live feature's knobs. Any omitted knob keeps its value; a knob the feature never declared, or a value of the wrong shape, is refused by name. The feature reads the result as ctx.settings from its next frame.

Parameters

handleOrRef any (optional) — A RenderFeatureHandle, its guid, or the feature's ref.
settings { [string]: any } — The knob values to apply.

Returns { [string]: any } — The feature's full settings now in force.

renderer.feature.configure(h, { intensity = 0.5, quality = "medium" })

globals/renderer/feature/create

renderer.feature.create(ref: any?, opts: (string | { guid: string?, settings: { [string]: any }? })?) -> any

Instantiate a render feature so the engine calls its render(ctx) hook every frame. ref is an AssetRef<renderFeature> whose init.luau returns { setup?, render, teardown? }. Returns a live RenderFeatureHandle (its guid is the stable id, same as mesh/texture handles); tear it down with renderer:destroy(handle). Pass guid to assign a specific id.

Parameters

ref any (optional) — An AssetRef<renderFeature>, or a string identity/guid resolved via asset.resolve(ref, "renderFeature").
opts (string | { guid: string?, settings: { [string]: any }? }) (optional)

Returns any — A RenderFeatureHandle.

local h = renderer.feature.create(asset.resolve("acrylic", "renderFeature"))
local h = renderer.feature.create("acrylic")
local h = renderer.feature.create("@builtin::systems.reflections.ssr", { settings = { intensity = 0.8 } })

globals/renderer/feature/destroy

renderer.feature.destroy(handleOrGuid: any?) -> boolean

Tear down a live render feature by its RenderFeatureHandle OR its guid string — the by-id path for when the handle was lost (e.g. across execute calls). Same effect as renderer.destroy(handle). Returns true if a feature was live under that id.

Parameters

handleOrGuid any (optional) — A RenderFeatureHandle or its guid string.

Returns boolean

renderer.feature.destroy("eb623975-d8bd-47a1-a0a4-5c0e56238e2f")

globals/renderer/feature/find

renderer.feature.find(ref: any?) -> any

The live handle of the feature running ref — an AssetRef or a renderFeature identity — or nil when none is. The check a caller makes before creating one, so a feature already live is configured rather than stacked.

Parameters

ref any (optional) — An AssetRef, a renderFeature identity, or a guid.

Returns any — The RenderFeatureHandle, or nil.

local h = renderer.feature.find("@builtin::systems.reflections.ssr")

globals/renderer/feature/list

renderer.feature.list() -> { { guid: string, identity: string, settings: { [string]: any } } }

List every render feature currently live (running its render(ctx) each frame). Each entry is { guid, identity, settings } — the guid is the same id a RenderFeatureHandle carries, so you can tear a feature down by guid even after losing its handle (e.g. across separate execute calls), and settings is every knob the feature declares at the value in force.

Returns { { guid: string, identity: string, settings: { [string]: any } } }

for _, f in renderer.feature.list() do print(f.identity, f.guid) end

globals/renderer/feature/schema

renderer.feature.schema(ref: any?) -> { [string]: any }

The settings a .renderFeature declares, with each knob's descriptor, read without running it: what a caller may pass to create / configure.

Parameters

ref any (optional) — An AssetRef or a renderFeature identity.

Returns { [string]: any } — { [name]: descriptor }, empty for a feature that declares none.

for name, desc in pairs(renderer.feature.schema("@builtin::systems.reflections.ssr")) do print(name, desc.default) end

globals/renderer/feature/settings

renderer.feature.settings(handleOrRef: any?) -> { [string]: any }?

A live feature's settings as they stand: every declared knob at its default or at what create / configure set. A copy — writing to it changes nothing; configure does.

Parameters

handleOrRef any (optional) — A RenderFeatureHandle, its guid, or the feature's ref.

Returns { [string]: any }? — The settings table, or nil when no such feature is live.

local s = renderer.feature.settings(h); print(s.intensity)

globals/renderer/feature/shaded

renderer.feature.shaded() -> { [string]: number }

How many pixels each fragment pass a render feature enqueued shaded on the last drawn frame, keyed by the pass's shader/effect name. A fragment pass draws one triangle over its target, so it shades the whole screen whatever its effect actually reaches — unless it declares bounds on the pass spec, the world-space box its effect stays inside, in which case it shades the rectangle that box projects into for the camera drawing it and is skipped for a camera that cannot see the box at all. This is the reading that says which of the two a pass is: it moves when the effect moves, and a pass absent from it shaded nothing. Summed over every camera the frame drew.

Returns { [string]: number } — { [shader: string]: number } — pixels shaded, last drawn frame.

local px = renderer.feature.shaded(); for name, n in pairs(px) do print(name, n) end

globals/renderer/featureTexture/configure

renderer.featureTexture.configure(width: number, height: number, layers: number)

Size the shared feature-texture array — the layers a surface shader reads through zero_feature_texture(uv, layer), and the layers a SpotLight projects through its cone via cookieLayer. Layers are rgba16f. A call for the size the array already has is left alone. One that changes the size reallocates, and the replacement is zeroed — so it empties every layer in the array, including the layers other features and other cookies own. renderer.featureTexture.state() reports the extent and the layers holding content, which is how a feature re-fills the layer a resize took from it.

Parameters

width number — Layer width in pixels.
height number — Layer height in pixels.
layers number — How many layers the array holds.

renderer.featureTexture.configure(512, 512, 4)

globals/renderer/featureTexture/setLayer

renderer.featureTexture.setLayer(layer: number, textureKey: string, x: number, y: number)

Copy a texture already on the GPU into one layer of the shared array, its top-left corner at (x, y) — GPU to GPU, with no readback. Several small images pack into one layer by calling this once per image at different offsets. The source must be rgba16f and fit at that offset.

Parameters

layer number — Which layer of the array to write into.
textureKey string — The source texture's name — the one it was created under. A compute.createStorageTexture2D target, a compute.createTextureHistory pair (its current side), and a texture a compute.copyBufferToTexture wrote all answer to the name they were given.
x number — Left edge of the destination rectangle, in pixels.
y number — Top edge of the destination rectangle, in pixels.

compute.createStorageTexture2D("gobo", { width = 256, height = 256, format = "rgba16f" })
renderer.featureTexture.setLayer(0, "gobo", 0, 0)

globals/renderer/featureTexture/state

renderer.featureTexture.state() -> {

What the shared feature-texture array is right now: the extent every layer carries, and filled, the ascending 0-based indices of the layers a setLayer has landed in since the array was last sized. One array is shared by every feature and every light cookie in the scene, and it has no allocator, so this is the call that tells a feature whether the array it sized and filled is still the array it is writing into — a configure that changed the size reallocates and zeroes every layer, and the layer it emptied leaves filled without it. Measured off the renderer at the end of the last rendered frame, so a configure or setLayer issued this frame reads back on a later one.

What this describes is the array a shader samples. The source texture a setLayer copied FROM is a GPU resource of its own and keeps the bytes it was written with for as long as it lives, so filled is the reading that answers whether the layer behind a cookieLayer is live right now.

fills counts, per layer, the setLayers that have landed in it since the array was last sized: fills[layer + 1] for a 0-based layer. A fill of a layer that already holds content leaves filled as it was, so the count moving is how a feature re-filling its own layer reads its fill arriving.

Returns { width, height, layers, filled, fills }

local ft = renderer.featureTexture.state()
print(("feature textures: %dx%d over %d layers"):format(ft.width, ft.height, ft.layers))
-- Re-fill the cookie layer this module owns if anything emptied it.
if ft.width ~= myWidth or table.find(ft.filled, myLayer) == nil then
print("refill the cookie layer", myLayer)
end

globals/renderer/framePacing

renderer.framePacing() -> FramePacing?

How far the CPU is allowed to run ahead of the GPU, and what holding it there cost the frame just finished. Submitting work to the GPU returns before the GPU has done it, and everything that submission holds — its staging allocations, its bind groups, its command buffer — stays alive until it completes. A frame that asks for more work than the GPU finishes in a frame's time therefore leaves that behind it, and unbounded that is memory growth rather than a lower frame rate.

framesInFlight is how many submitted frames have not reported done through the queue's completion signal, held under maxFramesInFlight: a device that keeps up reads under the bound, one that is behind reads at it. It counts submissions, which is its own quantity — how many presented images the swapchain permits in flight is a separate setting. mechanism names how that bound is enforced here: submission-wait waits for the frame that many frames back and reports the wait in waitMs, so a paced frame costs latency and still draws; submitted-work-done counts outstanding frames off the queue's completion signal and declines to start a frame while the bound is met, counting those in pacedFrames and leaving the last presented image up. submittedFrames counts the frames that were admitted and submitted, so it rises for as long as the renderer is producing frames — which is what tells a renderer running slowly under a tight bound from one that has stopped. stalled reads true while the pacer has given the bound up rather than hold the image indefinitely, on either mechanism: the counting one stands down when the completion signal goes quiet across a whole run of declined frames, and the waiting one when a wait on a frame reaches its ceiling with the GPU still holding it. Both clear on the first word the GPU sends after that. A signal that is merely late — a GPU finishing frames more slowly than the loop asks for them — keeps the bound, so stalled answers for a GPU that went silent. standDowns counts those over the session, which is how a session that lost the signal and got it back (stalled = false with standDowns above zero) reads apart from one that never lost it.

producing is whether the renderer is drawing frames at all. A headless renderer draws into an offscreen framebuffer that nothing presents, so its image reaches a reader only through something that copies it out: it draws while a consumer is asking — an MCP call in flight, a queued texture readback, a recording, a frame-egress session — and declines the frames between two asks, counting them in idleSkippedFrames. Every other renderer stat answers with the last frame that drew, so producing is what separates a live reading from a frozen one. A windowed renderer presents every frame it draws and reads producing = true throughout.

presentMode is what the surface presents with and presentModes what it offers; both are empty of meaning on a headless renderer, which never presents. A desktop presents mailbox unless told otherwise, and Android presents fifo, which shows each frame on a vsync of the panel and makes no frame the panel never shows.

The rest describe the rate frames run at. displayRefreshHz is the refresh rate of the display the frames are shown on, where the platform reads it: the phone's panel, the monitor a desktop window is on, the cadence of a browser's animation frames. A headless engine presents to no display and states none. frameRateCap is the cap renderer.setFrameRateCap set, targetFps the rate that cap delivers and refreshesPerFrame the refreshes each frame stays on screen under it, all three absent while no cap is set. deliveredFps is the rate the frame loop has run frames at over about the last twenty frames.

Returns FramePacing? — { framesInFlight, maxFramesInFlight, pacedFrames, submittedFrames, waitMs, mechanism, stalled, standDowns, producing, idleSkippedFrames, presentMode, presentModes, displayRefreshHz?, frameRateCap?, refreshesPerFrame?, targetFps?, deliveredFps? }, or nil before the renderer has drawn a frame

local p = renderer.framePacing()
print(("%d of %d frames in flight, paced by %s"):format(p.framesInFlight, p.maxFramesInFlight, p.mechanism))

globals/renderer/getRaytrace

renderer.getRaytrace() -> boolean

Whether ray tracing is currently enabled.

Returns boolean

globals/renderer/gpuMemory

renderer.gpuMemory() -> GpuMemory

Where the renderer's GPU memory went at the last completed frame — the call to reach for when something is holding memory and you do not know what.

Three figures answer three different questions, and they are meant to be read against each other:

The categories — shadow, textures, meshes, instances, compute, summing to categorised — are the renderer's own accounting of what it asked for on purpose. Always present, on every backend. textures is what the resident textures occupy, and on the texture-array backend also the slack its shared arrays hold beyond them (renderer.materialTable().bucketSlackBytes).
allocator is the device allocator's ledger, with a row per creation label largest first, which is what names an allocation no category claims. A render target's colour texture is reported under what holds it now: the target drawing into it, or released_scratch[<width>x<height>].color while it waits for the next target of its size. It exceeds categorised by the per-frame render targets and the scratch nothing categorises, and on the texture-array backend, for the frame after an array grows or is repacked, by the array it replaced. The allocator hands memory out from blocks it reserves whole from the device and returns a block only once nothing is left in it, so reservedBytes runs above allocatedBytes by what those blocks hold unused; blocks lists them emptiest first with the labels that keep each one alive, and emptyBytes plus slackBytes is that distance exactly — the pool held in empty blocks, and the room pinned inside blocks something still sits in.
driver.deviceLocalBytes is what the graphics driver charges this process, out of the kernel's own accounting. It is the biggest of the three and the one that fills a card, because it also holds the swapchain, the images the driver keeps on the renderer's behalf, and the rounding to whole pages and heap blocks that neither figure above sees. Read it when the question is how much of the machine's GPU this engine is using; read the two above when the question is what the engine spent it on. A platform with no per-process accounting reports available = false and the reason.
driver.outsideAllocatorBytes is that charge less everything the allocator reserved — what the driver holds on its own account, and the one figure here nothing releases: a dropped pipeline, another scene and renderer.collect() all leave it where it is, and it falls when the device is destroyed. Read it when a session's device memory has grown and no ledger row accounts for the growth.

compute is what the compute subsystem holds; compute.observe() names each of those resources and what it costs. arenas states the part of it the pages runs stand in hold, and what those pages' bytes are spent on. renderTargets counts the offscreen render targets the renderer holds at that frame, which is what says a renderer.destroy has been applied rather than queued. textureBudgetBytes is the texture budget in force and pressure the memory pressure level that shapes it: under "critical" the budget reads half the device's, and renderer.memoryPressure() says what the newest report released.

Returns GpuMemory — The accounting — see GpuMemory. The category figures are zeroed until the renderer has published its first frame; driver is read as the call runs and answers from the first.

local m = renderer.gpuMemory()
print(("shadows %.1f MiB"):format(m.shadow.total / 1024 / 1024))
-- how much of the card this engine is holding:
if m.driver.available then
print(("driver %.1f MiB"):format(m.driver.deviceLocalBytes / 1024 / 1024))
else
print("no driver figure: " .. m.driver.reason)
end
-- the biggest allocation in the engine:
local top = m.allocator and m.allocator.byLabel[1]
print(top and top.label, top and top.bytes)
-- the block held for the least, and what pins it:
local block = m.allocator and m.allocator.blocks[1]
if block and block.allocationCount > 0 then
print(block.size, block.allocatedBytes, block.byLabel[1].label)
end

globals/renderer/hold

renderer.hold(handleOrKind: any?, id: string?) -> boolean

Pin a runtime resource for the session. A held texture, material, mesh or render feature survives every collection — the one a root scene load runs and a direct renderer.collect() alike — until renderer.release lets it go or its destroy frees it. It is the way to keep an ad-hoc resource across the scenes that come and go under it. A hold keeps the resource in the registry; a mesh's GPU buffers are governed by what draws it, parked as a CPU definition when the last instance naming it goes and brought back when one names it again, so renderer.mesh.isResident(guid) is the separate question about the buffers.

Parameters

handleOrKind any (optional) — The resource's handle, or its kind ("texture", "material", "mesh", "feature") with the guid or key as the second argument.
id string (optional) — The guid or key, when the first argument is a kind.

Returns boolean true when the registry knows the resource.

renderer.hold(tex)
renderer.hold("material", "swatch")

globals/renderer/instanceData/clear

renderer.instanceData.clear(target: string | entityRef)

Drop every lane of an entity's per-instance shader data, so its draws read zero again — how a feature releases a subject it is still holding. Despawning an entity releases its block too, so this is for a subject that stays. It takes an entity that has already gone, which is when a feature releasing its subjects often runs, and does nothing for an entity holding no block.

Parameters

target string | entityRef — The entity — a proxy from entity(...) / entity.spawn(...), or an entity-id string.

renderer.instanceData.clear(subject)

globals/renderer/instanceData/laneCount

renderer.instanceData.laneCount() -> number

How many vec4 lanes each entity's per-instance block holds, so a lane index runs 0 .. laneCount() - 1. The same count a surface shader indexes input.shader_data against.

Returns number — lanes per entity.

for lane = 0, renderer.instanceData.laneCount() - 1 do
renderer.instanceData.set(subject, lane, 0)
end

globals/renderer/instanceData/set

renderer.instanceData.set(target: string | entityRef, lane: number, x: number, y: number?, z: number?, w: number?)

Write one vec4 lane of an entity's per-instance shader data — the channel that lets ONE material serve many entities that differ in a value. A surface shader reads the lane back as input.shader_data[lane], so a dissolve at its own progress per subject, an effect at its own age per firing, or a per-entity mask costs one material rather than one material per entity.

The engine attaches no meaning to a lane: a feature picks the lane indices it owns and packs whatever its shader agrees they carry. Name those indices in the module that writes them, so the writer and the shader read the block the same way.

The write reaches the block where it is called, so the entity it names is the one holding that id at that point in the tick, and the value is on the draw from the next frame. It is held until the lane is written again, the entity's block is cleared, or the entity is despawned — a despawned entity releases its whole block. A lane an entity was never given reads zero.

Parameters

target string | entityRef — The entity — a proxy from entity(...) / entity.spawn(...), or an entity-id string. It must name a live entity.
lane number — Which vec4 lane to write, 0 .. laneCount() - 1.
x number — The lane's .x.
y number (optional) — The lane's .y. Defaults to 0.
z number (optional) — The lane's .z. Defaults to 0.
w number (optional) — The lane's .w. Defaults to 0.

-- One dissolve material, each subject at its own progress.
local DISSOLVE_LANE = 0
for _, subject in ipairs(dying) do
renderer.instanceData.set(subject.entity, DISSOLVE_LANE, subject.progress)
end

globals/renderer/lighting

renderer.lighting() -> boolean

Whether the scene renders lit.

Returns boolean — true while lighting is on.

globals/renderer/loseDevice

renderer.loseDevice()

Destroy the render device on the next frame, so the engine meets a real device loss.

This is the one loss that can be caused on purpose, and it travels the same path a driver reset does: frames draw nothing until the rebuild lands, GET /engine/status reports the renderer as recovering while it does, engine.onDeviceRebuilt fires afterwards, and renderer.deviceGeneration() moves. Use it to prove that a world's content survives a device loss — anything it holds only on the GPU has to be remade from the rebuild hook, and this is how you find out whether it is.

renderer.loseDevice()
-- some frames later:
print(renderer.deviceGeneration()) -- one higher than before

globals/renderer/mainCameraView

renderer.mainCameraView() -> { number }?

The main camera's inverse view-projection (column-major, 16 numbers) followed by its world position (3 numbers) — {m0..m15, px,py,pz} — for reconstructing world positions from the depth buffer in a ray-tracing pass. Nil before the first render.

Returns { number }? 19 numbers, or nil.

globals/renderer/markBaseline

renderer.markBaseline() -> number

Record which render features are live and how the feature-texture array is configured, and make that the point renderer.resetToBaseline returns to.

A caller running work in an engine it shares — a sweep of test suites, a tool driving a session — leaves the renderer holding whatever that work set up: a feature a module created on first use, an array reconfigured to the size one pass wanted. This is how such a caller says where its own changes start. mark above the marks before it.

Returns number — The mark's generation, counting from 1 — what renderer.resetToBaseline takes to restore to this mark.

local mark = renderer.markBaseline() -- then run, then renderer.resetToBaseline(mark)

globals/renderer/material/animatedTexture

renderer.material.animatedTexture(texture: string | AssetRef, opts: { [string]: any }?) -> MaterialHandle

Build a material that PLAYS a layered texture: its layers bound as the frames, its timing bound beside them, and the engine's animatedTexture shader turning the clock into the layer showing now. One call from an imported animated image to a material an entity can wear.

The layer showing is resolved per pixel against the texture's own schedule, so frames of unequal length are shown for the lengths they were authored with, and the sequence loops. speed scales the clock (2 plays twice as fast, 0 holds the frame startTime lands in) and startTime offsets into the sequence, so two surfaces sharing one texture can run out of phase.

The clock is the engine's, and it runs in edit mode as much as in play and through a pause, so two screenshots of one surface taken moments apart are two different frames of it. speed = 0 holds one frame for as long as it is set, which is the state to compare two screenshots in.

The returned handle is what a surface wears — Model:applySessionMaterial takes it, and so does a Model's material field. The handle's guid is this material's REGISTRY KEY, the currency of setProperty, describe and destroy; a component field resolves an asset, so a bare key in one leaves the component waiting for an asset to register under that name.

The builtin plane mesh emits uv = (u, v) with v along its own +Z, so a quad pitched +90° about X (Transform.eulerToQuat(0, math.pi / 2)) shows the image upright to a camera on +Z, and -90° shows it first-row-last.

A texture whose layers carry no timing is rejected — there is nothing to play. renderer.texture.info(bytes).animated is the test.

Parameters

texture string | AssetRef — The texture — a guid, an identity, a name, a path, or a texture AssetRef.
opts { [string]: any } (optional) — { key?, speed?, startTime?, alphaCutoff?, baseColor?, uvScale?, uvOffset? }.

Returns MaterialHandle

local mat = renderer.material.animatedTexture("banner.texture")
local id = entity.spawn("billboard", { rotation = { Transform.eulerToQuat(0, math.pi / 2) } })
entity(id).component.add("Model", { model = "plane" })
entity(id).component.get("Model"):applySessionMaterial(mat)
renderer.material.setProperty(mat.guid, "speed", 2)

globals/renderer/material/create

renderer.material.create(content: MaterialContent, key: string) -> MaterialHandle

Parameters

content MaterialContent
key string

Returns MaterialHandle

globals/renderer/material/describe

renderer.material.describe(key: string | { [string]: any } | AssetRef) -> any

The recoverable definition ({ shader, properties, textures, name, render, features, bindings }) this module registered under key via renderer.material.create, or nil for keys registered elsewhere (e.g. material assets resolved by the assetType). properties and textures carry the material's current values: each setProperty / setTexture write lands on this record, a texture slot under the GPU key the slot binds by — these are the WRITES, held here whether or not the renderer took them up. renderer beside them is what the renderer holds for the same key: the program its prepared bind group was built against, the texture each slot of its row names (renderer.textures, the writes the renderer took up), the render state its draws are looked up under, whether a pipeline exists for that key, and the draws, instances and binds the last published frame counted for it. A frame counts those only while something observes the renderer (renderer.materialCost(), renderer.observe() and renderer.drawDiagnostics() each keep it observing for the next 120 drawn frames, and describe does not), so renderer.observed states whether they were counted: true with the counts, false with draws, instances, placeholderDraws, binds and bindsElided all nil. renderer is nil when the renderer holds no material under this key at all, and resident states the same fact as a boolean. Writes reach the screen through both halves: resident = false says the renderer holds nothing to put them in, and renderer.draws = 0 on a resident material says it holds them and no renderable is drawing with it. For a material that is resident AND drawn and still looks wrong, renderer.drawDiagnostics() names the renderable and the cause. A material that is not resident carries shaderStatus and reason, where its shader stands: "absent" when the shader it names resolves to no shader asset (nothing will compile it, so it stays unprepared until one arrives), "pending" while that shader has not compiled, "failed" with the compiler's message, or "compiled" when the shader is ready and the material has not been prepared yet. A resident material carries neither.

Parameters

key string | { [string]: any } | AssetRef — The material's registry key, the MaterialHandle from renderer.material.create, or an AssetRef from asset.resolve.

Returns any — MaterialContent? with resident: boolean, renderer: MaterialObservation?, shaderStatus: string?, reason: string? and, for a key a .material asset filed, asset (the asset it is) fields

globals/renderer/material/destroy

renderer.material.destroy(key: string | { [string]: any } | AssetRef) -> boolean

Drop a runtime material registered via renderer.material.create: clears its recoverable definition, unregisters its runtime-resource stamp so it is no longer swept into the material freeze/save flow, and frees the GPU record. Use for transient materials (e.g. a preview swatch) that must not outlive their use. The on-disk asset, if any, is untouched.

The reach is the registry: after this, describe and list stop answering for the key. A surface already wearing the handle goes on drawing what it was given — Model:restoreSessionMaterial is what puts a Model back on its authored material.

Parameters

key string | { [string]: any } | AssetRef — The material's registry key (the one passed to create), the MaterialHandle create returned, or an AssetRef from asset.resolve.

Returns boolean true when a definition was known under key.

renderer.material.destroy("__preview_swatch_" .. texGuid)

globals/renderer/material/list

renderer.material.list() -> { any }

Every runtime material currently registered, ordered by registry key. Each entry carries the key, where it came from, the shader it binds, and for a key a .material asset filed (guid@checksum) the asset it is. renderer.references("material", key) says what is still holding a row, and renderer.collect() releases the rows nothing holds.

Returns { any } — Array of { guid, origin, owner?, shader?, asset? }.

for _, m in ipairs(renderer.material.list()) do print(m.asset or m.guid, m.shader) end

globals/renderer/material/renderState

renderer.material.renderState(key: string | { [string]: any } | AssetRef) -> MaterialObservation?

What the renderer holds for a material, which is a different document from the values written to it. index is the material index its row in the material table is filed under, which an edit rewrites in place, and textures the texture each slot of that row names, by slot name, under the key the texture is resident as; a slot drawing its shader's declared default is absent, so a setTexture the renderer has taken up reads here as the key it bound. shader is the program its prepared bind group was built against, renderState the blend / cull / topology / queue / depth key its draws are looked up under, keyBuilt whether a pipeline exists for that key, and draws / instances / placeholderDraws / binds / bindsElided what it cost in the frame the renderer last published. A frame counts those only while something observes the renderer, which renderer.materialCost() and renderer.drawDiagnostics() start, so observed states whether they were counted: true with the five counts, false with all five nil. nil means the renderer holds no material under this key at all — the writes landed on a record nothing is drawing with.

Parameters

key string | { [string]: any } | AssetRef — The material's registry key, the MaterialHandle from renderer.material.create, or an AssetRef from asset.resolve.

Returns MaterialObservation?

local r = renderer.material.renderState("water"); print(r.renderState.blend, r.keyBuilt)

globals/renderer/material/sessionKeyFor

renderer.material.sessionKeyFor(entityId: string) -> string

The canonical registry key for an entity's SESSION material — the runtime material a system (e.g. GI baking) shows on an entity in place of its authored material for the lifetime of the engine session. One session material per entity: create it under this key, hand the handle to Model:applySessionMaterial, and the component re-adopts it across VM reloads by probing this key with describe. The key names the entity for as long as the entity stands: once it is gone the session store lets the handle go, and a collection releases the material and whatever its bindings were the last to hold.

Parameters

entityId string — The entity carrying the material.

Returns string — The registry key string.

local key = renderer.material.sessionKeyFor(entityId)

globals/renderer/material/setClock

renderer.material.setClock(key: string | { [string]: any } | AssetRef, seconds: number?, rate: number?) -> ()

Drive a material's surfaces from a clock of its own: the time its shader's vertex(), vertex_clip() and fragment() read. The clock reads seconds now and runs on by itself, rate of its seconds for every second of the world's gameplay time, so it is stated again only when that changes — a pause is a rate of 0, a seek is a new seconds, slow motion is a rate below 1. Every surface drawing the material follows it, motion vectors included. seconds = nil hands the material back to the world's time. A procedural graph's Generator drives the materials it realizes from its playhead this way, so an effect's shaders pause, slow and scrub with it.

Parameters

key string | { [string]: any } | AssetRef — The material's registry key, the MaterialHandle from renderer.material.create, or an AssetRef from asset.resolve.
seconds number (optional) — The time the material's surfaces read now, or nil to release it.
rate number (optional) — Seconds of the clock per second of gameplay time (default 1).

Returns ()

renderer.material.setClock(asset.resolve("frost_ice", "material"), 12.5, 0)

globals/renderer/material/setProperty

renderer.material.setProperty(key: string | { [string]: any } | AssetRef, name: string, value: any?) -> ()

Push one changed uniform property to a registered material's GPU record (frame-fast incremental update; no re-register). Keyed by the material's registry key. The value written becomes the material's current one: it is what describe reports, and — for a property the material's shader declares, which is what the uniform buffer is packed by — what a material AssetRef reads back through getProperty / getProperties and what the surface is drawn with. A write under any other name reaches the record describe reports, which is where it reads back.

Parameters

key string | { [string]: any } | AssetRef — The material's registry key, the MaterialHandle from renderer.material.create, or an AssetRef from asset.resolve.
name string — Property name.
value any (optional) — New value.

Returns ()

renderer.material.setProperty(asset.resolve("wall", "material"), "roughness", 0.2)

globals/renderer/material/setTexture

renderer.material.setTexture(key: string | { [string]: any } | AssetRef, slot: string, ref: string | { [string]: any } | AssetRef) -> boolean

Push one changed texture slot to a registered material's GPU record. Keyed by the material's registry key. The slot is one the material's shader declares as a texture: a slot it does not declare is refused with a warning naming the texture slots it does declare, and the material is left as it was. While the shader's vocabulary has not been read (a custom shader that has not loaded) every slot is taken.

Parameters

key string | { [string]: any } | AssetRef — The material's registry key, the MaterialHandle from renderer.material.create, or an AssetRef from asset.resolve.
slot string — Texture slot name ("base_color_texture", …).
ref string | { [string]: any } | AssetRef — Texture reference — a .texture guid / identity / name / path, the image path it was imported from, a color: / default: form, a live GPU handle, or a texture AssetRef carrying one. An asset reference is materialised (Disk→CPU→GPU) and bound by the key the upload lands under.

Returns boolean — true when the slot was bound, false when the shader declares no texture slot under that name.

renderer.material.setTexture("sky", "sky_texture", "panorama.texture")

globals/renderer/material/vocabulary

renderer.material.vocabulary(key: string | { [string]: any } | AssetRef) -> MaterialVocabulary?

A material's property names under props and its shader-declared and assigned texture slots under textureSlots, each a { [name] = true } set — with byRole beside them, the index a key spelled in another engine's convention resolves to a declared name through. This is what a write's key is routed against on the record, in the shape MaterialSchema.route takes. It is read off the definition this module holds and touches nothing the renderer holds, so a call costs a table lookup and belongs on a per-frame write path; describe is the read that carries the renderer's own state. One frozen table per record, shared by every key that reaches it, rebuilt on the first read after a write puts a new name on the record.

Parameters

key string | { [string]: any } | AssetRef — The material's registry key, the MaterialHandle from renderer.material.create, or an AssetRef from asset.resolve.

Returns MaterialVocabulary? — nil for a key registered elsewhere, and for a record carrying no property, assigned texture, or declared texture slot.

local MaterialSchema = require("@builtin::modules.material_schema")
local route, name = MaterialSchema.route(assert(renderer.material.vocabulary(handle)), "color", { 1, 0, 0, 1 })

globals/renderer/materialCost

renderer.materialCost() -> { MaterialObservation }

What each material cost the frame the renderer last drew, and the state it holds each one under. One row per material the renderer holds a prepared bind group for — a material an author wrote and the renderer never prepared is absent, which is itself the answer to "why is nothing I set reaching the screen". draws and instances cover that one frame; placeholderDraws is how many of those draws bound the magenta placeholder instead of this material's own program. Those three count what this material drew itself: a batch drawing several materials of one group adds to each material the draws and instances that wore it.

group is the material group the material draws in — the unit a pass binds a material's textures (group 1) and its shader's storage buffers (group 2) by — or nil while the material resolves no group, which leaves its draws nothing to bind. binds is how many group-1 and group-2 binds the frame's passes issued for draws of this material's shader in that group, and bindsElided how many times such a draw found the group already bound, which is what draw-key sorting buys. Those binds belong to the group, so every material of one group reports the same pair; a material with no group reports 0 for both, and a draw that fell back to the placeholder counts in placeholderDraws and in neither bind count.

uniformBytes is the size of the property block in the material's row of the material table: the shader's declared properties, packed at the offsets the shader reads them from and padded to a whole 16 bytes. The row also carries a 64-byte header and a 4-byte handle per texture slot, padded to a whole 16 bytes. renderer.materialTable().rowBytes is the space the table reserves for all rows — its row capacity, free rows included — plus its directories. renderer.drawDiagnostics() names WHICH renderable is not drawing what its material says, and why. material is the key the renderer holds the material under; asset names the .material asset a guid@checksum key is a version of.

Returns { MaterialObservation }

for _, m in renderer.materialCost() do print(m.asset or m.material, m.draws, m.binds, m.bindsElided) end

globals/renderer/materialIdentity

renderer.materialIdentity() -> MaterialIdentity

Which material each renderable draws with, as a number a shader can carry. A material is authored and bound by name, and no shader can read a string — so every renderable's per-instance record holds a material index instead. slots is the name → index table those indices are drawn from: an index is assigned the first time the renderer draws with that material and does not move while the renderer holds the material, so two renderables that differ only in material read different indices, and one renderable reads the same index frame after frame. A destroyed material's name leaves the table once nothing draws with it, and the next material takes its index, so the table holds a row per name the renderer holds a material under or something draws with. renderables is a row per renderable that owns a GPU slot — the entity it belongs to, that slot, and the index the record at it carries; populations is the same for an instanced draw, whose whole reserved run of slots carries the one material its registration named. That index is what a shader reads as zero_instance(slot).material_index, and the row a ray hit resolves through zeroMaterial(). A renderable draws with the material its entity references, so one whose entity names none carries index 0.

Returns MaterialIdentity — { slots: { [string]: number }, renderables: { { entity: string, slot: number, index: number, material: string } }, populations: { { slot: number, count: number, index: number, material: string } } }

local id = renderer.materialIdentity()
for _, r in id.renderables do print(r.entity, r.slot, r.index, r.material) end

globals/renderer/materialIndex

renderer.materialIndex(name: string) -> number?

The index standing for a material, or nil for a name the renderer holds no index for: one it has not drawn with yet, or a destroyed material nothing draws with any more. Pass it to a shader (or compare it against what a shader read out of zero_instance(slot).material_index) to tell which material a drawing instance carries.

Parameters

name string — string Material name, as renderer.material.create filed it.

Returns number?

local red = renderer.materialIndex("brick_red")

globals/renderer/materialTable

renderer.materialTable() -> MaterialTable

How this device resolves material textures, and what that storage costs. backend is "bindless" (a descriptor array indexed by each material's texture handles) or "buckets" (exact-size texture arrays, one per size, format and mip count, bound per material group). Bindless is taken on a Vulkan adapter that has texture binding arrays with non-uniform indexing and partially bound arrays, and room for descriptor arrays of at least 4096 textures; every other device — the browser build, Android, Metal, DirectX 12, and Vulkan without those — takes buckets.

materials is how many materials have a row, rowBytes the space the table reserves in the scene buffer (its row capacity, free rows included, plus the material and texture directories), textures how many distinct textures the rows name, descriptorCapacity the length of each bindless array (0 on buckets), and groups how many material groups the materials form. A group is the unit a pass binds a material's textures (group 1) and a shader's own storage buffers (group 2) by, and every material of a group draws under the same binds. On bindless every group binds the one texture group of the frame, so groups differ only by the set of storage buffers they bind. On buckets a group also names, per texture slot, the bucket its members' textures live in and the sampler those textures sample through: materials of one shader split into one group per distinct combination of (bucket and sampler per texture slot, storage buffers), and each group is its own draw.

buckets lists every texture array: its format, size, mip count, the layers textures occupy, its capacity, what it costs, the slack it holds beyond its live layers, whether it is frozen (a layer's view is held by something outside the material system — a UI image, a post-process input — or a copy out of a layer is taken this frame, such as a readback, so the array neither grows nor compacts), which shard of its key it is, and whether it is external (a one-layer bucket over a texture the GPU writes — a render target, a video frame, a compute output — which its producer owns). bucketBytes is what the shared buckets occupy together, and bucketSlackBytes the slack they hold together, which renderer.setTextureBucketSlack bounds; an external bucket counts in neither. slackBudgetBytes is the budget the buckets bound growth by. On bindless buckets is empty, both totals are 0, and slackBudgetBytes is the budget recorded, which bounds nothing.

Returns MaterialTable

local t = renderer.materialTable(); print(t.backend, t.groups, "groups over", t.materials, "materials")
for _, b in renderer.materialTable().buckets do print(b.format, b.width, b.height, b.layers, "/", b.capacity) end

globals/renderer/maxAnisotropy

renderer.maxAnisotropy() -> number

The highest anisotropy this device honours: 16 on hardware that filters anisotropically, 1 on hardware that does not, where a higher request would be downgraded to trilinear regardless. Read it to report quality honestly — renderer.setAnisotropy clamps for you, so a request never needs guarding.

Returns number — The device ceiling, 1 or 16.

local best = renderer.maxAnisotropy()

globals/renderer/maxViewportExtent

renderer.maxViewportExtent() -> number

The largest number of pixels this device draws a surface at, per axis — the bound renderer.setViewportSize refuses past. 0 before the first frame has drawn.

Returns number — pixels per axis.

local bound = renderer.maxViewportExtent()

globals/renderer/memoryPressure

renderer.memoryPressure() -> MemoryPressureReport

What memory pressure is doing to the engine. level is the level in force: "normal", "moderate", "low" or "critical". A report puts its level in force, whatever stood before, for holdSeconds after the last report, and then the level returns to normal. Reports come from the system (on Android, onTrimMemory: RUNNING_MODERATE is moderate, RUNNING_LOW and UI_HIDDEN low, RUNNING_CRITICAL, BACKGROUND, MODERATE and COMPLETE critical; Android 14 and later report only UI_HIDDEN, when the app leaves the screen, and BACKGROUND) and from renderer.signalMemoryPressure. signals counts the reports taken; lastLevel, lastSource and lastAgeSeconds describe the newest (absent before the first).

Each report releases at once what its level releases, and released says what the newest one freed: the textures taken to the level the frame samples and the textureBytes they gave up, and the render targets whose storage was freed. Moderate frees the mip levels the frame does not sample; low also frees render-target storage the last frame did not draw into; critical also halves the texture budget. While a level stands the residency plan keeps the textures there, and textureBudgetBytes is the budget the level leaves in force. engine.onMemoryPressure(cb) fires with each report once its release is done.

Returns MemoryPressureReport — { level, holdSeconds, signals, lastLevel?, lastSource?, lastAgeSeconds?, textureBudgetBytes, released }

local mp = renderer.memoryPressure()
print(mp.level, mp.released.textureBytes / 1048576, "MiB freed")

globals/renderer/mesh/bounds

renderer.mesh.bounds(mesh: string | { [string]: any } | AssetRef) -> Aabb?

The mesh's extent in its own space: the box its vertices span, or the one its author stated (aabbMin / aabbMax) for a mesh whose vertices live only on the GPU. It is what the mesh is culled against and what framing reads; place it in the world through whatever draws the mesh — an entity's transform, or instanceBounds for an instanced population.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh — a MeshHandle, a MeshCpuHandle, a guid, or a mesh AssetRef.

Returns Aabb? — The box, or nil for a mesh nothing is registered under or whose extent nobody knows.

local box = renderer.mesh.bounds(mesh)

globals/renderer/mesh/boundsSource

renderer.mesh.boundsSource(mesh: string | { [string]: any } | AssetRef) -> string

Where this mesh's culling bounds come from. "compute" once a compute pass has written its vertices: the engine reduces those vertices to an AABB every frame, so the mesh is culled against the geometry the pass produced wherever it puts it. "geometry" otherwise: the AABB of the geometry the mesh was created with.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh — a MeshHandle, a MeshCpuHandle, a guid, or a mesh AssetRef.

Returns string — "compute" or "geometry".

print(renderer.mesh.boundsSource(mesh))

globals/renderer/mesh/buildClusters

renderer.mesh.buildClusters(mesh: string | { [string]: any } | AssetRef) -> string?

Build a cluster-LOD DAG (Nanite-style virtualized geometry) for the static CPU mesh held under guid and return its serialized data.clusters bytes. Returns nil when the mesh is degenerate, and on an engine whose renderer.mesh.canBuildClusters reports false. Pair with renderer.mesh.uploadClusters.

Parameters

mesh string | { [string]: any } | AssetRef — A mesh loaded into the CPU store (renderer.mesh.loadCpu) — the MeshCpuHandle, a MeshHandle, a guid, or a mesh AssetRef.

Returns string? — Serialized cluster bytes, or nil.

local cb = renderer.mesh.buildClusters(cpu)

globals/renderer/mesh/canBuildClusters

renderer.mesh.canBuildClusters() -> boolean

Whether this engine bakes cluster-LOD hierarchies. It reads the binding the running engine registered: every target the engine ships on carries the builder, so a mesh loaded in a browser bakes its own clusters the same way one loaded natively does, and an engine built without it reports false and answers nil from renderer.mesh.buildClusters.

Returns boolean — True if renderer.mesh.buildClusters can bake on this platform.

if renderer.mesh.canBuildClusters() then ... end

globals/renderer/mesh/clusterBakeBudget

renderer.mesh.clusterBakeBudget(ms: number?) -> number

The wall time one frame may spend advancing scheduled cluster bakes, in milliseconds — set first when ms is given. A slice always runs at least one unit of the build, so the budget bounds what a frame spends by choice and the largest single unit a mesh imposes sets the floor under it.

Parameters

ms number (optional) — New per-frame budget in milliseconds, capped at 1000. A value that is not a positive, finite number raises.

Returns number — The budget in force after the call.

renderer.mesh.clusterBakeBudget(2)

globals/renderer/mesh/clusterBakes

renderer.mesh.clusterBakes() -> { [string]: any }

What the scheduled cluster bakes are costing. budgetMs is the slice a frame may spend, pending how many bakes are queued, completed how many have finished since the engine started, dropped how many left the queue because the geometry they were scheduled over stopped being readable or cannot be drawn as clusters, and heldBytes the source geometry the queue is holding across all of them — the vertex pool and index run the bake at the head is reading. inFlight is one row per queued bake — { guid, cpuMs, frames, slices, bytes, state, levels }: the wall time spent advancing it, the frames it has been queued for, the slices it has been advanced by, the geometry it is holding, "baking" for the one being advanced against "queued" for the ones waiting their turn, and levels the depth the bake builds, "dag" or "single".

Returns { [string]: any } — { budgetMs, pending, completed, dropped, heldBytes, inFlight }.

print(renderer.mesh.clusterBakes().heldBytes)

globals/renderer/mesh/clusterComponents

renderer.mesh.clusterComponents(clusterBytes: buffer | string) -> (ClusterComponents?, string?)

Split a cluster blob (from renderer.mesh.buildClusters) into its GPU-ready component byte pools — the cluster vertex pool, the geometry-addressing pool (every cluster's local→global vertex map, then every cluster's triangle bytes), and the per-cluster record array — plus their counts. A cluster's triangles address positions inside its own vertex map one byte at a time, and a record's vertexOffset indexes the geometry pool in u32 elements while its indexOffset indexes it in bytes, so ONE binding resolves a corner. A pure decode (no GPU work): upload the pools into buffers a compute shader owns (shaderRef:createBuffer + buf:writeBytes) to drive a cluster draw from Luau.

Parameters

clusterBytes buffer | string — Serialized cluster bytes (binary-safe).

Returns (ClusterComponents?, string?) — { vertices, geometry, records, vertexCount, vertexRefCount, triangleBytes, indexCount, clusterCount }, or (nil, err).

local c = renderer.mesh.clusterComponents(cb)

globals/renderer/mesh/clusterStore

renderer.mesh.clusterStore() -> {

What the geometry heap holds for cluster drawing. heapCapacity and heapLive are the heap's size and the bytes in use. The heap grows to place what does not fit, and gives memory back as meshes leave it: once what it holds has packed into a quarter of its size for 60 frames in a row, it shrinks to double that, never below the size it started at. A mesh destroyed and created again inside those frames finds the heap at the size it left it. vertexBytes and lightmapBytes are the vertex pools and lightmap-UV runs of the meshes drawn through clusters. bakedBytes is the cluster sets of meshes drawing a baked set; unbakedBytes is the sets of meshes still drawing lod0 plus those meshes' vertex pools — geometry drawn at full detail while its bake is outstanding. meshes counts meshes drawing any set, unbakedMeshes those on lod0. scriptSetBytes is CPU memory beside the heap: the sets scripts installed with renderer.mesh.uploadClusters, held so each mesh draws its set again after it returns to the device.

Returns { heapCapacity: number, heapLive: number, vertexBytes: number, lightmapBytes: number, bakedBytes: number, unbakedBytes: number, meshes: number, unbakedMeshes: number, scriptSetBytes: number }

local s = renderer.mesh.clusterStore(); print(s.unbakedBytes / 1048576, "MiB drawn at full detail while bakes are outstanding")

globals/renderer/mesh/clusters

renderer.mesh.clusters(mesh: string | { [string]: any } | AssetRef) -> { [string]: any }?

The cluster set a mesh draws. Every static mesh the geometry heap holds draws through clusters: kind is "lod0" while it draws the full-detail set built from its geometry as it uploaded, "single" for one baked level (clusterLod = false), and "dag" for the baked LOD hierarchy. clusterCount, levelCount and triangleCount describe that set, nodes how many node transforms its clusters lie in, and bytes its size in the heap. A mesh that is not drawn through clusters reads resident = false with reason: "heapFull" (the heap cannot grow to hold it), "spansNodes" (a triangle joins two independently moving parts) or "deformable" (a pass deforms its vertices). Nil for a mesh the renderer holds no GPU geometry for.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh to read — a MeshHandle, a MeshCpuHandle, a guid, or a mesh AssetRef.

Returns { [string]: any }? — { resident: boolean, reason: string?, kind: string?, clusterCount: number, levelCount: number, triangleCount: number, nodes: number, bytes: number }?

local c = renderer.mesh.clusters(gpu) ; if c and c.kind == "lod0" then print("still baking") end

globals/renderer/mesh/create

renderer.mesh.create(src: any?, guid: string?) -> MeshHandle

Create (or fetch) a GPU mesh resource and return its MeshHandle. src: a MeshCpuHandle from meshRef:load() (CPU→GPU upload under the asset's guid, idempotent — returns the resident handle if already uploaded); raw geometry {positions, indices, normals?, uvs?, colors?, uvs1?, unwrapUvs?, tangents?, skinning?, skins?} (a new runtime mesh — uvs1 is the lightmap UV set, unwrapUvs generates one, skinning/skins bind a skeleton); GPU compute buffers {vertexBuffer, indexBuffer, vertexCount, indexCount, aabbMin?, aabbMax?, prevVertexBuffer?} (size the vertex buffer at vertexCount * engine.vertexStride bytes, the engine's standard Vertex layout); or a MeshHandle (returned as-is). NEVER takes an AssetRef — load the CPU first.

prevVertexBuffer is a second buffer of the same size and layout holding those vertices as they stood on the previous frame. Naming it is what makes geometry a compute pass moves report a motion vector: the surface differences the two streams, so every consumer of screen-space velocity — motion blur, temporal reprojection — sees the movement. The engine fills it from the current vertices once per frame, ahead of that frame's compute dispatches, so a frame in which the pass does not run leaves the two streams equal and the geometry reports standing still.

morphTargets are the shapes the mesh can blend towards: a list of { name?, positions, normals? } records, each holding one offset per vertex from the base geometry, in the mesh's own vertex order. An entity blends them with ecs.MorphWeights, weight i scaling target i. A name makes the shape addressable as itself — renderer.mesh.morphTargets reads the names back and renderer.mesh.morphWeights drives them by name.

Raw geometry is read against the mesh type's conventions: indices count vertices from 0, and a triangle's FRONT face is the one whose vertices turn counter-clockwise as the viewer sees them — cross(v1 - v0, v2 - v0) points out of it. A material culls its back faces by default, so a triangle wound the other way draws nothing where it stands; reverse the index triple, or give the material render = { cull = "none" }, to draw that side. normals give the surface its outward direction and shade the face; the side that draws comes from the index order alone. uvs sample (0,0) at the image's top-left. Model space carries the world's basis: +X right, +Y up, -Z the direction transform.forward points. guides { path = "types/mesh" } has the whole table. A geometry src carrying keepCpu = true also keeps its geometry in the guid-keyed CPU store, so renderer.mesh.getVertices reads it and renderer.mesh.setVertices rewrites its positions in place — the per-frame deformation path, which sends positions alone where renderer.mesh.update re-sends the whole geometry. renderer.mesh.unloadCpu(mesh) releases that copy. Without it the geometry lives on the GPU alone and renderer.mesh.readback(mesh) is what brings it back.

A mesh this call builds lives while the returned handle, or another handle over the same guid, is reachable from a script, or while something else holds it (an entity or instanced draw displaying it, its owner, a renderer.hold). Once nothing holds it and the garbage collector has taken its last handle, the engine releases it mid-session: its definition, its buffers and its renderer.mesh.list row go, and its guid names nothing. A script that displays a mesh again later keeps its handle (or holds it) for as long as it means to.

A handle a script keeps names its mesh for the rest of the session: a module table or _SCRATCH holding it carries it across play/edit flips, scene loads and hot reloads of the components that use it, and a Model handed it after any of them draws the mesh. While nothing displays the mesh, the engine holds it off the device (renderer.mesh.heldOffDevice) and the next display puts it back. Only a destroy ends a mesh a handle still reaches (renderer.destroy, renderer.mesh.destroy): every handle a create returned over it then carries released, the verb that released it, and a point of use (ecs.Mesh, a Model's model field) refuses such a handle by name until a mesh stands under its guid again, which the handle then names and holds.

Parameters

src any (optional) — A MeshCpuHandle, geometry, compute buffers, or a MeshHandle.
guid string (optional) — Optional v4 guid for a NEW runtime mesh (minted Luau-side when absent). Ignored for the CPU-handle path (the asset's guid is used).

Returns MeshHandle

local gpu = renderer.mesh.create(meshRef:load())
local gpu = renderer.mesh.create({ positions = {...}, indices = {...} })
-- a runtime mesh whose positions are rewritten in place each frame
local m = renderer.mesh.create({ positions = P, indices = I, keepCpu = true })
renderer.mesh.setVertices(m, P)
-- a runtime mesh carrying a lightmap UV set (unwrapped at creation)
local gpu = renderer.mesh.create({ positions = {...}, indices = {...}, unwrapUvs = true })
-- a mesh with one shape to blend towards, driven by ecs.MorphWeights
local gpu = renderer.mesh.create({ positions = P, indices = I, morphTargets = { { positions = D } } })

globals/renderer/mesh/decode

renderer.mesh.decode(zmsh: buffer | string) -> (MeshGeometry?, string?)

Decode engine-native ZMSH bytes back into a MeshGeometry. Inverse of renderer.mesh.encode; each optional stream is present only when the blob carries it. Takes the bytes themselves — the geometry of a mesh the engine is holding comes from renderer.mesh.geometry(mesh).

Parameters

zmsh buffer | string — Engine-native ZMSH bytes (binary-safe).

Returns (MeshGeometry?, string?) the geometry, or (nil, errmsg).

local geom = renderer.mesh.decode(meshRef:getBytes())

globals/renderer/mesh/destroy

renderer.mesh.destroy(mesh: string | { [string]: any } | AssetRef) -> boolean

Release the GPU mesh mesh names, the release that pairs with renderer.mesh.create. Takes every form that names a mesh — the MeshHandle create returned, the guid renderer.mesh.list hands out, a MeshCpuHandle or a mesh AssetRef — and routes through renderer.destroy, the verb that releases any renderer resource by its kind. The CPU copy, if one was loaded, is freed separately by the CPU handle's :unload().

Parameters

mesh string | { [string]: any } | AssetRef — The mesh to release — a MeshHandle, a guid, a MeshCpuHandle or a mesh AssetRef.

Returns boolean true if a GPU mesh was known under the guid.

local m = renderer.mesh.create(cpu); renderer.mesh.destroy(m)
renderer.mesh.destroy(renderer.mesh.list()[1].guid)

globals/renderer/mesh/drawInstanced

renderer.mesh.drawInstanced(mesh: string | { [string]: any } | AssetRef, opts: any?) -> InstancedDraw

Draw one mesh instanceCount times in a single call, each copy placed by a world matrix read from a GPU buffer. The population is a renderable in its own right — it goes through the mesh's ordinary pipeline and the material's ordinary bind groups, so it appears in the deferred pass, the forward passes and the shadow maps exactly as an entity-backed draw of that mesh does.

The buffer holds instanceCount column-major 4x4 matrices, 64 bytes each, tightly packed — the layout a vertex shader reads as array<mat4x4<f32>>, which puts each matrix's translation in its LAST four floats (Lua indices 13/14/15 for x/y/z). Packing row-major transposes every instance.

The matrices are COPIED into the engine's transform slots, which is what buys that full-pass parity. Rewrite the buffer between frames and the instances move, with no re-registration and no re-upload. The copy is made on the frames that follow a write of the buffer (a :write, a compute pass whose shader binds it read_write, and every frame a material whose shader binds it read_write is live), so a population whose buffer nothing writes costs no copy per frame however many instances it holds; renderer.uploadStats() counts the copies a frame made.

material is what the population draws with, and it is required: a MaterialHandle (matRef:handle()), an AssetRef, or a registry key.

instanceDataBuffer names a second buffer, holding 64 bytes per instance — four vec4 lanes, tightly packed, in instance order. Those lanes arrive in the fragment stage as zero_object_data(in.instance_id, lane), the same read a per-entity __instancedata block answers, so the members of one population can differ in whatever their material's shader agrees the lanes carry. Copied like the transforms, on the frames that follow a write of the buffer, from a buffer a compute pass writes: the values never touch the CPU. Omit it and the lanes read zero.

reserveCount sizes the reservation above instanceCount so renderer.mesh.setInstanceCount can raise the drawn count later without re-registering; both buffers must back the reservation, not just the count.

mobility states whether the copies stand still — "static", or "movable" when it is left out. It is what a scene gather collecting geometry for precomputed lighting admits a population on, the same declaration Model.mobility makes for an entity: the transforms live in a buffer anything may rewrite between frames, so a population that says nothing is taken as one that moves.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh the population draws — a MeshHandle, the guid renderer.mesh.list hands out, a MeshCpuHandle or a mesh AssetRef. A registration holds the mesh on the device for as long as it lives, and takes a mesh that is currently held off the device — one nothing displays — back onto it.
opts any (optional) — { transformBuffer, instanceCount, material, instanceDataBuffer?, reserveCount?, renderLayer?, castsShadows?, receivesShadows?, mobility?, parent? }. castsShadows and receivesShadows, both true when left out, state whether the copies cast shadows and whether shadows darken them. parent stands the copies in an entity's space, as setInstanceParent does.

Returns InstancedDraw — An InstancedDraw handle for instanceInfo / setInstanceCount / dropInstanced.

local m = renderer.mesh.create({ positions = ..., indices = ... })
local buf = substrate.createBuffer({
name = "crowd.xf", type = "mat4", len = 64, kind = "gpu",
})
-- Column-major: translation lives at indices 13/14/15.
local xf = {}
for i = 0, 63 do
local m4 = { 1,0,0,0, 0,1,0,0, 0,0,1,0, i * 2, 0, 0, 1 }
for _, v in ipairs(m4) do xf[#xf + 1] = v end
end
buf:write(xf)
local rock = asset.resolve("rock", "material"):handle()
local draw = renderer.mesh.drawInstanced(m, { transformBuffer = "crowd.xf", instanceCount = 64, material = rock })

globals/renderer/mesh/dropClusters

renderer.mesh.dropClusters(mesh: string | { [string]: any } | AssetRef) -> boolean

Return a mesh to its lod0 set and cancel a bake still in flight for it. The inverse of renderer.mesh.uploadClusters.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh to return — a MeshHandle, a MeshCpuHandle, a guid, or a mesh AssetRef.

Returns boolean — True if the mesh drew a baked set or a bake was in flight.

renderer.mesh.dropClusters(gpu)

globals/renderer/mesh/dropInstanced

renderer.mesh.dropInstanced(draw: InstancedDraw) -> boolean

Release an instanced-draw registration and the transform slots it reserved. The mesh and the transform buffer outlive it — destroy those through renderer.destroy and the buffer handle's :destroy().

Parameters

draw InstancedDraw — The InstancedDraw to release.

Returns boolean — True if a registration was live under the handle.

renderer.mesh.dropInstanced(draw)

globals/renderer/mesh/encode

renderer.mesh.encode(geom: MeshGeometry) -> (string?, string?)

Encode raw geometry into engine-native ZMSH bytes (the on-disk mesh payload). The CPU codec behind the mesh assetType's onCreate. Every stream the format carries — including tangents, per-vertex skinning, and the skeleton — round-trips back through renderer.mesh.decode. This pair moves DATA the caller is holding; the geometry of a mesh the ENGINE is holding comes from renderer.mesh.geometry(mesh).

Parameters

geom MeshGeometry — MeshGeometry — flat per-vertex float / u32 arrays plus optional skinning and skins.

Returns (string?, string?) engine-native ZMSH bytes (binary-safe), or (nil, errmsg) naming what the geometry could not describe.

local bytes = renderer.mesh.encode({ positions = {...}, indices = {...} })
local bytes = renderer.mesh.encode(renderer.mesh.geometry(meshHandle))

globals/renderer/mesh/encodeCpu

renderer.mesh.encodeCpu(mesh: string | { [string]: any } | AssetRef) -> string

Encode a mesh's resident CPU copy into ZMSH bytes. Reads the ONE guid-keyed CPU store — meshRef:load() populates it for assets, and renderer.mesh.readback(mesh) populates it for a runtime mesh. Errors loudly when the mesh has no resident CPU copy.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh — a MeshHandle, a MeshCpuHandle, a guid, or a mesh AssetRef.

Returns string ZMSH bytes.

local bytes = renderer.mesh.encodeCpu(handle)

globals/renderer/mesh/geometry

renderer.mesh.geometry(mesh: string | { [string]: any } | AssetRef) -> MeshGeometry

The complete geometry of a mesh the engine is holding, as a MeshGeometry — the same shape renderer.mesh.create and renderer.mesh.encode take, carrying every stream the mesh has (positions, indices, and whichever of normals, uvs, colors, uvs1, tangents, skinning, skins it was built with). The read that pairs with create: hand it the MeshHandle create returned and get the vertex data back. Reads the resident CPU copy when there is one; for a runtime mesh that lives only on the GPU it reads the geometry back off the GPU first (yielding a frame or two) and leaves CPU residency as it found it. An optional stream is present only when the mesh carries one, so uvs1 == nil is the answer to whether it has a second UV set. The drawable mesh the renderer holds carries the tangent basis its positions, uvs and normals determine — supplied by the caller, or derived at the ingest that made it drawable — and that is what the GPU read gives back. The CPU store answers with the streams the bytes it decoded hold, so a .mesh written without a tangent stream reads back tangents == nil for as long as a CPU copy of it is resident.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh — a MeshHandle, a MeshCpuHandle, a guid, or a mesh AssetRef.

Returns MeshGeometry

local geom = renderer.mesh.geometry(renderer.mesh.create({ positions = P, indices = I }))
local tangents = renderer.mesh.geometry(handle).tangents

globals/renderer/mesh/getVertices

renderer.mesh.getVertices(mesh: string | { [string]: any } | AssetRef) -> { any }

Read the vertices of a mesh's resident CPU copy — one entry per vertex, { pos = {x,y,z}, normal = {x,y,z}, uv = {u,v} }. Reads the resident CPU store directly (no re-decode). Errors when the mesh has no resident CPU copy — renderer.mesh.geometry(mesh) is the read that works wherever the mesh lives, and returns the tangent, colour and skinning streams too.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh — a MeshHandle, a MeshCpuHandle, a guid, or a mesh AssetRef.

Returns { any } — { { pos: {x,y,z}, normal: {x,y,z}, uv: {u,v} }, ... } Each record carries pos, normal, uv and uv1 as named channels ({ x = , y = , z = }). Geometry going the other way — into renderer.mesh.create — is parallel flat arrays (positions, normals, uvs), and create accepts this record list under vertices so a mesh read back here can go straight into a new one.

globals/renderer/mesh/heldOffDevice

renderer.mesh.heldOffDevice(mesh: string | { [string]: any } | AssetRef) -> boolean

True while the engine holds this mesh off the device: nothing displays it, so its GPU buffers were given back and the engine keeps what rebuilds them. The next Model or instanced draw that names the mesh puts it back on the device, where renderer.mesh.isResident reads true again from the frame that bind lands in. A mesh is one of three things: resident, held off the device, or gone; a mesh neither this nor isResident answers true for names nothing the engine can draw.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh: a MeshHandle, a MeshCpuHandle, a guid, or a mesh AssetRef.

Returns boolean

local named = renderer.mesh.isResident(h) or renderer.mesh.heldOffDevice(h)

globals/renderer/mesh/instanceBounds

renderer.mesh.instanceBounds(draw: InstancedDraw, opts: { keep: number? }?) -> (Aabb?, string?)

The world box a registration's drawn copies occupy: every copy of the mesh's extent (renderer.mesh.bounds) where the engine composed it for the last frame drawn, carried through the entity the copies stand in when setInstanceParent named one. It reads the copies back from the GPU, so it waits a few drawn frames. A copy scaled to nothing draws nothing and is left out. A mesh whose extent nobody knows answers with the box around the copies' origins, and says so.

Parameters

draw InstancedDraw — The InstancedDraw to measure.
opts { keep: number? } (optional) — { keep? } — the share of the copies the box holds along each axis (default 1, all of them). Below 1, the copies reaching furthest out on each side are left out: keep = 0.9 is the box around where the body of a spray stands, without the few copies thrown far from it.

Returns (Aabb?, string?) — The box, or nil and the reason there is none; a box around origins alone comes with the note that says so.

local box, note = renderer.mesh.instanceBounds(draw)
local body = renderer.mesh.instanceBounds(draw, { keep = 0.9 })

globals/renderer/mesh/instanceInfo

renderer.mesh.instanceInfo(draw: InstancedDraw) -> InstancedDrawInfo?

What a live instanced-draw registration is drawing: which mesh, which transform buffer, which per-instance data buffer if it named one, how many instances, and how many slots it reserved. Returns nil once the registration has been dropped.

status is what the renderer did with it. The fields above it are the request, made a stage before the renderer sees it; status is the answer: "drawing" for a registration the renderer is drawing, "refused" for one it turned away — error carries its reason — and "pending" for the frame between the call and the renderer answering. So a registration whose copies are not being drawn says so here.

Parameters

draw InstancedDraw — The InstancedDraw to report on.

Returns InstancedDrawInfo? — The registration record, or nil.

print(renderer.mesh.instanceInfo(draw).instanceCount)
local info = renderer.mesh.instanceInfo(draw)
if info.status == "refused" then error(info.error) end

globals/renderer/mesh/instanceTransforms

renderer.mesh.instanceTransforms(draw: InstancedDraw) -> any

Read back the world matrices a registration's drawn copies are placed by: instanceCount matrices of 16 floats, column-major and tightly packed, in the layout the transform buffer holds them. The read is of the buffer as it stands when it runs, so a population a compute pass rewrites every frame answers with the placement of the frame the read lands in.

Parameters

draw InstancedDraw — The InstancedDraw whose copies to locate.

Returns any — A Readback to poll — :ready() then :result() — or nil for a registration that is no longer live.

local pending = renderer.mesh.instanceTransforms(draw)
while not pending:ready() do task.wait() end
local floats = pending:result()

globals/renderer/mesh/instanceWorlds

renderer.mesh.instanceWorlds(draw: InstancedDraw) -> any

Read back where a registration's drawn copies stand in the world, as the engine composed them for the last frame drawn: each copy's matrix from the transform buffer, carried through the world matrix of the entity the copies stand in when they have one (setInstanceParent). instanceCount column-major 4x4 matrices, 16 floats each. Async — poll the handle it returns the same way as instanceTransforms.

Parameters

draw InstancedDraw — The InstancedDraw to read.

Returns any — A pending readback: :ready(), then :result() for the floats.

local pending = renderer.mesh.instanceWorlds(draw)
while not pending:ready() do task.wait() end
local x = pending:result()[13]

globals/renderer/mesh/isCpuResident

renderer.mesh.isCpuResident(mesh: string | { [string]: any } | AssetRef) -> boolean

True if this mesh has a resident CPU copy in the guid-keyed CPU store.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh — a MeshHandle, a MeshCpuHandle, a guid, or a mesh AssetRef.

Returns boolean

if renderer.mesh.isCpuResident(handle) then ... end

globals/renderer/mesh/isResident

renderer.mesh.isResident(mesh: string | { [string]: any } | AssetRef) -> boolean

True if a GPU mesh is resident under this mesh's guid — the device holds its buffers, or the upload queued this frame is still going to hand them over. This is the store the draw paths are gated on, so a mesh this reports resident is one renderer.mesh.drawInstanced and a Model can draw — from the frame after the upload commits: a render asked for in the same tick the upload was queued does not carry the mesh yet. renderer.mesh.onDevice is the answer that waits for that commit.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh — a MeshHandle, a MeshCpuHandle, a guid, or a mesh AssetRef.

Returns boolean

print(renderer.mesh.isResident(handle))

globals/renderer/mesh/list

renderer.mesh.list() -> { any }

Every mesh currently registered, ordered by guid — the ones a script created and the ones that reached the device through an asset alike. Answers "which mesh is this?" when all that is known is a size: each entry carries the guid, the vertex/index counts it was created with, where it came from (origin is "asset" for a mesh the asset path uploaded), whether the GPU still holds it, and where its culling bounds come from (boundsFrom is "compute" for a mesh a compute pass writes). A resident entry also carries the bytes its buffers cost. bytes is the mesh's whole VRAM footprint and is the sum of the THREE buffer columns beside it — vertexBytes + vertexStorageBytes + indexBytes, where the storage column is the same vertices bound as a storage buffer for the passes that read them that way. Summing only the vertex and index columns understates a mesh by its vertex size. The bytes column is what sums to the meshes category of renderer.gpuMemory(). renderer.references("mesh", guid) says what is still holding a row, and renderer.collect() releases the rows nothing holds.

Returns { any } — Array of { guid, vertexCount?, indexCount?, origin, owner?, resident, boundsFrom, bytes?, vertexBytes?, vertexStorageBytes?, indexBytes?, primitives?, revision? }.

for _, m in ipairs(renderer.mesh.list()) do print(m.guid, m.bytes) end

globals/renderer/mesh/listInstanced

renderer.mesh.listInstanced() -> { InstancedDrawInfo }

Every instanced-draw registration this engine is drawing, in registration order. Each record is what instanceInfo answers with, and carries a draw handle of its own — so a population whose handle its caller no longer holds is reached here and released, resized or read like any other.

Returns { InstancedDrawInfo } — An array of registration records; empty when nothing is registered.

for _, pop in ipairs(renderer.mesh.listInstanced()) do
renderer.mesh.dropInstanced(pop.draw)
end

globals/renderer/mesh/loadCpu

renderer.mesh.loadCpu(ref: string | AssetRef) -> MeshCpuHandle

Load a .mesh asset's geometry into the ONE guid-keyed CPU store (the Disk→CPU step) and return a CPU handle. The handle holds NO geometry — only the guid plus counts and the per-handle read/encode/unload ops (which read the Rust-side store). Called by meshRef:load(). DEFAULT lifecycle: upload to the GPU then handle:unload(); the store is populated only by this call.

Parameters

ref string | AssetRef — A mesh AssetRef (carries .guid and reads its primary via getBytes), or any string asset.ref resolves to one — the guid encodeCpu takes, an identity, a name or a source path.

Returns MeshCpuHandle

local cpu = meshRef:load(); local gpu = renderer.mesh.create(cpu); cpu:unload()
local cpu = renderer.mesh.loadCpu(gpuMesh.guid)

globals/renderer/mesh/morphTargets

renderer.mesh.morphTargets(mesh: string | { [string]: any } | AssetRef) -> { string }

The names of the shapes this mesh blends towards, in the order an entity's ecs.MorphWeights addresses them — weight i drives the target named at i. An imported model carries the names its source file gave its blend shapes, so content drives a face by the shape it means rather than by the ordinal that shape happened to import at (which moves when the model is re-exported). A target the source never named reads as an empty string.

Empty for a mesh with no morph targets. Errors when the mesh is neither GPU- nor CPU-resident — materialise it first (meshRef:handle()).

Parameters

mesh string | { [string]: any } | AssetRef — The mesh — a MeshHandle, a MeshCpuHandle, a guid, or a mesh AssetRef.

Returns { string } one name per morph target, in target order.

for i, name in renderer.mesh.morphTargets(meshRef) do print(i, name) end

globals/renderer/mesh/morphWeights

renderer.mesh.morphWeights(mesh: string | { [string]: any } | AssetRef, weights: { [string]: number }) -> { number }

Turn weights named by shape into the ordered weight array ecs.MorphWeights takes — the drive-a-face-by-name call. Every target the mesh carries gets a slot; the ones weights names take their value and the rest are 0, so the returned array always describes the whole mesh and a shape left out is a shape at rest.

A name the mesh does not carry is an error listing the names it does: a mistyped viseme that silently moved nothing would be indistinguishable from a rig that never had it.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh — a MeshHandle, a MeshCpuHandle, a guid, or a mesh AssetRef.
weights { [string]: number } — { [string]: number } — how strongly to blend each named shape.

Returns { number } one weight per morph target, in target order.

local w = renderer.mesh.morphWeights(meshRef, { Eyes_Blink = 1, Mouth_Smile = 0.4 })
ecs.set(face, ecs.MorphWeights { weights = w })

globals/renderer/mesh/onDevice

renderer.mesh.onDevice(mesh: string | { [string]: any } | AssetRef) -> boolean

True once the device holds this mesh's buffers — the upload queued for it has committed — so the next render asked for carries the mesh. renderer.mesh.isResident answers true one frame earlier, while the upload is still queued; a reader about to render and read the picture back waits on this one.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh — a MeshHandle, a MeshCpuHandle, a guid, or a mesh AssetRef.

Returns boolean

while not renderer.mesh.onDevice(guid) do task.wait() end

globals/renderer/mesh/readback

renderer.mesh.readback(mesh: string | { [string]: any } | AssetRef, into: string?) -> MeshCpuHandle

Read a runtime GPU mesh's geometry back to CPU and return a MeshCpuHandle for it — the GPU→CPU half of the runtime-mesh freeze path. A mesh made with renderer.mesh.create keeps no CPU copy, so persisting it (:encode() → asset.create("mesh", …)) reads it back here first. Yields until the readback completes (a frame or two). After it returns the geometry is resident in the guid-keyed CPU store: :getTriangles, :getVertices, :getBounds, :geometry, :encode, :unload all work. Errors if the mesh never becomes resident in the vertex pool.

The CPU store is keyed by guid and shared by everything that reads the mesh, and whatever owns the mesh can retire the copy under its own guid: renderer.mesh.update from compute buffers unloads it, which an ocean surface does every frame it steps. into lands the copy under a key of the caller's own instead: the handle answers for that key, and it stays resident until the caller unloads it, whatever happens to the mesh.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh — the MeshHandle renderer.mesh.create returned, a guid, or a mesh AssetRef.
into string (optional) — The CPU-store key the geometry lands under. Defaults to the mesh's own guid.

Returns MeshCpuHandle

local cpu = renderer.mesh.readback(handle); local zmsh = cpu:encode(); cpu:unload()
local snap = renderer.mesh.readback(handle, "bake:" .. handle.guid); local tris = snap:getTriangles(); snap:unload()

globals/renderer/mesh/readbackPosed

renderer.mesh.readbackPosed(requests: { { entity: string, mesh: any } }) -> { [string]: MeshCpuHandle }

Read the POSED geometry of skinned entities back to CPU: for each request, the vertices the skinning pass wrote for that entity this frame, joined by the indices of the mesh it is posed from. A skinned surface's world-space triangles are produced on the GPU from the entity's joint matrices, so the mesh asset holds the bind pose and only this reads where the surface actually is. The posed vertices are in model space, so the entity's own world transform still places them — the same transform the raster draw uses.

Takes a LIST and answers a map, because the readbacks are queued together and polled together: a scene's worth of characters costs the frames of one readback rather than one entity's after another. Each posed mesh lands in the CPU store under a guid of its own, derived from the entity, so compute.buildBvh, meshcpu.* and every other guid-keyed reader takes it like any other mesh. Call handle:unload() when done with it.

An entity the map omits holds no live pose — nothing skinned it this frame, which is also what makes its draws read the source mesh, so its bind-pose geometry is what stands for it.

The vertices read are ones the skinning pass wrote for the pose the entity is bound to. An entity bound to a pose that pass has not written yet — the frame it was posed, before the pass reached the queue — is asked again on the next frame rather than read from a slice still holding what stood there before.

Parameters

requests { { entity: string, mesh: any } } — { { entity = <id>, mesh = <mesh> } } — the entity to read, and the mesh it is posed from (a guid, MeshHandle or mesh AssetRef).

Returns { [string]: MeshCpuHandle } — A map from entity id to the MeshCpuHandle holding that entity's posed geometry.

local posed = renderer.mesh.readbackPosed({ { entity = id, mesh = ecs.get(id, ecs.Mesh).mesh } })
local tris = posed[id]:getTriangles()

globals/renderer/mesh/scheduleClusters

renderer.mesh.scheduleClusters(mesh: string | { [string]: any } | AssetRef, opts: { levels: string? }?) -> boolean

Queue a cluster bake over the mesh's geometry, off the frame loop. The mesh keeps drawing its current cluster set while the bake runs, and the set the bake produces replaces it when it lands. opts.levels is "dag" for the LOD hierarchy or "single" for one level of clusters at full detail; when omitted, a mesh AssetRef bakes the depth its clusterLod setting asks for and any other argument bakes "dag". The CPU mesh may be unloaded on the very next line. Geometry with a triangle joining two independently moving parts leaves the queue unbaked with a warning; a mesh renderer.mesh.clusters reports as "deformable" or "heapFull" leaves it unbaked without one.

One bake is advanced per frame — the one at the head of the queue — and the geometry is read once a bake gets there and the mesh's GPU upload has landed, from the definition the engine holds for the mesh. The queue therefore holds one mesh's geometry rather than one per mesh, whatever its depth, and the set installs only on the geometry it was read from: a mesh whose geometry is replaced before its set lands keeps the set of its new geometry. renderer.mesh.clusterBakes().heldBytes reports what the queue is holding, and its inFlight rows report which bakes it is holding for. This is what the .mesh assetType materialisation path uses; reach for renderer.mesh.buildClusters when you want the bytes in hand instead. Scheduling the same mesh again replaces the bake already in flight for it.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh to bake — a MeshHandle, a MeshCpuHandle, a guid, or a mesh AssetRef, loaded into the CPU store (renderer.mesh.loadCpu).
opts { levels: string? } (optional) — { levels: "dag" | "single" }?

Returns boolean — True if a bake was queued; false for a mesh with no GPU mesh registered or being created under its guid.

renderer.mesh.scheduleClusters(cpu, { levels = "single" }) ; cpu:unload()

globals/renderer/mesh/setInstanceCount

renderer.mesh.setInstanceCount(draw: InstancedDraw, count: number) -> InstancedDraw

Change how many of a registration's instances draw. Constant time — the reservation, the transform buffer and the pipeline all stay put, so this is the verb for a population whose size changes per frame. The new count must fit the reservation drawInstanced was given.

Parameters

draw InstancedDraw — The InstancedDraw to reconfigure.
count number — Instances to draw, at least 1 and within the reservation.

Returns InstancedDraw — The same InstancedDraw.

renderer.mesh.setInstanceCount(draw, visibleCount)

globals/renderer/mesh/setInstanceParent

renderer.mesh.setInstanceParent(draw: InstancedDraw, parent: any?) -> InstancedDraw

Stand a registration's copies in an entity's space: every copy's matrix is taken relative to the entity, so the population moves, turns and scales with it from the next frame on, with its transform buffer untouched. nil puts the copies back in world space. It is the verb for a population that belongs to something — a generator's output, an asset stood on an entity.

Parameters

draw InstancedDraw — The InstancedDraw to reconfigure.
parent any (optional) — The entity the copies stand in (a proxy or its id), or nil for world space.

Returns InstancedDraw — The same InstancedDraw.

renderer.mesh.setInstanceParent(draw, entity.find("Forest"))

globals/renderer/mesh/setInstanceRenderLayer

renderer.mesh.setInstanceRenderLayer(draw: InstancedDraw, renderLayer: number) -> InstancedDraw

Change which render layers a registration's copies belong to. Constant time — the reservation, the transform buffer and the pipeline all stay put, and the next frame drawn tests the copies against the new membership. It is the verb for a population that follows something whose membership moves: a camera or a capture including the layer draws the copies, one excluding it does not.

Parameters

draw InstancedDraw — The InstancedDraw to reconfigure.
renderLayer number — The membership bitmask, the same value drawInstanced takes as renderLayer. At least one bit must be set.

Returns InstancedDraw — The same InstancedDraw.

renderer.mesh.setInstanceRenderLayer(draw, mask)

globals/renderer/mesh/setVertices

renderer.mesh.setVertices(mesh: string | { [string]: any } | AssetRef, positions: { number })

Replace a mesh's resident CPU vertex positions (flat { x,y,z, ... }) IN PLACE — indices, normals/uvs, and skinning are preserved, the AABB recomputes, and the GPU re-fetches the new geometry so it shows on screen. The positions alone travel, so this is the per-frame deformation path where renderer.mesh.update re-sends the whole geometry. The mesh must be CPU-resident: renderer.mesh.create({ ..., keepCpu = true }) keeps a copy from the start, renderer.mesh.readback(mesh) recovers one from the GPU, and meshRef:load() loads one for a .mesh asset. Errors with the reason otherwise, or when the vertex count doesn't match.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh — a MeshHandle, a MeshCpuHandle, a guid, or a mesh AssetRef.
positions { number } — Flat { x,y,z, ... } — one xyz per vertex; count must match the mesh.

local m = renderer.mesh.create({ positions = P, indices = I, keepCpu = true })
renderer.mesh.setVertices(m, P) -- P mutated in place each frame

globals/renderer/mesh/unloadCpu

renderer.mesh.unloadCpu(mesh: string | { [string]: any } | AssetRef)

Drop a mesh's resident CPU copy from the guid-keyed CPU store. The explicit release for a runtime geometry mesh's recoverable definition.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh — a MeshHandle, a MeshCpuHandle, a guid, or a mesh AssetRef.

globals/renderer/mesh/update

renderer.mesh.update(mesh: string | { [string]: any } | AssetRef, src: any?) -> MeshHandle

Overwrite the GPU resource mesh names IN PLACE, under the same guid, from new geometry or compute buffers. Never writes a .mesh file — the play-mode mutate path. A Model bound to the guid reflects the change with no re-bind. Takes every form that names a mesh — the MeshHandle create returned, the guid renderer.mesh.list hands out, a MeshCpuHandle or a mesh AssetRef. Returns a handle carrying the bounds the new geometry has: the handle it was given, refreshed, and a handle over the guid otherwise.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh to update — a MeshHandle, a guid, a MeshCpuHandle or a mesh AssetRef.
src any (optional) — New geometry {positions, indices, ...} or compute buffers {vertexBuffer, indexBuffer, vertexCount, indexCount, prevVertexBuffer?}.

Returns MeshHandle — A MeshHandle for the updated mesh.

globals/renderer/mesh/uploadClusters

renderer.mesh.uploadClusters(mesh: string | { [string]: any } | AssetRef, clusters: string) -> boolean

Replace the cluster set a mesh draws with a serialized set (bytes from renderer.mesh.buildClusters), built from the geometry the mesh holds as the call is made. A set whose vertex count, vertex maps or nodes do not describe the mesh, or one whose mesh's geometry is replaced before it lands, is refused with a warning, and the mesh keeps its current set and the set it was asked for. Once the set is installed, a bake still in flight for the mesh is cancelled, and the engine keeps the bytes on the CPU (renderer.mesh.clusterStore().scriptSetBytes) so the mesh draws this set again when it returns to the device after nothing displayed it, or after the device is rebuilt. The bytes go when the mesh's geometry is redefined, when renderer.mesh.dropClusters returns it to lod0, and when the mesh is destroyed.

Parameters

mesh string | { [string]: any } | AssetRef — The mesh the clusters belong to — a MeshHandle, a MeshCpuHandle, a guid, or a mesh AssetRef.
clusters string — Serialized cluster bytes (binary-safe).

Returns boolean — True if the upload was queued; false while the mesh's GPU upload has not landed yet.

renderer.mesh.uploadClusters(gpu, cb)

globals/renderer/minScreenSize

renderer.minScreenSize() -> number

The on-screen radius, in pixels, an object must reach to be drawn. 0 while the cutoff is off.

Returns number

local px = renderer.minScreenSize()

globals/renderer/morphStats

renderer.morphStats() -> {

The morph state the last frame drew with. A mesh carries the shapes it can blend towards and an entity carries how strongly each is blended (ecs.MorphWeights); where both are present, the vertex stage adds the weighted deltas to the base geometry.

instances is how many render slots that happened at, and blends how many single-target blends those slots carry between them: a slot contributes one per target its weights move, or that they moved the frame before, so the number of targets a mesh can be given is bounded by the buffer the blends live in. meshes is how many meshes hold a delta block and targets how many targets those blocks cover between them; deltaBytes is what the shared buffer they are appended into holds, and deltaBufferBytes the size of that buffer, which the vertex stage binds whole.

deltaBudgetBytes is the most that buffer may hold on this device: the largest storage buffer the device binds at once (128 MiB on many mobile browsers). A block costs 32 bytes per vertex per target, over every vertex of the mesh. A mesh whose block would take the resident blocks past the budget is refused by name in the engine log and draws in its base shape, and refusedMeshes counts the meshes refused that way; the refusal holds until the mesh is uploaded again. A morph-target mesh whose weights are all zero reads a meshes above zero beside an instances and blends of zero.

Returns { instances: number, blends: number, meshes: number, targets: number, deltaBytes: number, deltaBufferBytes: number, deltaBudgetBytes: number, refusedMeshes: number }

local m = renderer.morphStats()
print(("%d instances carrying %d blends over %d meshes"):format(m.instances, m.blends, m.meshes))

globals/renderer/observe

renderer.observe() -> RenderObservation

Everything the renderer knows about the frame it last drew: what program it bound for each renderable, the render state it holds each material under, and what each program has cost in pipeline builds. renderables is one row per renderable in the renderer's draw list, carrying the program its material named (requestedProgram) beside the one that was bound (boundProgram) — __error__ wherever the lookup missed and the draw went ahead on the magenta placeholder — plus substituted, the outcome (drew / drewPlaceholder / skipped / notDrawn), the reason that forced it and the compiler's own detail for a failed compile. observed says which of two answers a row is: true for a resolution a geometry pass took as it drew, false for the renderer's own resolution of a renderable this frame drew nowhere, which is what a renderable outside every camera's frustum or layer mask reports. materials is one row per material the renderer holds a prepared bind group for; shaders is one row per program pipelines have been built for. frame names the frame every per-frame count covers; retainedFrames how many frames a resolution a pass took is kept for after the last frame that drew it; window and costWindow state both in the document itself. Recording is armed by the first read, so this waits for the frame that first records rather than answering empty.

Returns RenderObservation

local o = renderer.observe(); print(o.placeholderDraws, "renderables drew the placeholder in frame", o.frame)

globals/renderer/occlusionCulling

renderer.occlusionCulling() -> boolean

Whether occlusion culling is currently enabled.

Returns boolean

globals/renderer/passSchedule

renderer.passSchedule() -> {

The schedule check over this frame's enqueued render passes. Passes declare what they read (inputs) and what they write (output / outputs / storage), and the frame runs them in phase order and, inside a phase, in order order. violations holds every input bound to a resource the frame produces LATER: that read samples the resource as it stands ahead of that pass, which is the previous frame's contents for a render target that persists, an empty target for one just created, and the scene draw's own output for a @scene.* buffer — and the pass renders either way. The frame's own buffers are checked on the same terms as a render target: bind @scene.motion at a phase ahead of the pass that writes it and the read is reported, naming the buffer and its writer. A pass reading a resource ahead of that write on purpose declares that slot in its enqueue's readsPrevious and drops out of the list; unboundPrevious holds declared slots the pass binds no such resource to, which cover nothing. A resource no queued pass writes is not reported — a camera rendering to texture and compute.dispatch both fill targets outside the pass queue, and the scene draw fills the @scene.* buffers every frame. A read the frame has only one order for is not reported either: where the writing pass consumes something the reading pass produces, the reader runs first or the writer has nothing to write, which is what a pass reading a buffer into a target of its own and a second pass copying that target back over the buffer forms. unreachable holds passes at a phase that does not run their kind: every phase drains its fragment and compute passes, while afterLighting is the one that draws geometry, draw and splat passes, so one of those enqueued elsewhere sits in the queue and never runs. Each finding is also stated in the engine log the first time it appears.

Returns { violations, unboundPrevious, unreachable }

local s = renderer.passSchedule()
for _, v in s.violations do print(v.message) end

globals/renderer/pendingWork

renderer.pendingWork() -> PendingWork?

What the renderer still owes the picture — the work it has taken on and has not yet put in the frame.

A frame draws what the renderer holds at the moment it runs, and several of the things it holds arrive over more than one frame. A pipeline is built off the frame thread and bound the frame its result is taken in; a texture is uploaded under a per-frame time budget; a released texture's GPU object is freed on the frame the release queue is drained; a streamable texture climbs toward the level its screen size asks for one level at a time. pipelineBuilds, textureUploads, textureReleases and textureResidency are those four. textureTransfers counts the textures the texture cache is still moving over several frames: a large upload written a band at a time, a promotion waiting on its finer levels to be read or decoded, a demotion waiting on its levels to be read back. Each lands as a new GPU texture on a frame of its own, and the frames between two landings replace nothing, so textureResidency alone reads zero on them. outstanding is the sum of every count.

settled is true when outstanding is zero, and says the renderer owes this picture nothing: the frames after this one draw the same picture unless something changes the scene. That is the renderer's own word for what a test measuring a picture holds for before it reads one.

epoch counts the times the renderer has gone from owing nothing to owing something since the process started. It is what a pair of readings needs: a run of work that begins AND ends between two samples leaves both of them reading settled and leaves epoch higher, so settled at both ends with the same epoch at both ends is the statement that nothing arrived across the whole span, rather than at its two endpoints alone. Test.settleReading is built on exactly that pair of conditions.

frame names the drawn frame the reading was published on. Every count is published from the frame thread on every frame the renderer runs, so a headless engine declining frames answers as readily as a windowed one. deviceRebuild is one while the engine is between render devices: the picture is owed the device itself, every frame in that window draws nothing, and what content puts back on the new device lands on the frames after it. computeCommands counts the compute commands a drain held over: a dispatch waiting on the program it names and every command over the buffers it would write queued behind it, each of which runs or drops on a later drain. clusterSets counts the meshes whose cluster set is still being built, or is owed again after a device rebuild: each lands as the set the mesh draws, and the geometry a cut of it draws changes on the frame it does. materialRows counts the materials the frame does not yet draw as they ask: one whose raster or resolve pipeline is still building draws nothing until it lands, one whose shader's resolve gate has not reached a verdict resolves through the error shader until it does, and one whose routing changed since the frame last published it is still drawn by the previous routing. Each changes the picture on the frame it lands. skippedDraws counts the renderables the last drawn frame left undrawn because a pipeline or a material group they draw with was not built yet; each appears on the frame what it waits on lands. shaderCompiles counts the shader compiles running off the frame thread that no frame has landed: each lands as the program its keys name, and the materials drawing with those keys change on the frame it does.

Returns PendingWork? — { settled, outstanding, pipelineBuilds, textureUploads, textureReleases, textureTransfers, textureResidency, deviceRebuild, computeCommands, clusterSets, materialRows, skippedDraws, shaderCompiles, epoch, frame }, or nil on a build with no renderer attached

if renderer.pendingWork().settled then print("the picture is what it is going to be") end
local w = renderer.pendingWork(); print(w.outstanding, "arrivals still owed, over", w.epoch, "runs of work")

globals/renderer/pipelineCache

renderer.pipelineCache() -> PipelineCache?

What the driver's compiled-pipeline store held, built, and wrote back. A pipeline is machine code the GPU driver compiles from the shader bound into it, and that compile is what a launch pays before the first frame drawing with each pipeline can appear. The store keeps that compiled code across runs, so a launch whose shaders have not changed reads back what the previous one compiled.

restoredBytes is what a previous run left for this GPU and this launch read; pipelinesBuilt counts the pipelines built since startup and buildMs is what they cost together, which is the number the store lowers. saves and savedBytes describe writing it back — deferred until a burst of builds settles, so one launch is one write — and dirty is true while pipelines have been built that the file does not hold, including after a write that failed, which lastError then names. path is the file, named after the GPU it belongs to.

supported is false where the platform holds no store a program can carry: a browser keeps its own and hands none out, and an adapter can lack the capability. reason says which, and the build count and timing still read true there. lastError names a read or write failure; a failed store costs the saved compile and never the frame, since every pipeline is built from its source either way.

A pipeline a material, the sky, a depth view, a post-process pass or a compute program a script registers needs after boot is built off the frame thread — the shader module behind that compute program with it — on buildWorkers worker threads on native, behind the browser's asynchronous entry point on the web, where buildWorkers reads 0. So a burst of new materials costs a frame nothing beyond drawing the placeholder until each pipeline lands, and a burst of effects costs it nothing beyond each effect starting on the frame its pipeline does. offFrameBuilt and offFrameBuildMs are the pipelines and the milliseconds that never held a frame; pipelinesBuilt - offFrameBuilt is what the frame thread built, which is the renderer's own — the set it stands up before its first frame, and the few it stands up on the frame a pass first asks for them. buildsSubmitted, buildsFinished, buildsFailed and buildsInstalled count build requests (shader modules and pipelines) made, completed, refused and taken back by a frame. buildsAbandoned counts the requests a renderer still held when a device rebuild let it go, which no frame will take back, and buildsPending is what the renderer drawing now has yet to take back: asked for and not yet in use, a finished result included until a frame installs it. So every request made is installed, abandoned or pending, and a reader waiting on buildsPending to reach zero reads pipelinesBuilt with every build it waited for counted, and can tell a queue draining from one that is stuck. modulesBuilt and moduleBuildMs are the shader modules compiled off the frame thread, parse and validation included.

pipelinesBuilt and buildMs are engine-wide totals; renderer.shaderCost() is the same cost broken down per program, with each one's permutation count.

Returns PipelineCache? — { supported, reason, path, restoredBytes, pipelinesBuilt, buildMs, offFrameBuilt, offFrameBuildMs, buildsSubmitted, buildsFinished, buildsFailed, buildsInstalled, buildsAbandoned, buildsPending, modulesBuilt, moduleBuildMs, buildWorkers, saves, savedBytes, dirty, lastError }, or nil before the renderer has drawn a frame

local c = renderer.pipelineCache()
print(("%d pipelines in %.1f ms, %d bytes restored"):format(c.pipelinesBuilt, c.buildMs, c.restoredBytes))

globals/renderer/presentedFrames

renderer.presentedFrames(opts: { scope: string?, frames: number?, interval: number?, velocity: boolean? }?) -> { PresentedFrame }

Parameters

opts { scope: string?, frames: number?, interval: number?, velocity: boolean? } (optional)

Returns { PresentedFrame }

globals/renderer/projectionJitter

renderer.projectionJitter() -> { { number } }

The offsets renderer.setProjectionJitter stated, each {x, y} in pixels from the pixel centre.

Returns { { number } } — The offsets; empty while every render samples pixel centres.

local offsets = renderer.projectionJitter()

globals/renderer/projectionOffset

renderer.projectionOffset() -> (number, number)

The sub-pixel projection offset in force for the main camera, in NDC.

Returns (number, number) — The x and y offset, both 0 when the projection samples pixel centres.

local ox, oy = renderer.projectionOffset()

globals/renderer/qualityTier

renderer.qualityTier() -> QualityTier

The device's quality tier and the renderer defaults it chose. The engine assigns every device one of "low", "medium" and "high" at boot, from its GPU family and limits, its memory and its display, so a world authored on a desktop plays on a phone at the phone's defaults without knowing it runs on one. A desktop, and any device whose GPU is outside the handheld families, is "high": the renderer as a desktop runs it, one scene pixel per display pixel. A phone renders the scene below a dense display's resolution with UI and text at full resolution, and spends less on shadows, textures and filtering the lower its tier. A lower tier chosen on a desktop renders the scene within that tier's pixel budget the same way, so renderer.setQualityTier alone is a whole graphics setting on every platform.

Every default the tier states stands wherever the world's settings name no value of their own (render.anisotropy), and the matching renderer call overrides any one of them: renderer.setResolutionScale, renderer.setTextureBudget, renderer.setAnisotropy. renderer.setQualityTier chooses another tier altogether.

Returns QualityTier: { tier, deviceTier, overridden, reason, tiers, device, defaults }

local q = renderer.qualityTier()
print(q.tier, q.reason, q.defaults.resolutionScale)

globals/renderer/raycast

renderer.raycast(origin: vec3, direction: vec3, maxDistance: number?, exclude: (string | { string })?) -> RenderRayHit?

Cast a ray against the geometry the renderer DRAWS and return the nearest surface it meets. Every visible mesh answers, whether or not anything gave it a rigid body — so a terrain, a procedurally generated mesh, or any plain Model reports the surface at a point, which is what a camera station, a prop, a sound source or a scatter standing on the ground needs to know. The answer is the nearest triangle of the mesh, so a sloped or terraced surface reports its height where it was asked rather than the extent of its bounding box.

distance is measured from origin along the direction given, so it is a world-space distance whenever that direction is a unit vector, and it is directly comparable to a physics.raycast distance along the same ray. normal is a unit vector turned to face back along the ray. exact is true when the answer is a triangle and false when it is the object's bounding box, which is what a mesh whose vertices live only in GPU buffers answers with. The triangles are the mesh's own, placed by the entity's transform and by the mesh's bind pose, so a surface a skinning or morph pass deforms on the GPU answers as the geometry the mesh holds.

EVERYTHING drawn is in scope — the ground you meant, and equally a character standing on it, a prop, a placeholder floor. The hit names its entity in entityId, exclude steps over the ones you do not want, and renderer.raycastAll hands back the whole column so you can pick the surface yourself. A height you did not expect is usually a nearer surface you did not mean to ask about, so read entityId before trusting a number.

Parameters

origin vec3 — vec3 ray start in world space
direction vec3 — vec3 ray direction; any length, the engine normalises
maxDistance number (optional) — number? how far the ray reaches, in world units. Default 1000
exclude (string | { string }) (optional) — (string | { string })? entity id, or ids, to step over

Returns RenderRayHit?

local hit = renderer.raycast({ x, 100, z }, { 0, -1, 0 }, 200)
if hit then entity(cameraId).position = { x, hit.point.y + 1.7, z } end

globals/renderer/raycastAll

renderer.raycastAll(origin: vec3, direction: vec3, maxDistance: number?, maxHits: number?, exclude: (string | { string })?) -> { RenderRayHit }

Cast a ray against the geometry the renderer draws and return every surface along it, nearest first. One entry per renderable the ray crosses — the nearest intersection with each — so a stack of surfaces reads as the order they stand in, and a caller after one particular surface finds it by entityId rather than hoping it is the nearest. Each entry carries the fields renderer.raycast returns.

Parameters

origin vec3 — vec3 ray start in world space
direction vec3 — vec3 ray direction; any length, the engine normalises
maxDistance number (optional) — number? how far the ray reaches, in world units. Default 1000
maxHits number (optional) — number? how many surfaces to return. Default 32
exclude (string | { string }) (optional) — (string | { string })? entity id, or ids, to step over

Returns { RenderRayHit }

for _, hit in renderer.raycastAll(eye, look, 500) do print(hit.entityId, hit.distance) end

globals/renderer/raytraceCapability

renderer.raytraceCapability() -> string

The active ray-tracing backend: "hardware" (GPU ray query) or "compute" (software traversal — the path on devices without hardware ray query, e.g. the web). The same ray-tracing features work on both.

Returns string "hardware" | "compute"

if renderer.raytraceCapability() == "hardware" then ... end

globals/renderer/raytraceStats

renderer.raytraceStats() -> { [string]: any }

What the ray-tracing acceleration structure holds, and what this session's frames have spent building it. A ray walks a structure built over the scene's geometry, and keeping it current is work a frame pays before it traces anything. On the "compute" backend geometry that has stood still long enough is filed under a static partition the frames after it leave alone: staticTriangles + dynamicTriangles = triangles, nodes is the hierarchy over them, fullRebuilds / partialRebuilds / reusedFrames count what the session's frames did, and trianglesRebuilt is what those rebuilds re-emitted, summed. On the "hardware" backend blas is the bottom-level structures cached, blasBuilt how many the last frame built, and tlasInstances what the top-level structure names. The counters are cumulative — sample, run the scene, sample again.

Returns { [string]: any } — table {backend, triangles, staticTriangles, dynamicTriangles, nodes, fullRebuilds, partialRebuilds, reusedFrames, trianglesRebuilt, blas, blasBuilt, tlasInstances}

local before = renderer.raytraceStats().trianglesRebuilt

globals/renderer/references

renderer.references(handleOrKind: any?, id: string?) -> RuntimeResourceStatus?

What holds a runtime resource right now — the answer a root scene load reads before releasing it. references names each live consumer the engine found: { by = "entity", id } for an entity wearing the material or mesh, "material" for a material whose slot names the texture, "instancedDraw", "camera", "sky", "lightmap", "ui" (a screen drawing it) and "postProcess" (an effect sampling it). handleHeld says whether a script still reaches a handle to it, assetBacked whether an asset stands behind it, ownerLive whether the component instance, scene load or feature that created it still stands, and held whether a hold pins it. origin reads "device" for a GPU texture the device holds that no script created — the one the cache loaded for an asset, the atlas the engine built — whose holders are the references, a handle and the asset. Runs a full garbage collection first, the same one renderer.collect runs, so a handle nothing reaches counts as let go and the row says what the next collection does with the resource. Yields for the frame the census runs on.

Parameters

handleOrKind any (optional) — The resource's handle, or its kind with the id second.
id string (optional) — The guid or key, when the first argument is a kind.

Returns RuntimeResourceStatus? — The resource's status, or nil for a key the registry does not record and the device holds no texture under.

local s = renderer.references(mat) for _, r in s.references do print(r.by, r.id) end

globals/renderer/reflectionEnvironment

renderer.reflectionEnvironment() -> {

What a reflective surface is reflecting. probes is how many reflection probes the shading blends; they are gathered highest priority first, each rank taking the coverage the ranks above it left, so a small interior probe ranked above the large exterior one it sits inside wins outright wherever it reaches full weight. ranks is the priority each of those probe slots was published with, in slot order. sky is whether the sky fallback is armed: with it, coverage no probe claims reflects the captured sky, and without it a surface outside every probe's radius falls back to the nearest probe alone. skyCaptured is whether the sky slot holds a capture — arming is refused until it does, since an uncaptured slot reflects black. skyStale is whether the sky changed after that capture, so the slot still depicts the sky from before the change until the engine's recapture, or an explicit environment.captureSky(), retakes it. The engine retakes a sky swapped for another material or turned off on the first frame the new sky can be drawn, and a sky whose look moved once a pacing window has passed since the last capture. skyPosition is where the capture was taken from, absent before one. skyCaptures counts the sky captures that have landed since the renderer was built, so one retaken from the same position reads as new, and skyMaterial names the material the sky pass draws from, absent when no sky is drawn. slots is how many cube slots the environment array holds right now: the sky's alone, at index skySlot, until a probe is captured into it, then that one, the slot a camera's stated environment is held in, and one per probe. maxProbes is how many of them probes may take, and resident whether the array has grown past the sky's single slot. Capture the sky with environment.captureSky().

Returns { probes, ranks, sky, skyCaptured, skyStale, skyPosition?, skyCaptures, skyMaterial?, resident, slots, skySlot, maxProbes }

local env = renderer.reflectionEnvironment()
print(("%d probes, sky fallback %s"):format(env.probes, tostring(env.sky)))

globals/renderer/release

renderer.release(handleOrKind: any?, id: string?) -> boolean

Let go of the hold renderer.hold placed. The resource stays until nothing else holds it and a collection releases it — the one a root scene load runs, or a direct renderer.collect(). A runtime mesh no script reaches a handle to is judged again once its hold goes, and released with no collection when nothing else holds it.

Parameters

handleOrKind any (optional) — The resource's handle, or its kind with the id second.
id string (optional) — The guid or key, when the first argument is a kind.

Returns boolean true when the registry knows the resource.

renderer.release(tex)

globals/renderer/releaseBaseline

renderer.releaseBaseline(generation: number) -> boolean

Drop the mark generation names, and every mark taken after it, once the caller that took it is done resetting to it. It touches the marks alone.

Parameters

generation number — The mark to drop, as renderer.markBaseline returned it.

Returns boolean — True when the mark stood.

renderer.releaseBaseline(mark) -- the run is over

globals/renderer/renderTargetLimits

renderer.renderTargetLimits() -> {

The size a render target may be on this device. maxDimension is the device's own maximum 2D texture dimension — the largest either side of a render target may take. maxPixels is how many pixels one render target may hold, so the RGBA8 image it reads back as fits in a single buffer on every platform the engine runs on, and maxSquare is the largest square that budget buys. A capture, a renderer.texture.create({ width, height }) or a render-to-texture camera past either bound is refused at the call with the reason, so ask here for the size to request.

Returns { maxDimension, maxPixels, maxSquare }

local lim = renderer.renderTargetLimits()
local w = math.min(want, lim.maxSquare)

globals/renderer/renderTargets

renderer.renderTargets() -> {

Every render target the renderer owns and what each one costs, measured from the texture that is allocated. One row per target, each carrying its name, whether it is resident, the bytes it holds while it is, its width/height/layers/mipLevels, and onDemand. An onDemand target exists only while something needs it: a target nothing writes into reads resident = false and bytes = 0 and appears again the frame something writes it, and one sized by content (the reflection-probe cube array) holds the slots content asked for. The scene's occlusion and reflection channels and its lit colour history (scene_occlusion, scene_reflection, scene_color_history) are resident at the scene's extent while a render-feature pass writes or reads them, and shadow_page_pool while a light casts into the virtual shadow map, with shadow_transmittance_pool beside it while a translucent caster casts into it. The scratch the draws into a render target have needed is reported as camera[<handle>].* rows: depth and motion vectors under any rasterized pass, and the occlusion, reflection and colour-history channels, the scene copies and the G-buffer over them under a camera's scene render. A draw builds what it needs, or takes the scratch a released target of the same size left. A target releases its set to the next target of its size when no live camera names it and when it is destroyed or resized, and frees it once sixty frames pass without a draw, so a target nothing draws into carries no such row. Released scratch waits as released_scratch[<width>x<height>].* rows for the next target of that size, held newest first within 64 MiB and freed once sixty frames pass without a target taking it. The colour image drawn into belongs to the texture cache while its target stands; a destroyed or resized target's colour texture waits on the same terms as a released_scratch[<width>x<height>].color row for the next target made at that size and format, which starts from it cleared to transparent black. totalBytes is what the resident targets hold together. allocations counts the render-target allocations the renderer has made since it was built, the colour textures of the live targets included, each on the first frame whose report holds it: a target taken back at a size it held before adds nothing, so the difference between two readings is what the frames between them allocated. Measured at the end of the last rendered frame.

Returns { targets, totalBytes, residentCount, allocations }

local rt = renderer.renderTargets()
print(("render targets: %.1f MiB over %d resident"):format(rt.totalBytes / 1048576, rt.residentCount))
for _, t in rt.targets do
if t.onDemand then print(t.name, t.resident, t.bytes) end
end

globals/renderer/resetToBaseline

renderer.resetToBaseline(generation: number?) -> number

Parameters

generation number (optional)

Returns number

globals/renderer/resolutionScale

renderer.resolutionScale() -> number

The fraction of the display resolution the scene is currently rendered at. Until something sets it, the scale the device's quality tier chose for its display (renderer.qualityTier().defaults.resolutionScale): 1 on a desktop at the high tier, below 1 on a dense handheld display and on a desktop at a lower tier.

Returns number

local s = renderer.resolutionScale()

globals/renderer/resolutionScaleMode

renderer.resolutionScaleMode() -> string

Where the render scale resolves: "present" or "reconstruct". "present" until something sets it.

Returns string

local m = renderer.resolutionScaleMode()

globals/renderer/setAnisotropy

renderer.setAnisotropy(level: number) -> number

Set the maximum anisotropy material textures are sampled with. Takes effect on the next frame for content already on screen — no reload, no texture re-upload. 1 is plain trilinear.

Parameters

level number — One of 1, 2, 4, 8, 16. Any other value is an error.

Returns number — The EFFECTIVE level after clamping to renderer.maxAnisotropy(), so asking for more than the device offers reports what was actually applied.

renderer.setAnisotropy(16)

globals/renderer/setBlendedBatching

renderer.setBlendedBatching(enabled: boolean) -> ()

Whether neighbours in a view's back-to-front blended order draw together. On by default: alpha-blended geometry is submitted farthest-first, and a stretch of neighbours in that order sharing a mesh, a material, a shader and a pose is submitted as one instanced draw over those neighbours, which puts the same members on screen in the same order out of a single submission. A run stops wherever a differently-drawn renderable sorts between two of its members, and a mesh of several primitives keeps a draw per renderable — both would otherwise move fragments through each other. Off, every blended renderable draws on its own at its own slot, so a transparent crowd costs a draw per member. The image is the same either way, which is what makes this the comparison a frame suspected of being formed by the batching is made against; renderer.drawStats().draws counts the difference.

Parameters

enabled boolean — boolean

Returns ()

renderer.setBlendedBatching(false)  -- a draw per blended renderable

globals/renderer/setFrameRateCap

renderer.setFrameRateCap(fps: number?) -> number?

Hold the engine to at most fps frames per second, or lift the cap with 0 or nil. The whole frame waits for its turn: scripts, physics, audio mixing and networking run once a frame as the drawing does, so a game capped at 30 runs at 30, and a phone spends no battery on frames past it.

Frames come due on an even grid. Where the display's refresh rate is known (a phone's panel, a desktop window's monitor, a browser's animation frames), the grid is a whole number of refreshes, the fewest whose rate stays inside the cap, so every frame stays on screen for the same number of vsyncs: on a 120 Hz display a cap of 30 shows each frame for 4 refreshes and 60 for 2, and a cap the refresh rate does not divide evenly delivers the next rate below it (50 on a 120 Hz display delivers 40). On Android the system compositor is told the rate as well, so a panel with more than one refresh rate can run at one the cap divides; with no cap it is told the panel's highest rate.

Read renderer.framePacing() for the effect: targetFps is the rate the cap delivers, refreshesPerFrame the refreshes each frame holds, displayRefreshHz the display it was worked out against, and deliveredFps the rate frames actually ran at. Takes effect on the next frame. frame_rate_cap under [render] in the world's .world_settings seeds it (0 for no cap), so a world states its cap once and every session opens under it; a call here stays in force until that value changes.

Parameters

fps number (optional) — number? Frames per second, 1 through 1000; 0 or nil lifts the cap.

Returns number? — The cap in force, or nil when frames are not capped.

renderer.setFrameRateCap(30) -- a steady 30 on any display 30 divides
print(renderer.framePacing().targetFps, "fps delivered under the cap")
renderer.setFrameRateCap(nil) -- every frame the display shows

globals/renderer/setGpuMemoryTracking

renderer.setGpuMemoryTracking(frames: number?) -> number

Set how often the GPU allocator sampler reads — one reading every frames frames — or turn it off with 0. It starts at 60, a reading a second at 60 Hz, so renderer.gpuMemory().allocator answers without anything arming it. Building the ledger walks every live allocation, which is why it is sampled rather than read every frame; the category figures cost nothing either way, and a reader between samples sees the most recent ledger, so a slow interval still answers.

Called with no argument it reports the interval in force and changes nothing, which is how something that retimes the sampler puts it back afterwards instead of restoring a number it assumed was the default.

Parameters

frames number (optional) — number? Frames between readings; 0 turns the sampler off. Omit to read the interval without changing it.

Returns number — The interval now in force.

renderer.setGpuMemoryTracking(60)
local was = renderer.setGpuMemoryTracking()
renderer.setGpuMemoryTracking(1)
-- ... take a tight reading ...
renderer.setGpuMemoryTracking(was)

globals/renderer/setLighting

renderer.setLighting(enabled: boolean) -> ()

Turn the scene's lighting on or off. Off, every lit surface renders as its own colour with no light on it, emissive included: the editor's lighting toggle, for reading a scene's materials without its lights. With no light applied there is no shadow either, so the shadow maps draw nothing while it is off, and the shadow switches apply again once lighting is back on. The sky and post-processing stay as they are.

Parameters

enabled boolean — true renders the scene lit, false unlit.

Returns ()

renderer.setLighting(false); print(renderer.lighting())

globals/renderer/setMaxFramesInFlight

renderer.setMaxFramesInFlight(frames: number) -> number

Set how many frames of GPU work may be outstanding before the renderer stops running ahead. One is the least overlap this can express — a frame's work is waited for as soon as the next frame has been submitted — which is the lowest latency and the lowest throughput; higher values let a slow frame build a longer backlog, and that backlog is memory. Takes effect on the next frame.

Answers the bound after clamping to [1, 8], so asking for more than the renderer honours reports what you actually got.

Parameters

frames number — number Frames of GPU work that may be outstanding, 1 through 8.

Returns number — The bound that took effect, after clamping.

renderer.setMaxFramesInFlight(1) -- lowest latency
print(renderer.setMaxFramesInFlight(99), "was clamped")

globals/renderer/setMaxTextureSize

renderer.setMaxTextureSize(px: number?) -> { px: number, devicePx: number }

The longest side a texture loaded from an asset may be resident at on this client, in texels. Such a texture larger than the cap is served from the first level of its mip chain that fits, so a 4096² texture under a 1024 cap holds its 1024² level and the levels below it, and the texture budget never takes it finer than that level. A changed cap moves the resident textures over the next frames; a raised one reads their finer levels back from their assets. A texture made from pixels with no asset behind it keeps every level it was made with, because the device holds its only copy. The device's own maximum 2D texture dimension bounds the cap from above. Call with no argument to lift the cap to the device's limit.

Parameters

px number (optional) — number? — texels, a whole number of at least 1; nil for the device's own limit

Returns { px: number, devicePx: number } — { px, devicePx } — the cap in force and the device's limit

renderer.setMaxTextureSize(2048)
renderer.setMaxTextureSize()  -- the device's own limit

globals/renderer/setMinScreenSize

renderer.setMinScreenSize(pixels: number) -> ()

Stop drawing an object once its on-screen radius falls below this many pixels. A few pixels across, an object carries no detail a viewer can resolve while still costing a full vertex and submission pass, and the cutoff drops it from the camera's draws entirely — 0, the default, keeps every object however small it lands. Measured from the object's own bounds against the camera's projection, so the same threshold means the same apparent size at any distance or field of view. Shadow casters have their own threshold in renderer.setShadowCasterCutoff.

Parameters

pixels number — number — smallest on-screen radius still drawn; 0 disables.

Returns ()

renderer.setMinScreenSize(4) -- drop anything under 4 px of radius
print(renderer.drawStats().compactedDrawn, "instances survived it")

globals/renderer/setOcclusionCulling

renderer.setOcclusionCulling(enabled: boolean) -> ()

Enable or disable occlusion culling. While enabled the renderer reduces the depth the visibility raster lays into a pyramid each frame and tests every renderable that cleared the frustum against it, dropping the ones another surface entirely covers before their geometry is submitted. The pyramid describes the frame being drawn, so an object that becomes visible this frame is never held back a frame.

Parameters

enabled boolean — boolean

Returns ()

renderer.setOcclusionCulling(true); print(renderer.cullStats().occlusionCulled)

globals/renderer/setPresentMode

renderer.setPresentMode(mode: string) -> string

Set how a presented frame reaches the display. fifo queues every frame and shows it on a vertical blank, which never tears and never drops one; mailbox replaces the queued frame with the newest, which does not tear and does not hold the renderer to the refresh rate; immediate presents as soon as a frame is ready and can tear; fifo_relaxed is fifo that tears rather than stall when a frame misses its blank; auto_vsync and auto_no_vsync leave the choice to the backend.

A surface that does not offer the mode presents fifo instead, so read renderer.framePacing().presentMode for what took effect and .presentModes for what this surface offers. Takes effect on the next frame.

Parameters

mode string — string One of "fifo", "fifo_relaxed", "mailbox", "immediate", "auto_vsync", "auto_no_vsync".

Returns string — The canonical spelling of the request — renderer.framePacing().presentMode is what the surface presents with, and differs when the surface does not offer the request.

renderer.setPresentMode("mailbox")
print(renderer.framePacing().presentMode, "is what the surface took")

globals/renderer/setProjectionJitter

renderer.setProjectionJitter(offsets: { { number } })

State where inside the pixel consecutive renders of a view sample, for the frames until it is stated again: a list of {x, y} offsets in pixels from the pixel centre, each within half a pixel. Every view the renderer draws (the viewport, each render-to-texture camera, each capture) walks the list by its own count of renders in a row, so a camera drawn beside the viewport takes nothing out of the viewport's sequence: a view's first render, and its first after a cut, samples the pixel centre, and its next renders take the offsets in order, wrapping. Each offset is applied in the view's own pixels. Velocity (@scene.motion) is measured against the offset-free projection, and a history target's state buffer carries the offset each render took. renderer.setProjectionOffset, while it is not (0, 0), and an upscale backend's own jitter both stand over the list on the viewport. An empty list samples pixel centres. An offset outside half a pixel is refused, naming it, and leaves the list in force.

Parameters

offsets { { number } } — The offsets, each {x, y} in pixels from the pixel centre.

renderer.setProjectionJitter({ { 0.25, -0.25 }, { -0.25, 0.25 } })

globals/renderer/setProjectionOffset

renderer.setProjectionOffset(x: number, y: number)

Offset the main camera's projection by a sub-pixel amount, in NDC, for the frames until it is set again. The offset is in NDC because that is the space it is constant in: one pixel is 2.0 / width across, so half a pixel is 1.0 / width. Velocity (@scene.motion) is measured against the offset-free projection, so a still scene reports no motion however the samples are placed — and picking resolves a click to the same ray either way. (0, 0) samples pixel centres.

Parameters

x number — Horizontal offset in NDC. One pixel is 2.0 / width.
y number — Vertical offset in NDC. One pixel is 2.0 / height.

renderer.setProjectionOffset(0.5 * 2 / w, -0.25 * 2 / h)

globals/renderer/setQualityTier

renderer.setQualityTier(tier: string?) -> QualityTier

Render at another quality tier than the one the device was assigned: "low", "medium" or "high", or nil / "default" for the device's own. A change of tier stands the new tier's defaults at once wherever the world's settings state none, replacing what an earlier renderer call set for the render scale, the shadow budgets and the anisotropy; choosing the tier already in force changes no setting. A call made after it, in the same script or later, overrides that one default again. A budget renderer.setTextureBudget names stands over the tier's texture cap until it is returned to the device's own. The choice holds for the process, across scene loads and restarts of the scene, until it is set again. A settings screen offering the player a quality choice is this call.

Parameters

tier string (optional) — string?: "low", "medium", "high", or nil / "default" for the device's own.

Returns QualityTier: the reading renderer.qualityTier() answers, for the tier now in force.

renderer.setQualityTier("low")
renderer.setQualityTier()  -- back to the device's own

globals/renderer/setRaytrace

renderer.setRaytrace(enabled: boolean) -> ()

Enable or disable GPU ray tracing. While enabled the engine builds the scene acceleration structure each frame so ray-tracing render features can trace against it; disabling stops the build (so it costs nothing until a ray-traced effect is active). Required before any ray-traced shadows / AO / reflections render.

Parameters

enabled boolean — boolean

Returns ()

renderer.setRaytrace(true); renderer.feature.create("rt_shadows")

globals/renderer/setResolutionScale

renderer.setResolutionScale(scale: number) -> number

Render the scene at a fraction of the display's resolution. Shading cost scales with pixel count and with nothing else, so this trades sharpness for frame time without taking anything out of the scene: at 0.5 the scene rasterizes a quarter of the pixels. What the smaller raster MEANS for everything composed on top of it — the post chain, the UI, every post-scene pass — is renderer.setResolutionScaleMode. The scene rows in profiler.gpuFrame() are what move.

Parameters

scale number — number — fraction of the display resolution, clamped to [0.25, 1].

Returns number — the scale in force after clamping.

renderer.setResolutionScale(0.7)

globals/renderer/setResolutionScaleMode

renderer.setResolutionScaleMode(mode: string) -> string

Where in the frame the render scale is resolved — what a smaller raster MEANS. "present" (the default) composes the whole frame at the raster: the scene, every post-process layer, the UI and every post-scene pass, carried up to the size it is shown once it is finished. A smaller raster is then a smaller picture, so an effect whose output is a function of pixel size — a dither ladder, a posterize, a pixel grid — resolves at the pitch it was authored at, and the frame carries one composite resolution throughout. "reconstruct" carries the scene up to the size it is shown BEFORE the post-scene phases and composes everything from there, which is what a temporal upsampler needs: it refines the upscale it is handed and has nowhere to put the detail it recovers unless the image it accumulates into is the display's. Either reading is honoured identically whether a viewport panel or the swapchain shows the frame, so what an editor draws and what a shipped runtime draws are the same picture of the same scene.

Parameters

mode string — string — "present" or "reconstruct".

Returns string — the mode in force.

renderer.setResolutionScaleMode("present")

globals/renderer/setShadowCaching

renderer.setShadowCaching(enabled: boolean) -> ()

Whether a shadow page that nothing changed is kept rather than drawn again. On by default: every light's pages — the sun's, and each spot, area and point light's — are drawn again only where a caster that changed reaches them. Off, every requested page is invalidated and drawn on every pass, which is what a shadow suspected of holding a stale image is compared against.

Parameters

enabled boolean — boolean

Returns ()

renderer.setShadowCaching(false)  -- draw every requested shadow page, every frame

globals/renderer/setShadowCasterCutoff

renderer.setShadowCasterCutoff(cfg: {
    minRadiusPx: number?,
    maxDistance: number?,
}) -> ShadowCasterCutoff

Set the shadow-caster cutoff. An omitted field keeps its current value, so a call can adjust one threshold without restating the other. Both are measured against the camera the frame draws from rather than against each light, so one setting covers every light's pages, and a caster that stops casting is one whose shadow the viewer could not have resolved. maxDistance is measured to the near side of the caster's bounding sphere, so a large object keeps casting while any part of it is in range. 0 releases a threshold; releasing both draws the casters the frame drew before either was set.

Parameters

cfg { minRadiusPx: number?, maxDistance: number?, } — The fields to change — minRadiusPx and/or maxDistance.

Returns ShadowCasterCutoff — The cutoff now in force.

renderer.setShadowCasterCutoff({ minRadiusPx = 2 })
renderer.setShadowCasterCutoff({ minRadiusPx = 1.5, maxDistance = 120 })

globals/renderer/setShadowConfig

renderer.setShadowConfig(cfg: {
    distance: number?,
    fadeFraction: number?,
    softness: number?,
    punctualMipFloor: number?,
    detailBias: (number | string)?,
    budgetMegabytes: (number | string)?,
}) -> ShadowConfig

Set the sun's shadow settings. Any omitted field keeps its current value, so a call can adjust one knob without restating the rest. Values are clamped: distance >= 0, fadeFraction [0, 1], softness [0, 1]. A distance of 0 shadows out to the camera's far plane, and a positive one caps it, which is what a scene bounding its shadow cost states; the sun's clipmap holds the levels that reach it. punctualMipFloor (a whole number, 0 to 5) caps how fine a spot, area or point light's map renders; it is refused outside that range, and every punctual light's map is drawn again under a new floor. detailBias ([-8, 8] levels) trades shadow detail for pages: each level coarser asks a quarter as many. budgetMegabytes (>= 1) bounds the depth the pages hold. Each takes "default" to return to the device's bias or the quality tier's budget, and a change to either sizes the pool again, which draws every page again.

Parameters

cfg { distance: number?, fadeFraction: number?, softness: number?, punctualMipFloor: number?, detailBias: (number | string)?, budgetMegabytes: (number | string)?, } — The fields to change — see ShadowConfig.

Returns ShadowConfig — The full config now in force.

renderer.setShadowConfig({ softness = 0.65, fadeFraction = 0.28 })
renderer.setShadowConfig({ distance = 240 })
renderer.setShadowConfig({ detailBias = 1, budgetMegabytes = 8 })
renderer.setShadowConfig({ detailBias = "default", budgetMegabytes = "default" })

globals/renderer/setShadowHero

renderer.setShadowHero(entity: string, padding: number?) -> ()

Give one caster a finer sun shadow, in a box around its world bounds.

The sun's virtual shadow map resolves every receiver at the level its screen footprint selects, so a character standing in the view gets the texel density every surface at its distance gets. A registered hero widens its bounds by padding into a box, and every receiver inside it requests and samples a level renderer.shadowHero().levelBias levels finer — each level halving the world a texel covers — where that level's window holds it, crossing back to its footprint's level at the box's edge. Nothing else about the shadow changes: the same casters reach it through the same filter.

Parameters

entity string — The entity whose renderables the box is made around.
padding number (optional) — How much room the box leaves around those bounds — for a pose that leaves the bind-pose box and for the filter that samples outside a silhouette. 1.0 holds them exactly.

Returns ()

local before = time.drawnFrameCount()
renderer.setShadowHero(player.id)
local h = renderer.shadowHero()
while h.frame <= before do
task.waitDrawnFrames(1)
h = renderer.shadowHero()
end
print(("hero shadow: %d levels finer over %.1f units"):format(h.levelBias, h.worldExtent))

globals/renderer/setShadowProxy

renderer.setShadowProxy(mesh: string, proxy: string) -> ()

Rasterize proxy in place of mesh in every shadow page. A shadow is a silhouette resolved at the resolution of a shadow map, so the triangles that carry a mesh's close-up detail write depth no reader can resolve — a decimated version of the shape, a level of its own LOD chain, or a hand-built hull casts the same shadow for a fraction of the geometry.

The registration is keyed by MESH, so one call covers every instance of it — entities and GPU-driven populations alike — and a crowd sharing that mesh stays one draw. The proxy is placed by whatever places the caster, its instance's own transforms, so it stands where the caster stands, at the caster's scale.

An entity caster keeps its own geometry where a stand-in could not be placed or deformed correctly: it is skinned (it rasterizes the post-skinned vertices written for its own mesh), it blends morph targets (whose deltas describe its own mesh and are read by vertex id), or its proxy would be placed by a different node of its model than the source mesh is. Either caster keeps it where the renderer holds no geometry under the proxy's guid. Each of those is counted in renderer.shadowProxies().

Nothing else in the scene draws a proxy, so this call is what brings it onto the GPU, and it raises where it cannot. A proxy already resident there is registered as it stands.

Parameters

mesh string — The mesh a caster draws, as a guid or any mesh reference.
proxy string — The mesh it rasterizes into shadow pages instead.

Returns ()

renderer.setShadowProxy(statueMesh, statueHullMesh)
print(renderer.shadowProxies().triangles, "vs", renderer.shadowProxies().sourceTriangles)

globals/renderer/setSkinnedBatching

renderer.setSkinnedBatching(enabled: boolean) -> ()

Whether skinned instances holding one pose draw together. On by default: instances of one mesh wearing one material and posed alike read the same post-skinned vertices, so the camera's colour passes submit them as a single instanced draw. Off, each skinned instance draws on its own at its own slot in every pass that rasterizes it. The image is the same either way, which is what makes this the comparison a frame suspected of being formed by the batching is made against — renderer.drawStats().draws counts the difference and renderer.skinningStats().poses says how many distinct poses it holds.

Parameters

enabled boolean — boolean

Returns ()

renderer.setSkinnedBatching(false)  -- a draw per skinned instance

globals/renderer/setSkinningPoseHold

renderer.setSkinningPoseHold(enabled: boolean) -> ()

Whether a pose the skinning pass already wrote is read as it stands. On by default: the pass produces an instance's vertices from its joint matrices, its node transforms, its blend weight and its blend model, so the slice holding a pose already holds what running the pass over those same inputs would write. A frame binding a pose whose slice still holds it reads the slice and dispatches nothing, and skinning costs what the frame's poses CHANGED — a paused clip, a skeleton nothing drives, a cast standing in one pose, each cost compute the frame the pose arrived and nothing after it. Off, every pose a frame binds is dispatched again, which is the comparison a frame suspected of reading a slice that no longer holds its pose is made against; the image is the same either way and renderer.skinningStats() counts the difference as dispatches against held. A mesh whose vertices a compute pass writes is dispatched every frame however this stands.

Parameters

enabled boolean — boolean

Returns ()

renderer.setSkinningPoseHold(false)  -- dispatch every pose, every frame

globals/renderer/setTextureBucketSlack

renderer.setTextureBucketSlack(bytes: number?) -> TextureBucketSlack

Bound the GPU memory texture-array buckets may hold beyond the textures in them. A bucket without room doubles while the total slack stays under this budget, and past it grows to the largest capacity whose slack the budget still holds, at least by what the arrival needs. A growth copies every layer the bucket held, so a large one that would gain few layers for what it copies opens another bucket of the same size instead. The default follows the device (a sixty-fourth of the system's RAM where the GPU draws from it, otherwise a sixteenth of the memory scale it reports; between 32 MiB and 512 MiB either way); nil restores it. bytes is a positive number of bytes, 1 or more — 0, a negative number or a non-finite one raises an error naming that range. A budget bounds the next growth; slack a bucket already holds stays until a compaction takes it. On the bindless backend no bucket exists, so the budget is recorded and reads back but bounds nothing. The textures themselves are bounded by renderer.setTextureBudget; read what the buckets hold with renderer.materialTable().

Parameters

bytes number (optional) — number? — the slack budget in bytes, or nil for the device's

Returns TextureBucketSlack — { slackBudgetBytes, deviceSlackBudgetBytes } — the budget now in force, and the device's own

renderer.setTextureBucketSlack(16 * 1024 * 1024)
renderer.setTextureBucketSlack(nil) -- the device's budget

globals/renderer/setTextureBudget

renderer.setTextureBudget(opts: TextureBudgetOpts?) -> TextureBudget

Bound the VRAM textures occupy, by keeping only the mip levels the frame is actually sampling once the textures outgrow the budget. Every device starts at its own budget and a bias of 0. Where the GPU draws from the system's RAM (every Android phone, an integrated GPU, a software rasterizer) and the system states how much, the budget is an eighth of that RAM; otherwise it is half the memory scale the adapter reports; between 256 MiB and 2 GiB either way, and held to 768 MiB on a medium-tier and 384 MiB on a low-tier device (renderer.qualityTier()). renderer.textureStreaming() names the rule (budgetRule) and the RAM it read (deviceMemoryBytes), and memory pressure narrows what the budget keeps (renderer.memoryPressure()). Pass { megabytes = 256 } to override the budget for the process; megabytes is a positive number, 1 to 65536, and 0 or a negative number is an error. A field set to the string "default" returns that one value to the device's own, a field the table leaves out keeps the value it has, and a field holding anything else — a boolean, another string — is an error naming what it holds. nil restores the device's budget and bias together. The budget belongs to the device the world is played on.

Below the budget every texture keeps its whole mip chain, whatever its surfaces cover. Mips stream down only while the whole chains of the streamable textures exceed the budget. Then each frame measures how many screen pixels ONE traversal of a texture's coordinate range covers on the surface that spans it widest, and asks for the mip level that serves that span one texel per pixel — the level the GPU picks from the fragment's own derivatives. A material with uvScale = 8 lays eight copies of its texture across a surface, so each copy spans an eighth of the surface and asks for three levels coarser than the surface's own size would. A shader that declares // @uv_space: world advances its coordinate over world units rather than over the mesh's UVs, so how many copies a surface carries follows how large that surface is. The textures whose surfaces cover the fewest pixels give up levels until the set fits. Detail climbs one level per frame, from the image already on screen, so a surface the camera approaches sharpens rather than popping, and no texture is taken below the level whose longest side is 64 texels.

bias shifts every measurement by whole mip levels either way — negative for finer than the sampling implies, positive for coarser — over a look that wants a different trade than one texel per pixel. It is clamped to [-8, 8].

The plan moves a texture whose demand the frame can measure: one at least 256 texels on its narrowest side, worn by a surface an entity draws. A texture a UI image, a post-process property or a render feature holds a view of stays whole, because nothing measures how much of the screen those cover.

Which textures the budget governs follows the surfaces the frame draws. A texture whose asset still holds its bytes is enrolled the frame a measured surface wears it, whenever it loaded, because a level change reads the levels it needs back from the asset; when the last such surface goes it leaves the set whole, at the level it uploaded at, and a surface reaching it again takes it back up. A texture a script uploaded has its pixels on the device alone, so it holds no copy while it is at its whole chain; the first time the budget takes it coarser its chain is read back off the device before any level is freed, and it holds that copy (renderer.textureMemory().streamSourceBytes) while it stays coarser. Once a surface has worn it and gone it leaves the set whole and stays out for the rest of the process. renderer.textureMemory().pinnedTextures counts those, together with the textures whose asset could not be read back and the ones a UI image, a post-process property or a render feature holds a view of.

Parameters

opts TextureBudgetOpts (optional) — { megabytes: (number | "default")?, bias: (number | "default")? }? — nil for the device's budget and bias

Returns TextureBudget — { megabytes, bias, deviceMegabytes, deviceBias } — the budget and bias now in force, and the device's own

renderer.setTextureBudget({ megabytes = 128 })
renderer.setTextureBudget({ megabytes = 128, bias = -1 })  -- one level finer everywhere
renderer.setTextureBudget({ megabytes = "default" })  -- the device's budget, the bias as it stands
renderer.setTextureBudget()  -- back to the device's budget and bias

globals/renderer/setTransmissionShadows

renderer.setTransmissionShadows(enabled: boolean) -> ()

Let translucent casters tint the sunlight they block instead of blocking it outright. A shadow map holds one depth per texel and is compared as a yes-or-no test, so stained glass, water and thin fabric all project the same black silhouette a wall does. With this on, a caster whose material declares transmission draws into a light-space transmittance map instead, and the colour it lets through multiplies into the directional light reaching whatever stands behind it. Stacked casters compose. A surface whose material blends writes no depth shadow at all and throws its shadow through this map alone: a transparent or premultiplied blend lets 1 - alpha of the light through untouched and its transmission of the rest tinted by its base colour, so red stained glass throws red light; a multiply blend lets its base colour through; an additive one blocks nothing. Opaque casters are unaffected, and a scene with no translucent caster allocates nothing and records no pass.

Parameters

enabled boolean — boolean

Returns ()

renderer.setTransmissionShadows(true)  -- stained glass tints the floor

globals/renderer/setUpscaleBackend

renderer.setUpscaleBackend(backend: string, quality: string?) -> string

Request a backend for the upscale seam — the step that brings a scene rendered below display resolution back up to it. "blit" stretches; "dlss" reconstructs from motion and history through NGX DLSS Super Resolution on an adapter that has it. The renderer resolves the request against its own capability and reports the outcome through renderer.upscaleBackend() — a request for "dlss" on an adapter renderer.upscaleCapabilities() omits it from resolves to "blit" there, answered truthfully rather than echoed back here. quality names the reconstruction quality mode a reconstructing backend derives its render scale from; the blit keeps whatever renderer.setResolutionScale() set and reads no value from quality. @builtin::systems.temporalUpscale.temporalUpscale is the content-facing surface built on this primitive — it carries the policy this raw setter carries none of: refusing "dlss" where unavailable, refusing quality alongside an explicit render scale, and keeping its own reconstruction from stacking with a reconstructing backend.

Parameters

backend string — string — "blit" or "dlss".
quality string (optional) — string? — "quality" | "balanced" | "performance" | "ultraPerformance", read only by a reconstructing backend.

Returns string — the backend now requested, echoed back before the renderer resolves it against this adapter's capability.

renderer.setUpscaleBackend("dlss", "balanced")

globals/renderer/setViewportSize

renderer.setViewportSize(width: number, height: number) -> { width: number, height: number }

Draw at this many pixels for the rest of the task that asks. The engine's drawing surface is resized to it, and everything measured against that surface follows within a frame: getViewportSize(), ui.screenSize(), the layout every UI screen rebuilds from it, each camera's aspect, and what a capture taken in that task encodes. This is how one session checks a responsive layout at a second shape: a HUD written against ui.screenSize() is re-laid-out at the new size, so an anchored element is drawn where that shape puts it rather than scaled from where the boot size put it.

The size is held for the task that asked for it (an execute call is one task), and when that task ends, however it ends, the surface goes back to the size it had before. So a check reads what it needs at the new shape inside the call that set it, and the window or browser canvas a person is looking at is handed back as it was. While several tasks hold a size the surface draws at the one asked for last, and when that task ends it draws at the size the task before it holds. Asked for outside any task, from a component's lifecycle, the size stands until another is asked for.

Who honours the size depends on who owns the surface. A headless engine and a browser canvas own theirs and are resized exactly; a browser page that lays its canvas out at a new size meanwhile ends every hold, and the page's size stands. Where an OS window owns it, the window manager is asked and has the last word: a tiled or maximized window keeps the size it has. Read renderer.surfaceSize() on a later frame for what was realized.

The new size is in force from the NEXT frame, so wait a frame before reading it back: renderer.surfaceSize() read straight after the call still reports the size the surface had.

The logical UI space is normalised to about 1280 points wide, so a resize to a size of the same aspect moves ui.pixelRatio() and leaves ui.screenSize() where it was, while a resize that changes the aspect changes that shape too. Both resize the surface.

A size larger than the device draws is refused naming the bound, which renderer.maxViewportExtent() reports.

Parameters

width number — number — width in pixels, at least 1.
height number — number — height in pixels, at least 1.

Returns { width: number, height: number } — the size asked for, in whole pixels.

renderer.setViewportSize(390, 844); task.waitFrames(2); print(ui.screenSize()) -- the phone shape, then the surface goes back when the call ends

globals/renderer/shaderCache

renderer.shaderCache() -> ShaderCache?

What the shader compile gate's store of baked WGSL held, answered and wrote back. Compiling a .shader wraps the author's body in its framework, expands every #include, and hands the result to naga to parse and validate — work that is a pure function of the text going in, and that a launch would otherwise repeat for every shader it draws with. The store keeps that baked text across launches.

restoredEntries and restoredBytes are what a previous launch left that this one read back. hits counts the compiles answered out of the store and misses those that ran in full; savedMs sums what each hit's own recorded compile had cost, against compileMs, what the misses spent. stale counts the misses whose key was held but whose #included modules had changed underneath — an entry records every module its expansion consumed, so editing a module invalidates exactly the shaders that included it and leaves the rest. The store also keeps each resolve gate pass a surface shader's resolve module earned, so a launch does not validate the same resolve text again: verdictHits counts the verdicts answered out of it and verdictMisses the ones whose gate ran, apart from the compile counts above.

entries and bytes are what the store now holds, evictions how many a write dropped to stay inside its bounds, and saves / savedBytes / dirty describe writing it back, deferred until a burst of compiles settles. persistent is false where a launch has nowhere to keep artifacts and reason says why; location is the file, or the browser store, they are kept in. restoreState is how the read of what a previous launch left has gone — pending while it is still out (a browser answers through a promise, so a launch reaches its first frames before it lands), restored once entries came back, empty when there were none to come back, failed when what was there could not be read, and none where a launch keeps nothing. A cold, missing or corrupt store leaves every shader compiling from source with identical output, and lastError then names what went wrong.

Returns ShaderCache? — { persistent, reason, restoreState, location, restoredEntries, restoredBytes, hits, misses, stale, verdictHits, verdictMisses, entries, bytes, compileMs, savedMs, saves, savedBytes, dirty, evictions, lastError }, or nil on a build with no renderer

local c = renderer.shaderCache()
print(("%d hits, %d misses, %.1f ms saved"):format(c.hits, c.misses, c.savedMs))

globals/renderer/shaderCost

renderer.shaderCost() -> { ShaderCost }

What each program has cost in pipeline builds, beside the compile gate's most recent word about it. variants is how many pipelines this engine has built for it — one per (target format, vertex layout, render-state key) permutation reached — and buildMs what those builds cost, both summed since engine start. A pipeline the driver's own store restored is not built and so is not counted, so a second launch on the same adapter reports less than the first. status is compiled, failed or pending, and error carries the compiler's message for a failure. Ordered by cost, most expensive first. shader is the key the program is registered under; asset names the .shader asset a guid@checksum key is a version of.

Returns { ShaderCost }

local top = renderer.shaderCost()[1]; print(top.asset or top.shader, top.variants, top.buildMs)

globals/renderer/shaderVariants

renderer.shaderVariants() -> { ShaderVariants }

Every shader that declares optional features, and the programs its materials have made it compile. Each row carries the features the shader declares, the base program it ships as, and one entry per variant with the features that variant holds — so the permutation count a scene's materials are spending is a number to read rather than something to infer from compile time. A shader whose variants reach budget compiles no more; the materials asking for further feature sets draw with the base program.

Returns { ShaderVariants } — An array of ShaderVariants, one per feature-declaring shader.

for _, s in renderer.shaderVariants() do print(s.shader, #s.variants, s.budget) end

globals/renderer/shadingOf

renderer.shadingOf(subject: string | { [string]: any }) -> ShadingReading

What the renderer is shading ONE subject with, taken from the document the renderer publishes — the call a system holding a handle makes to find out whether what reaches the screen is its own material or the magenta placeholder standing in for it, without reading the engine log. subject is an entity that draws or the registry key of a material. state reads itsMaterial where the renderer bound the program the material names, errorMaterial where it bound the placeholder instead, stalePipeline where the pipeline drawing it was built before that program's most recent compile, nothingBound where the renderer resolved no pipeline for it, pending where this call is the one that armed per-draw recording and the frame after it publishes, and unknown where the renderer holds a resolution under no such subject. A fault state carries the renderer's own reason from the closed set renderer.drawDiagnostics() names — plus materialNotPrepared, which a material subject reads where the renderer prepared nothing under that key — the compiler's detail, the program the material asked for and the bound one; means states the reading in a sentence. A material subject answers from the renderables drawing with it, and from the renderer's record for the material itself where a draw registered against the material carries no row of its own; a subject that several renderables draw answers with a refused one wherever there is one. The reading follows the renderer, so a program that compiles on a later edit puts the subject back on itsMaterial from the frame the renderer draws it with again.

Parameters

subject string | { [string]: any } — The entity — a proxy from entity(...) or an entity-id string — or the material, as its registry key or the MaterialHandle renderer.material.create returned.

Returns ShadingReading

local s = renderer.shadingOf(emitter); if s.state ~= "itsMaterial" then print(s.means) end

globals/renderer/shadowCacheStats

renderer.shadowCacheStats() -> {

What the main camera's newest counted run did with the shadow pages it asked for, in pages, by the kind of light that owns them. A page is drawn again only when something it holds changed: its light moved, a caster over it moved or appeared or vanished, a caster's geometry or material changed, or the caster changed pose or moved the nodes its parts are placed by. Anything else keeps the depth already in the pool, so a scene that stops moving reads rendered 0 while its pages stay cached. A caster moving on one side of a point light dirties only the pages of the faces it reaches. A mesh whose vertices a compute pass writes — a population, or a mesh built from a compute buffer — re-renders the pages it stands over every frame. spot counts spot and area lights.

These are totals over every light of a kind. renderer.shadowViews() is the same run one light at a time.

Returns { directionalRendered, directionalCached, spotRendered, spotCached, pointRendered, pointCached, rendered, cached }

local s = renderer.shadowCacheStats()
print(("shadow pages: %d drawn, %d cached"):format(s.rendered, s.cached))

globals/renderer/shadowCaching

renderer.shadowCaching() -> boolean

Whether a shadow page may keep the depth it already holds.

Returns boolean

globals/renderer/shadowCasterCutoff

renderer.shadowCasterCutoff() -> ShadowCasterCutoff

How small, and how far away, a caster may get before it stops writing depth into any shadow page. A shadow page rasterizes a caster's whole triangle count whatever the shadow it produces ends up covering, so an object the viewer resolves a fraction of a pixel of, and one past the range the scene cares about, each cost a full depth draw per page it reaches for detail nothing reads. Both thresholds are 0 — released — until something sets them.

Returns ShadowCasterCutoff — The cutoff in force — see ShadowCasterCutoff.

local c = renderer.shadowCasterCutoff(); print(c.minRadiusPx, c.maxDistance)

globals/renderer/shadowConfig

renderer.shadowConfig() -> ShadowConfig

The sun's shadow settings now in force. distance bounds how far from the camera the sun shadows (the world's render.shadow_distance until renderer.setShadowConfig sets it); fadeFraction and softness shape how the shadow is sampled. The sun's virtual shadow map follows each receiver's screen footprint, so its resolution follows the view. punctualMipFloor is the finest mip a spot, area or point light's map renders. detailBias and budgetMegabytes are what a caller stated, or "default"; what the renderer sized the pages with is renderer.shadowSizing(). The table is one renderer.setShadowConfig takes back.

Returns ShadowConfig — The full config — see ShadowConfig.

print(renderer.shadowConfig().distance)

globals/renderer/shadowHero

renderer.shadowHero() -> ShadowHeroReport

The registered hero caster and what the last drawn frame's box was. A frame that made no box says why in decline.

entity is the live registration; everything else is the box of the drawn frame frame names, and stands until the renderer draws another. So a reading taken straight after setShadowHero or clearShadowHero describes the registration before the call: read the drawn frame count as the change is made, and take the report once its frame is past it.

Returns ShadowHeroReport — See ShadowHeroReport.

local h = renderer.shadowHero()
print(h.active, h.levelBias, h.decline)
local before = time.drawnFrameCount()
renderer.setShadowHero(player.id)
task.waitDrawnFrames(1)
print(renderer.shadowHero().frame > before)

globals/renderer/shadowMemory

renderer.shadowMemory() -> {

How much GPU memory the virtual shadow map's pools hold right now, in bytes. One pool serves every light — the sun, and every spot, area and point light that casts. pool is the depth page pool with the sun's static cache laid out as it, held while any light casts; transmittance is the transmittance pool, held while a translucent caster casts into it (renderer.renderTargets() names them shadow_page_pool, shadow_static_cache and shadow_transmittance_pool). Both read 0 while released.

Returns { pool, transmittance, total } in bytes

local m = renderer.shadowMemory()
print(("shadows: %.1f MiB"):format(m.total / 1024 / 1024))

globals/renderer/shadowPageBreakdown

renderer.shadowPageBreakdown() -> { { [string]: any } }

The breakdowns renderer.watchShadowPages gathered since the last call, oldest first, one per run of the main camera. levels has a row per clipmap level: texelWorld (world units per texel), the pages it requested, held (resident), drew (rendered) and had dirtied by the frame's caster changes (invalidated), the pages those changes reach in its window (reached) and the smallest size a change projects to there (smallestRecordTexels). records lists each caster change with its light-space sizeWorld and, per level, the texels it covers and the pages it reaches. cull is what the page cull walked (slotsWalked caster slots, ncMembers casters drawn from their own vertex buffers) and kept (pairs, casters tested, admitted and cutOff, chunks, visible clusters, ncPairs, batchPages). pages has a row per batch page: its level, the casters it drew and how many changed this frame (dynamicCasters), its chunks, clusters and triangles with the changed casters' shares, and its pairs of casters drawn from their own vertex buffers (ncPairs, dynamicNcPairs).

Returns { { [string]: any } } — An array of breakdowns, empty when none was read since the last call.

renderer.watchShadowPages(30)
task.wait(1)
for _, run in renderer.shadowPageBreakdown() do print(run.frame, #run.pages) end

globals/renderer/shadowPageStats

renderer.shadowPageStats() -> {

What the sun's virtual shadow map did on the main camera's newest run the GPU counted. The sun's shadow lives in pages around the camera; a run marks the pages the receivers on screen ask for and draws the ones that are missing or that a moved caster dirtied. A still camera over a still scene runs nothing, so ran reads false and runs stands still while frames climbs. The counts come back from the GPU a few frames after the run, so frame names the frame they describe. proxyPairs counts the (caster, page) pairs a shadow proxy stood in for, proxyTriangles the triangles those pairs drew from the proxy and sourceTriangles what the casters' own meshes would have drawn there. castersDropped counts the casters drawn outside the cluster lists (a deforming mesh, a population whose mesh has no cluster set) that the newest frame could not list, which cast no shadow that frame.

Returns { frame, ran, runs, frames, requested, rendered, pending, starved, resident, clusters, batchPages, proxyPairs, proxyTriangles, sourceTriangles, invalidated, invalidateOverflow, castersDropped }, or nil before the first run is read back

local s = renderer.shadowPageStats()
if s then print(("%d pages asked for, %d drawn"):format(s.requested, s.rendered)) end

globals/renderer/shadowProxies

renderer.shadowProxies() -> ShadowProxyReport

The shadow proxies in force and what the shadow pages did with them. triangles and sourceTriangles are what the main camera's newest counted run submitted through proxies and what the same (caster, page) pairs would have submitted from the source meshes — the before/after of every registration, equal while nothing is proxied.

Returns ShadowProxyReport — See ShadowProxyReport.

local p = renderer.shadowProxies()
print(("shadow triangles: %d of %d"):format(p.triangles, p.sourceTriangles))

globals/renderer/shadowSizing

renderer.shadowSizing() -> ShadowSizing

What the sun's virtual shadow map was sized with. Its pool of pages holds about twice the pages a frame at the render resolution asks for, within the depth budget (the quality tier's: 64, 12 and 8 MiB at high, medium and low, or the one renderer.setShadowConfig states). It is sized again only when the render resolution, the detail bias or the budget changes. A run that finds no slot for a page it asked for raises overflowBias by a level, up to 4, until the pool is sized again.

Returns ShadowSizing — See ShadowSizing.

print(renderer.shadowSizing().depthMegabytes)

globals/renderer/shadowViews

renderer.shadowViews() -> ShadowViewReport?

Every shadow-casting spot, area and point light's virtual shadow map on the main camera's newest counted run, one row each.

renderer.shadowCacheStats() and renderer.shadowPageStats() sum these counts over every light; this is the same run one light at a time. The sun's pages are reported by renderer.shadowPageStats().

Each row names its light and says what its pages did: how many its receivers asked for, how many it holds, how many were drawn, dirtied, starved or left for a later run, and what the page cull did with the casters it walked for the light. camera carries the main camera's share of the instanced-compaction counters.

The counts come back from the GPU a few frames after the run, and frame names the frame they belong to. A light that stops casting leaves views.

The frame names its lights only while something is reading them, so this call asks the frames after it to publish theirs and waits out the first one. Nil on an engine that renders no frame at all.

Returns ShadowViewReport? — See ShadowViewReport.

local report = renderer.shadowViews()
for _, view in report.views do
print(("%s %d (%s): %d of %d pages resident, %d drawn"):format(
view.role, view.index, view.light, view.pagesResident,
view.pagesRequested, view.pagesRendered))
end

globals/renderer/signalMemoryPressure

renderer.signalMemoryPressure(level: string) -> ()

Report memory pressure the way the system does: the engine releases what level releases on its next frame, and engine.onMemoryPressure subscribers hear it with source "script". Content about to need a lot of memory (a large scene load, a big decode) frees room ahead of it with this, and "normal" lifts a level a script put in force. level is "normal", "moderate", "low" or "critical"; anything else is an error naming the four. Read the outcome with renderer.memoryPressure() once a frame has passed.

Parameters

level string — "normal", "moderate", "low" or "critical"

Returns ()

renderer.signalMemoryPressure("moderate") -- free unsampled mips before a big load

globals/renderer/skinnedBatching

renderer.skinnedBatching() -> boolean

Whether skinned instances holding one pose draw together.

Returns boolean

globals/renderer/skinningPoseHold

renderer.skinningPoseHold() -> boolean

Whether a pose already written into its slice skips its dispatch.

Returns boolean

globals/renderer/skinningStats

renderer.skinningStats() -> {

What the last frame's skinned instances cost. A skinned instance is posed by a compute pass that writes its vertices into a shared pool, and instances holding the same pose read one slice of that pool and the single dispatch that fills it. instances is how many were posed, poses how many distinct poses they held, and dispatches how many dispatches those poses cost this frame — so a crowd whose members move together costs what its poses cost rather than what its head count does, while members at different animation times each hold their own pose and pay for it.

held is how many of the frame's poses cost no dispatch at all. The pass produces a slice from what the pose is made of, so a slice an earlier frame filled already holds what running it again would write, and a pose still wearing that slice is read as it stands. Skinning is paid for by the poses that CHANGED: a cast standing still reads dispatches 0 beside a held equal to its poses, and the two add up to poses in any frame. A slice counts as filled once a render carrying its dispatch has been drawn: the frame's own view, or a render of the scene into a target, such as a camera drawing into its render texture. A frame whose view the renderer declines (a headless one nothing is consuming, a browser behind its GPU) and that renders the scene into no target records nothing, so a pose that arrives over such frames is counted in dispatches on each of them and reads as held from the frame after the first one that records it. A wait for a crowd to settle is therefore counted in drawn frames (task.waitDrawnFrames).

reusedSlices is how many of the frame's poses took a slice the pool already held — one a retired pose gave back, or one a pose nothing has asked for this frame was holding — rather than one cut from pool the engine had never used. A scene whose poses keep changing reads a non-zero count beside a poolBytes that stays where it was.

liveBytes is what the slices holding this frame's poses occupy, against unsharedBytes — what the same instances would occupy with a slice each. poolBytes is what the pool holds; a previous-position buffer of the same size rides alongside it so skinned deformation reaches motion vectors.

The same pass draws a soft body's model through the body's elements (its cage). cages is how many soft bodies it draws that way, cageBytes what their buffers occupy, and cageUploads how many of them wrote their elements' maps to the GPU this frame: a soft body at rest writes nothing, and a removed one takes its cage with it.

Returns { instances: number, poses: number, dispatches: number, held: number, reusedSlices: number, liveBytes: number, unsharedBytes: number, poolBytes: number, cages: number, cageBytes: number, cageUploads: number }

local s = renderer.skinningStats()
print(("%d instances over %d poses"):format(s.instances, s.poses))
print(("%d poses dispatched, %d read as they stood"):format(s.dispatches, s.held))
print(("skinned vertex storage: %d B of an unshared %d B"):format(s.liveBytes, s.unsharedBytes))

globals/renderer/splat/components

renderer.splat.components(bytes: any?, convention: string?) -> (SplatComponents?, string?)

Decode a Gaussian splat capture — a Niantic .spz (gzipped or raw) or a 3DGS .ply — into the GPU-ready byte pools a render feature uploads. records is the packed splat array at recordBytes per splat (position, log scale, quaternion, DC colour + opacity); sh is the quantized higher-order spherical-harmonics pool at shStrideWords u32 words per splat, empty at degree 0. A pure decode (no GPU work): upload the pools with shaderRef:createBuffer + buf:writeBytes and draw them with a kind = "splat", channel = "gaussian" pass.

Parameters

bytes any (optional) — Capture bytes — .spz or .ply, as a buffer or a binary string.
convention string (optional) — Source axis convention: "rightDownFront" (the default, what COLMAP-trained captures use) or "engineNative" for a capture already in engine space.

Returns (SplatComponents?, string?) — { records, sh, count, shDegree, shStrideWords, recordBytes, boundsMin?, boundsMax?, antialiased, format }, or (nil, err).

local c = renderer.splat.components(vfs.read("/source/captures/ceramic.spz"))

globals/renderer/surfaceSize

renderer.surfaceSize() -> { width: number, height: number }

The whole drawing surface in pixels — the window, the browser canvas or the headless framebuffer. This is the size renderer.setViewportSize sets and the size every UI screen is laid out over, so it is what says whether a resize was realized, on every layout including an editor one whose viewport panel holds a smaller rect than the window. { width = 0, height = 0 } before the first frame has drawn.

Returns { width: number, height: number } in pixels.

local s = renderer.surfaceSize()

globals/renderer/temporal/held

renderer.temporal.held() -> boolean

Whether a hold is pinning the per-frame clock right now.

Returns boolean — True while at least one renderer.temporal.hold stands.

if renderer.temporal.held() then print("frame is pinned") end

globals/renderer/temporal/hold

renderer.temporal.hold(at: number?, options: TemporalHoldOptions?) -> () -> ()

Pin the clock every per-frame effect draws itself against, and return the release. While the hold stands, renderer.temporal.now answers at instead of the running clock, so film grain and every other field redrawn each frame is redrawn as the same field. Two renders taken under holds at the same instant therefore agree pixel for pixel wherever the scene itself has not moved, which is what makes one frame comparable with another. Holds nest: the innermost names the instant, and the clock runs again once the last release is called. Each release takes its own hold off the stack whatever order the releases come in, so two callers holding at once — two captures in flight together — each end their own hold and the clock runs again when both have. exclusive takes the clock for the owner key the call states: while that hold stands, a hold is admitted only when it states the same key, and every other one is refused with an error naming the key and the instant holding it. That is what lets one caller wind the clock to the second it means to photograph and keep it there while another agent drives the same engine. The key is what an owner presents to take a nested hold of its own, and what renderer.temporal.release hands the clock back by. A capture taken while the hold stands renders at the held instant, and a deterministic capture states that key as its own owner option to be admitted under the hold and photograph the instant it pinned.

Parameters

at number (optional) — The instant to pin the clock at, in seconds. Two holds that state the same instant produce the same field; the default 0 is that shared instant.
options TemporalHoldOptions (optional) — owner is the key this hold is taken under, and an exclusive hold states one. A hold that states no key is labelled with the caller the call is attributed to, which names the calling session where the transport attested one and the ZeroMind account otherwise. exclusive takes the clock for the stated key until the hold is released.

Returns () -> () — A function that releases this hold. Calling it twice releases once.

local release = renderer.temporal.hold() ; local picture = tools.use("capture", "fromPosition", { position = { 0, 2, 8 } }) ; release()
local release = renderer.temporal.hold(46.0, { exclusive = true, owner = "stage-air" })

globals/renderer/temporal/now

renderer.temporal.now() -> number

The instant a per-frame effect should draw itself at: the innermost hold's instant while one stands, and seconds since boot otherwise. A system that redraws a field every frame reads this rather than the running clock, and a capture asking for a repeatable frame then gets one.

Returns number — Seconds — pinned while a hold stands, running otherwise.

local params = { grainTime = renderer.temporal.now() }

globals/renderer/temporal/onChange

renderer.temporal.onChange(listener: (number) -> ()) -> () -> ()

Register a listener called with the pinned instant whenever it changes — a hold taken, a hold released — and return the unsubscribe. A system whose shader reads the clock out of a GPU buffer registers here, so the buffer carries the pinned instant before the frame that hold was taken on is drawn rather than a frame later.

Parameters

listener (number) -> () — Called with the instant now in force, in seconds.

Returns () -> () — A function that removes this listener.

local stop = renderer.temporal.onChange(function(t) print("temporal clock", t) end)

globals/renderer/temporal/owner

renderer.temporal.owner() -> { id: string?, name: string?, at: number, exclusive: boolean }?

The hold naming the instant the clock answers right now: who took it, what instant it pinned, and whether it took the clock exclusively. Several agents drive one engine at once and a hold any of them takes moves the clock every registered field is redrawn against, so this is how a caller sees that another agent holds it before its own instant is quietly replaced — and, when exclusive is true, id is the key a hold of its own states to be admitted, and the key renderer.temporal.release hands the clock back by. id and name are nil for a hold that stated no key and that the engine attributes to no agent.

Returns { id: string?, name: string?, at: number, exclusive: boolean }? — { id, name, at, exclusive } for the standing hold, or nil when the clock is running.

local who = renderer.temporal.owner()
if who ~= nil and who.exclusive then print(who.name, "holds the clock at", who.at) end

globals/renderer/temporal/release

renderer.temporal.release(owner: string) -> number

Hand the clock back by the key its holds were taken under, and report how many came off. A hold stands until its release is called, and the release is a closure the call that took the hold holds: a caller that takes a hold in one call and comes back in another, and a task that ends between the two, both leave the clock pinned with nobody holding a release for it. Naming the key is how the clock runs again, and how a caller refused by an exclusive hold takes one over.

Parameters

owner string — The key the holds to release were taken under — what owner stated when they were taken, which renderer.temporal.owner reports.

Returns number — How many holds came off the stack.

renderer.temporal.release("stage-air")

globals/renderer/texture/capture

renderer.texture.capture(texture: string | { [string]: any } | AssetRef) -> string

Request a CPU readback of the GPU texture texture names (e.g. a camera's rendered output). Returns a result key to pass to a TextureCpuHandle's :encode() once the readback completes. Takes every form that names a texture — the TextureHandle create returned, the guid renderer.texture.list hands out, a TextureCpuHandle or a texture AssetRef.

Parameters

texture string | { [string]: any } | AssetRef — The texture to read back — a TextureHandle, a guid, a TextureCpuHandle or a texture AssetRef.

Returns string The capture result key.

local key = renderer.texture.capture(cameraTarget)

globals/renderer/texture/cpuCreate

renderer.texture.cpuCreate(width: number, height: number, fill: any?) -> TextureCpuHandle

Allocate a blank CPU image (RGBA8) filled with a solid colour and return a TextureCpuHandle. Compose into it with canvas:blit(src, x, y, w, h), then canvas:encodeJpeg() / :encodePng() for the bytes; :unload() drops it.

Parameters

width number — number Canvas width in pixels.
height number — number Canvas height in pixels.
fill any (optional) — Optional { r, g, b, a } (0-255) solid fill; defaults to opaque white.

Returns TextureCpuHandle

local c = renderer.texture.cpuCreate(1024, 576, { 255, 255, 255, 255 })

globals/renderer/texture/cpuFromBytes

renderer.texture.cpuFromBytes(bytes: buffer | string, encodeOpts: any?) -> TextureCpuHandle

Load engine-native ZTEX bytes — or an encoded image (png / jpg / webp) — into the CPU store under a fresh guid and answer the CPU handle, for pixels that come from somewhere other than a texture asset: a data.ztex read as a file, a payload held in memory. The pixels stay at the format they were encoded in. DEFAULT: handle:unload() once done with them.

Parameters

bytes buffer | string — The ZTEX or image bytes.
encodeOpts any (optional) — { colorSpace?, compression?, generateMipmaps?, maxDimension?, filter?, textureType?, alpha?, alphaIsTransparency? } — the opts renderer.texture.encodeFromImage takes — applied when the bytes are an encoded image and need the engine-native encode. maxDimension bounds ZTEX bytes too: a payload that stores a mip chain is read from the first level whose longest side is within it, that level alone, so a thumbnail of a 4096² texture decodes a level near the thumbnail's size.

Returns TextureCpuHandle

local cpu = renderer.texture.cpuFromBytes(assert(vfs.read(path))); local png = cpu:encodePng(); cpu:unload()
local thumb = renderer.texture.cpuFromBytes(assert(vfs.read(texPath .. "/data.ztex")), { maxDimension = 256 })

globals/renderer/texture/create

renderer.texture.create(src: any?, guid: string?) -> TextureHandle

Create (or fetch) a GPU texture resource and return its TextureHandle. src: a TextureCpuHandle from texRef:load() (CPU→GPU under the asset's guid, idempotent); raw pixels {rgba, width, height, srgb?, format?} (a flat widthheight4 byte payload, 0-255, row-major, top-to-bottom, RGBA — a buffer, a binary string, or a number array; format = "rgba16f" uploads an HDR texture instead, where rgba carries float channel values); a TextureHandle (returned as-is); or render-target dimensions {width, height, name?, format?} with no pixel source — an empty GPU texture a render pass writes into (camera output, UI surface) and that samples like any other texture. format names the colour format the target is allocated in, and the passes drawing into it are built for that format: "rgba8unorm" / "bgra8unorm" (the two eight-bit channel orders, either of which a surface may carry), "rgba16f" / "rgba32f", "rg16f" / "rg32f", "r16f" / "r32f". Each also answers to its spelled-out width ("rgba16float", "r32float", and so on), in any case. Omit it and a screen target takes the scene's own format ("rgba16f") — a screen-sized target holds a picture of the scene, and that is the format the scene is held in: linear light with headroom above the display's white and no 256-step floor under a dark gradient. A target that is not screen-sized and names no format takes the surface's. "rgba8" is what a mask, a weight or an occlusion term asks for, where 0..1 is the whole range; hand one back over @scene.color and the frame loses every highlight above white and bands every gradient, and the engine says so in the log. A float format carries what eight bits quantize — positions, velocities, HDR. Any other format raises an error naming every name that works, so a target is allocated in the format it was asked for or not at all. A render target takes filter the way raw pixels do: "nearest" keeps its own pixels square wherever something draws it larger than it is — a viewport widget, a magnified capture — which is what an image whose pixels ARE the subject needs, since a 64x32 panel holds no detail between its pixels to interpolate; "linear" (the default) smooths between them. It also takes screen (the engine keeps it the size of the image being drawn), screenScale (the fraction of that size it takes) and screenSpace ("scene", the default, or "composite" — the image the post-scene phases draw into, which is the display's own resolution while the renderer presents the viewport itself and the scene's size while a UI viewport panel owns the presentation). A scene-space target is resized for every render target drawn and cleared before an offscreen one; a composite-space target follows the presented frame alone, which is what lets a pass keep an accumulation in it. One scene-space screen target is therefore one resource every render target draws through in turn, so its guid holds the last one's image at the last one's size, and a value read back from it belongs to whichever render target was drawn last. A screen target a live Camera draws into is that camera's output instead: the frame's other render targets neither resize nor clear it, so what the camera drew there stays for as long as the camera does. A reading that has to be the viewport's own comes from screenSpace = "composite", or from a target created without screen. history = true makes a history target instead, the target an accumulation across frames lives in: it is screen-sized and keeps one pair of images per view a frame draws (the viewport, every render-to-texture camera, every capture). For the length of each view's render, the handle's guid names the image that render writes and previous.guid the image that view's previous render wrote, so a pass binds the two like any other targets and reads its own view's history wherever it runs; no other view resizes, clears or overwrites them. Before each render of a view the engine writes the compute buffer named by the handle's state (one vec4<f32>, bound through a pass's buffers): x is 1 when previous holds that view's previous image and 0 when it does not (the view's first render, its first after a cut or a change of size, or a render of the view that ran without the target, and then previous reads transparent black), y how many renders in a row have carried the history, zw the render's sub-pixel projection offset in pixels (renderer.setProjectionJitter). A view's images are freed when the view stops drawing, and all of them with the handle. NEVER takes an AssetRef — load the CPU first.

Parameters

src any (optional) — A TextureCpuHandle, raw pixels, a TextureHandle, or render-target dimensions.
guid string (optional) — Optional v4 guid for a NEW runtime texture — the asset identity the texture is filed under, which a material's texture slot resolves through. Minted when absent. Ignored for the CPU-handle and render-target paths.

Returns TextureHandle

local gpu = renderer.texture.create(texRef:load())
local gpu = renderer.texture.create({ rgba = pixels, width = 16, height = 16 })
local px = buffer.create(16 * 16 * 4); local gpu = renderer.texture.create({ rgba = px, width = 16, height = 16 })
local rt = renderer.texture.create({ width = 512, height = 256, name = "panel_rt" })
local hdr = renderer.texture.create({ width = 512, height = 256, name = "cam_rt", format = "rgba16f" })
local led = renderer.texture.create({ width = 64, height = 32, name = "panel", filter = "nearest" })
local hist = renderer.texture.create({ width = 1, height = 1, history = true, name = "accum" })

globals/renderer/texture/createFromAsset

renderer.texture.createFromAsset(ref: string | AssetRef, encodeOpts: any?, keepCpu: boolean?) -> TextureHandle

Put a .texture asset on the GPU under its own guid and answer its handle at once. The asset's bytes are decoded off the frame and the texture lands on the device when the decode finishes, a frame or more later: a material naming the guid draws the shader's default for that slot until then and rebinds when it arrives, and renderer.texture.isResident reports the arrival. The decoded pixels are dropped once uploaded unless keepCpu holds them in the CPU store for textureRef:load()-style reads. An asset the device already holds is answered from the shape the device reports, without reading the asset's bytes and without a second decode.

Parameters

ref string | AssetRef — A texture AssetRef, or a string naming one (guid, identity, name or source path).
encodeOpts any (optional) — { colorSpace?, compression?, generateMipmaps?, maxDimension?, filter?, textureType?, alpha?, alphaIsTransparency? } — the opts renderer.texture.encodeFromImage takes — applied when the primary is an encoded source image and needs the engine-native encode (a .ztex primary is decoded as-is).
keepCpu boolean (optional) — Keep the decoded pixels in the CPU store after the upload.

Returns TextureHandle

local h = renderer.texture.createFromAsset(texRef) -- bind h.guid on a material; it draws once resident

globals/renderer/texture/decode

renderer.texture.decode(bytes: buffer | string, opts: { straightAlpha: boolean? }?) -> (any, any, any, any)

Decode a texture payload to its pixel buffer. Takes the two shapes the renderer's own texture loader takes, told apart by their leading bytes:

an engine-native ZTEX payload — handed back at the texel format the payload was written in, so a height field read back here keeps every bit it was authored with. A ZTEX holding block-compressed or verbatim source-image levels decodes to "rgba8".
source image bytes — png, jpeg, gif or webp, straight off disk or out of a capture — decoded to "rgba8" at whatever colour type, bit depth or interlacing the file was written with. This is the call that reads the pixels of a screenshot.

The fourth return names the format the buffer came back in: "rgba8" (4 bytes/texel, channels 0-255), "rgba16" (8 bytes/texel, 16-bit unsigned normalized channels 0-65535) or "rgba32f" (16 bytes/texel, float channels).

The pixels are the texels the payload stores. With opts.straightAlpha = true, a payload encoded with alpha = "premultiplied" has its colour divided back by alpha, in the colour space it was premultiplied in — the straight-alpha pixels renderer.texture.encode takes. A texel whose alpha is 0 keeps the black premultiplying left in it.

Parameters

bytes buffer | string — A ZTEX payload or source image bytes.
opts { straightAlpha: boolean? } (optional) — { straightAlpha? }

Returns (any, any, any, any) — (string?, number?, number?, string?) pixels, width, height, format — or (nil, errmsg) where errmsg is in the 2nd slot.

local pixels, w, h = renderer.texture.decode(vfs.read("/source/tmp/shot.png"))
local straight = renderer.texture.decode(vfs.read(texPath .. "/data.ztex"), { straightAlpha = true })

globals/renderer/texture/destroy

renderer.texture.destroy(texture: string | { [string]: any } | AssetRef) -> boolean

Release the GPU texture texture names. For an empty render-into texture (camera output, UI surface) this also frees its render scratch; for an uploaded runtime texture it drops the GPU resource (and any CPU shadow). After this, renderer.texture.list stops answering for the guid. Takes every form that names a texture — the TextureHandle create returned, the guid the listing hands out, a TextureCpuHandle or a texture AssetRef.

Parameters

texture string | { [string]: any } | AssetRef — The texture to release — a TextureHandle, a guid, a TextureCpuHandle or a texture AssetRef.

Returns boolean true when a texture was known under the guid.

renderer.texture.destroy(rt)
renderer.texture.destroy(renderer.texture.list()[1].guid)

globals/renderer/texture/encode

renderer.texture.encode(rgba: any?, width: number, height: number, opts: any?) -> (string?, string?)

Encode raw pixels into an engine-native ZTEX payload (the on-disk texture content). The CPU codec behind the texture assetType's onCreate. Called from a task (an execute chunk, a component hook the engine lets wait, task.spawn), a "uastc" or "bc7" encode runs where the frame loop keeps stepping: on a worker where the build has one, and otherwise a bounded slice of its blocks per frame (__texture.encodeBudget, reported by __texture.encodeJobs). The call returns on the frame the payload lands. Code that cannot yield, and a coroutine the script resumes itself (coroutine.create / coroutine.wrap), encode in place, and every other compression is encoded on the call. The payload is the same bytes either way. opts.compression selects what the texels are stored as: "rgba8" (default, 8 bits/channel, rgba is widthheight4 bytes), "uastc" (UASTC 4x4 blocks, a quarter of the RGBA8 size, transcoded on each device to the block format it holds; rgba is widthheight4 bytes and each side is resampled to the nearest power of two, halfway rounding up), "bc7" (BC7 blocks, a quarter of the RGBA8 size; rgba is still widthheight4 bytes and both sides a multiple of 4) or the high-precision data compressions "rgba16" (16-bit unsigned normalized, widthheight8 bytes) / "rgba32f" (32-bit float, widthheight16 bytes) — for height/displacement fields, baked lightmaps, and other data rasters 8 bits quantize visibly. The two high-precision compressions store rgba verbatim, reject opts.generateMipmaps / opts.maxDimension, and take opts.colorSpace = "linear" only. opts.colorSpace is "srgb" (default, colour data) or "linear" (normal maps, masks, heights). opts.input ("rgba8" / "rgba16" / "rgba32f") names the texel format rgba arrives in when it is not the one opts.compression stores — the pixels renderer.texture.decode hands back at their own width, say — and the encoder carries each channel across as the same normalized value, natively, before storing it. opts.textureType records what the texture holds — "color" (default), "normalMap" (when the encode generates mips, every level of the chain, the base included, is renormalised to unit length), "mask" or "sprite". opts.alpha is what the encode does to alpha: "input" (default, kept as given), "premultiplied" (RGB multiplied by alpha — in linear light under srgb, as stored under linear — the source a material with the premultiplied blend mode composites) or "none" (every texel opaque, RGB kept). opts.alphaIsTransparency = true fills each fully transparent texel with the average colour of its non-transparent neighbours, spreading outward, alpha kept at 0 — before the alpha mode and the mips, so filtering never pulls black into an edge. All three are recorded in the payload and reported by renderer.texture.info. The sampler opts are recorded the same way: opts.filter ("linear", the default, or "nearest"), opts.wrapU / opts.wrapV ("repeat", the default, "clamp" or "mirror" — how the texture is sampled past its edges) and opts.anisotropy ("world", the default, following the world's render.anisotropy, or a fixed "1" / "2" / "4" / "8" / "16", also taken as the number). opts.effort ("fastest", the default, "fast" or "thorough") is how long a uastc encode searches for each block: every effort writes a payload the same size on the GPU, and a higher one lands closer to the source. opts.uastcRdo ("none", the default, "low", "medium" or "high") is how strongly a uastc encode trades accuracy for a payload that compresses smaller. A block-compressed encode (uastc, bc7) called from a task that can yield runs off the frame and the task waits for it, so the world keeps running while it cooks.

Parameters

rgba any (optional) — Pixel payload at opts.input's (else opts.compression's) native byte width — a buffer, a binary string, or a number array.
width number — number
height number — number
opts any (optional) — { colorSpace?, compression?, uastcRdo?, generateMipmaps?, maxDimension?, filter?, wrapU?, wrapV?, anisotropy?, textureType?, alpha?, alphaIsTransparency?, input? }

Returns (string?, string?) ZTEX bytes, or (nil, errmsg).

globals/renderer/texture/encodeFromImage

renderer.texture.encodeFromImage(bytes: buffer | string, opts: any?) -> (string?, string?)

Encode source image bytes (png/jpg/webp/…) into an engine-native ZTEX payload. Used by the texture importer / assetType onChange. A "uastc" or "bc7" encode called from code that can yield runs where the frame loop keeps stepping, as renderer.texture.encode describes. opts.compression is "rgba8" (default), "uastc", "bc7", "rgba16" or "rgba32f", opts.colorSpace "srgb" (default) or "linear", and opts.uastcRdo as renderer.texture.encode takes it. "rgba8" with generateMipmaps = false, a maxDimension the image fits within (or none), alpha = "input" and no alphaIsTransparency stores the source image bytes verbatim. "rgba16" and "rgba32f" decode a still image at its own bit depth and store one linear level: a 16-bit PNG keeps every bit, an 8-bit channel v is stored as v * 257 (rgba16) or v / 255 (rgba32f). The frames of an animated image (GIF, APNG, animated WebP) are decoded at 8 bits, so a 16-bit APNG's frames are stored as v * 257 of their 8-bit values. Like encode, they reject opts.generateMipmaps and a maxDimension the image exceeds. opts.textureType records what the texture holds — "color" (default), "normalMap" (when the encode generates mips, every level of the chain, the base included, is renormalised to unit length), "mask" or "sprite". opts.alpha is what the encode does to alpha: "input" (default, kept as given), "premultiplied" (RGB multiplied by alpha — in linear light under srgb, as stored under linear — the source a material with the premultiplied blend mode composites) or "none" (every texel opaque, RGB kept). opts.alphaIsTransparency = true fills each fully transparent texel with the average colour of its non-transparent neighbours, spreading outward, alpha kept at 0 — before the alpha mode and the mips, so filtering never pulls black into an edge. All three are recorded in the payload and reported by renderer.texture.info, as are the sampler opts filter, wrapU, wrapV and anisotropy, which take the values renderer.texture.encode takes.

Parameters

bytes buffer | string — source image bytes.
opts any (optional) — { colorSpace?, compression?, uastcRdo?, generateMipmaps?, maxDimension?, filter?, wrapU?, wrapV?, anisotropy?, textureType?, alpha?, alphaIsTransparency? }

Returns (string?, string?) ZTEX bytes, or (nil, errmsg).

globals/renderer/texture/frameSchedule

renderer.texture.frameSchedule(texture: string | AssetRef) -> { number }?

The times at which each layer of a timed texture stops being shown, in seconds from the start of the sequence — the running total of the layer display times, so the last entry is the length of one pass.

This is the form a sampler reads a sequence through: a time is turned into a layer by finding the first entry it has not passed, whatever the individual layer times are. It is what the schedule slot of the builtin animatedTexture shader holds, one entry per layer.

A texture whose layers carry no timing — a still image, a sprite sheet, a LUT stack — has no schedule and answers nil.

Parameters

texture string | AssetRef — The texture — a guid, an identity, a name, a path, or a texture AssetRef.

Returns { number }? one cumulative end time per layer, or nil.

local ends = renderer.texture.frameSchedule(texRef) -- {0.1, 0.15, 0.35}

globals/renderer/texture/info

renderer.texture.info(ztex: buffer | string) -> (any, any)

Read the header of an engine-native ZTEX payload without copying the pixels. Returns its format, its compression ("rgba8", "bc7", "rgba16" or "rgba32f" — what the texels are stored as) and colorSpace ("srgb" or "linear" — how they are read when sampled), dimensions, mip count, the sampler recorded in the payload from the asset's settings — filter ("nearest" or "linear"), wrapU and wrapV ("repeat", "clamp" or "mirror") and anisotropy ("world", or a fixed "1" / "2" / "4" / "8" / "16"); a payload written before wrap and anisotropy were recorded reads repeat / repeat / world — what the payload was encoded as — textureType ("color", "normalMap", "mask" or "sprite"), alpha ("input", "premultiplied" or "none") and alphaIsTransparency (boolean); a payload written before these were recorded reads color / input / false — and the payload's layer shape.

layers counts the array layers the payload carries and isArray is true past one — the answer to "am I about to sample a texture_2d_array?", available before anything samples it. animated is true when those layers are a sequence in time; then frameDelaysMs lists each layer's display time in milliseconds in display order, and durationMs totals one pass. An animated image imports as one layer per frame, so layers is its frame count. A still texture reports layers = 1, isArray = false.

Parameters

ztex buffer | string — ZTEX bytes.

Returns (any, any) — (table?, string?) { format, compression, colorSpace, textureType, alpha, alphaIsTransparency, width, height, mipCount, filter, wrapU, wrapV, anisotropy, layers, isArray, animated, frameDelaysMs?, durationMs? }, or (nil, errmsg).

local i = renderer.texture.info(bytes); if i.animated then print(i.layers, "frames", i.durationMs, "ms") end

globals/renderer/texture/isCpuResident

renderer.texture.isCpuResident(texture: string | { [string]: any } | AssetRef) -> boolean

True if this texture has a resident CPU copy in the guid-keyed CPU store.

Parameters

texture string | { [string]: any } | AssetRef — The texture — a TextureHandle, a TextureCpuHandle, a guid, or a texture AssetRef.

Returns boolean

if renderer.texture.isCpuResident(handle) then ... end

globals/renderer/texture/isResident

renderer.texture.isResident(texture: string | { [string]: any } | AssetRef) -> boolean

True if a GPU texture is resident under this texture's guid.

Parameters

texture string | { [string]: any } | AssetRef — The texture — a TextureHandle, a TextureCpuHandle, a guid, or a texture AssetRef.

Returns boolean

print(renderer.texture.isResident(handle))

globals/renderer/texture/keepCpuOnLanding

renderer.texture.keepCpuOnLanding(texture: string | { [string]: any } | AssetRef, keep: boolean) -> boolean

Set whether the decode on its way to this texture — scheduled by textureRef:handle() or renderer.texture.createFromAsset and not yet on the device — keeps its decoded pixels in the CPU store once it lands. A texture's keepCpu setting changing while its decode is in flight reaches the texture through this.

Parameters

texture string | { [string]: any } | AssetRef — The texture — a TextureHandle, a TextureCpuHandle, a guid, or a texture AssetRef.
keep boolean — Keep the CPU copy once the decode lands.

Returns boolean — true when a decode was on its way; false changes nothing.

renderer.texture.keepCpuOnLanding(texRef, true)

globals/renderer/texture/list

renderer.texture.list() -> { any }

Every texture currently registered, ordered by guid — the ones a script created and the ones that reached the device through an asset alike. Each entry carries the guid, where it came from (origin is "asset" for a texture the asset path uploaded), and whether the GPU still holds it. A resident entry also carries the bytes it costs, its dimensions, its texel format and the mip levels the device holds (mipLevels), so the listing sums to renderer.textureMemory(). A streamable one carries streamOrigin — "asset" when a level change reads the levels it needs back from the asset, "device" when no asset answers for it and its whole chain on the device is the only copy, "retained" when it is resident coarser and holds the levels read back off the device. A resident entry carries how a material binding it samples it, as the texture records it: filter ("linear" / "nearest"), wrapU and wrapV ("repeat" / "clamp" / "mirror") and anisotropy ("world" or a fixed "1".."16"), and samplerAnisotropy, the level the texture's sampler takes: anisotropy resolved against renderer.anisotropy() and the device's ceiling, and 1 for a nearest texture — the entry of the renderer's sampler table the texture's handle names. A script-created entry also carries held — whether renderer.hold pins it for the session — and scene, the load that created it. renderer.references("texture", guid) says what is still holding a row, and renderer.collect() releases the rows nothing holds.

Returns { any } — Array of { guid, origin, owner?, scene?, held?, resident, bytes?, width?, height?, mipLevels?, format?, filter?, wrapU?, wrapV?, anisotropy?, samplerAnisotropy?, compressed?, streamable?, streamOrigin? }.

for _, t in ipairs(renderer.texture.list()) do print(t.guid, t.bytes) end

globals/renderer/texture/loadCpu

renderer.texture.loadCpu(ref: string | AssetRef, encodeOpts: any?) -> TextureCpuHandle

Load a .texture asset's pixels into the ONE guid-keyed CPU store (the Disk→CPU step) and return a CPU handle for per-pixel access (no GPU readback). The handle holds NO pixels — only the guid, dims and texel format plus the read/write/encode/unload ops (which read the Rust store). The pixels stay at the format they were authored in: handle.format is "rgba8", "rgba16" or "rgba32f", and :readPixel reports channels in that format's own units. Called by texRef:load(). DEFAULT: upload to the GPU then handle:unload().

Parameters

ref string | AssetRef — A texture AssetRef (carries .guid, reads its primary via getBytes), or any string asset.ref resolves to one — a guid, an identity, a name or a source path.
encodeOpts any (optional) — { colorSpace?, compression?, generateMipmaps?, maxDimension?, filter?, textureType?, alpha?, alphaIsTransparency? } — the opts renderer.texture.encodeFromImage takes — applied when the primary is an encoded source image and needs the engine-native encode (a .ztex primary is loaded as-is).

Returns TextureCpuHandle

local cpu = texRef:load(); local r,g,b,a = cpu:readPixel(3, 4); cpu:unload()

globals/renderer/texture/readback

renderer.texture.readback(texture: string | { [string]: any } | AssetRef) -> TextureCpuHandle

Read a runtime GPU texture's pixels back to CPU and return a TextureCpuHandle for them — the GPU→CPU half of the runtime-texture freeze path. A texture made with renderer.texture.create keeps no CPU copy, so persisting it (:encode() → asset.create("texture", …)) reads it back here first. Yields until the readback completes (a frame or two). After it returns the pixels are resident in the guid-keyed CPU store: :readPixel, :writePixel, :getInfo, :encode, :unload all work. Errors if the texture never becomes GPU-resident.

A SCENE-space screen-sized render target is one resource shared by every render target drawn — the viewport, an offscreen capture, a camera rendering into a texture — resized and re-derived for each of them in turn. The copy is taken ahead of all of them for the frame, so what a readback of its guid answers is the content of the last frame the renderer drew: the presented view's own image at the presented resolution, since the presented view is the sink that draws last. A request made while the renderer is holding frames back is carried to the next frame it draws rather than being answered from a target another sink left standing, so a readback can wait a frame longer than the copy itself takes.

Parameters

texture string | { [string]: any } | AssetRef — The texture — the TextureHandle renderer.texture.create returned, a guid, or a texture AssetRef.

Returns TextureCpuHandle

local cpu = renderer.texture.readback(handle); local ztex = cpu:encode(); cpu:unload()

globals/renderer/texture/reencode

renderer.texture.reencode(ztex: buffer | string, opts: any?) -> (string?, string?)

Re-encode an engine-native ZTEX payload against new opts. A "uastc" or "bc7" re-encode called from code that can yield runs where the frame loop keeps stepping, as renderer.texture.encode describes. opts is the complete encode spec, exactly as for renderer.texture.encode: every field it leaves out takes its default, so textureType, alpha, alphaIsTransparency, filter, wrapU, wrapV and anisotropy come from opts — not from the payload's header — and a payload re-encoded with { compression = "uastc" } alone loses a nearest filter or a premultiplied alpha it carried. renderer.texture.withSampler changes how a payload is sampled without re-encoding it. The payload's layers and their display times are kept, whatever compression either side is in. Each layer's texels are read back with straight alpha (a payload encoded with alpha = "premultiplied" has its colour divided back by alpha), carried to the width the new compression stores as the same normalized value, and encoded exactly as renderer.texture.encode would encode them; a payload that embeds its source image is re-encoded from that image, read at its own bit depth. What an earlier encode discarded stays discarded: after alpha = "none" the alpha is opaque. The texture assetType re-encodes a container that holds no source image through this call when its settings change. A block-compressed encode called from a task that can yield runs off the frame while the task waits.

Parameters

ztex buffer | string — A ZTEX payload — a buffer or a binary string.
opts any (optional) — { colorSpace?, compression?, uastcRdo?, generateMipmaps?, maxDimension?, filter?, wrapU?, wrapV?, anisotropy?, textureType?, alpha?, alphaIsTransparency? }

Returns (string?, string?) ZTEX bytes, or (nil, errmsg).

local wide = renderer.texture.reencode(vfs.read(texPath .. "/data.ztex"), { compression = "rgba16", colorSpace = "linear", generateMipmaps = false })

globals/renderer/texture/tone

renderer.texture.tone(histogram: any?) -> TextureTone

Reduce a histogram to what the picture's tone IS: where its darkest and brightest pixels sit, where the body of it sits, and how much of it is standing on the floor or the ceiling — all in code values on the 0-255 scale the pixels were delivered at.

span (max - min) is the whole range including a single stray pixel; spread (p95 - p5) is the range the body of the picture occupies, which is the reading that says whether a shot is legible. A frame whose subject is modelled and shaded but delivered inside a few code values reads a large mean and a tiny spread, and no mean alone can tell that apart from a frame with a subject in it.

crushed and clipped are the shares of the picture at code 0 and at code 255, each 0..1 — what a shot loses to the floor and to the ceiling.

Parameters

histogram any (optional) — A histogram from cpu:histogram().

Returns TextureTone

local t = cpu:tone(); if t.spread < 24 then error("the shot is flat") end

globals/renderer/texture/unloadCpu

renderer.texture.unloadCpu(texture: string | { [string]: any } | AssetRef)

Drop this texture's resident CPU copy from the guid-keyed CPU store. The explicit release for a texture asset's keepCpu setting turning off.

Parameters

texture string | { [string]: any } | AssetRef — The texture — a TextureHandle, a TextureCpuHandle, a guid, or a texture AssetRef.

renderer.texture.unloadCpu(texRef)

globals/renderer/texture/update

renderer.texture.update(texture: string | { [string]: any } | AssetRef, src: any?) -> TextureHandle

Overwrite the GPU texture texture names IN PLACE, under the same guid, from new pixels. Never writes a .texture file — the play-mode mutate path. Takes every form that names a texture — the TextureHandle create returned, the guid renderer.texture.list hands out, a TextureCpuHandle or a texture AssetRef. src is raw pixels, or a TextureCpuHandle (texRef:load(), renderer.texture.loadCpu, renderer.texture.readback), whose pixels arrive with everything they carry: texel format, colour space, mip chain, layers and filter. Every material, UI image and post-process effect sampling the guid draws the new texture. Returns a handle carrying the new dimensions: the handle it was given, refreshed, and a handle over the guid otherwise.

Parameters

texture string | { [string]: any } | AssetRef — The texture to update — a TextureHandle, a guid, a TextureCpuHandle or a texture AssetRef.
src any (optional) — New raw pixels {rgba, width, height, srgb?, format?} — rgba as a buffer, a binary string, or a number array — or a TextureCpuHandle.

Returns TextureHandle — A TextureHandle for the updated texture.

globals/renderer/texture/withSampler

renderer.texture.withSampler(ztex: buffer | string, opts: { filter: string?, wrapU: string?, wrapV: string?, anisotropy: (string | number)? }) -> (string?, string?)

Change how an engine-native ZTEX payload is sampled without re-encoding it. opts names any of filter ("linear" / "nearest"), wrapU and wrapV ("repeat" / "clamp" / "mirror") and anisotropy ("world", or "1" / "2" / "4" / "8" / "16", also taken as the number); a field it omits keeps the payload's value. The result is the same bytes with only the sampler bits of the header replaced, so every level is byte-identical to the input — the cheap path the texture assetType takes when a settings change touches only these four fields. An unrecognised value raises.

Parameters

ztex buffer | string — A ZTEX payload — a buffer or a binary string.
opts { filter: string?, wrapU: string?, wrapV: string?, anisotropy: (string | number)? } — { filter?, wrapU?, wrapV?, anisotropy? }

Returns (string?, string?) ZTEX bytes, or (nil, errmsg).

local clamped = renderer.texture.withSampler(vfs.read(texPath .. "/data.ztex"), { wrapU = "clamp", wrapV = "clamp" })

globals/renderer/textureMemory

renderer.textureMemory() -> {

What the GPU texture cache holds, split by whether the texture is block-compressed. compressedBytes and uncompressedBytes are what those textures cost in VRAM, measured from each texture's own format and mip chain — so a .texture whose settings name compression = "uastc" appears in the compressed columns at a quarter of what the same image costs as RGBA8. blockCompressionSupported is whether this adapter samples any block-compressed format; where it is false a block-compressed payload is uploaded decoded and lands in the uncompressed columns instead, so the texture is present everywhere and compressed where the hardware allows it. blockFormat names the format block-compressed textures are resident in on this adapter ("bc7" on desktop, "astc4x4" on a phone, "etc2" on a phone without ASTC) and is nil where it samples none. A texture stored as UASTC is transcoded to blockFormat; one stored as BC7 keeps its blocks where the adapter samples BC7 and is re-encoded into blockFormat where it does not, each at the same one byte per texel. Measured at the end of the last rendered frame. streamableTextures is how many of them the texture budget can move the base mip level of, split by where a level change reads the levels it needs from: assetStreamedTextures are read back from the asset they came from and hold nothing in system memory; the rest no asset answers for, so the device holds their only copy. Such a texture holds no copy of its own while it is at its whole mip chain; the first time the budget takes it coarser its chain is read back off the device before any level is freed, and it holds that copy while it stays coarser. retainedTextures counts the ones holding a copy, and streamSourceBytes is what those copies occupy in system memory — bytes that are not VRAM. pinnedTextures counts the textures big enough to stream that stand at a level nothing can move: no asset answers for them and they left the streamable set, the asset behind them could not be read back, or a UI image, a post-process property or a render feature holds a view of them. A texture out of the streamable set only because no measured surface wears it stands in neither count: a surface reaching it takes it back up, so its level moves again as soon as there is a footprint to move it by. byFormat names every GPU format a cached texture is held in, keyed by the format's name (Bc7RgbaUnormSrgb, Astc { block: B4x4, .. }, Rgba8UnormSrgb, ...), with how many textures it holds and their bytes. heldAboveCap lists the textures resident larger than renderer.setMaxTextureSize allows: one whose mip chain has no level within the cap, and one with no asset behind it, which keeps every level it was made with. decodedFallback lists the UASTC textures this device holds as texels because it samples no block format.

Returns { blockCompressionSupported, blockFormat?, compressedTextures, compressedBytes, uncompressedTextures, uncompressedBytes, streamableTextures, assetStreamedTextures, retainedTextures, pinnedTextures, streamSourceBytes, byFormat, heldAboveCap, decodedFallback }

local tm = renderer.textureMemory()
print(("%d compressed textures hold %.1f MiB"):format(tm.compressedTextures, tm.compressedBytes / 1048576))
print(("%d textures stream from their asset, %d hold %.1f MiB of pixels"):format(
tm.assetStreamedTextures, tm.retainedTextures, tm.streamSourceBytes / 1048576))

globals/renderer/textureStreaming

renderer.textureStreaming() -> TextureStreaming

What the last frame's texture-residency plan decided. budgetBytes is the budget in force: the device's own, or the one setTextureBudget names, as much of it as the memory pressure in force leaves (half at "critical"). deviceBudgetBytes is the device's own budget (what setTextureBudget names deviceMegabytes): memoryBudgetBytes, held to the quality tier's cap where the tier states one. memoryBudgetBytes is what the device's memory affords and budgetRule the rule it came from: "systemMemory", an eighth of the RAM the system states (deviceMemoryBytes, nil where it states none), on a GPU that draws from that RAM, or "bufferSize", half the adapter's max_buffer_size; between 256 MiB and 2 GiB either way. pressure is the memory pressure level in force; from "moderate" up the plan holds every texture at the level its surfaces sample, below the budget as well as past it (renderer.memoryPressure()). streamable is how many textures the plan can move. residentBytes is what those textures occupy now, measured from the textures that are allocated; demandedBytes is what the frame asks for before the sweep — every texture's whole chain while those fit the budget, and the level each surface asked for once they exceed it — so the two part exactly where the budget is doing something. starved counts the textures left coarser than the frame asked for, promoted the ones that climbed a level this frame, and changed the ones whose GPU texture was replaced. A camera approaching a surface reads promoted above zero for a few frames and then zero once it settles.

textures is one row per streamable texture, ordered by key, carrying the level each one was asked for and the measurement that asked. Two byte totals can agree while a single texture sits several levels off what its surface samples, so read the row when the question is which level a texture holds and why.

Returns TextureStreaming — { budgetBytes, deviceBudgetBytes, memoryBudgetBytes, budgetRule, deviceMemoryBytes?, pressure, streamable, residentBytes, plannedBytes, demandedBytes, starved, promoted, changed, textures }

local ts = renderer.textureStreaming()
print(("textures: %.1f MiB resident of %.1f MiB demanded, %d starved"):format(
ts.residentBytes / 1048576, ts.demandedBytes / 1048576, ts.starved))

globals/renderer/transmissionShadows

renderer.transmissionShadows() -> boolean

Whether translucent casters tint the directional light they block.

Returns boolean

globals/renderer/uploadStats

renderer.uploadStats() -> {

What the last frame the renderer drew spent re-describing its renderables to the GPU, counted from the frame it drew before that, so a read after task.waitDrawnFrames describes the frame waited for even where the frame pacer declines frames between two drawn ones. Every renderable owns a slot in the per-instance data a draw reads (its world matrix, the bounds the culler tests it by, and the flags that decide which passes and which culling stages see it), and a frame uploads only the slots whose contents changed. A renderer.mesh.drawInstanced population's slots are filled by copies from its own buffers, counted the same way: a frame that copies a population's transforms or shader data counts one write per copy and the bytes it moved. bytes is what those uploads and copies carried, fullBytes what re-sending every slot would have cost, and writes how many buffer writes carried it. The three numbers cover that per-renderable data alone, so a scene standing still reads bytes = 0 against a fullBytes that grows with the scene, and the ratio says how much of it the scene's own churn, rather than its size, is paying for.

Returns { writes: number, bytes: number, fullBytes: number }

local u = renderer.uploadStats()
print(("instance upload: %d B of %d B in %d writes"):format(u.bytes, u.fullBytes, u.writes))

globals/renderer/upscaleBackend

renderer.upscaleBackend() -> string

The backend actually filling the upscale seam this frame — "blit" or "dlss". Reports what the renderer resolved the last renderer.setUpscaleBackend() request into, which reads "blit" after a request for "dlss" on an adapter renderer.upscaleCapabilities() omits it from.

Returns string — "blit" or "dlss".

local backend = renderer.upscaleBackend()

globals/renderer/upscaleCapabilities

renderer.upscaleCapabilities() -> { string }

The backend names this engine can actually run for the upscale seam, from a boot-time probe of the adapter. "blit" always answers, since the stretch needs no adapter capability; "dlss" joins it only where NGX DLSS Super Resolution is available. Read this before renderer.setUpscaleBackend("dlss", ...) to know ahead of the request what it resolves to.

Returns { string } — at least { "blit" }.

local caps = renderer.upscaleCapabilities()

globals/renderer/variantSource

renderer.variantSource(program: string) -> string?

The WGSL one of the programs renderer.shaderVariants() lists holds, exactly as the shader compiler received it. program is the program field of a row's base or of one of its variants. Reading a base alongside a variant shows what a feature set selected: each program's text holds the code its own features guard. The variant-report spelling of renderer.compiledSource, which answers the same for every other shader.

Parameters

program string — A program name from renderer.shaderVariants().

Returns string? — The compiled WGSL, or nil for a name no compile has run under.

local row = renderer.shaderVariants()[1]
local base = renderer.variantSource(row.base.program)

globals/renderer/viewportSize

renderer.viewportSize() -> { width: number, height: number }

The rect the scene was drawn into in the frame just drawn, in pixels. That is the whole drawing surface in a runtime window, a browser canvas and a headless engine, and the editor's viewport panel — smaller than the surface — in an editor layout. Follows a renderer.setViewportSize, a window drag and a browser page resize alike, so it reports what is realized rather than what was asked for.

Returns { width: number, height: number } in pixels.

local v = renderer.viewportSize()

globals/renderer/watchShadowPages

renderer.watchShadowPages(runs: number)

Gather a breakdown of what the sun's virtual shadow map does on each of the main camera's next runs runs. Each run's counts come back from the GPU a few frames after it, and renderer.shadowPageBreakdown() hands them over. Nothing is gathered while no reader has asked. 0 stops a gathering under way.

Parameters

runs number — How many of the main camera's runs to gather.

renderer.watchShadowPages(120)

globals/require

require(path) -> module

Load a Lua module by VFS path. Cached after first load. Use for shared libraries.

Parameters

path string — VFS module path (e.g. '@libs/mylib')

Returns any — The module's return value

globals/scopes/current

scopes.current() -> string?

The scope the calling code registers a resource under right now: the module whose body is running, the task scope its work runs in, the component instance whose lifecycle hook is on the stack, or the chunk of this call. Nil when the caller registers under no context.

Returns string? — The scope tag, or nil.

print("resources I register follow", scopes.current())

globals/scopes/list

scopes.list() -> { LiveResource }

Every live resource that follows an owning context, across every subsystem holding them. A resource registered with no context above it, the engine's own, is not listed, because no scope reaches it.

Returns { LiveResource } — Array of live resources.

for _, r in scopes.list() do print(r.scope, r.kind, r.detail) end

globals/scopes/open

scopes.open(options: ScopeOptions?) -> Scope

Open a task scope without waiting for it: start work in it with s:spawn, end it with s:cancel(), or let what it is lived inside end it ({ entity = id } ties it to an entity). Wait for its work with s:join() from a task.

Parameters

options ScopeOptions (optional) — { entity = id } ties the scope to an entity; { name = "..." } names it.

Returns Scope — The scope.

local s = scopes.open({ entity = npc }); s:spawn("patrol", patrol)

globals/scopes/release

scopes.release(scope: string) -> { Released }

End every resource registered under scope, across every subsystem: tasks are cancelled, jobs dropped. Reaches contexts no seam does, such as the chunk of an execute call that started something and ended without ending it. An execute call runs as a task listed under its own exec: scope, so releasing that scope from the same call cancels the call too: release it from a later call.

Parameters

scope string — A scope tag, as the scope field of a list() row carries it.

Returns { Released } — One row per subsystem that ended something, with how many it ended.

local ended = scopes.release("exec:__exec_12")

globals/scopes/run

scopes.run(fn: (Scope) -> ...any, options: ScopeOptions?) -> ...any

Run fn(s) in a new task scope and wait until every piece of work started in it is finished or cancelled. Returns what fn returned. The first error raised by fn or by any work in the scope cancels the rest, and is raised again here, as it was raised, once their cleanup has run. Waits, so it must be called from a task: in code that cannot wait (a component hook called directly, a callback) it raises, naming that.

Parameters

fn (Scope) -> ...any — The function to run; it receives the scope, and starts work with s:spawn.
options ScopeOptions (optional) — { entity = id } ties the scope to an entity; { name = "..." } names it.

Returns ...any — What fn returned.

scopes.run(function(s) s:spawn("a", fa); s:spawn("b", fb) end)

globals/screenToPlanePoint

screenToPlanePoint(sx, sy, plane_y) -> x, y, z

Unproject screen coords to the intersection point on a horizontal Y-plane (the common cursor-on-ground case). Returns three numbers (x, plane_y, z) on hit, or no values when the ray is parallel to the plane, the plane is behind the camera, or no main camera has rendered yet — so local x, y, z = screenToPlanePoint(...) cleanly yields nil, nil, nil on miss.

Parameters

sx number — Screen X in pixels
sy number — Screen Y in pixels
plane_y number — World-space Y of the horizontal plane to intersect

Returns number, number, number — World-space hit point (x, plane_y, z). Returns no values on miss.

globals/screenToRay

screenToRay(sx, sy) -> { origin: vec3, direction: vec3 } | nil

Unproject screen coordinates to a world-space ray. Returns a single table { origin = {x,y,z}, direction = {x,y,z} } where direction is a unit vector, or nil if no main camera has rendered yet. The ray is computed against the main viewport, so offscreen passes (capture, shadow, post-process) do not disturb it.

Parameters

sx number — Screen X in pixels
sy number — Screen Y in pixels

Returns table | nil — Ray table { origin = {x,y,z}, direction = {x,y,z} }, or nil if unavailable

globals/select

select(index, ...) -> values

Select from varargs. select('#', ...) returns count.

globals/service/authenticated

service.authenticated() -> boolean

Whether a platform identity (JWT) is available to attach to service calls. Returns only a boolean — never the token.

Returns boolean — True if a caller identity is available.

if not service.authenticated() then error("link ZeroMind") end

globals/service/balance

service.balance() -> PromiseHandle?

Read the caller's credit balance from ZeroMind. Returns a promise handle for task.await() resolving the balance JSON, or nil when the gateway is unconfigured or no caller identity is available.

Returns PromiseHandle? — Promise handle for task.await(), or nil if not ready.

local h = service.balance(); local raw = h and task.await(h)

globals/service/gatewayConfigured

service.gatewayConfigured() -> boolean

Whether the ZeroMind service gateway has been configured. Service handlers use this to distinguish "gateway not configured" from "not signed in" when invoke returns nil.

Returns boolean — True if the gateway base URL is set.

if not service.gatewayConfigured() then error("no gateway") end

globals/service/invoke

service.invoke(offering: string, endpoint: string, opts: InvokeOpts?) -> PromiseHandle?

Invoke a provider offering's logical endpoint through ZeroMind. Returns a promise handle for task.await() resolving the InvokeResponse JSON, or nil when the gateway is unconfigured or no caller identity is available. The JWT and real upstream URL are never exposed to Luau.

Parameters

offering string — Fully-qualified offering identity provider/name (e.g. "origozero/mesh_gen").
endpoint string — Logical endpoint name (e.g. "create_preview").
opts InvokeOpts (optional) — { params?, headers?, body?, idempotency_key? }.

Returns PromiseHandle? — Promise handle for task.await(), or nil if not ready.

local h = service.invoke("origozero/mesh_gen", "create_preview", { body = { prompt = p } })

globals/service/invokeStream

service.invokeStream(offering: string, endpoint: string, opts: StreamOpts) -> ServiceStream?

Invoke a provider offering's logical endpoint as a stream. ZeroMind relays the upstream's server-sent events as they arrive, then ends with a zeromind.settled event whose data is {"charged":N,"balance":M}, or a zeromind.failed event whose data is {"reason":...} when the relay stopped early. A call it refuses before the stream starts arrives as one zeromind.refused event carrying the HTTP status and the JSON answer as data; a call that could not be made, or a stream that broke or went quiet, arrives as zeromind.failed. Returns a stream at once, or nil when the gateway is unconfigured or no caller identity is available, exactly as invoke. Take events with stream:poll() (never waits) or stream:read() (waits for the next event; nil at the end), and release the connection with stream:close(). opts takes what invoke takes, plus a required idempotency_key that is unique to this call (ZeroMind refuses a key a streamed call already used) and an optional timeout_ms bounding the wait for the response head and for each next chunk (default 120000). The JWT and the real upstream URL are never exposed to Luau.

Parameters

offering string — Fully-qualified offering identity provider/name (e.g. "origozero/llm").
endpoint string — Logical endpoint name (e.g. "chat").
opts StreamOpts — { params?, headers?, body?, idempotency_key, timeout_ms? }.

Returns ServiceStream? — A stream, or nil if not ready.

local s = service.invokeStream("origozero/llm", "chat", { body = { messages = msgs }, idempotency_key = id })
while s do local ev = s:read(); if ev == nil then break end; print(ev.event, ev.data) end

globals/service/jobStatus

service.jobStatus(jobId: string) -> PromiseHandle?

Poll a submitted service job. Returns a promise handle for task.await() resolving the JobStatusResponse JSON { job_id, status, result?, error? }: status walks pending/running -> succeeded (with result, the same InvokeResponse invoke returns) or failed (with error). nil when the gateway is unconfigured or no caller identity is available.

Parameters

jobId string — Job id returned by submitJob.

Returns PromiseHandle? — Promise handle resolving the job status JSON, or nil if not ready.

local h = service.jobStatus(jobId); local raw = h and task.await(h)

globals/service/submitJob

service.submitJob(offering: string, endpoint: string, opts: InvokeOpts?) -> PromiseHandle?

Submit a durable async invocation of an offering endpoint. Same arguments as invoke, but the provider round-trip runs server-side (off this connection), so a slow synchronous provider or a dropped link no longer loses the result. Returns a promise handle for task.await() resolving { job_id, status }; poll it with jobStatus. nil when the gateway is unconfigured or no caller identity is available.

Parameters

offering string — Fully-qualified offering identity provider/name (e.g. "origozero/mesh_gen").
endpoint string — Logical endpoint name (e.g. "create_preview").
opts InvokeOpts (optional) — { params?, headers?, body?, idempotency_key? }.

Returns PromiseHandle? — Promise handle resolving { job_id, status }, or nil if not ready.

local h = service.submitJob("origozero/mesh_gen", "create_preview", { body = { prompt = p } })

globals/setmetatable

setmetatable(table, mt) -> table

Set a table's metatable.

globals/settings/all

settings.all() -> { [string]: any }

Snapshot of the entire settings document (parsed). Modifying the returned table does NOT propagate — call set or setMany to persist. Useful for editors/inspectors that render every section.

Returns { [string]: any } — A nested table mirroring the TOML document.

for section, keys in pairs(settings.all()) do

globals/settings/get

settings.get(key: string) -> any

Look up a value by dotted key. Returns whatever the file holds at that path — string / number / boolean / array / table — or nil if missing.

Parameters

key string — Dotted-path key, e.g. "render.culling_mode".

Returns any — The raw value at that key, or nil.

local mode = settings.get("render.culling_mode")

globals/settings/getBool

settings.getBool(key: string, default: boolean?) -> boolean

Boolean-typed accessor. Returns the value when present and boolean-typed; falls back to default (or false) on missing key or type mismatch.

Parameters

key string — Dotted-path key.
default boolean (optional) — Optional fallback boolean.

Returns boolean — The boolean value or the fallback.

if settings.getBool("render.shadows", true) then ... end

globals/settings/getNumber

settings.getNumber(key: string, default: number?) -> number

Number-typed accessor. Returns the value when present and number-typed; falls back to default (or 0) on missing key or type mismatch.

Parameters

key string — Dotted-path key.
default number (optional) — Optional fallback number.

Returns number — The number value or the fallback.

local g = settings.getNumber("physics.gravity", -9.81)

globals/settings/getString

settings.getString(key: string, default: string?) -> string

String-typed accessor. Returns the value when present and string-typed; falls back to default (or "" if omitted) on missing key or type mismatch.

Parameters

key string — Dotted-path key.
default string (optional) — Optional fallback string.

Returns string — The string value or the fallback.

local mode = settings.getString("render.culling_mode", "gpu")

globals/settings/set

settings.set(key: string, value: any?)

Set a value by dotted key, then serialize and write the file. In play mode the write fails like any other source-file write — call wld.edit() first to unlock.

Parameters

key string — Dotted-path key.
value any (optional) — Replacement value (string / number / boolean / table).

settings.set("render.culling_mode", "cpu")

globals/settings/setMany

settings.setMany(updates: { [string]: any })

Apply many key/value updates in one batched write — fewer serialize+write round-trips than calling set per-key. Same lock semantics as set.

Parameters

updates { [string]: any } — A table of dotted-key → value pairs.

settings.setMany({

globals/shader/compile

shader.compile(keys: string | { string }, opts: { [string]: any }) -> boolean

Compile a zero-scaffolding SURFACE shader: the author wrote only vertex() / fragment() and declared its material properties, and the engine generates the group(1) material interface plus every render-mode entry point. Compiles once and registers the result under its key.

Parameters

keys string | { string } — The shader key (guid@checksum of the version compiled), alone or as a one-element array.
opts { [string]: any } — { source, domain?, properties?, replaces? } — the author's WGSL, its @domain, the declared property schema, and the key this shader compiled under before its source changed, whose materials move to the new key.

Returns boolean — True when the compile was queued.

local m = asset.meta(ref); shader.compile(m.guid .. "@" .. m.checksum, { source = wgsl, properties = props })

globals/shader/forget

shader.forget(keys: string | { string }) -> boolean

Drop what a removed shader registered under its key: its programs and their variants, the composed WGSL, the compile status and the diagnostic. The pipelines built from those programs are evicted and the materials drawing through them re-prepare.

Parameters

keys string | { string } — The shader key (guid@checksum), alone or in an array.

Returns boolean — True when the removal was queued.

local m = asset.meta(ref); shader.forget(m.guid .. "@" .. m.checksum)

globals/shader/registerModule

shader.registerModule(keys: string | { string }, source: string) -> boolean

Register a block of WGSL other shaders include. Every key names the same source, so a shader includes it by whichever name it holds — its guid, its identity, or an alias. Registering again replaces it, and the shaders that include it recompile.

Parameters

keys string | { string } — One key, or the array of keys this module answers to.
source string — The module's WGSL.

Returns boolean — True when the registration was queued.

shader.registerModule({ ref.guid, ref.identity }, wgsl)

globals/shader/status

shader.status(name: string) -> (string, string?)

A shader's latest compile outcome, without reading the engine log: "compiled", "failed" (with the compiler error second), or "pending". Compilation is async, so a "pending" straight after a write means ask again next frame.

Parameters

name string — The key it compiled under (guid@checksum), or a reference to a shader asset (identity, path, guid), which reaches the key of the version that reference resolves to now.

Returns (string, string?) — Status, and the compiler error when it failed.

local status, err = shader.status("myShader")

globals/shell/run

shell.run(command: string, sessionKey: string?) -> ShellResult

Execute a command in the engine's emulated Unix shell and return once it has completed. This is the same shell as the MCP bash tool — 60+ builtins (ls, cat, grep, find, echo, ...) operating on the virtual scene filesystem. A command that runs Luau (run, luau, zm, zero) needs the engine's frame loop, and a command whose work is a wait (sleep) cannot take its interval while holding it, so from a coroutine the whole line is queued to run off the frame loop and this yields until it finishes; everything else runs inline. The queued run starts the line again from the top, so a line that had already written a file, sent a request, read input or started a background job comes back with the explanation in stderr and shell.runAsync as the way to run it whole.

Parameters

command string — Shell command to execute.
sessionKey string (optional) — Which caller's shell session the command runs in. A session holds what a command leaves behind: the directory cd moved to, exports, shell functions, set -e. It hands that to the next command run under the same key. Omit it and the command runs in the session of the caller this code runs on behalf of, the one that caller's bash commands also use; code with no attested caller behind it runs in the session every caller that names none shares.

Returns ShellResult — Command result { stdout, stderr, exitCode, ok }.

local r = shell.run("ls /zero/source")

globals/shell/runAsync

shell.runAsync(command: string, sessionKey: string?) -> string

Asynchronous version of shell.run. Returns a promise ID that resolves to a JSON-encoded result string. Use with task.await().

Parameters

command string — Shell command to execute.
sessionKey string (optional) — Which caller's shell session the command runs in. A session holds what a command leaves behind: the directory cd moved to, exports, shell functions, set -e. It hands that to the next command run under the same key. Omit it and the command runs in the session of the caller this code runs on behalf of, the one that caller's bash commands also use; code with no attested caller behind it runs in the session every caller that names none shares.

Returns string — Promise ID — pass to task.await() to get the JSON result.

local json = task.await(shell.runAsync("find /zero -name '*.luau'"))

globals/skeleton/applyPose

skeleton.applyPose(sinkHandle: number, poseBuffer: Substrate.TypedBuffer) -> boolean

Snapshot the buffer's first layout.total_floats values and queue a pending apply for the next ECS drain. Returns false on unknown sink/buffer or buffer too small for the layout. The Buffer is unchanged.

Parameters

sinkHandle number — Sink handle from bindPose.
poseBuffer Substrate.TypedBuffer — The pose buffer to apply.

Returns boolean — True on success.

globals/skeleton/bindClip

skeleton.bindClip(zanimBytes: buffer | string, boneOrder: { string }) -> ClipBindInfo?

Decode a zanim payload and bind it to boneOrder, precomputing which of the clip's channels feed each bone so per-frame sampleClip is allocation-free. Returns { handle, matched, total, duration }, or nil on a malformed payload / empty bone order. Check matched: 0 means the clip drives none of these bones.

Parameters

zanimBytes buffer | string — The clip's data.zanim payload bytes (binary-safe).
boneOrder { string } — Output bone names — one stride-10 record per bone.

Returns ClipBindInfo? — { handle, matched, total, duration }, or nil.

globals/skeleton/bindPose

skeleton.bindPose(entityId: (string | entityRef)?, opts: SkeletonLayout) -> number?

Register a pose sink targeting entityId. The opts table carries the layout: boneOrder is the bone-name array ({"hip", "spine", ...}), stride defaults to 10 (translation.xyz + rotation.xyzw + scale.xyz). Pass entityId as nil to use the current component's owning entity.

Parameters

entityId (string | entityRef) (optional) — Engine entity id or proxy, or nil for the current entity.
opts SkeletonLayout — { boneOrder, stride }.

Returns number? — Sink handle, or nil.

local h = skeleton.bindPose(nil, { boneOrder = bones, stride = 10 })

globals/skeleton/clipBones

skeleton.clipBones(zanimBytes: buffer | string) -> { string }?

Decode a zanim payload and return its bone-name array. Pure: build a bind order or a retarget map from a clip without binding a sampler. Returns nil on bytes that aren't a valid zanim payload.

Parameters

zanimBytes buffer | string — The clip's data.zanim payload bytes (binary-safe).

Returns { string }? — Bone names referenced by the clip, or nil.

local names = skeleton.clipBones(assert(vfs.read(path .. "/data.zanim")))

globals/skeleton/clipDecode

skeleton.clipDecode(zanimBytes: buffer | string) -> string?

Decode a zanim payload to its readable JSON form ({ name, duration, channels, bone_names }), the clip as text to read, store or diff. The binary parse is the engine's. Code that transforms a clip's channels in Luau (e.g. the retarget bake) reads the same clip as a table with clipToTable, which skips the text. Returns nil on bytes that aren't a valid zanim payload. Inverse of clipEncode.

Parameters

zanimBytes buffer | string — The clip's data.zanim payload bytes (binary-safe).

Returns string? — The clip as a JSON string, or nil.

local json = require("@builtin::modules.json")
local clipJson = skeleton.clipDecode(bytes)
local clip = if clipJson then json.decode(clipJson) else nil

globals/skeleton/clipEncode

skeleton.clipEncode(jsonString: string) -> string?

Encode a clip's JSON form (the shape clipDecode returns) back to a zanim payload — the bytes a .animation stores and bindClip/sampleClip consume. Inverse of clipDecode. A clip held as a table encodes with clipFromTable, which skips the text. Returns nil on invalid JSON.

Parameters

jsonString string — A clip JSON document.

Returns string? — The clip's zanim payload bytes, or nil.

local json = require("@builtin::modules.json")
local bytes = skeleton.clipEncode(json.encode(clip))

globals/skeleton/clipFromTable

skeleton.clipFromTable(clip: { [string]: any }) -> (string?, string?)

Encode a clip table (the shape clipToTable returns) to a zanim payload, the bytes a .animation stores and bindClip/sampleClip consume. Every field is required and every number finite. Inverse of clipToTable. Returns nil and the reason, naming the field, when the table is not a clip.

Parameters

clip { [string]: any } — A clip table: { name, duration, channels, bone_names }.

Returns (string?, string?) — The clip's zanim payload bytes, or nil and the reason.

local bytes = skeleton.clipFromTable(clip)

globals/skeleton/clipPack

skeleton.clipPack(packed: PackedClip) -> (string?, string?)

Encode a clip in packed form (the shape clipUnpack returns) into zanim bytes. Channels may share one run of times by naming the same keyOffset; interpolation defaults to Linear; with no boneNames the clip names its bones in the order its channels first drive them. times and values take a compute readback's resultBytes() directly. Returns nil and the reason, naming the channel, when a channel's keys fall outside the runs or its times run backwards.

Parameters

packed PackedClip — The packed clip.

Returns (string?, string?) — The clip's zanim payload bytes, or nil and the reason.

local bytes = skeleton.clipPack(packed)

globals/skeleton/clipToTable

skeleton.clipToTable(zanimBytes: buffer | string) -> { [string]: any }?

Decode a zanim payload to a table: { name, duration, channels = { { target_bone, path, keyframes_in, keyframes_out, interpolation } }, bone_names }, the shape clipDecode writes as JSON, built by the engine in place. path is Translation, Rotation or Scale; interpolation is Linear, Step or CubicSpline. Inverse of clipFromTable. Returns nil on bytes that aren't a valid zanim payload.

Parameters

zanimBytes buffer | string — The clip's data.zanim payload bytes (binary-safe).

Returns { [string]: any }? — The clip as a table, or nil.

local clip = skeleton.clipToTable(bytes)

globals/skeleton/clipUnpack

skeleton.clipUnpack(zanimBytes: buffer | string) -> PackedClip?

Decode a zanim payload into its packed form: a channel table ({ bone, path, interpolation, keyOffset, keyCount, valueOffset }, offsets 0-based element indices) and two buffers of little-endian f32s, times and values, holding every channel's keys channel after channel (3 floats a key for Translation and Scale, 4 for Rotation, three times that for CubicSpline). The runs go straight into a typed buffer (buf:writeBytes(packed.values)) for a compute pass. Returns nil on bytes that aren't a valid zanim payload. Inverse of clipPack.

Parameters

zanimBytes buffer | string — The clip's data.zanim payload bytes (binary-safe).

Returns PackedClip? — The packed clip, or nil.

local packed = skeleton.clipUnpack(bytes)

globals/skeleton/jointTransforms

skeleton.jointTransforms(entityId: string | entityRef) -> table

Read a skinned entity's per-joint world transforms for the current animated pose.

Parameters

entityId string | entityRef — Engine entity id or proxy of a skinned entity.

Returns table — Array of joint transforms: { position, matrix, parent, name }.

local joints = skeleton.jointTransforms(meshId)

globals/skeleton/sampleClip

skeleton.sampleClip(handle: number, time: number, poseBuffer: Substrate.TypedBuffer) -> boolean

Sample the bound clip at time (clamped to [0, duration]) and write one stride-10 pose record per bound bone into the Buffer, starting at index 0. Bones the clip does not drive are written as identity. Returns false on unknown handle/buffer or a buffer too small for the bone count.

Parameters

handle number — Sampler handle from bindClip.
time number — Sample time in seconds.
poseBuffer Substrate.TypedBuffer — The stride-10 pose buffer written into.

Returns boolean — True on success.

globals/skeleton/unbindClip

skeleton.unbindClip(handle: number) -> boolean

Drop the bound clip sampler from the registry.

Parameters

handle number — Sampler handle to remove.

Returns boolean — True if the sampler existed.

globals/skeleton/unbindPose

skeleton.unbindPose(sinkHandle: number) -> boolean

Remove the sink from the registry.

Parameters

sinkHandle number — Sink handle to remove.

Returns boolean — True if the sink was present.

globals/sky/get

sky.get() -> { [string]: any }

Get all current sky configuration as a table. Returns the same fields as sky.set accepts, plus read-only fields like material_name and type. Color values are returned as positional arrays [r, g, b].

Returns { [string]: any } — Full sky configuration table.

local cfg = sky.get(); print(cfg.time_of_day)

globals/sky/getTimeOfDay

sky.getTimeOfDay() -> number

Get the current time of day in hours (0-24).

Returns number — Current time of day.

local t = sky.getTimeOfDay()

globals/sky/installFallback

sky.installFallback()

Register the engine fallback sky material and install it as the engine-level fallback (rendered when a scene has no sky entity). Idempotent; requires a live renderer — the scene loader calls it.

sky.installFallback()

globals/sky/preset

sky.preset(name: string)

Apply a named sky preset. Available: clear_day, sunset, sunrise, overcast, night, studio, none. Raises a Luau error for unrecognized names — wrap in pcall if uncertain.

Parameters

name string — Preset name (case-sensitive).

sky.preset("sunset")

globals/sky/set

sky.set(opts: SkyOpts)

Configure the sky system. All fields are optional — only provided fields are updated. Color fields accept both named {x=r, y=g, z=b} and positional {r, g, b} forms. color is an alias for solid_color.

Parameters

opts SkyOpts — Sky configuration properties.

sky.set({ type = "procedural", time_of_day = 14, sync_sun_to_light = true })

globals/sky/setSunDirection

sky.setSunDirection(dir: SkyColor)

Set an explicit sun direction and disable time-based sun positioning. The directional light is updated to match.

Parameters

dir SkyColor — Normalized sun direction vector.

sky.setSunDirection({ 0.5, -1, 0.3 })

globals/sky/setTimeOfDay

sky.setTimeOfDay(time: number)

Set the time of day (0-24 hours). 0 = midnight, 6 = sunrise, 12 = noon, 18 = sunset.

Parameters

time number — Time of day in hours.

sky.setTimeOfDay(18.5)

globals/spawn

spawn(...): any

High-level spawn helpers — auto-injected by the prelude from @builtin::modules.spawn. Provides spawn.cube, spawn.sphere, etc.

globals/stream/accept

stream.accept(listener: string) -> string?

Take the connection that has waited longest on the listener, as a stream handle that reads, writes, and closes exactly like one stream.open returned. Returns nil when nothing is waiting, so call it in a loop each tick to take every peer that arrived. stream.listenerStatus(listener).pending is how many are still waiting.

Parameters

listener string — Listener handle from stream.listen.

Returns string? — The connection's stream handle, or nil when none is waiting.

while true do local h = stream.accept(listener); if not h then break end; table.insert(peers, h) end

globals/stream/close

stream.close(handle: string) -> boolean

Finish whatever handle names — a stream or a listener — and drop it from the registry.

Closing a stream refuses every later write and carries the bytes already queued to the peer before the connection ends, so a write and a close in the same tick deliver — the shape a request answered with one response has. A peer that has stopped reading altogether holds that finish for thirty seconds; past that the connection ends and what is still queued ends with it, so a caller that must know its bytes went out watches stream.status(handle).pending reach zero before it closes.

Closing a listener stops it answering new peers and closes the connections nobody took; the connections stream.accept already handed over keep running until they are closed themselves.

Parameters

handle string — Stream handle from stream.open or stream.accept, or listener handle from stream.listen.

Returns boolean — True if a stream or listener was closed, false if handle already named none.

stream.write(peer, response); stream.close(peer)

globals/stream/listen

stream.listen(url: string, opts: StreamListenOpts?) -> PromiseHandle

Hold the address url names and answer the peers that dial it — the other direction from stream.open, for when the thing you are talking to starts the conversation and restarts on its own schedule. Returns a promise handle: task.await() it to get the listener handle once the address is held, or it raises the reason a malformed url, a scheme that cannot listen, or a failed bind was refused with. Take the connections with stream.accept.

The host in the url is the interface bound, and the whole of what decides who can reach it. tcp://127.0.0.1:9000 answers only programs on this same machine. tcp://0.0.0.0:9000 answers any host that can route to this machine on that port — every device on the wifi, and anything beyond it the network lets through. Write the one you mean; there is no default, and stream.listenerStatus reports which of the two you got. A port of 0 asks the operating system to choose one, which that same status then reports.

opts.inboundCapacity and opts.outboundCapacity bound each answered connection (65536 bytes each by default); opts.backlog bounds the connections held for stream.accept before the listener stops taking them from the operating system, which leaves the rest queued in the kernel rather than answered and forgotten (16 by default, and at least 1).

The listener belongs to the chunk that opened it — the chunk whose own code called stream.listen, which is the module holding that line even when something else called into it. When that chunk runs again — a module hot-reload, a cleared require cache — the listener and the connections it answered are closed, and the new run binds the address for itself. Peers see the connection close and dial again. stream.listeners() names that chunk as each entry's owner.

Parameters

url string — Listen URL — scheme://host:port.
opts StreamListenOpts (optional) — Per-connection capacities and the accept backlog (optional).

Returns PromiseHandle — Promise handle for task.await().

local pending = stream.listen("tcp://127.0.0.1:9000"); local listener = task.await(pending)

globals/stream/listenerStatus

stream.listenerStatus(listener: string) -> ListenerStatus?

Report what the listener holds and has handed over. address is the address the operating system resolved the bind to, port included — the one to hand a peer. reach says who can connect to it: "thisMachine" when it is a loopback address and only programs on this machine can, "network" when any host that can route here can. accepted counts the connections stream.accept handed over, pending the ones still waiting, and capacity the value pending may reach before the listener stops taking connections from the operating system. nil when handle names no open listener.

Parameters

listener string — Listener handle from stream.listen.

Returns ListenerStatus? — Listener status, or nil when handle names no open listener.

local s = stream.listenerStatus(listener); print(s.address, s.reach, s.pending)

globals/stream/listeners

stream.listeners() -> { OpenListener }

Every listener this engine currently holds an address for, in the order they were opened. Each entry is what stream.listenerStatus reports about it, plus the handle it is addressed by and the owner chunk its life follows.

This is how an address is reached again once nothing holds its handle: filter on address for the port you want and close the entry by its handle, rather than guessing at handles.

Returns { OpenListener } — An array of open listeners.

for _, l in stream.listeners() do if l.address == want then stream.close(l.handle) end end

globals/stream/open

stream.open(url: string, opts: StreamOpenOpts?) -> PromiseHandle

Open a byte stream at url (scheme://target[?k=v]). loopback carries written bytes back out of the same stream and works on every platform; tcp dials host:port; tty opens a serial device node — /dev/ttyACM0 or /dev/ttyUSB0 for a USB CDC board such as an ESP32, /dev/rfcomm0 for a Bluetooth controller paired over classic SPP (both present as a tty on Linux, so one transport serves either peer), COM5 on Windows. tty query parameters: baud (default 115200), dataBits (5-8, default 8), parity (none | odd | even, default none), stopBits (1 or 2, default 1).

ble connects to a Bluetooth Low Energy device over GATT, on a desktop engine and in a browser alike — the wireless transport a web world reaches a device through: ble://<device>?service=<uuid>&write=<uuid>&notify=<uuid>. The device is the name it advertises, * any device offering the service, a trailing * a name prefix (Paw*). write is the characteristic this engine writes to and notify the one it subscribes to, which on a Nordic UART peripheral are that peripheral's RX and TX; a module with one bidirectional characteristic names it for both. UUIDs may be 16-bit (ffe0), 32-bit, or full. Optional: chunk (bytes per packet, 1-512 — otherwise what the connection carries), writeMode (withResponse | withoutResponse, default withResponse), timeout (seconds to find and connect to the device, default 15).

opts bounds the stream's undrained inbound buffer and in-flight outbound bytes (default 65536 each). Returns a promise handle: task.await() it to get the stream handle once the transport is open, or it raises the reason a malformed url, an unknown or unsupported scheme, or a failed connect was refused with. A ble stream resolves as soon as it exists and reports the rest as state — watch stream.status(handle).state go opening, permissionPending while the browser asks the person at the machine to pick a device, then open; writes made meanwhile are queued and go out when it connects. Check stream.transports() first to tell a mistyped scheme from one this build does not carry.

Parameters

url string — Stream URL — scheme://target[?k=v&k=v].
opts StreamOpenOpts (optional) — Buffer capacities (optional).

Returns PromiseHandle — Promise handle for task.await().

local pending = stream.open("loopback://echo"); local handle = task.await(pending)
local paw = task.await(stream.open("ble://Paw*?service=ffe0&write=ffe1&notify=ffe1"))

globals/stream/read

stream.read(handle: string, max: number?) -> string

Drain up to max buffered inbound bytes from the stream.

Parameters

handle string — Stream handle from stream.open.
max number (optional) — Maximum bytes to drain (optional). Omit to drain everything buffered.

Returns string — Drained bytes, byte-safe. "" when none buffered or handle names no open stream.

local chunk = stream.read(handle)

globals/stream/serialPorts

stream.serialPorts() -> SerialPorts

Every serial device this machine has, for picking the one to open. ports is an array ordered by path. Each entry carries the path the device is at (/dev/ttyACM0 on Linux, COM3 on Windows), the url that opens it, the kind of bus it attaches by, and — for a USB device — the vendorId, productId, serialNumber, manufacturer and product it advertises.

A device's path moves with enumeration order: a board that came up at /dev/ttyACM0 is at /dev/ttyACM1 once something else is plugged in first, and moves across COM3-COM5 on Windows. What the device advertises holds still across those moves, so match on vendorId/productId — or on serialNumber to tell two of the same board apart — and open the url that entry carries, appending the port settings stream.open documents.

Three answers are distinct. supported false with a reason means this platform has no serial bus to enumerate at all. error set means it has one and the operating system refused this enumeration, so a later call may answer. An empty ports with neither means the machine has no serial device attached, which is an ordinary result.

Returns SerialPorts — { supported, reason, error, ports } — the platform's answer, this enumeration's, and the devices it found.

for _, p in stream.serialPorts().ports do if p.vendorId == 0x303A then print(p.url, p.product) end end

globals/stream/status

stream.status(handle: string) -> StreamStatus?

Report what the stream has carried and lost. state is where the stream is in its life: opening, permissionPending while the platform asks the person at the machine to allow the connection, open, denied when that permission was refused, and closed when it is finished. pending is bytes accepted and not yet handed to the peer; capacity is the value pending may reach before a write is refused. error holds the most recent transport failure and the refusal a denied stream carries, retained for the life of the stream. nil when handle names no open stream.

Parameters

handle string

Returns StreamStatus? — Stream status, or nil when handle names no open stream.

local s = stream.status(handle); print(s.pending, s.capacity)

globals/stream/streams

stream.streams() -> { OpenStream }

Every open stream, dialled or answered, in the order they were opened. Each entry is what stream.status reports about it, plus the handle it is addressed by and the owner chunk its life follows — a connection stream.accept handed over carries the owner of the listener that answered it.

Returns { OpenStream } — An array of open streams.

for _, s in stream.streams() do print(s.handle, s.transport, s.pending, s.owner) end

globals/stream/transports

stream.transports() -> { [string]: TransportSupport }

Every stream scheme this build knows about — a capability probe, in both directions. supported answers stream.open and listen answers stream.listen, since a scheme can carry one and not the other. Each reason is nil when its direction works, otherwise it names why not: an unbuilt transport names its own absence, a transport this platform lacks (tcp and tty on wasm; ble in a browser without Web Bluetooth or with the radio off, which the page itself answers) names that, and a loopback stream, whose peer is itself, names that nothing dials it. A typo'd scheme is absent from this table entirely, which is what tells it apart from a real transport this build lacks.

Returns { [string]: TransportSupport } — Map of scheme name to { supported, reason, listen, listenReason }.

local t = stream.transports(); if not t.tcp.listen then warn(t.tcp.listenReason) end

globals/stream/write

stream.write(handle: string, bytes: string) -> WriteOutcome

Queue bytes for the stream's peer. Never blocks. "accepted" means the bytes were queued. "full" means the outbound queue has no room right now — backpressure, not failure: the peer is alive and draining slower than this call is producing, so a retry after it catches up can succeed. Compare pending against capacity on stream.status() to see it coming before a write is refused. "closed" means the stream is finished, or handle names no open stream — reopen to continue, retrying never succeeds. "tooLarge" means bytes is bigger than the stream's whole outbound capacity, so it can never fit at any queue depth — retrying the same write returns this again.

Parameters

handle string — Stream handle from stream.open.
bytes string — Bytes to queue, byte-safe.

Returns WriteOutcome — "accepted" | "full" | "closed" | "tooLarge"

local outcome = stream.write(handle, data)

globals/streaming/cells

streaming.cells() -> { [string]: any }

What the spatial-streaming store has resident: the configured radii and budget, the counters the store keeps, and one row per cell with how many of its groups are standing, what it costs, and whether a release wrote it to a file it now reads back from.

Returns { [string]: any } — { config, stats, sources, cells, proxies }.

local s = streaming.cells()

globals/streaming/levels

streaming.levels(scene: any?) -> { [string]: any }

Which level every mesh-LOD receiver is drawing at, and the screen fraction that selection was measured from. A receiver whose entity the scene no longer holds is reported as standing = false: the chain is registered and there is nothing left for it to draw.

Parameters

scene any (optional) — The scene walk to read against. Omitted, the call takes its own.

Returns { [string]: any } — { count, receivers }.

local l = streaming.levels()

globals/streaming/observe

streaming.observe() -> { [string]: any }

The whole reading in one document: terrain, voxel, streaming cells and mesh LOD, plus the totals those rows sum to.

Built by this call and published as its last act, so /zero/runtime/observations/streaming serves the same document rather than a second derivation of it.

Returns { [string]: any } — { terrain, voxel, cells, levels, totals }.

local r = streaming.observe()

globals/streaming/reasons

streaming.reasons() -> { string }

Every reason whyNotDrawn can answer with, so a caller can enumerate the set rather than meeting it one failure at a time.

Returns { string } — Sorted array of reason names.

local r = streaming.reasons()

globals/streaming/terrain

streaming.terrain(scene: any?) -> { [string]: any }

What each terrain entity is drawing: whether a heightfield is bound to it, the LOD cut it settled on, what that cut costs in indices and in the vertex pool, and the eye the cut was refined under.

Parameters

scene any (optional) — The scene walk to read against. Omitted, the call takes its own, which is what makes a whole reading one walk rather than four.

Returns { [string]: any } — { count, entities } — one row per entity carrying a Terrain.

local t = streaming.terrain()

globals/streaming/voxel

streaming.voxel(scene: any?) -> { [string]: any }

What became of every chunk of every voxel world: how many are meshed, queued, failed or empty, and one row per chunk carrying the state, the engine's reason when a build failed, and what the build reserved on the device.

Parameters

scene any (optional) — The scene walk to read against. Omitted, the call takes its own.

Returns { [string]: any } — { count, worlds } — one entry per entity carrying a VoxelWorld.

local v = streaming.voxel()

globals/streaming/whyNotDrawn

streaming.whyNotDrawn(subject: any?) -> { [string]: any }

Why a piece of a world's detail is not on screen, as one reason from the closed set streaming.reasons() enumerates, with a detail line naming what that reason is about.

The subject picks which system answers:

an entity ref, id or name — whichever of the four systems holds it
{ entity = ..., chunk = "cx_cy_cz" } — one chunk of a voxel world
{ entity = ..., level = n } — one level of a mesh-LOD chain
{ cell = "x_z" } — one cell of the spatial-streaming store

Parameters

subject any (optional) — The entity, chunk, level or cell to answer about.

Returns { [string]: any } — { kind, reason, detail }.

local w = streaming.whyNotDrawn({ entity = "Vox", chunk = "0_0_0" })

globals/stringx/scanNumbers

stringx.scanNumbers(s: string, pos: number?) -> ({ number }, number)

Read the run of numbers starting at pos — separated by commas and/or whitespace — and report where the run ended.

The run stops at the first character that neither continues a number nor separates two of them (], }, a quote, a letter), and nextPos is that character's index, so the caller's own parser resumes exactly there. A token that is not a valid number also ends the run, with nextPos left ON it rather than past it, so nothing is skipped without the caller seeing it.

Parameters

s string — The text to read.
pos number (optional) — 1-based index to start at. Defaults to 1.

Returns ({ number }, number) — The numbers found, and the 1-based position just past them.

-- A JSON array of numbers, in one crossing instead of one per token.
local values, nextPos = stringx.scanNumbers(payload, afterBracket)
-- A whitespace-separated block (OBJ, PLY, a matrix dump).
local m = stringx.scanNumbers("1 0 0 0  0 1 0 0", 1)

globals/subscriptions/cancel

subscriptions.cancel(id: string) -> boolean

Cancel a subscription by id: disconnects the live connection immediately and marks the row cancelled. Returns true when a live subscription was cancelled, false for an unknown or already-disconnected id.

Parameters

id string — Subscription id to cancel.

Returns boolean — True when a live subscription was disconnected.

subscriptions.cancel(conn.id)

globals/subscriptions/get

subscriptions.get(id: string) -> SubscriptionRow?

One subscription row by id, or nil when the id is unknown (never tracked, or evicted after its publisher was destroyed).

Parameters

id string — Subscription id (conn.id, or a /zero/runtime/events/subscriptions/ entry).

Returns SubscriptionRow? — The subscription row, or nil.

local row = subscriptions.get(conn.id); print(row and row.deliveries)

globals/subscriptions/list

subscriptions.list(filter: SubscriptionFilter?) -> { SubscriptionRow }

Every tracked subscription row, optionally filtered by publisher instance id, publisher entity id, event name, and/or connected state.

Parameters

filter SubscriptionFilter (optional) — Optional filter table.

Returns { SubscriptionRow } — Array of subscription rows.

for _, s in ipairs(subscriptions.list({ connected = true })) do print(s.id, s.event, s.deliveries) end

globals/subscriptions/publishers

subscriptions.publishers() -> { PublisherRow }

Every live event publisher: component instance, entity, and per-event fire stats (fires happen whether or not anyone subscribes) plus current subscriber ids.

Returns { PublisherRow } — Array of publisher rows.

for _, p in ipairs(subscriptions.publishers()) do print(p.component, p.entityName) end

globals/substrate/bindFrames

substrate.bindFrames(key: string, id: string, slots: { any }) -> (boolean, string?)

Bind the play key to the sequence id through slots, so showFrame(key, f) places its elements from frame f. slots[s] is { kind = "instances", key, transforms, data?, capacity, draw, interval? } for elements an instanced draw copies from transforms — the scatter binding key places up to capacity of them there and the draw draw draws the frame's count — or { kind = "entities", key, binding?, interval? } for elements standing in entities: the scatter binding key, whose entities scatterTransforms has already set, places them, and the frame's fields are written across the entity binding binding.

Parameters

key string — The play's name.
id string — The sequence the play shows.
slots { any } — What each numbered slot places, for this play.

Returns (boolean, string?) — True when the binding was queued, and why it was refused when not.

substrate.bindFrames("spark#1", "spark", { { kind = "instances", key = "spark#1", transforms = "spark#1.xf", capacity = 64, draw = drawId } })

globals/substrate/bufferSource

substrate.bufferSource(name: string) -> string?

The chunk that created the GPU buffer name holds, as debug.info names a chunk — a module's identity, a test suite's, an execute body's.

The record is the name's for as long as a buffer holds the name, across a device rebuilt underneath it. A name created again after its buffer went answers for the call that created it again. A buffer the engine made itself, or one made through another surface, has no record.

Parameters

name string — The name the buffer was created under.

Returns string? — The creating chunk's name, or nil.

if substrate.bufferSource("debug_viz.vtx") then print("made through the substrate") end

globals/substrate/clearScatterTransforms

substrate.clearScatterTransforms(key: string) -> boolean

Drop the device transform binding under key. Its entities go back to being placed by whatever writes their Transform. Queued.

Parameters

key string — The key the binding was made under.

Returns boolean — True when the drop was queued.

substrate.clearScatterTransforms("wave")

globals/substrate/createBuffer

substrate.createBuffer(opts: BufferOpts) -> TypedBuffer?

Allocate a typed buffer and return its handle.

A "gpu" buffer is storage a compute shader binds; usage adds "vertex", "index", "indirect" or "readback" on top of the storage it always has. A "cpu" buffer lives in the scripting heap and reads back as a flat array of floats.

The handle's write answers whether the words landed: a payload whose end falls past the end of the buffer is refused whole on both kinds, so the buffer keeps what it held and the call answers false. writeU32 and writeBytes answer the same way, against the same extent.

A "cpu" buffer holds at most 1 GiB of storage: 268435456 f32 words, 16777216 mat4 records. A len past that raises an error naming the bytes it needs and the limit, and a len within it whose storage the heap cannot supply raises an error naming the bytes. Neither allocates anything.

A "gpu" buffer holds at most what the device creates in one buffer, its max_buffer_size. A len whose bytes pass it raises an error naming the records, the bytes they need, that limit and the most records of the type one buffer holds; nothing is allocated and the name stays free.

class makes a "gpu" buffer a run: it stands in a range of a page the engine's device arenas share among many runs of the class, rather than in an allocation of its own, so thousands of short-lived buffers cost a handful of device allocations. "effect" is for runs made and let go of as an effect plays, "persistent" for runs held until let go of. A run binds, writes, reads and frees like any "gpu" buffer under the name its handle carries, which the engine mints, so a run takes no name; a compute program declared with ranges = true binds any number of runs of one page through one bind group. A run carries "readback" usage and no other: a vertex, index or indirect buffer is one of its own. owner names whom a run is held for, and substrate.releaseOwner(owner) lets go of every run created with it at once.

A job on a worker thread makes a run the same way. The handle answers at once, with the type and length asked for and a name the worker mints; substrate.listBuffers lists the run at once; and the engine thread places it in its arenas ahead of every write, dispatch and read the job makes over it afterwards, so one job can make a run, fill it, dispatch over it and read it back. A class, usage or len the engine thread refuses raises at the call in the job too. When the engine thread cannot place the run, the handle's next method call raises why. substrate.setOwner and substrate.releaseOwner answer in a job as they do on the engine thread, counting the job's own changes at once.

Parameters

opts BufferOpts — { type, len, kind?, usage?, name?, class?, owner? } — type is "f32", "vec3", "vec4", "quat" or "mat4"; kind is "cpu" (the default) or "gpu". name is the name a dispatch binds a "gpu" buffer by, and the name substrate.getBuffer and substrate.destroyBuffer reach it under. class ("effect" or "persistent") makes a "gpu" buffer a run; owner names whom the run is held for.

Returns TypedBuffer? — The buffer handle, or nil when the allocation failed — an unknown type or kind, a zero length, or a name that already holds a GPU buffer of another shape. A name holding a buffer of the SAME type and length hands that buffer back, contents and all; substrate.destroyBuffer frees a name whose buffer is the wrong shape.

local pose = substrate.createBuffer({ type = "mat4", len = boneCount })
local field = substrate.createBuffer({ type = "vec3", len = 4096, kind = "gpu" })
local column = substrate.createBuffer({ type = "f32", len = n, kind = "gpu", class = "effect", owner = "burst:" .. id })
local values = pose:read(0, 16):result()

globals/substrate/defineFrames

substrate.defineFrames(id: string, frames: { any }) -> (boolean, string?)

Hold a sequence of frames on the engine under id, for plays bound to it (bindFrames) to show by index (showFrame) with no script running per frame. frames[f] is { scatters = {...}, fields = {...} }. A scatter { slot, positions, rotations?, scales?, lanes?, identities?, count } names the buffers the slot's elements are placed from in that frame and how many elements they hold; lanes[k] is a buffer name or false. A field { slot, component, field, values } gives the value the field takes on each of the slot's entities, in binding order. Slots are numbered from 1. The buffers stay the caller's and must stand for as long as the sequence is held. Defining id again replaces it.

Parameters

id string — The sequence's name.
frames { any } — The frames, in order.

Returns (boolean, string?) — True when the sequence was queued, and why it was refused when not.

substrate.defineFrames("spark", { { scatters = { { slot = 1, positions = P:bufferName(), count = 64 } } } })

globals/substrate/destroyBuffer

substrate.destroyBuffer(name: string) -> boolean

Free the GPU buffer name denotes, whatever else still holds a handle to it.

The allocation goes and the name is free to be created again at any type and length; every handle that pointed at it answers :alive() false. This is what releases a name whose creating handle is gone, so a build that re-runs at a different size gets its name back.

Parameters

name string — The name the buffer was created under.

Returns boolean — True when a GPU buffer under that name was freed.

substrate.destroyBuffer("env.town.xf")

globals/substrate/forgetFrames

substrate.forgetFrames(id: string) -> boolean

Drop the sequence held under id. A play still bound to it shows nothing further.

Parameters

id string — The sequence's name.

Returns boolean — True when the drop was queued.

substrate.forgetFrames("spark")

globals/substrate/getBuffer

substrate.getBuffer(name: string) -> TypedBuffer?

The GPU buffer name denotes, as a handle you now hold.

A name is how a dispatch binds a buffer, so the name is what an owner asks by once the handle it created with has gone out of scope — a .module that hot-reloaded, a build that ran in an earlier execute. The handle carries everything createBuffer's does and releases its reference with :destroy().

Parameters

name string — The name the buffer was created under.

Returns TypedBuffer? — The buffer handle, or nil when no GPU buffer holds that name.

local xf = substrate.getBuffer("env.town.xf")
local shape = xf and { xf:type(), xf:length() }

globals/substrate/gpuReadback

substrate.gpuReadback(key: string?) -> Readback?

Wrap the key an FFI read handed back as the Readback that polls it. Every GPU→CPU read reaches the caller through this, so a texture's read and a buffer's read answer with the same thing.

Parameters

key string (optional) — The key the read returned.

Returns Readback? — The Readback, or nil when the read did not start.

local pending = substrate.gpuReadback(compute.readTexture3D(handle))

globals/substrate/listBuffers

substrate.listBuffers() -> { NamedBuffer }

Every named GPU buffer the engine holds, in name order.

Each record states id, name, type ("F32", "Vec3", "Vec4", "Quat", "Mat4"), len in records, refs — how many holders it has — holder, the owner a run is held for when one was named (createBuffer's owner), class, the lifetime class of a run ("effect" or "persistent"; absent for a buffer of its own), and bytes, the device bytes it takes: the range a run stands in, or the extent of a buffer of its own. This is what states which names are taken, at what shape, and what an owner holds.

Returns { NamedBuffer } — Array of { id, name, type, len, refs, holder?, class?, bytes }.

for _, b in ipairs(substrate.listBuffers()) do print(b.name, b.type, b.len) end

globals/substrate/losses

substrate.losses() -> number

How many times a buffer has been freed by name or by owner: destroyBuffer, releaseOwner and the reset to a baseline. A holder that checked that the buffers it names exist at one count knows they still do while the count is the same, whatever else it asks.

Returns number — The count, which only grows.

local at = substrate.losses()

globals/substrate/markBaseline

substrate.markBaseline() -> number

Record which named GPU buffers the engine holds right now, and make that the point substrate.resetToBaseline returns to.

A caller running work in an engine it shares — a sweep of test suites, a tool driving a session — leaves buffers behind under the names the work chose: a view's geometry built on its first capture, a kernel's storage. This is how such a caller says where its own allocations start.

Returns number — The mark's generation, counting from 1 — what substrate.resetToBaseline takes to restore to this mark.

local mark = substrate.markBaseline() -- then run, then substrate.resetToBaseline(mark)

globals/substrate/releaseBaseline

substrate.releaseBaseline(generation: number) -> boolean

Drop the mark generation names, and every mark taken after it, once the caller that took it is done resetting to it. It touches the marks alone.

Parameters

generation number — The mark to drop, as substrate.markBaseline returned it.

Returns boolean — True when the mark stood.

substrate.releaseBaseline(mark) -- the run is over

globals/substrate/releaseOwner

substrate.releaseOwner(owner: string) -> number

Let go of every run created with owner (createBuffer's class and owner), whatever else still holds a handle to it.

Each run's range goes back to its arena and every handle to it answers :alive() false. This is what something that makes runs as it plays — an effect, a cook, a cache — calls when it ends, so everything it made goes with it at once, runs it handed on to others included.

Parameters

owner string — The owner the runs were created with.

Returns number — How many runs were let go of.

substrate.releaseOwner("burst:" .. id)

globals/substrate/resetToBaseline

substrate.resetToBaseline(generation: number?) -> number

Free every named GPU buffer created since substrate.markBaseline, leaving the ones it recorded where they are.

A module that created one of these holds its own handle to it. Put that module's state back alongside this (engine.restoreModuleState) or it goes on addressing storage this freed. is touched.

Parameters

generation number (optional) — The mark to restore to, as substrate.markBaseline returned it; the newest mark when omitted. The marks taken after it go with the reset.

Returns number — How many buffers were freed.

local freed = substrate.resetToBaseline(mark)

globals/substrate/scatterTransforms

substrate.scatterTransforms(key: string, entityIds: { string }?, positions: string, rotations: string?, scales: string?, lanes: { string? }?, interval: number?, identities: string?) -> boolean

Place entities from the buffers that already hold their transform, on the device, instead of from a Transform written per entity.

entityIds[i] is placed from row i of every named buffer, so the order is the buffers' own. Pass nil for entityIds to keep the entities a binding already has and replace only the buffers — what a re-cook does, since a field chain mints a fresh run each time while its membership stands still, and sending the element list again each time would put back the per-element crossing the binding exists to remove.

Re-binding the same key with a list replaces the previous binding. Queued: it takes effect on the next frame, and an entity whose renderable has not landed yet is placed on the frame it does.

Parameters

key string — Stable identity of the binding.
entityIds { string } (optional) — The entities in buffer order, or nil to rebind the buffers alone.
positions string — Name of a buffer holding three f32 per element.
rotations string (optional) — Name of a buffer holding four f32 per element.
scales string (optional) — Name of a buffer holding three f32 per element.
lanes { string? } (optional) — Per-instance shader data, by lane: lanes[k + 1] names a buffer holding one vec4 per element, written into lane k of each element's block — what its material reads back as input.shader_data[k]. A lane left out is left to whatever else writes it.
interval number (optional) — Seconds between the presents of the domain the buffers hold. Given, every frame draws each element between the last two presents — position, rotation, scale and lanes blended — the way a fixed-rate simulation is drawn between its ticks, so its motion is continuous at any frame rate while the domain presents at its own. An element new to a slot starts where it was presented. Omitted or 0, each present is drawn as it lands.
identities string (optional) — Name of a buffer holding one f32 per element naming which element stands in that row at each present — for a domain whose rows are taken by different elements over time, as a particle's slot is when it dies and another is born into it. A row whose identity moved since the last present starts where it was presented rather than blending from the element that held it before. -1 names a row no element stands in. Omitted, each row's element is the entity placed from it.

Returns boolean — True when the binding was queued.

substrate.scatterTransforms("wave", ids, positions:bufferName())
substrate.scatterTransforms("wave", nil, positions:bufferName())
substrate.scatterTransforms("wave", nil, positions:bufferName(), nil, nil, { colors:bufferName() })
substrate.scatterTransforms("sparks", ids, positions:bufferName(), nil, nil, nil, 1 / 30)
substrate.scatterTransforms("sparks", nil, positions:bufferName(), nil, scales:bufferName(), nil, 1 / 30, ids:bufferName())

globals/substrate/scatterTransformsInto

substrate.scatterTransformsInto(key: string, into: { transforms: string, data: string?, count: number }, positions: string, rotations: string?, scales: string?, lanes: { string? }?, interval: number?, identities: string?) -> boolean

Place elements that stand in no entity from the substrate buffers that hold their transform, on the device: row i of the columns is composed into matrix i of into.transforms, and lane k of block i of into.data — the two buffers an instanced draw (renderer.mesh.drawInstanced with transformBuffer and instanceDataBuffer) copies into its slots every frame. It is scatterTransforms for a population: the same columns, the same blend between presents when interval is given, the same identities. Binding the same key into the same buffers again replaces only the columns it reads. Queued: it takes effect on the next frame, ahead of the instanced copies that frame records.

Parameters

key string — Stable identity of the binding.
into { transforms: string, data: string?, count: number } — { transforms, data?, count }: a buffer of count column-major mat4, a buffer of count 64-byte blocks of four vec4 lanes when lanes are bound, and how many elements there are.
positions string — Name of a buffer holding three f32 per element.
rotations string (optional) — Name of a buffer holding four f32 per element.
scales string (optional) — Name of a buffer holding three f32 per element.
lanes { string? } (optional) — lanes[k + 1] names a buffer holding one vec4 per element, written into lane k of each element's block in into.data.
interval number (optional) — Seconds between the domain's presents; given, each element is drawn between the last two presents.
identities string (optional) — Name of a buffer holding one f32 per element naming which element stands in that row at each present; -1 names a row no element stands in.

Returns boolean — True when the binding was queued.

substrate.scatterTransformsInto("sparks", { transforms = "sparks.xf", count = 4096 }, P:bufferName())

globals/substrate/scatterWorlds

substrate.scatterWorlds(key: string) -> any

Read back where the device drew the elements of the transform binding under key in the last frame drawn — the device's own answer, since the CPU never learns where an element it places stands. 16 floats per placed element, column-major world matrices, in the binding's element order, leaving out elements whose renderable has not landed. A binding drawn between its presents answers the blended pose that frame drew.

Parameters

key string — The key the binding was made under.

Returns any — A pending readback: :ready(), then :result() for the floats.

local pending = substrate.scatterWorlds("sparks")
while not pending:ready() do task.wait() end
local x = pending:result()[13]

globals/substrate/setOwner

substrate.setOwner(name: string, owner: string?) -> boolean

Hold the run name names for owner from now on, in place of the owner it was created with, or for no one when owner is nil.

substrate.releaseOwner(owner) then lets go of it with the rest of what owner holds, and releasing the owner it was made for leaves it standing. What hands a run on to whatever keeps it longer than the one that made it — a cache several makers read from.

Parameters

name string — The name the run binds under (its handle's name).
owner string (optional) — The owner to hold it for, or nil.

Returns boolean — Whether name names a run (a buffer created with a class).

substrate.setOwner(column.name, "cache#3")

globals/substrate/showFrame

substrate.showFrame(key: string, frame: number) -> boolean

Show frame frame (from 1) of the sequence the play key is bound to: each slot is placed from that frame's buffers, each instanced draw draws that frame's count, and that frame's fields are written. Showing the frame already shown changes nothing. Queued.

Parameters

key string — The play's name.
frame number — The frame, from 1.

Returns boolean — True when the frame was queued.

substrate.showFrame("spark#1", 12)

globals/substrate/unbindFrames

substrate.unbindFrames(key: string) -> boolean

Drop the play key's binding to its sequence. What it placed stays as the last frame shown left it.

Parameters

key string — The play's name.

Returns boolean — True when the drop was queued.

substrate.unbindFrames("spark#1")

globals/text/alive

text.alive(handle: any?) -> boolean

Whether the text system still holds this handle — true between text.create and the text.destroy that released it.

Parameters

handle any (optional) — Text handle from text.create.

Returns boolean — True while the handle is live.

if not text.alive(h) then h = text.create({ content = "again" }) end

globals/text/count

text.count() -> number

How many text objects the text system is holding — the number that moves when text.create and text.destroy are called.

Returns number — The live text-object count.

local before = text.count()

globals/text/create

text.create(options: table) -> any

Create a text handle from an initial content + style table. The handle owns a runtime GPU texture (see text.textureGuid); pass it to every other call.

Parameters

options table — Table of content plus style fields (fontSize, color, alignment, richText, maxWidth, ...).

Returns any — An opaque text handle, or nil if the text system is unavailable.

local h = text.create({ content = "Hello", fontSize = 48 })

globals/text/destroy

text.destroy(handle: any?) -> boolean

Destroy a text handle and release its raster + glyph layout.

Parameters

handle any (optional) — Text handle from text.create.

Returns boolean — True when the text system held the handle and released it; false for a handle it did not have.

text.destroy(h)

globals/text/face

text.face(handle: any?) -> any

Which font face one handle actually shaped with, and whether that is the family its style selects. requested is the name the style wrote, selected is the family that name selects — itself for a family the shaper's own database carries, and the family a font file declares for a name registered onto it — resolved is the face that answered, matched says whether the face is the family the name selected and reason says why when it is not — one of text.faceReasons(). A style that named no family reports noFamilyRequested: it got the default because it asked for nothing, so reason rather than matched is what an alert switches on. faces lists every face the shaper used, most glyphs first, so a fallback that covered part of the string is visible alongside the face that covered the rest. missing lists the characters no face drew (the missing glyph stood in); a character an on-demand face covers requests it, and the object is shaped again once the face is installed (font.onDemand()).

Parameters

handle any (optional) — Text handle from text.create.

Returns any — { requested, selected, resolved, postScriptName, matched, reason, faces, glyphCount, missing }, or nil for a handle the text system does not hold.

local r = text.face(h).reason; if r == "familyUnknown" or r == "familyNotSelectable" then print(r) end

globals/text/faceReasons

text.faceReasons() -> { string }

Every reason the face readings give for a label or a family not being in the family a style named, nearest cause first. text.face gives them for one label; font.reconcile() also gives familyCoveredNoGlyph, which it can only reach by laying the family out under its own weights and over several scripts.

Returns { string } — Array of reason strings.

for _, r in ipairs(text.faceReasons()) do print(r) end

globals/text/isDirty

text.isDirty(handle: any?) -> boolean

Whether the handle was laid out again since it was last rasterised: its content or style changed, or a font arriving, leaving or changing the names it answers to changed the face it draws with. Rasterise it again when this answers true to show what it now says.

Parameters

handle any (optional) — Text handle from text.create.

Returns boolean — True when the layout changed since the last text.rasterize.

if text.isDirty(h) then text.rasterize(h, text.textureGuid(h)) end

globals/text/listFonts

text.listFonts() -> { string }

List the font families currently available to the text system.

Returns { string } — Array of font-family name strings.

local fonts = text.listFonts()

globals/text/loadFont

text.loadFont(ref: any?) -> any

Load a font from an asset reference so it becomes available to setStyle's fontFamily.

Parameters

ref any (optional) — Font asset reference or path.

Returns any — The loaded font-family name, or nil on failure.

text.loadFont(asset.ref("fonts.inter", "font"))

globals/text/measure

text.measure(handle: any?) -> any

Measure the rasterised text in pixels without producing a texture.

Parameters

handle any (optional) — Text handle from text.create.

Returns any — Table with width and height in pixels.

local size = text.measure(h)

globals/text/observe

text.observe() -> any

Everything the text system is holding right now. count is the live text objects; objects is one row each, carrying its content, the style it was laid out with, its measured extent, whether it is dirty, the face the shaper actually used, the owner entity whose component created it with whether that entity is still there, and the raster texture its last rasterisation landed in with the bytes it costs. orphans is the subset whose owning entity is gone, fonts the families the shaper can resolve, and raster the glyph-raster bytes with the pool they belong to named. Built when you ask, so it costs nothing per frame and reads the same in edit mode as in play.

Returns any — { count, objects, orphans, fonts, dirty, raster }.

local live = text.observe().count

globals/text/orphans

text.orphans() -> { any }

The text objects whose owning entity no longer exists — a quad the engine is still holding for something that has been despawned. Each row is the same shape text.observe().objects carries.

Returns { any } — Array of text-object rows with a dead owner.

print(#text.orphans() .. " labels outlived their entity")

globals/text/rasterMemory

text.rasterMemory() -> any

The glyph-raster bytes, broken out of the runtime GPU texture pool. bytes is summed off the same map renderer.gpuMemory().textures is totalled from, so shareOfPool is a share of that number rather than a second count of the same memory.

Returns any — { pool, bytes, textures, poolBytes, shareOfPool }.

local r = text.rasterMemory(); print(r.bytes .. " of " .. r.poolBytes)

globals/text/rasterize

text.rasterize(handle: any?, texture: any?, scale: number?) -> any

Rasterise the handle's current text + style into the given runtime GPU texture. Bind that texture's guid as a material's base_color_texture to display the text; re-rasterising the same texture overwrites it in place.

Parameters

handle any (optional) — Text handle from text.create.
texture any (optional) — Destination GPU texture handle (renderer.texture.create) or its guid string — WHERE the raster lands.
scale number (optional) — World/pixel scale factor for the raster (default 1.0).

Returns any — Table with width and height (in pixels), or nil if nothing rasterised.

local tex = renderer.texture.create({ width = 256, height = 64 })
local r = text.rasterize(h, tex, 1.0)

globals/text/setStyle

text.setStyle(handle: any?, style: table) -> boolean

Replace the handle's style. Fields not present keep their current value.

Parameters

handle any (optional) — Text handle from text.create.
style table — Style table (fontSize, color, alignment, outline, ...).

Returns boolean — True when the text system held the handle and took the style; false when it did not.

text.setStyle(h, { fontSize = 64, color = "yellow" })

globals/text/setText

text.setText(handle: any?, content: string) -> boolean

Replace the handle's text content.

Parameters

handle any (optional) — Text handle from text.create.
content string — New text string.

Returns boolean — True when the text system held the handle and took the content; false when it did not, which is how a caller learns its handle went away.

if not text.setText(h, "HP: 100") then h = text.create({ content = "HP: 100" }) end

globals/text/textureGuid

text.textureGuid(handle: any?) -> string?

The runtime GPU texture guid this handle rasterises into — bind it as a material texture (base_color_texture) to display the text.

Parameters

handle any (optional) — Text handle from text.create.

Returns string? — The texture guid string, or nil for a handle the text system does not hold.

entity(id).component.get("Material"):setTexture("base_color_texture", text.textureGuid(h))

globals/toml/encode

toml.encode(root: { [string]: any }) -> string

Encode a Luau table as canonical TOML bytes. Top-level string-keyed sub-tables become section headers ([name]); deeper string-keyed tables become dotted sections ([a.b]). A sequence whose every element is a non-empty string-keyed table becomes one [[a.b]] block per element, after the section's scalars. Other sequence tables are emitted as inline arrays, and string-keyed tables in value position (inside an array or a block) as inline tables ({ k = v }). Section + key order is alphabetical so the same input always produces the same bytes.

Parameters

root { [string]: any } — The table to encode. Must be string-keyed at the root.

Returns string — A TOML-formatted string suitable for vfs.write.

local body = toml.encode({ render = { culling_mode = "gpu" } })

globals/toml/parse

toml.parse(src: string) -> { [string]: any }

Parse a TOML document into a nested Luau table. Sections ([a.b]) become nested tables; key/value pairs become entries on the current section (or root if before any section header). Arrays of tables ([[a.b]]) become sequences of tables, one per block, and a section after a block ([a.b.c]) is a sub-table of that block's row. Throws with the line number on syntax errors.

Parameters

src string — TOML source bytes as a string.

Returns { [string]: any } — The parsed root table. Sub-tables are plain Luau tables; arrays are 1-indexed sequence tables.

local t = toml.parse('[a]\nx = 1\ny = "hi"\n')

globals/toml/replaceArrayTables

toml.replaceArrayTables(src: string, name: string, rows: { { [string]: any } }) -> string

Replace every [[name]] block in a TOML document with rows. A block is its [[name]] header and the lines after it up to the next header that is neither [name.…] nor [[name.…]] — those sub-tables go with their row. Comment lines between a block's last key and the header that ends it stay, with that header; a line inside a multi-line array value belongs to that value. Every line outside the blocks is kept byte for byte, including its line ending. When nothing is removed and rows is empty the document is returned unchanged; otherwise blank lines at the end are dropped, and each row is appended as a [[name]] block after one blank line, in the document's own line ending (CRLF when it uses CRLF).

Parameters

src string — The document's text.
name string — The top-level array-of-tables name, e.g. "presets".
rows { { [string]: any } } — The rows, each a string-keyed table.

Returns string — The new document text.

local path = "/zero/source/.world_settings"

globals/tonumber

tonumber(value, base?) -> number | nil

Convert value to number.

globals/tools/__useVia

tools.__useVia(via: string, toolbox: string, tool: string, ...: any?) -> ...any

tools.use for a call site written as editorTools("<toolbox>").<tool>(...): the toolbox the site recorded under via is the one reached.

Parameters

via string — The reference kind the call site recorded ("editorTools" or "tools.use").
toolbox string — The toolbox name the site wrote.
tool string — The tool name.
... any (optional)

Returns ...any — Everything the tool returned. Raises on an unknown toolbox / tool or a tool-reported failure.

__useVia("editorTools", "entityOps", "spawn", { Model = { model = "cube" } })

globals/tools/bind

tools.bind(identity: string, positional: { any }?, named: { [string]: any }?, opts: BindOpts?) -> BindResult

Resolve a call's arguments against a tool's declared parameters, turning named arguments into the positional call the tool actually takes. This is the binder behind zero <toolbox> <tool> --name value and the named args an MCP caller passes, so a name resolves the same way whichever surface the caller reached for. Reads the schema from tools.get, matches each name to a parameter (exactly first, then case-insensitively), and reports the first unresolvable name, a name that a positional argument already filled, a call longer than the signature, and a required parameter skipped over while a later one is filled. Resolves the call only — running it is tools.use.

Parameters

identity string — Tool identity ("<toolbox>.<name>", with or without the leading tools.).
positional { any } (optional) — Arguments already given by position, filling slots from 1.
named { [string]: any } (optional) — Arguments given by name, { [parameterName]: value }.
opts BindOpts (optional) — order — the order to visit named in, so the first problem reported is the caller's first (defaults to sorted, for a stable answer). prefix — written in front of every argument name in the failure message, "--" for a shell flag. whole — read named as ONE table argument when not one of its keys names a parameter and the first parameter can hold a table, as the first argument written inline when only some of them do, and as arguments named against the parameter list when no parameter could hold the table — where an empty named then resolves to a call with no arguments at all. positionalCount — how many slots positional fills, for a caller that passed an explicit nil and so cannot be measured by length.

Returns BindResult — { ok, call, count, failure? }. Call the tool with table.unpack(call, 1, count); on failure call is empty and failure carries the reason.

tools.bind("camera.lookAt", {}, { target = { 0, 5, 0 } })
tools.bind("MaterialAuthor.fromColor", {}, { color = { 1, 0, 0 } }, { whole = true })

globals/tools/callable

tools.callable(toolbox: string, tool: string) -> string?

The identity a registered tool is called by — <toolbox>.<tool>, the pair tools.use takes — or nil when no toolbox registered under toolbox carries tool. Reads the bound toolbox proxies, so it answers for what this engine can call rather than for what the asset tree holds: a .tool folder whose source never registered a function answers nil. A name several toolboxes share reaches no one tool and answers nil too; tools.toolboxesNamed names the toolboxes it reaches.

Parameters

toolbox string — The toolbox's registered name ("zm", a library toolbox's scoped identity "@lib::ns", or a toolbox's canonical identity "pkg.blender").
tool string — The tool's leaf name ("status").

Returns string? — "<toolbox>.<tool>", or nil when the registry carries no such tool.

local callable = tools.callable("zm", "status") -- "zm.status"

globals/tools/create

tools.create(args: { [string]: any }) -> { ok: boolean, path: string?, identity: string?, signature: string?, error: string? }

Create a new tool on disk inside an EXISTING toolbox. Scaffolds the .tool/ folder via asset.create("tool", …), then writes the supplied code wrapped in a documented typed function into init.luau (the --!desc/--!arg/--!return/ --!example doc block + signature built from the structured metadata), the tags into .metadata, and a brief README.md — so the authored tool is indistinguishable from a builtin. Errors cleanly if the parent toolbox doesn't exist — call tools.createToolbox first so the toolbox starts out with a real description instead of the placeholder template body. The engine's normal hot-reload pipeline picks up the new files and binds the tool's global on the next pass. Every toolbox created via tools.createToolbox ships a shared.module/ (the template default has ok/fail result-envelope constructors; users can extend or replace it via sharedCode at create time or by editing the module later). The generated init.luau automatically declares local shared = require(".shared") before your function body, so the code you pass can reference shared.ok(...) / shared.fail(...) (or whatever the toolbox's custom helpers expose) directly. If a toolbox doesn't have a shared.module/ for some reason, the require is skipped so there's no dangling import to fail.

Parameters

args { [string]: any } — Structured tool definition. Required: name (leaf, e.g. "hello"), toolbox (parent toolbox, e.g. "mytools"), code (the Luau function body — not a full module, just the statements that will become the function's body), description (one or more sentences describing what the tool does). Optional but recommended: args (array of {name, type, description} argument records), returns ({type, description} for the return value), examples (array of call-site code strings), tags (array of strings for search).

Returns { ok: boolean, path: string?, identity: string?, signature: string?, error: string? } — { ok, path?, identity?, signature?, error? }. path — the new .tool/ folder's VFS path. identity — the tool's <toolbox>.<name> identity. signature — the derived name(args) -> ret signature.

tools.create({ name = "hello", toolbox = "mytools", description = "Greet.", code = "return 'hi'" })

globals/tools/createToolbox

tools.createToolbox(args: { [string]: any }) -> { ok: boolean, path: string?, identity: string?, error: string? }

Create a new toolbox folder. A toolbox is the namespace container for tools — .toolbox/ on disk; once tools are authored inside it they are invoked as tools.use("<name>", "<toolName>", …). This call scaffolds the .toolbox/ folder via asset.create("toolbox", …) and overwrites the placeholder README.md with a real description so the toolbox doesn't ship the template stub. Optionally seeds shared.module/ if the caller supplies cross-tool helper code. Authoring a tool inside this toolbox is the separate tools.create call.

Parameters

args { [string]: any } — Structured toolbox definition. Required: name (the toolbox's leaf name, e.g. "mytools" — becomes <name>.toolbox/ on disk and the namespace for tool identities), description (one or more sentences describing the surface this toolbox exposes — what problem space the tools cover and who calls them). Optional: sharedCode — replacement body for the toolbox's shared.module/init.luau. The toolbox template ALWAYS ships a shared.module/ with default ok/fail result-envelope constructors so tools in this toolbox can require(".shared") from day one. Supply sharedCode only when you want to override that default with custom cross-tool helpers (parsers, registries, …); the module body is left untouched if you omit it. path — explicit VFS destination (/zero/source/... form). Defaults to /zero/source/tools/<name> if omitted; pass an explicit path to author a library-scoped or package-internal toolbox.

Returns { ok: boolean, path: string?, identity: string?, error: string? } — { ok, path?, identity?, error? }. path — the new .toolbox/ folder's VFS path. identity — the toolbox's registered identity (<name>).

tools.createToolbox({ name = "mytools", description = "Custom tooling for my workflow." })

globals/tools/delete

tools.delete(identity: string) -> { ok: boolean, path: string?, error: string? }

Remove a previously-authored tool by identity. Deletes the .tool/ folder and its contents from the VFS via vfs.remove(..., { recursive = true }). The engine drops the tool's global on the next hot-reload pass. Refuses to operate on a path that doesn't exist (returns { ok = false } with an explanatory error).

Parameters

identity string — Tool identity ("<toolbox>.<name>").

Returns { ok: boolean, path: string?, error: string? } — Table shaped { ok, path?, error? }.

tools.delete("mytools.hello")

globals/tools/get

tools.get(identity: string) -> ToolMeta?

Read back a tool's assembled metadata — description, typed signature, per-argument docs, return, examples, and tags — gathered from the four canonical sources in its .tool/ folder. Individual tools aren't entries in the asset index (only their parent toolboxes are), so this resolves the toolbox via asset.resolve(toolbox, "toolbox") and reads the tool relative to it. Returns nil when the toolbox or tool is missing.

Parameters

identity string — Tool identity ("<toolbox>.<name>", with or without the leading tools.).

Returns ToolMeta? — The assembled metadata { name, signature?, description, args, varargs, returns?, examples, typeDefs, tags }, or nil if not found. Each entry in args is { name, type, description, optional } in signature order, so a caller can tell a required parameter from an optional one without parsing the rendered signature. varargs is true when the tool accepts trailing arguments beyond the named parameters. typeDefs carries { name, definition } for every named type the signature refers to, resolved transitively, so a signature reading spawn(opts: SpawnOpts) can be called without opening the tool's source.

local meta = tools.get("entityOps.modify"); print(meta.signature)

globals/tools/list

tools.list(tier: (number | string)?, toolbox: string?) -> { stdout: string, value: any }

Discover registered code-mode tools, grouped by toolbox. Default tier returns just { <toolbox> = { name, name, … } } plus a formatted stdout listing every toolbox on one line with its tools — compact enough that listing the whole catalogue doesn't flood agent context. Higher tiers enrich each entry with its one-line description (tier 2) or its full assembled metadata — signature, args, returns, examples (tier 3). Pass a toolbox name to restrict the output to one toolbox.

Parameters

tier (number | string) (optional) — Verbosity level: 1 (default) toolbox summary, 2 adds one-line descriptions, 3 adds the full assembled metadata. Pass a string instead to restrict to that toolbox at tier 1.
toolbox string (optional) — Optional toolbox name to restrict the output to (e.g. "entityOps"). Passing a string as the first arg also works.

Returns { stdout: string, value: any } — Table shaped { stdout: string, value: <grouped> }. For tier 1, value is { [toolbox] = { toolName, … } }. For tier ≥ 2, value is { [toolbox] = { { name, identity, signature?, description, … }, … } }.

tools.list()              -- default: every toolbox, names only
tools.list(2)             -- include first-line descriptions
tools.list("entityOps")     -- only the `entityOps` toolbox at tier 1
tools.list(2, "entityOps")  -- only `entityOps`, with descriptions

globals/tools/miss

tools.miss(toolbox: string, tool: string?) -> string

What every tool surface says when a toolbox and tool reach no single tool: a toolbox name several toolboxes share names each toolbox it reaches, a toolbox with no tool of that name lists the tools it holds with the nearest one, a tool's name written as the toolbox names the toolboxes holding that tool, and any other name reaching no toolbox lists the nearest toolboxes. tools.use, the describe_tool MCP tool and zero in the engine shell all refuse with this reason, so the same name is refused the same way wherever it is written.

Parameters

toolbox string — The toolbox name the caller wrote.
tool string (optional) — The tool name the caller wrote, when there is one.

Returns string — The reason, as one sentence the caller can act on.

tools.miss("blender", "probe")

globals/tools/registeredToolbox

tools.registeredToolbox(name: string) -> string?

The name a drill-in lists a registered toolbox by, or nil when the registry holds no toolbox under name. Reads the bound toolbox proxies, the same ones tools.callable reads, so it answers for what this engine can call rather than for what the asset tree holds. A name several toolboxes share reaches no one toolbox and answers nil; tools.toolboxesNamed names the toolboxes it reaches.

Parameters

name string — The toolbox's registered name ("zm", a library toolbox's scoped identity "@lib::ns", or a toolbox's canonical identity "pkg.blender").

Returns string? — name, or nil when the registry holds no such toolbox.

local box = tools.registeredToolbox("zm") -- "zm"

globals/tools/search

tools.search(query: string?, opts: { toolbox: string?, limit: number?, mode: string?, from: string? }?) -> { stdout: string, value: any, total: number?, containers: { any }? }

Search registered code-mode tools, the canonical in-engine tool-discovery entry point — callable from execute Luau so an agent can find the tool it needs without leaving the engine. A query is answered by @builtin::modules.discovery (searchTools), the question search_tools { query } asks. In semantic mode on an engine bound to a world, the search core ranks the tools, toolboxes and workflows of the whole published library, then this world's own matches follow; a hit falls by how closely it means what was asked, so a tool is found by what it does rather than by the words it happens to use. In keyword mode, and on an engine bound to no world, the corpus is the files this engine holds. In keyword mode a hit falls by what the query's words the asset carried as words of their own weigh, a word weighing more the fewer of the assets read carry it, then by the most of them it carried side by side in the query's order, then by how many it carried anywhere, then by the order its type declares the role that matched, then by name (discovery.keyword). With an empty query the toolbox named is listed, or the whole catalogue when none is. Each row's identity is the tool's <toolbox>.<tool> identity; invoke it with tools.use("<toolbox>", "<tool>", ...), from Luau or inside a tool_exec script, the form each row's examples are rendered in, or in the engine shell as zero <toolbox> <tool> [args].

Parameters

query string (optional) — Free-text search string, matched against every file each type's indexing recipe names — for a tool, its README and its source. Empty or nil lists the catalogue, or the toolbox opts.toolbox names.
opts { toolbox: string?, limit: number?, mode: string?, from: string? } (optional) — Optional filters table. toolbox — keep only rows whose owning toolbox exactly matches. limit — maximum rows to return (default 10; pass a larger value to return more). mode — "semantic" ranks by what a tool means, "keyword" by the words it carries. Omit it for discovery.defaultMode(): semantic wherever a world is bound, keyword where none is and the only corpus is the files this engine holds. from — WHERE results may come from. 'all' (default) is everything published. 'bound' is the current world and its installed libraries, answered by this engine with no published-library leg; mode still picks how the query is matched, and mode='keyword' reads this engine's own files, unpushed ones and a private world's included.

Returns { stdout: string, value: any, total: number?, containers: { any }? } — Table shaped { stdout: string, value: { SearchRow } }. Each row is { identity, name, assetType, description, toolbox?, registered?, signature?, tags?, examples?, matched?, guid?, installed?, use?, install?, p?, container_key?, system_key?, from? }. A query row says whether this world holds the hit (installed) and the one call that acts on it (use: the tool_exec call that runs it through tools.use, a search_tools { toolbox } drill-in, the workflow toolbox's start through tool_exec, or for a hit this world does not hold bash { command: "zm ls-tree @<guid>" }, the shell read of its implementation, with install the zeromind_install call for the candidate the reading shows fits); from = "world" marks a match from this world's own content after a library ranking. A tool row's identity is the <toolbox>.<tool> a call names; a toolbox row's is the toolbox's name, which toolbox repeats and a drill-in takes. registered says whether the registry holds what identity names: the function a call would run, or the toolbox a drill-in lists. matched is the role and file a query hit matched in and what it ranked on (weight, terms, words, together); a listing carries none. stdout is a one-line summary followed by one line per row. total, on a query naming no toolbox, is how many matched in all, of which value holds the best limit. containers, beside rows the search core ranked, is the tier their container_key and system_key point into: each entry { key, guid, name, assetType, parent_key?, description? } is a container a row ships in, naming the one above it by parent_key, so a tool's container_key is its toolbox and its system_key the package that toolbox ships in.

tools.search("spawn camera")
tools.search("material", { limit = 5 })
tools.search("", { toolbox = "physics" })   -- list a toolbox
tools.search("")                             -- list everything
tools.search("spawn camera", { mode = "keyword" })

globals/tools/toolboxes

tools.toolboxes() -> { stdout: string, value: { { toolbox: string, purpose: string, toolCount: number } } }

Discover the registered toolboxes as a grouped overview — one row per toolbox with its one-line purpose and tool count. The toolbox-first entry point to tool discovery: an agent navigates by DOMAIN (which toolbox), then drills into a toolbox's tools with tools.search("", { toolbox = "<name>" }). purpose is sourced from the toolbox's README.md (its first descriptive line), falling back to the .metadata description.

Returns { stdout: string, value: { { toolbox: string, purpose: string, toolCount: number } } } — Table shaped { stdout: string, value: { { toolbox, purpose, toolCount } } }. value is toolbox-name-sorted; each entry is { toolbox = "<name>", purpose = "<one line>", toolCount = <n> }.

tools.toolboxes()

globals/tools/toolboxesNamed

tools.toolboxesNamed(name: string) -> { string }

The toolboxes a toolbox name reaches, each spelled the way a call reaches it alone. A toolbox is bound under its folder's name, so two toolbox folders with the same name — one in each of two packages, say — share it: that name reaches both, and each is then reached by its own canonical identity (pkgA.blender, pkgB.blender) instead. Every tool surface reads this answer: the listing names each toolbox by it, and a call or lookup through a name reaching more than one is refused naming each (tools.miss states the refusal).

Parameters

name string — A toolbox name ("entityOps"), a library toolbox's scoped key ("@lib::ns"), or any other name the toolbox asset answers to — its canonical identity or its path.

Returns { string } — The names a call reaches each toolbox by, sorted: one entry when the name reaches one toolbox, several when it reaches more, none when it reaches no toolbox holding a tool.

tools.toolboxesNamed("blender") -- { "pkgA.blender", "pkgB.blender" }

globals/tools/tryUse

tools.tryUse(toolbox: string, tool: string, ...: any?) -> ToolCall

Invoke one code-mode tool and report the outcome as a value. Takes the same arguments as tools.use and resolves the toolbox the same way, and returns { ok, value, error, toolbox, tool } for every outcome — an unknown toolbox, an unknown tool, a tool that reported failure, and a tool that raised all arrive as ok = false with the reason in error. The ok / value / error fields are the ones a tool_exec answer reports, so a script comparing several calls in one pass reads the same names it would over MCP, and reads them without wrapping each call in pcall.

Parameters

toolbox string — The owning toolbox — an ambient toolbox by its name ("entityOps") or a library toolbox by its scoped identity ("@lib::ns").
tool string — The tool's leaf name within that toolbox ("spawn").
... any (optional)

Returns ToolCall — { ok, value, error, toolbox, tool }. value carries the tool's own value when ok is true; error carries the reason when it is false.

tools.tryUse("entityOps", "spawn", { Model = { model = "cube" } })

globals/tools/use

tools.use(toolbox: string, tool: string, ...: any?) -> ...any

Invoke one code-mode tool by naming its toolbox and tool explicitly. The same call runs from engine Luau (editor panels, shipped modules, execute) and inside an MCP tool_exec script (return await tools.use("<toolbox>", "<tool>", ...)), and zero <toolbox> <tool> [args] is its engine-shell spelling. Resolves the toolbox from the runtime store (builtin, library, and user-authored runtime toolboxes all work), calls the named tool with the remaining args, and returns the tool's value directly — unwrapping the ZmToolResult envelope and RAISING a Luau error when the tool fails. Because every call names both toolbox and tool, it can never read as a tools.<box> namespace. tools.tryUse takes the same arguments and reports the outcome as a value instead of raising.

Parameters

toolbox string — The owning toolbox — an ambient toolbox by its name ("entityOps") or a library toolbox by its scoped identity ("@lib::ns").
tool string — The tool's leaf name within that toolbox ("spawn").
... any (optional)

Returns ...any — Everything the tool returned, in the order it returned it — a tool declaring (path, reason, frame) hands back all three, so a second and third value the tool states are read the way the tool's own signature says. Raises on an unknown toolbox / tool or a tool-reported failure. These are the tool's OWN values, not an { ok, value } envelope; reach for tools.tryUse when the call's outcome is what you want.

tools.use("entityOps", "spawn", { Model = { model = "cube" } }, { position = {0, 2, 0} })

globals/tostring

tostring(value) -> string

Convert value to string.

globals/transform

transform: any

Lowercase alias of Transform. Same identity. Prefer this in new code so the convention matches entity, scene, cam, physics, ...

globals/type

type(value) -> string

Returns type name of a value.

globals/typeof

typeof(value) -> string

Returns detailed type name.

globals/ui/blur

ui.blur()

Surrender keyboard focus from whichever widget currently holds it, in the editor's UI and the game's alike.

globals/ui/bringAreaToFront

ui.bringAreaToFront(id: string)

Raise a movable area to the top of the window stacking order — the programmatic equivalent of clicking it. Areas sharing a stacking band order by interaction, so this is the call that brings one forward from code: use it when a taskbar button, focus change, or app launch should raise a window. Moving a screen to a higher layer band raises it over the bands below.

Parameters

id string — Area widget id.

globals/ui/captureWindow

ui.captureWindow(screen: string, window: string, opts: CaptureOpts?) -> CaptureResult?

Render a single Window widget to its own offscreen texture and write the result as PNG at /runtime/render_surfaces/<rtHandle>.png. The screen does NOT need to be visible. Returns { rtHandle, texturePath } or nil on invalid inputs (width/height clamped to [1, 8192], defaults 600x400).

Parameters

screen string — Screen id containing the target Window.
window string — Widget id of the Window.
opts CaptureOpts (optional) — { width, height } (optional).

Returns CaptureResult? — { rtHandle, texturePath } or nil.

globals/ui/click

ui.click(callbackId: string, value: any?) -> boolean

Simulate a widget click / interaction by its callback id. The call carries no screen, so an id that names widgets on several screens reaches every component that declared it, once each. The callback id is the one a widget emits: the string its onClick / onChange / other on* prop names, or its element id where it leaves that prop unnamed (a button clicked without an onClick, a slider changed without an onChange). Given the element id of a widget whose interactions emit other ids, it raises an error naming those ids, and clicks nothing.

Parameters

callbackId string — Callback id the widget emits (its onClick / onChange / on* prop).
value any (optional) — Optional value the callback receives as data.value: a number, a string, a boolean, or a table with the fields a structured interaction delivers (a code editor's { token, line, col }, a canvas drag's deltas).

Returns boolean — true when a widget on a registered screen emits the id; false when none does, and the click then still reaches every ui.registerCallbackEnv env and every component and scene entrypoint that receives callbacks no screen claims.

globals/ui/copyText

ui.copyText(text: string)

Put text on the host clipboard — the operating system's clipboard on a native build, the browser's in a web build. What a "Copy path" or "Copy reference" action calls.

Parameters

text string — The text to copy.

ui.copyText("/zero/source/scenes/main.scene")

globals/ui/defineStyle

ui.defineStyle(name: string, style: StyleProps)

Define a named style. Style keys follow <widgetType>.<className> (e.g. "label.h1", "button.primary") or bare <className> to apply across widget types. Widgets reference styles via the classes (or class) prop.

Parameters

name string — Style name.
style StyleProps — Style properties table.

globals/ui/defineStyles

ui.defineStyles(styles: { [string]: StyleProps })

Define multiple named styles at once.

Parameters

styles { [string]: StyleProps } — Map of style name to style properties.

globals/ui/defineWidget

ui.defineWidget(name: string, builderFn: (WidgetTree, { WidgetTree }) -> WidgetTree)

Register a custom widget kind. When a tree contains { type = name, props = ..., children = ... }, the decoder calls builderFn(props, children) at register / update time and substitutes the returned widget table in place. Errors surface through ui.lastValidation() with codes widget-builder-error / widget-builder-bad-return / decode-recursion-depth-exceeded.

Parameters

name string — Custom widget kind name.
builderFn (WidgetTree, { WidgetTree }) -> WidgetTree — Builder closure (props, children) -> widgetTable.

globals/ui/diagnose

ui.diagnose(widgetId: string) -> WidgetPaint?

Why one widget did or did not reach the last frame. Returns that widget's row from ui.observe() — the same fields, resolved against the same reading. An id no registered screen carries reads noSuchWidget, which is how a misspelling separates from a widget whose screen is hidden and from one the frame laid out no box for.

Parameters

widgetId string — The id the widget records layout under.

Returns WidgetPaint? — The widget's row, or nil before the UI has published a frame.

local row = ui.diagnose("hud-healthbar")

globals/ui/displayScale

ui.displayScale() -> DisplayScale

The physical scale the UI is sized at. density is the physical pixels one density-independent point covers on a handheld display (a phone, a tablet, a page in a touch browser): a point of the UI covers at least that many pixels whatever the screen's width or orientation, so text and buttons keep their physical size. It is nil on a desktop window, whose UI is sized by its width alone. textScale is the user's text size setting, which multiplies every font size the UI typesets (labels, buttons, inputs, editor chrome; canvas text is drawn at the size it names). touchTarget is the smallest target, in points, a touch reaches an interactive widget within: 48 on a handheld display, 0 on a desktop. pixelRatio is ui.pixelRatio() of the same frame. ui.simulateDisplayScale previews a phone's on any engine.

Returns DisplayScale — { density?, textScale, touchTarget, pixelRatio }.

local scale = ui.displayScale()
local buttonHeight = math.max(32, scale.touchTarget)

globals/ui/dragState

ui.dragState() -> { payload: string, x: number, y: number, label: string?, ghost: boolean }?

The in-flight drag-and-drop payload while a dragPayload widget is being dragged, else nil. x/y are the pointer's position in the logical space ui.getLayoutInfo rects live in, so the reading resolves directly against widget rects. label is the text the drag ghost beside the pointer reads (dragLabel, else the source's text; nil for an outline ghost), and ghost whether the engine draws one (dragGhost) — a capture never shows the ghost, so this is how a drag's name-at-cursor is checked. Poll during a drag to drive live feedback; the drop itself still lands through the target's onDrop. Snapshotted each frame.

Returns { payload: string, x: number, y: number, label: string?, ghost: boolean }? — { payload, x, y, label?, ghost } during a drag, nil otherwise.

globals/ui/elementTree

ui.elementTree(screenName: string) -> ElementNode?

Introspect a screen's rendered widget hierarchy with each element's layout rect. Every node the renderer draws appears, nested exactly as the widgets nest, under the id it records layout against: the id set on the node when the author gave it one, otherwise <screen>/<type>@<path>. bounds is that element's rect — the same table ui.getLayoutInfo(id) returns — and appears once the element has been measured. Kinds registered through ui.defineWidget appear expanded into the primitives they build. Feeds the gui.captureElement tool: list the tree, pick the ids to frame, capture their region.

Parameters

screenName string — Screen id passed to ui.registerScreen.

Returns ElementNode? — An ElementNode tree, or nil when no screen is registered under that name.

globals/ui/focus

ui.focus(widgetId: string)

Programmatically request keyboard focus on a widget. Queued as a one-shot; the next render of the matching widget calls response.request_focus().

Parameters

widgetId string — Widget id to focus.

globals/ui/focusedWidget

ui.focusedWidget() -> (string?, string?)

Return the widget id of whichever widget currently holds keyboard focus, or nil, and the space it draws in: "window" for the editor's UI around the game view, "camera" for the game's own UI in the game view. Snapshotted post-render each frame.

Returns (string?, string?) — Focused widget id or nil, and its space.

local id, space = ui.focusedWidget()

globals/ui/getAreaPos

ui.getAreaPos(id: string) -> AreaPos?

Read the current pivot position of an area widget, including any user drag deltas. Returns { x, y } or nil if the area didn't render this frame.

Parameters

id string — Area widget id.

Returns AreaPos? — { x, y } or nil.

globals/ui/getAreaSize

ui.getAreaSize(id: string) -> AreaSize?

Read the measured size of an area widget, including any user resize-grip drags if the area is resizable. Returns { w, h } or nil if the area didn't render this frame.

Parameters

id string — Area widget id.

Returns AreaSize? — { w, h } or nil.

globals/ui/getDockLayout

ui.getDockLayout(id: string) -> string?

Read the current serialized layout (split/tab arrangement) of a dockArea widget as a JSON string. Returns nil if the dockArea didn't render this frame. Persist the string and pass it back via the dockArea's layout prop to restore the arrangement.

Parameters

id string — DockArea widget id.

Returns string? — Serialized DockState JSON string, or nil.

globals/ui/getLayoutInfo

ui.getLayoutInfo(widgetId: string?) -> LayoutInfo?

Get layout info (position, size, content bounds) for UI containers. If widgetId is given, returns info for that widget only; otherwise returns all.

Parameters

widgetId string (optional) — Optional widget id to query.

Returns LayoutInfo? — Layout info table or nil.

globals/ui/getPersistentDockLayout

ui.getPersistentDockLayout(id: string) -> string?

Read a dockArea's arrangement as a JSON string with every tab whose dockPanel is transient left out — the form to save and restore across sessions. Returns nil if the dockArea didn't render this frame.

Parameters

id string — DockArea widget id.

Returns string? — Serialized DockState JSON string without transient tabs, or nil.

globals/ui/getScreenTree

ui.getScreenTree(screenName: string) -> WidgetTree?

Return the last widget tree table passed to registerScreen / updateScreen for screenName.

Parameters

screenName string — Screen name to query.

Returns WidgetTree? — Widget tree or nil.

globals/ui/getTheme

ui.getTheme() -> string

Get the name of the currently active theme.

Returns string — Active theme name.

globals/ui/getToken

ui.getToken(name: string) -> string?

Look up a single design token value from the active theme.

Parameters

name string — Token name (without $ prefix).

Returns string? — Token value or nil.

globals/ui/getTokens

ui.getTokens() -> { [string]: string }

Get all design tokens from the active theme as a key-value map.

Returns { [string]: string } — Token map.

globals/ui/getWidgetProps

ui.getWidgetProps(typeName: string) -> { WidgetPropDescriptor }?

Get the property definitions for a widget type.

Parameters

typeName string — Widget type name.

Returns { WidgetPropDescriptor }? — Array of property descriptors, or nil if type not found.

globals/ui/getWidgetTypes

ui.getWidgetTypes() -> { string }

Get all available widget type names that can be used in widget trees.

Returns { string } — Array of widget type names.

globals/ui/hideScreen

ui.hideScreen(name: string) -> boolean

Hide a registered screen, and report whether a screen by that name is registered. The engine applies the hide later in the frame; listScreens reflects it from the next call onwards.

Parameters

name string — Screen identifier to hide.

Returns boolean — True when a screen by this name is registered.

globals/ui/hitTest

ui.hitTest(x: number, y: number) -> PaintHitTest?

Which widget a pointer at (x, y) reaches, and the stack beneath it. Coordinates are logical points on the window, the space getLayoutInfo and gui.bounds rects are in wherever a widget's screen draws (a screen in the Scene view's picture included). A widget that takes no pointer, a canvas declaring pointerEvents = "none", is passed over: the answer is the widget beneath it.

Parameters

x number — Logical X.
y number — Logical Y.

Returns PaintHitTest? — { widget, screen, stack, x, y } — widget nil when the point is over no UI — or nil before the UI has published a frame.

local hit = ui.hitTest(640, 360)

globals/ui/idCollisions

ui.idCollisions() -> { [string]: { string } }

Element ids more than one screen recorded, each mapped to the screens that recorded it. An id addresses one element: two screens sharing one leave ui.getLayoutInfo and the gui element tools answering for whichever painted last, and the gui tools refuse such an id. Empty while every id is unique.

Returns { [string]: { string } } — Colliding id -> the screens that recorded it.

for id, screens in ui.idCollisions() do print(id, table.concat(screens, ", ")) end

globals/ui/invisibilityReasons

ui.invisibilityReasons() -> { string }

Every verdict ui.diagnose can report, as a closed list.

Returns { string } — The reason names.

globals/ui/lastRegistration

ui.lastRegistration() -> { name: string, layer: number? }?

The name and layer passed to the most recent ui.registerScreen call, recorded synchronously at call time. A host that mounts a nested app reads this immediately after the mount to learn which screen the nested code registered, without intercepting the ui table.

Returns { name: string, layer: number? }? — { name, layer } for the last registration, or nil if none yet.

globals/ui/lastValidation

ui.lastValidation(screenName: string?) -> any

Validation diagnostics produced at the most recent registerScreen / updateScreen, plus what the render stage found while painting — unknown-font-family reports a style.fontFamily that named no registered font family, once per family per screen, wrap-label-no-width reports a label with props.wrap whose box came out narrower than its own longest word, once per label per screen, and svg-not-as-declared reports an svg widget whose document declared something the picture does not hold — a <text> whose font-family selected no registered family, a <text> the picture holds no glyph of, or a shape whose url(#id) paint resolved to nothing — once per finding per widget per screen. With no args returns a { [screen] = entry } map; with a name returns that screen's entry or nil. Validation gated by world setting ui.validation = "off" | "warn" | "strict" (default "warn").

Parameters

screenName string (optional) — Optional screen name.

Returns any — Validation entry, full map, or nil.

globals/ui/listFonts

ui.listFonts() -> { FontFamilyInfo }

Every font family a style.fontFamily can select. Read from the registry the UI text renderer resolves a family token through, so a family this returns is one a label renders in. family and every name in aliases are accepted as a fontFamily, case-insensitively; aliases carries the web-font names, CSS generic families and face names that select the same group. faces names the concrete face in each weight/style slot, so a fontWeight = 700 against a family with no bold face gets a synthesised heavy. more lists the other faces registered in the family's slots, each drawing the codepoints the slot's face lacks: the installed chunks of an on-demand face (font.onDemand() lists them all). system = true marks a family taken from the host OS — present on this machine, absent on one without it, and absent on WASM — so a UI that must look the same everywhere picks a family with system = false. A fontFamily naming nothing in this list is reported as an unknown-font-family warning through ui.lastValidation(screen) once the screen paints, and the text renders in the default proportional face.

Returns { FontFamilyInfo } — Array of { family, aliases, faces, more, system }, by family.

for _, f in ui.listFonts() do print(f.family) end

globals/ui/listScreens

ui.listScreens() -> { ScreenSummary }

List every registered screen with its current visibility, z-order layer, renderLayer (the render-layer membership mask it draws under: the mask setScreenRenderLayer stated, else the one the surface setScreenSpace named implies, else the ui bit), editorChrome (whether the mask setScreenRenderLayer stated carries the EditorUI bit, which makes the screen the editor's own chrome; a screen placed on the window without stating it is not chrome), whether the screen has a populated root widget tree, and the space it draws in ("window" for the editor's UI, "camera" for the game's), including the register / show / hide / unregister calls the running script has already made. Sorted by layer ascending, then name.

Returns { ScreenSummary } — Array of screen summaries.

globals/ui/listThemes

ui.listThemes() -> { string }

List all registered theme names.

Returns { string } — Array of theme names.

globals/ui/lockOrientation

ui.lockOrientation(lock: OrientationLock)

Hold the display to the orientations lock allows: "landscape" (both landscape directions, so turning the device over turns the app with it), "portrait" (both portrait directions) or "any" (the display follows the device). A world states the orientation it plays in as world.orientation, which the engine locks to in play mode and releases in edit mode, so the editor follows the device; call this to change it while the world runs. Android turns the app at once, a browser grants the lock only to a page in fullscreen in most cases (the page asks again each time it enters fullscreen), and a desktop window presents in the shape it is given. ui.orientation() reports the lock and whether the platform holds it from the next frame. Any other value is refused by name.

Parameters

lock OrientationLock — "any", "landscape" or "portrait".

ui.lockOrientation("portrait")
ui.lockOrientation("any")

globals/ui/missingGlyphs

ui.missingGlyphs() -> { string }

The characters the UI is showing without a glyph: each appears in text a UI surface (the window, the game view, a camera's UI overlay, a screen drawn into a texture) drew in what it last painted, and no face of the family it was drawn in has a glyph for it, so it shows as the replacement box. Every family falls back glyph by glyph to every face the UI holds, so a character here is one no installed face covers. A character a declared on-demand face covers (font.onDemand()) requests that face the frame it is met, and leaves this list once the face is installed and the next frame draws it.

Returns { string } — Sorted array of single-character strings; empty when every character drawn had a glyph.

if #ui.missingGlyphs() > 0 then print(table.concat(ui.missingGlyphs(), " ")) end

globals/ui/observe

ui.observe(screenName: string?) -> PaintObservation?

What the last UI frame painted. Returns { generation, viewport, pointer, pointerOverUi, pointerWidget, widgets } with one widgets row per widget any registered screen holds — its layout box, the clip chain it painted under, the part of that box which reached the frame (visible), the order it painted in (paintIndex), and its reason from the closed set ui.invisibilityReasons() lists. Every rect, the reading's viewport (the window) and pointer are in logical points on the window; a row's viewport is the region the pass that drew it paints into (the whole window, or the rect a Scene panel presents the camera's picture in), and offViewport is judged against it. generation advances once per re-rendered frame, so two calls reporting the same number describe the same frame.

Parameters

screenName string (optional) — Narrow the rows to one screen. Omit for every screen.

Returns PaintObservation? — The reading; nil before the UI has published a frame, and nil for a screenName no registered screen answers to.

local reading = ui.observe("hud")

globals/ui/onOrientationChange

ui.onOrientationChange(callback: (reading: OrientationReading) -> ()) -> () -> ()

Call callback with the new reading every time ui.orientation() changes: the display turns between landscape and portrait or to another rotation, the lock changes, or the platform starts or stops holding it. Each listener is watched by a task of the context that registered it, so a listener a component registered ends when that component is removed, and every other listener goes on hearing changes.

Parameters

callback (reading: OrientationReading) -> () — Receives the reading, as ui.orientation() returns it.

Returns () -> () — A function that removes this listener.

local off = ui.onOrientationChange(function(o) print(o.current, o.rotation) end)

globals/ui/orientation

ui.orientation() -> OrientationReading

The orientation the display presents the app in, and the lock it is held to. current is "landscape" (wider than tall) or "portrait": the shape of the window the app presents in. lock is what ui.lockOrientation last asked, "any" when the display follows the device. enforced is whether the platform holds the display to it: Android always does, a browser does once it grants the Screen Orientation lock, and a desktop window never does, presenting in the shape it is given; reason names why when it does not. rotation is the display's rotation from its natural orientation in degrees (0, 90, 180, 270) where the platform states one, and nil on a desktop window. revision advances every time any of these changes; ui.onOrientationChange calls back on each change.

Returns OrientationReading — { current, lock, enforced, reason?, rotation?, revision }.

if ui.orientation().current == "portrait" then print("lay the HUD out for a tall screen") end

globals/ui/paintOrder

ui.paintOrder(a: string, b: string) -> number?

Which of two widgets paints later: -1 when a paints before b, 1 when after, 0 when level. This is what separates two widgets whose rects are identical.

Parameters

a string — First widget id.
b string — Second widget id.

Returns number? — -1, 0, 1, or nil when the reading holds no row for one of them.

local order = ui.paintOrder("panel-a", "panel-b")

globals/ui/pixelRatio

ui.pixelRatio() -> number

Physical pixels per logical point — the factor between the logical space ui.screenSize() / getLayoutInfo rects live in and the physical space input.mousePosition, the camera viewport rect and input.simulateMouse* coordinates live in. Multiply a layout coordinate by this to aim a simulated pointer at a widget.

Returns number — Physical pixels per logical point (1.0 when unscaled).

globals/ui/pointerWidget

ui.pointerWidget() -> PointerRead?

Whether the UI is consuming the pointer, and which widget holds it — the pointer counterpart of ui.focusedWidget(). x, y are in logical points on the window, the space getLayoutInfo rects are in.

Returns PointerRead? — { x, y, overUi, widget, screen }, or nil before the UI has published a frame.

globals/ui/registerBackgroundShader

ui.registerBackgroundShader(shaderHandle: any?, width: number?, height: number?)

Register a screen-domain .shader as a UI background, drawn via the backgroundShader style. Takes the shader's asset handle from asset.resolve. A shader whose compile failed (shaderHandle:compileStatus() reads failed), or whose program the compile gate refuses, is not registered: the call raises an error naming the shader and the compiler's message, the one compileStatus().error reads. A shader whose compileStatus() still reads pending raises too: register it once it reads compiled.

Parameters

shaderHandle any (optional) — The screen .shader's asset handle, from asset.resolve.
width number (optional) — Render target width (default 1280).
height number (optional) — Render target height (default 720).

globals/ui/registerCallbackEnv

ui.registerCallbackEnv(key: string, env: { [string]: any })

Register an environment table to receive widget-callback broadcasts: its global onCallback(id, value) fires for any widget callback not owned by a specific component instance — the same broadcast a component's onCallback receives. Keyed by key; re-registering the same key replaces the previous env. A component instance is folded into the callback dispatch automatically, so reach for this from a non-component context that hosts a UI surface (a scene entrypoint registering its own screen). Pair with ui.unregisterCallbackEnv(key) so the ref is released.

Parameters

key string — Stable identifier for this registration (re-register replaces).
env { [string]: any } — Environment table whose onCallback receives the broadcasts.

globals/ui/registerScreen

ui.registerScreen(name: string, widgetTree: WidgetTree, layer: number?)

Register a named UI screen with a widget tree. Optional layer controls z-ordering (higher = on top), in bands: below 0 behind everything, 0-99 ordinary app depth, 100-999 always-on-top chrome, 1000+ menu and popup depth. A screen in a higher band covers one in a lower band whatever their roots are; inside a band a floating area or window root sits over ordinary content, and a modal root sits over the whole stack. Tag-based grouping lives in Z.tags (Z.tags.set(name, { "editor" }) after register). A screen belongs to the component instance whose own code registers it (its hooks and methods, and any closure they create, such as a listener it subscribes): that instance's onCallback receives its callbacks, and the screen is unregistered when the instance ends. Code that is not an instance's own (a scene entrypoint, a module, a listener either subscribed) registers a screen no instance owns, which stays until it is unregistered, whichever component's hook ran that code. topics/ui has the rule.

Parameters

name string — Unique screen identifier.
widgetTree WidgetTree — Root widget table.
layer number (optional) — Z-order layer (optional).

ui.registerScreen("hud", tree)

globals/ui/registerTheme

ui.registerTheme(name: string, theme: ThemeDefinition)

Register a theme from a flat Luau table. Most callers should use Z.theme.register(name, table) which runs the cascade for them.

Parameters

name string — Theme name to register.
theme ThemeDefinition — Flat-resolved theme table.

globals/ui/removeScreen

ui.removeScreen(name: string) -> boolean

Alias for ui.unregisterScreen.

Parameters

name string — Screen identifier to remove.

Returns boolean — True when a screen by this name was registered.

globals/ui/resetAreaSize

ui.resetAreaSize(id: string)

Clear a resizable area's remembered size (from a grip drag or ui.setAreaSize) so its declared — or content — size takes over again.

Parameters

id string — Area widget id.

globals/ui/response

ui.response(widgetId: string) -> WidgetResponse?

Per-widget interaction snapshot for the most recent frame. Returns { clicked, hovered, focused, changed, value } where clicked / changed mark transitions and hovered / focused mark current state.

Parameters

widgetId string — The widget id (NOT the onClick / onChange callback id).

Returns WidgetResponse? — WidgetResponse or nil.

globals/ui/safeArea

ui.safeArea() -> SafeArea

The part of the screen no system bar, display cutout or on-screen keyboard covers, in the logical points ui.screenSize() reports. top / right / bottom / left are how far in from each edge the covered band reaches; x / y / width / height are the rect they leave; cutouts lists each display cutout (a camera hole or a notch) as { x, y, width, height }. Screens lay out inside this rect, so anchors, top-level % sizes and a window's default position already keep clear of it; read it to place something by absolute pos, or to run a background under the bars while its controls stay inside. revision advances every time the area changes (a rotation, the system bars shown or hidden, the keyboard opening, a window resize): compare it with the one you last laid out against. All zero on a desktop window; ui.simulateSafeArea lays the UI out as a phone would on any engine.

Returns SafeArea — { top, right, bottom, left, x, y, width, height, cutouts, revision } in logical UI points.

local safe = ui.safeArea()
ui.setAreaPos("hud", safe.x + 8, safe.y + 8)

globals/ui/screen

ui.screen(name: string) -> { [string]: any }?

Get a screen proxy with methods like setResolution and rasterize.

Parameters

name string — Screen name.

Returns { [string]: any }? — Screen proxy table, or nil.

globals/ui/screenSize

ui.screenSize() -> { width: number, height: number }

The UI coordinate space as { width, height } (logical points). This is the space area pos, anchors, and getLayoutInfo rects use — and it is NOT the pixel size of a capture screenshot, which may be downscaled. Use this for absolute area positioning (e.g. pinning a menu above a bottom taskbar) instead of guessing the size from a capture image.

It follows the size the engine draws at, and renderer.setViewportSize changes that size for the task that asks, which is how a layout written against this is checked at a second shape without rebooting the engine, read in the call that set the size before the surface goes back. The pixel size behind these points is renderer.surfaceSize() — the whole drawing surface a screen is laid out over, which an editor layout makes larger than the rect renderer.viewportSize() draws the scene into — and ui.pixelRatio() is the factor between the two spaces.

Returns { width: number, height: number } — { width, height } in logical UI points.

globals/ui/scroll

ui.scroll(deltaX: number, deltaY: number)

Simulate a mouse-wheel scroll event on the UI.

Parameters

deltaX number — Horizontal scroll delta.
deltaY number — Vertical scroll delta.

globals/ui/setAreaPos

ui.setAreaPos(id: string, x: number, y: number)

Programmatically move a movable area widget to (x, y). Applied for one frame; subsequent frames let drag tracking take over.

Parameters

id string — Area widget id.
x number — Target pivot x (screen coords).
y number — Target pivot y (screen coords).

globals/ui/setAreaSize

ui.setAreaSize(id: string, w: number, h: number)

Programmatically set a resizable area's size (the user-size override) — for maximize / restore / tile. Persists until the area's declared width/height changes or ui.resetAreaSize(id) clears it.

Parameters

id string — Area widget id.
w number — Target width (screen coords).
h number — Target height (screen coords).

globals/ui/setDockWindowRect

ui.setDockWindowRect(dockId: string, panelId: string, x: number, y: number, width: number, height: number)

Place the floating window of a dockArea panel at (x, y) with size (width, height). Applies once the panel occupies a window — a request made before then waits for it.

Parameters

dockId string — DockArea widget id.
panelId string — Id of the panel held by the window to place.
x number — Window left edge (screen coords).
y number — Window top edge (screen coords).
width number — Window width (screen coords).
height number — Window height (screen coords).

globals/ui/setScreenRenderLayer

ui.setScreenRenderLayer(name: string, mask: number)

Set a screen's render-layer membership bitmask. A screen draws into a camera or capture only when this mask intersects the camera's include mask, the same rule geometry follows. Content UI carries the ui bit; the editor places its chrome on EditorUI so agent captures can drop it, and a screen carrying that bit is part of the window rather than the camera's picture. Stating a mask here fixes the membership: a screen left to itself takes the one the surface it is placed on implies, EditorUI for the window and ui for the camera's picture. __renderLayers.bit(name) answers with a layer's bit INDEX, so a mask sets the bit at that index and __renderLayers.names(mask) reads it back as the names it holds.

Parameters

name string — Screen identifier.
mask number — Render-layer membership bitmask — the OR of bit32.lshift(1, __renderLayers.bit(name)) over the layers the screen is on.

ui.setScreenRenderLayer("hud", bit32.lshift(1, __renderLayers.bit("ui")))

globals/ui/setScreenSpace

ui.setScreenSpace(name: string, space: string)

Set which surface a screen is part of. "camera" is the picture the viewport camera produces: the screen lays out over the camera's target and is drawn into it, so it sits inside the Scene panel presenting the scene and appears in every capture of the camera — the default for content UI. "window" is the window itself, whatever presents the scene — where the editor's chrome lives, and the default for a screen on the EditorUI layer. The two are one statement made from either end, so a screen placed on the window takes the EditorUI membership unless setScreenRenderLayer states another: the window goes on drawing it while a capture that drops the authoring layers, which is what an agent takes by default, leaves it out. Where the scene presents on the window, the two surfaces coincide. A name that is neither raises.

Parameters

name string — Screen identifier.
space string — "camera" | "window".

ui.setScreenSpace("agent-banner", "window")

globals/ui/setScrollPosition

ui.setScrollPosition(widgetId: string, offsetY: number)

Set the scroll offset of a scrollArea widget.

Parameters

widgetId string — Scroll area widget id.
offsetY number — Vertical scroll offset in pixels.

globals/ui/setShaderUniforms

ui.setShaderUniforms(name: string, uniforms: { [string]: number })

Set uniform values on a registered background shader.

Parameters

name string — Shader name identifier.
uniforms { [string]: number } — Map of uniform name to number value.

globals/ui/setTheme

ui.setTheme(name: string)

Switch the active global theme by name.

Parameters

name string — Theme name to activate.

globals/ui/showScreen

ui.showScreen(name: string) -> boolean

Make a registered screen visible, and report whether a screen by that name is registered. The engine applies the show later in the frame; listScreens reflects it from the next call onwards.

Parameters

name string — Screen identifier to show.

Returns boolean — True when a screen by this name is registered.

globals/ui/simulateDisplayScale

ui.simulateDisplayScale(scale: DisplayScaleSpec?)

Size the UI as a display stating scale would, in place of what the platform states. A density sizes it as a handheld display of that many physical pixels per point (omit it for a desktop display sized by its width), and textScale multiplies every font size the UI typesets (1 when omitted). Previews a phone's UI on a desktop or headless engine, and makes a layout test independent of the device it runs on. ui.displayScale(), ui.screenSize() and ui.pixelRatio() report it from the next frame. Call it with no argument (or nil) to return to the platform's own scale. A value that is not a number above 0 and at most 8 is refused by name.

Parameters

scale DisplayScaleSpec (optional) — { density?, textScale? }; nil returns to the platform's scale.

ui.simulateDisplayScale({ density = 2.625, textScale = 1.3 })
ui.simulateDisplayScale()

globals/ui/simulateSafeArea

ui.simulateSafeArea(area: SafeAreaSpec?)

Lay the UI out as if the system covered the screen by area (logical points) in place of what the platform states: each inset defaults to 0 and cutouts is a list of { x, y, width, height }. Previews a phone's status bar, camera cutout and navigation bar on a desktop or headless engine, and makes a layout test independent of the device it runs on. ui.safeArea() reports it from the next frame. Call it with no argument (or nil) to return to the platform's own safe area. A negative or non-number inset is refused by name.

Parameters

area SafeAreaSpec (optional) — { top?, right?, bottom?, left?, cutouts? } in logical points; nil returns to the platform's safe area.

ui.simulateSafeArea({ top = 40, bottom = 30, cutouts = { { x = 600, y = 0, width = 80, height = 40 } } })
ui.simulateSafeArea()

globals/ui/unregisterCallbackEnv

ui.unregisterCallbackEnv(key: string)

Remove an environment registered with ui.registerCallbackEnv. Its onCallback stops receiving broadcasts. No-op if key isn't registered.

Parameters

key string — The key passed to ui.registerCallbackEnv.

globals/ui/unregisterScreen

ui.unregisterScreen(name: string) -> boolean

Remove a screen from the registry entirely. Unlike hideScreen, this deletes the entry so it no longer appears in listScreens or render iteration.

Parameters

name string — Screen identifier to unregister.

Returns boolean — True when a screen by this name was registered.

globals/ui/unregisterWidget

ui.unregisterWidget(name: string)

Drop a registered custom widget kind. Subsequent references produce an unknown-widget-type diagnostic.

Parameters

name string — Custom widget kind name.

globals/ui/updateScreen

ui.updateScreen(name: string, widgetTree: WidgetTree)

Replace the widget tree of an already-registered screen.

Parameters

name string — Screen identifier to update.
widgetTree WidgetTree — New root widget table.

globals/ui/useStyles

ui.useStyles(themeName: string)

Apply a registered style file's classes additively without changing the active theme.

Parameters

themeName string — Name of the registered style / theme asset.

globals/ui/widgetState

ui.widgetState(widgetId: string, key: string, default: any?) -> any

Read per-widget cross-frame state. Returns the value previously written via widgetStateSet, or default (or nil). State is keyed by widget id and persists across re-renders within a screen's lifetime; cleared automatically when the owning screen is unregistered.

Parameters

widgetId string — Widget id whose state to read.
key string — State key.
default any (optional) — Value to return when nothing has been written.

Returns any — Stored value, default, or nil.

globals/ui/widgetStateClear

ui.widgetStateClear(widgetId: string, key: string)

Remove a per-widget state entry.

Parameters

widgetId string — Widget id whose state to clear.
key string — State key.

globals/ui/widgetStateSet

ui.widgetStateSet(widgetId: string, key: string, value: any?)

Write per-widget cross-frame state. Replaces any existing value under (widgetId, key). Tables are stored by reference.

Parameters

widgetId string — Widget id to scope the state under.
key string — State key.
value any (optional) — Value to store (must be non-nil).

globals/unpack

unpack(table, i?, j?) -> values

Unpack table elements as return values.

globals/userfile/pick

userfile.pick(opts: PickOpts?) -> PickResult

Open the user's system file picker and bring the chosen file(s) into the engine. Yields until the user finishes (call from a coroutine / task, like any task.await) and returns { cancelled, files = {{ name, mime, size, bytes?, vfsPath? }} }. Without writeTo each file carries bytes (a binary-safe string); with writeTo each carries vfsPath (read it with vfs.read). Cancelling returns { cancelled = true, files = {} }; a genuine failure (e.g. a lost browser user-activation gesture) raises an error.

Parameters

opts PickOpts (optional) — Picker options (optional): multiple, folder, title, filters, writeTo.

Returns PickResult — The decoded result table.

local r = userfile.pick({ filters = {{ name = "Images", extensions = {"png","jpg"} }} })
if not r.cancelled then vfs.write("/source/textures/wall.png", assert(r.files[1].bytes)) end

globals/userfile/pickFolder

userfile.pickFolder(opts: PickOpts?) -> PickResult

Convenience for userfile.pick({ folder = true }) — pick a whole directory tree. Yields until the user finishes and returns the same result table as pick. On the web this degrades to a multi-file selection.

Parameters

opts PickOpts (optional) — Picker options (optional); folder is forced true.

Returns PickResult — The decoded result table.

local r = userfile.pickFolder({ writeTo = "/source/imported/" })

globals/userfile/save

userfile.save(path: string, opts: SaveOpts?) -> (boolean, string, SaveReport?)

Save a file the engine holds onto the device of the person at the screen: the way a person keeps a recording, a capture or an export. The person decides where it lands. In the browser the file downloads at once when the save rides a click or key press of the person at the screen (call it from a button's click handler), and otherwise the page shows them an offer they save with a click of their own, since a browser starts a download only for a person who just acted. A native engine opens the host's save dialog in their downloads folder, with the name filled in, and writes the file where they choose. An engine with no window has no person at its screen and refuses. The call waits for the first answer, so run it where it can yield (task.spawn from a callback). Returns true, a sentence saying where the file went, and the report { state, detail, name, size, path } whose state is "saved" (the file is on the device) or "offered" (it is in front of the person, as the page's offer or the save dialog, and lands when they act); or false and the reason the save was refused: no file at the path, a folder, a name with a folder in it, no window, or a file too large for the browser to hold (which the page also shows).

Parameters

path string — VFS path of the file to save.
opts SaveOpts (optional) — { name }, the file name the device gets (the file's own when omitted).

Returns (boolean, string, SaveReport?) — Whether the device took the file. Where it went, or why it was refused. The save report, when the device took the file.

local ok, where = userfile.save("/zero/runtime/recordings/take.mp4")
local ok, where, report = userfile.save("/zero/source/export/level.json", { name = "level-3.json" })

globals/vfs/claimWalks

vfs.claimWalks() -> number

Lifetime count of walks over every asset the world offers: the part of an enumeration's cost that grows with how many assets exist. asset.list reads a snapshot of every registered claim, built once for each change to what the world offers and shared by every list after it; asset.categories reads the category registry and walks nothing. Read it around a burst to learn what that burst cost in walks, a number that reads the same on a loaded machine and a quiet one.

Returns number — Count of claim walks this session. Monotonic.

local at = vfs.claimWalks()
vfs.write("/zero/source/probe.module/init.luau", "return {}")
asset.list("module")
print(vfs.claimWalks() - at)

globals/vfs/clearPlayShadow

vfs.clearPlayShadow() -> boolean

Forget the entire play-shadow set after a bulk promote or discard. Tracking only — never touches the bytes.

Returns boolean — Always true.

vfs.clearPlayShadow()

globals/vfs/contentAddress

vfs.contentAddress(path: string, opts: VfsOpts?) -> string?

The content address (sha-256 hex) of the bytes vfs.read(path) returns, for a path whose bytes the blob store holds, answered synchronously and without the bytes. Two answers that agree name the same bytes, and a write that changes them changes it. Nil for a path whose bytes are held in memory (text under the inline size threshold, a play shadow, a kept read), which a synchronous vfs.read already returns, and for a path nothing is stored at. Keep it beside a copy of a file you read to tell, from a caller that cannot yield, whether the copy is still what the path holds.

Parameters

path string — VFS path.
opts VfsOpts (optional) — { root = "/source/" }.

Returns string? — The sha-256 hex of the blob-held bytes at path, or nil.

local address = vfs.contentAddress("/zero/source/scenes/main.scene/scene.json")

globals/vfs/contentVersion

vfs.contentVersion(path: string) -> number

A number that moves every time the file at path changes this session: each write, and each move onto it, stamps it with a version no file has held this session, and while nothing is stored at path (before it is written, after it is removed or moved away) it reads 0. It moves the moment any change lands, from every origin: a script's write, a peer's, a dev-mode disk pickup, a world-load or branch-swap write made before the world's content is ready, a world's unmount, both ends of a move, and a directory removal. Keep the value you saw when you last read the file, and read it again only when the number moved.

Parameters

path string — The file's path, in any form vfs.read takes.

Returns number — The path's content version.

local seen = vfs.contentVersion("/zero/source/.world_settings")

globals/vfs/copy

vfs.copy(src: string, dst: string) -> (boolean, (WriteReport | string)?)

Copy a file OR directory from src to dst, cp -r style. A directory recurses — every descendant is replicated at the same relative path under dst, .refs sidecars included. .meta sidecars are minted fresh, so a copy is a distinct asset with its own identity. Both paths are absolute. The source may live in any layer (writable, library mount, builtin, runtime-generated); the destination must be a writable route.

Parameters

src string — Source absolute VFS path (file or directory).
dst string — Destination absolute VFS path.

Returns (boolean, (WriteReport | string)?) — True and a WriteReport on success. A copy is a write at its destination, so it reports what vfs.write reports for the same bytes at that path: where they went, and what the static passes read in the Luau among them, per landed leaf when the copy took a folder. A module that is advisory where it sat reads as a script error once it sits on a component path, and the copy is what put it there. (false, errmsg) on failure.

vfs.copy("/zero/runtime/recordings/take1.mp4", "/zero/source/clips/take1.mp4")
local ok, report = vfs.copy(mod, "/zero/source/game/Vent.component")

globals/vfs/currentAuthor

vfs.currentAuthor() -> { id: string, stable: string, connection: string, agent: string?, name: string?, account: string? }?

The caller this call is attributed to: the author a /source write made right now would be recorded under in the play shadow. id names the calling session where the transport attested one, so two agent sessions driving one engine read as two callers even when one linked ZeroMind account minted both their tokens. stable is the key the caller had before the world's roster named it, which a rename leaves where it was; one of several agents sharing a connection, which names itself with the as argument every tool takes, has a stable of its own. connection is the key of the connection the caller called over, the one the caller's staging area and the writer its dirty rows record are kept under, so the agents of one connection stage and commit one another's work. agent is the name the world's roster holds for the caller now; name is its account's username, which every agent of the account shares; account is that shared account. Nil when the engine can name no caller at all, which is the case for engine-authored work and for a caller that presented neither a session nor a token. Compare its id against vfs.playShadowAuthors() to separate your own pending edits from a co-author's.

Returns { id: string, stable: string, connection: string, agent: string?, name: string?, account: string? }? — { id, stable, connection, agent, name, account } for the acting caller, or nil when unattributed.

local me = vfs.currentAuthor()
print(if me ~= nil then me.id else "unattributed")

globals/vfs/dirSink

vfs.dirSink(path: string, opts: DirSinkOpts) -> PluginHandle

A sink for new files under the folder at path, subfolders included, to pass a plugin in a call. The plugin makes whatever files it needs within opts.maxFiles files and opts.maxBytes bytes in all. Only the world's own files take a sink.

Parameters

path string — The folder to write under.
opts DirSinkOpts — { maxFiles = <n>, maxBytes = <bytes>, extensions = { "png" }? }.

Returns PluginHandle — The folder sink.

local frames = vfs.dirSink("/source/frames", { maxFiles = 240, maxBytes = 200 * 1024 * 1024, extensions = { "png" } })

globals/vfs/discardPlayShadow

vfs.discardPlayShadow() -> { discarded: { string }, owed: { { path: string, reason: string } } }

Throw away every play-shadow edit: the discard a play-exit takes when the play session's work is not kept. Each path reverts to its pre-play copy, as vfs.revertPlayShadow does. Every path it throws away is returned in discarded, posted as a notice and logged by name. A path whose pre-play copy cannot be put back yet (the world's blob, still being pulled; the world's manifest row, read once the manifest has taken the update it is applying; or a copy whose write back is refused) leaves the play shadow all the same and is OWED: it holds the play bytes until the copy lands and its write is taken, and the copy is put back then, unless the path has changed in the meantime. Every owed path is returned in owed, posted as a notice and logged by name. Calling it again asks for every owed restore again, pulling a copy whose pull failed.

Returns { discarded: { string }, owed: { { path: string, reason: string } } } — discarded: every path the discard took off the play shadow, sorted, owed ones included. owed: every path it left owed, sorted by path, with why its restore has not landed; empty when every path reverted.

local result = vfs.discardPlayShadow()
for _, path in ipairs(result.discarded) do
print("discarded", path)
end
for _, owed in ipairs(result.owed) do
print(owed.path, owed.reason)
end

globals/vfs/durability

vfs.durability(paths: string | { string }, opts: VfsOpts?) -> Durability

Did that write save to disk, and if not how is it kept? What became of the bytes just written at paths — one path, or an array of them answered as ONE write. durable is true when they are where the call filed them and the next session reads them back, and false when the play shadow took any of them (live in this session and undone by a guarded play-exit unless kept) or any of them is in /source/tmp/ scratch (this session's working space: it reads back while the engine runs and is gone when it stops or restarts, in edit mode and in play alike). /source/local/ is durable: the machine keeps it. durable is present whatever the answer is, so its absence is never a reading. A non-durable answer carries survivesRestart (true when a restart of this engine keeps the bytes, because the bound world syncs them as they land; false when a restart ends them, which it always does for scratch) and warning (the state and what a restart does with it, for a reader scanning values rather than checking a field; it opens with PLAY SHADOW or SCRATCH). When the shadow holds a path the answer also carries playShadow (the routes to disk and what each costs a session other people are running in) and shadowed (which of the given paths the shadow has, or under pending would take); when a path is in scratch it carries scratch (which of them are in scratch).

This is the answer, off the same shadow set and in the same words, that the write_file / edit_file / capture tools attach to their own results and that asset.create reports as its second return value. Ask it here at any other site that lands files, so every write surface states where the bytes went in one set of terms.

Parameters

paths string | { string } — One VFS path, or an array of paths answered together as one write. A path resolves the way vfs.write resolves its own — absolute or @-rooted as it stands, a bare one under /source/ — so the answer is about the file that write landed. A value that is not a path — an AssetRef, a record, a number, a string with nothing in it — raises rather than being answered off the empty set it reads as; a list with no entries names nothing and answers durable.
opts VfsOpts (optional) — { root = "/source/", pending = false }. pending = true asks the same question of a write that has NOT landed yet: where bytes written at these paths would go, read off the same play-lock verdict that will decide it. That is what a surface arranging a write it files later reports from — av.record opens a take here and the film is written when the take ends — because until those bytes land the shadow is holding nothing under the path and the answer would be durable at the one moment the caller can still act on it. The fields and the words are the same either way; the state names the tense it is true in.

Returns Durability — { durable, survivesRestart?, warning?, playShadow?, shadowed?, scratch? }: survivesRestart and warning present exactly when durable is false, playShadow and shadowed when the play shadow holds a path, scratch when a path is in /source/tmp/, which reads back while the engine runs and is gone when it stops or restarts.

local ok = vfs.write(path, body)
local d = vfs.durability(path)
if not d.durable then log.warn(d.warning) end
local ahead = vfs.durability("/zero/source/clips/take.mp4", { pending = true })

globals/vfs/evict

vfs.evict(path: string, opts: VfsOpts?) -> boolean

Drop the in-memory bytes for path from the writable MemFs layer without removing the asset. Use after processing large binaries to reclaim RAM.

Parameters

path string — VFS path whose bytes should be evicted.
opts VfsOpts (optional) — { root = "/source/" }.

Returns boolean — True if MemFs bytes were dropped; false otherwise.

vfs.evict("/zero/source/textures/imported_big.png")

globals/vfs/exists

vfs.exists(path: string, opts: VfsOpts?) -> boolean

Is the path known to the VFS? Checks the Stage-1 metadata (.meta sidecar / ManifestView) — NOT "are the bytes locally cached?". Use vfs.read(path) ~= nil to confirm bytes are reachable.

Parameters

path string — VFS path to check.
opts VfsOpts (optional) — { root = "/source/" }.

Returns boolean — True if the path is known regardless of byte cache state.

assert(vfs.exists("@builtin/models/Cube"))

globals/vfs/folderHashWalks

vfs.folderHashWalks() -> number

Lifetime count of category-folder checksum walks — the part of a write's cost that grows with how much content already exists. A folder's checksum is the tree hash over the descendants that describe the asset, so deriving one reads every key under it, and a write re-derives it for each enclosing category folder whose bytes that write reaches. Content the checksum is made of therefore moves this once per enclosing folder per write; bytes that reach no checksum — a managed sidecar, a scene's autosave overlay — move it by nothing. Read it around a burst to learn what that burst cost in walks, a number that reads the same on a loaded machine and a quiet one.

Returns number — Count of folder checksum walks this session. Monotonic.

local at = vfs.folderHashWalks()
vfs.write("/zero/source/scenes/probe.scene/scene_dirty/entities/e1.json", "{}")
print(vfs.folderHashWalks() - at)

globals/vfs/holdContentSync

vfs.holdContentSync(held: boolean)

Hold (true) or release (false) the content sync of the world this engine binds. While it is held, a take of the world's content (the initial sync a binding runs, or the one a session that lost its connection runs when it subscribes again) applies no row: the rows the backend sends wait unapplied, in order, and are applied the moment the hold is released. The binding stands and writes to authored source are taken and uploaded meanwhile, so the engine sits between binding a world and being shown its content for as long as a test needs it there: world.syncStatus() reads content_synced = false and content_sync_held = true, and the world's load waits at its content step. A take that already finished carries on applying live changes; the hold stops the next take at its first row. A boot started with --hold-content-sync (holdContentSync=1 in a browser launch) raises the hold before the world binds, which is how the initial sync of a fresh engine is held.

Parameters

held boolean — True to hold the content sync, false to release it.

vfs.holdContentSync(true)
vfs.holdContentSync(false)

globals/vfs/installedFrom

vfs.installedFrom(path: string, opts: VfsOpts?) -> InstalledFrom?

Where the asset holding path comes from, when the library carries it by reference. The library's source then holds an install record naming the asset, the world that published it and the commit its bytes are, and the engine serves the asset's files from those. path is the asset's own folder or anything inside it. Nil for a path whose bytes the library or the world holds itself.

Parameters

path string — VFS path of the asset or of anything inside it.
opts VfsOpts (optional) — { root = "/source/" }.

Returns InstalledFrom? — The origin, or nil when the path is not inside an asset carried by reference.

local from = vfs.installedFrom(ref.path) if from then print(from.world, from.commit) end

globals/vfs/isDirectory

vfs.isDirectory(path: string, opts: VfsOpts?) -> boolean

Is ONE path a directory? Answers from reality — the writable layer's children, a resolver-served folder listing, an explicit empty-directory marker — so a loose file whose extension collides with an assetType name (notes.json) reads as the file it is while a real <name>.<type>/ folder reads as a folder. Costs the same whatever the containing folder holds; use vfs.list when you want every entry's kind, this when you hold one path.

Parameters

path string — VFS path to classify.
opts VfsOpts (optional) — { root = "/source/" }.

Returns boolean — True if the path is a directory.

if vfs.isDirectory("/zero/source/Goblin.dynamicAsset") then print("folder asset") end

globals/vfs/isSaveExcluded

vfs.isSaveExcluded(path: string, opts: VfsOpts?) -> boolean

Does this path hold content the machine keeps to itself? /source/tmp/ is session scratch and /source/local/ is this machine's own durable content — each directory itself included, and everything under it. Both are writable, hot-reloadable and enumerable like the rest of /source/; what separates them is where they stop. The engine filters them out of every world save and every peer broadcast, so they reach no world, carry no manifest row there, and a staging verb handed one refuses it by name. The match reads a whole path segment, so /source/tmpfoo/ is ordinary content. Ask here whenever your code has to agree with what a world can hold.

Parameters

path string — VFS path to classify.
opts VfsOpts (optional) — { root = "/source/" }.

Returns boolean — True when the path is held back from world saves and sync.

if not vfs.isSaveExcluded(p) then table.insert(publishable, p) end

globals/vfs/list

vfs.list(path: string?) -> { VfsListEntry }

List entries in a VFS directory. A directory whose current listing the next frame assembles (/zero/runtime/sync/entities after a stretch nothing read the replication surface) is waited for where the calling code can yield (coroutine.isyieldable()), as vfs.read waits; where it cannot, the listing raises VFS_READ_CANNOT_WAIT, naming the path.

Parameters

path string (optional) — Directory path (defaults to /zero).

Returns { VfsListEntry } — Array of { name, isDirectory, path } tables; vfs.stat(entry.path) gives an entry's size and modification time.

for _, e in ipairs(vfs.list("/zero/source")) do print(e.name) end

globals/vfs/memResident

vfs.memResident() -> { { path: string, bytes: number, kind: string } }

List the MemFs entries that are NOT resident-by-default — the writable in-memory layer's binary blobs and its large text files (text at or above the inline-text size threshold). These are the bytes vfs.evict can reclaim: the ones kept in RAM rather than left to fall through to the on-disk BlobStore cache. Small text (resident by default) is omitted. The audit counterpart to vfs.evict and to reading with { keep = true } — use it to see what encoded bytes are held in RAM, and why.

Returns { { path: string, bytes: number, kind: string } } — Array of { path, bytes, kind }; kind is "binary" for non-text content and "large-text" for oversized text. Empty when the VFS isn't up.

for _, e in ipairs(vfs.memResident()) do print(e.path, e.bytes, e.kind) end

globals/vfs/mkdir

vfs.mkdir(path: string, opts: VfsOpts?) -> boolean

Create a directory. mkdir -p semantics — idempotent. Errors if a file already exists at the same path. While play is running an authored /source directory waits for the lock to lift and the refusal RAISES with the reason; writing a file under the path creates it as part of that write, and scratch under /source/tmp/ creates as in edit mode.

Parameters

path string — Directory path.
opts VfsOpts (optional) — { root = "/source/" }.

Returns boolean — True if the directory exists after the call. A creation the running play session refuses raises with the whole reason.

vfs.mkdir("/zero/source/scenes/")

globals/vfs/move

vfs.move(src: string, dst: string, opts: { quiet: boolean? }?) -> (boolean, (WriteReport | string)?)

Move a file from src to dst. By default fires the destination's write side effects; pass opts.quiet = true to suppress them. A quiet move still returns the destination's report and still posts the resolve rule the publish gate reads, which is what vfs.write does with the same option: a publish takes no override, so the finding is stated at the line that carries it. While play is running a move takes the source away, so authored /source content that predates play is refused and the refusal RAISES with the reason; content this play session created moves and stays tracked on the play shadow.

Parameters

src string — Source absolute VFS path.
dst string — Destination absolute VFS path.
opts { quiet: boolean? } (optional) — Optional { quiet: boolean? }.

Returns (boolean, (WriteReport | string)?) — True and a WriteReport on success. A move IS a write at its destination, so it reports what vfs.write reports for the same bytes at that path: where they went, and what the static passes read in the Luau among them, per landed leaf when the move took a folder. (false, errmsg) when the paths themselves refuse it. A move the running play session refuses raises with the whole reason instead.

vfs.move("/zero/source/a.luau", "/zero/source/b.luau")
local ok, report = vfs.move(mod .. "/init.luau", comp .. "/init.luau")

globals/vfs/movedReferenceNames

vfs.movedReferenceNames(path: string?, opts: VfsOpts?) -> { MovedReferenceSource }

The world sources that spell a reference by a name its target no longer has. A source records, beside each name it uses, the guid of the asset the name reached; when that asset later sits at another path (it was moved, or the engine's library files it elsewhere), the reference keeps resolving through the guid while the text keeps the old name. This lists those sources and changes nothing. Writing an entry's currentText back with vfs.write(source, currentText) brings the text up to date without changing what any reference resolves to. Only sources the world authors are considered: the engine library and /source/deps/ keep the text they were published with.

Parameters

path string (optional) — One source to ask about; every source the world authors when absent.
opts VfsOpts (optional) — { root = "/source/" }.

Returns { MovedReferenceSource } — One entry per source holding moved names, in path order; empty when none does.

for _, s in ipairs(vfs.movedReferenceNames()) do print(s.source, #s.names) end

globals/vfs/mutationSeq

vfs.mutationSeq() -> number

Lifetime count of VFS mutations the engine has APPLIED — the drain's clock. A write queues its side effects (an asset's content reload, the assetType's onChange, a component or scene registration) and a later frame runs them; this number advances as each one completes. Read it, write, then poll for a larger value to learn the queue has moved past the point you wrote at — instead of waiting a guessed number of frames. It counts every mutation kind, so it answers about the pipeline rather than about one file; asset.reloadSeq(ref) is the per-asset reading.

Returns number — Count of applied VFS mutations this session. Monotonic.

local at = vfs.mutationSeq()
vfs.write("/zero/source/tmp/note.txt", "hi")
repeat task.wait() until vfs.mutationSeq() > at

globals/vfs/open

vfs.open(path: string, opts: VfsOpts?) -> PluginHandle

A reader of the file at path, to pass a plugin in a call. The plugin reads it by range and reaches nothing else through it; it is refused, raising the reason, where this script could not read. A reader whose file leaves its path answers the plugin that the asset is gone.

Parameters

path string — The file to read.
opts VfsOpts (optional) — { root = "/source/" }.

Returns PluginHandle — The reader.

local weights = vfs.open(asset.containing(__FILE__).path .. "/weights.gguf")

globals/vfs/openDir

vfs.openDir(path: string, opts: VfsOpts?) -> PluginHandle

A reader of the folder at path and everything under it, to pass a plugin in a call. The plugin lists it and opens readers of its files, by paths relative to the folder that cannot leave it.

Parameters

path string — The folder to read.
opts VfsOpts (optional) — { root = "/source/" }.

Returns PluginHandle — The folder reader.

local model = vfs.openDir(asset.containing(__FILE__).path .. "/model")

globals/vfs/pendingWrites

vfs.pendingWrites() -> { string }

List the /source paths with a local write or removal the synced manifest has not reflected yet: everything this client still owes the server. A just-written or just-removed file appears here until its upload round-trips and the synced dirty state catches up; world.vcsStatus unions these so a fresh edit reads back as dirty immediately, and world.syncStatus().owed_paths reports the same list. Empty when fully synced.

Returns { string } — Array of VFS paths with a pending (unconfirmed) local write or removal.

for _, p in ipairs(vfs.pendingWrites()) do print(p) end

globals/vfs/playDiscardsOwed

vfs.playDiscardsOwed() -> { { path: string, reason: string } }

The play edits a discard threw away whose pre-play copy has not been put back yet. Each of these paths holds the play session's bytes until its restore lands, which happens on its own once the copy arrives and its write is taken. A path that changes first keeps its new content and leaves this list.

Returns { { path: string, reason: string } } — Every owed path, sorted by path, with why its restore has not landed.

for _, owed in ipairs(vfs.playDiscardsOwed()) do
print(owed.path, owed.reason)
end

globals/vfs/playHoldsInPlace

vfs.playHoldsInPlace(path: string, opts: VfsOpts?) -> boolean

Whether the running play session holds path where it stands: vfs.move of it, and vfs.remove of it without durable = true, are refused while play runs, because what they would take away is content that predates play (or that a durable write made canonical). Content only the play shadow holds (this play session wrote it) answers false and moves freely, and so does every path once play ends or pauses, a path in /source/tmp/ or /source/local/, and a path outside authored source. Ask it before the first step of a change that ends by taking path away, so a change the lock would refuse at its last step is not started: the importers ask it before relocating a source into the asset they produce, and owe the import until play ends.

Parameters

path string — The VFS path: absolute, @-rooted, or bare under /source/.
opts VfsOpts (optional) — { root = "/source/" }.

Returns boolean — True while a move or ambient removal of path would be refused.

if vfs.playHoldsInPlace(src) then log.warn(src .. " moves once play ends") end

globals/vfs/playShadowAuthors

vfs.playShadowAuthors() -> { [string]: { id: string, stable: string, connection: string, agent: string?, name: string?, account: string? } }

The caller behind each currently-shadowed /source path, in the shape vfs.currentAuthor() answers in: the id the write was attributed to, the name the world's roster holds for that caller now (agent, the one to show), its account's username, and the ZeroMind account it authenticated as. Several agents drive one engine at once and every one of their in-play source edits sits in the same shadow set, so this is how a review, a refusal or a verdict tells one agent's pending work from another's. id names the calling session where the call carried one, so two agent sessions on a machine linked to one ZeroMind account are two authors. A path written by a caller the engine could not name carries no entry, because it belongs to nobody in particular, and stays settleable by anyone.

Returns { [string]: { id: string, stable: string, connection: string, agent: string?, name: string?, account: string? } } — Map of shadowed path to { id, stable, connection, agent, name, account }.

local mine = vfs.currentAuthor()
for path, who in pairs(vfs.playShadowAuthors()) do
if mine == nil or who.stable ~= mine.stable then print(path, "belongs to", who.agent or who.name) end
end

globals/vfs/playShadowPaths

vfs.playShadowPaths() -> { string }

Every source write made during play that is not durable yet — the set to keep or discard before leaving play. Lists the /source paths edited during running play that the play shadow holds, each beside the pre-play copy a guarded discard puts back: the universal play shadow-copy set. These are the in-play edits persist promotes over the originals on confirm, or drops on a guarded discard. Empty outside play or when nothing was edited.

Returns { string } — Array of normalized VFS paths currently shadowed.

for _, p in ipairs(vfs.playShadowPaths()) do print(p) end

globals/vfs/promotePlayShadow

vfs.promotePlayShadow(path: string | { string }) -> ((string | { string })?, { string }, { string })

KEEP a source write made while play was running — the call that saves a play-mode edit onto disk and makes it durable. Promotes play-shadow edits into canonical writes: re-asserts the live overlay bytes through the full write pipeline, then unmarks each path. The bytes stay in the engine end to end, so binary content promotes exactly. It answers while play is RUNNING and leaves the mode, the clock and every shadow entry it did not name exactly where they stood. Takes ONE path, or an ARRAY of them — a single folder-asset write shadows the entry file, the README and the metadata together, so a slice is tens of paths, and naming them keeps the call to the caller's own work on an engine other sessions are running in. Every named path is attempted; one that refuses does not stop the ones after it. A path with nothing on the play shadow at it or beneath it settles nothing and raises nothing, since no play edit stands there: the third return value names it. A promotion that cannot happen raises with the reason: the path is a plain folder covering shadowed edits (the refusal names the first five assets and loose files under it by the paths that settle them, and counts the rest), the workspace is read-only, or the write-through failed. A shadow entry whose bytes are gone is dropped as promoted, so the path comes back with nothing written for it.

A folder asset is kept whole: a file of one promotes with every other shadowed file of the same asset, so the files its type requires reach source with whichever of them was named, and naming the folder asset's own path (Gold.material, controls.inputMap) promotes every shadowed file of it, the assets nested in it included. An asset nested inside another settles with its own files, and takes the own files of every enclosing asset the play session made with it, since it stands on source only inside them. A path an earlier call settled, by name or carried this way, answers as settled when it is named again, so a walk over a vfs.playShadowPaths() snapshot one path at a time settles every path in it.

Parameters

path string | { string } — A shadowed VFS path to promote, or an array of them (from vfs.playShadowPaths()).

Returns ((string | { string })?, { string }, { string }) — The path the call settled (nil when nothing was on the play shadow there), or the array of settled paths when an array went in; then the array of every path the call settled, a folder asset's carried files included; then the array of named paths that held nothing on the play shadow. None of the named paths are shadowed any more.

local promoted = vfs.promotePlayShadow("/zero/source/cover.jpg")
local _, settled = vfs.promotePlayShadow("/zero/source/Hero.module/init.luau")
local _, kept = vfs.promotePlayShadow("/zero/source/game/Gold.material")
local named, _, unshadowed = vfs.promotePlayShadow("/zero/source/maybe_written.md")
local mine = {}
for _, p in ipairs(vfs.playShadowPaths()) do
if string.find(p, "/zero/source/mine/", 1, true) == 1 then table.insert(mine, p) end
end
local settled = vfs.promotePlayShadow(mine)

globals/vfs/read

vfs.read(path: string, opts: VfsOpts?) -> (string?, string?)

Read a file from the virtual filesystem. Binary-safe. Returns file contents as a string. A read that finds no bytes returns nil and the reason: one sentence naming the path and why, the text the read_file tool and the VFS HTTP route give for the same path (/zero/source/a.txt: not found, a directory's refusal, a /runtime node's account of what it holds). Handed straight to another call, the read passes that reason on as the next argument too; wrap it in parentheses, Json.decode((vfs.read(path))), to pass the bytes alone. Relative paths resolve under opts.root (default /source/). When the bytes aren't locally cached and the calling code can yield (coroutine.isyieldable(): a task, awake, start, update, onEnable, onOwnerChanged, onSyncReceived, a replicated event's handler), the read waits while the lazy fetch runs and answers the bytes. Code that cannot yield (onDisable, onDestroy, onPropertyChanged, a received synced-function call, a vfs.watch callback, a module's top level while require runs it, a metamethod, a table.sort comparator) cannot wait: a read there of bytes still on their way (a file vfs.exists knows, a render surface being captured, a /runtime/sync node whose current reading the next frame assembles) raises VFS_READ_CANNOT_WAIT, naming the path and where to read it.

Parameters

path string — VFS path.
opts VfsOpts (optional) — { root = "/source/" }.

Returns (string?, string?) — File contents (binary-safe); nil and the reason (the path and why it holds no bytes to read) for a read that finds none.

local src, why = vfs.read("/zero/source/notes.txt"); if src == nil then print(why) end

globals/vfs/readAsync

vfs.readAsync(path: string, opts: VfsOpts?) -> string

Asynchronous binary-safe read. Returns a promise ID that resolves to the file contents. Useful for reading render textures from the main thread without blocking.

Parameters

path string — VFS path.
opts VfsOpts (optional) — { root = "/source/" }.

Returns string — Promise ID — pass to task.await().

local data = task.await(vfs.readAsync("/zero/runtime/screenshots/last.png"))

globals/vfs/readBounded

vfs.readBounded(path: string, maxBytes: number) -> (string?, string?)

Read a file from the virtual filesystem, at most maxBytes long. Binary-safe, and read the way vfs.read reads it. A file longer than maxBytes answers nil, "too_large"; where the layer holding it states its length (a file held in memory, a synced or staged file) its bytes are never read. Relative paths resolve under /source/. Where the calling code can yield, a file whose bytes are not cached is fetched and held to the same bound; where it cannot, a known file whose bytes have not arrived raises VFS_READ_CANNOT_WAIT, as vfs.read does. A read that finds no bytes answers nil and the reason vfs.read gives.

Parameters

path string — VFS path.
maxBytes number — The most bytes the read answers, a whole number 0 or more.

Returns (string?, string?) — File contents (binary-safe); nil and "too_large" for a file longer than maxBytes; nil and the reason vfs.read gives for a read that finds no bytes.

local raw, why = vfs.readBounded("/zero/source/notes.json", 65536)

globals/vfs/reload

vfs.reload(modulePath: string?) -> boolean

Clear entries from the require() cache so the next require(name) re-runs the module's source. Pass a single module identity to drop only that entry; call with no arguments to drop every cached module of the world's own. A module of the built-in library runs once for the process — its body registers into registries that outlive it — and stands through a reload with no argument. The next require runs the source the module's file holds when the reload is called, whichever write brought those bytes there, a quiet one included.

Parameters

modulePath string (optional) — Module identity to reload (omit to reload every module of the world's own).

Returns boolean — For a single identity, whether a module was cached under that name and has now been dropped — false says the name matched nothing. The no-arg form returns true.

vfs.reload("@mylib/utils.helpers")

globals/vfs/remove

vfs.remove(path: string, opts: VfsOpts?) -> (boolean, string?)

Remove a file. Refuses to remove directories unless opts.recursive = true. Refuses protected system roots. While play is running, authored /source content that predates play is refused and the refusal RAISES with the reason; content this play session created is removable, a folder included. Pass opts.durable = true to say the removal IS a change to the source: it takes the route edit mode takes with play still running and the session still in play, the counterpart of vfs.write(path, bytes, { durable = true }) for taking a path away rather than replacing its bytes. Every other refusal stands in either terms — a protected root, a read-only route, a plain directory without recursive.

Parameters

path string — VFS path to remove.
opts VfsOpts (optional) — { root = "/source/", recursive = false, durable = false }.

Returns (boolean, string?) — True on success; (false, errmsg) when the path itself refuses the removal. A removal the running play session refuses raises with the whole reason instead, so a pcall around the call reads it — and the lock answers ahead of whether the path is there, so a locked /source path raises whether or not it holds anything.

vfs.remove("/zero/source/scratch.luau")

globals/vfs/revertPlayShadow

vfs.revertPlayShadow(path: string | { string }) -> ((string | { string })?, { string }, { string })

DISCARD a source write made while play was running, keeping nothing — the opposite of promoting it. Reverts play-shadow edits: restores the pre-play copy captured at the first play-mode write (the last edit-mode state, unstaged edits included) into the live slot — or remove the file when it did not exist at that moment — then unmark the path. Hot-reload picks the original back up, so the running session actually reverts. It answers while play is RUNNING and takes ONE path or an ARRAY of them, on the same terms as vfs.promotePlayShadow. A revert that cannot happen raises with the reason. A file of a folder asset reverts with every other shadowed file of the same asset, as a promotion carries them, and a file of an asset the play session made reverts with every shadowed file beneath it, the assets nested in it included. Naming the folder asset's own path reverts every shadowed file of it. A path with nothing on the play shadow settles nothing and raises nothing, as for vfs.promotePlayShadow. A path an earlier revert settled answers as settled when it is named again; one an earlier promotion settled raises, naming that promotion.

Parameters

path string | { string } — A shadowed VFS path to revert, or an array of them (from vfs.playShadowPaths()).

Returns ((string | { string })?, { string }, { string }) — The path that is now reverted (nil when nothing was on the play shadow there), or the array of reverted paths when an array went in; then the array of every path the call reverted, a folder asset's carried files included; then the array of named paths that held nothing on the play shadow. None of the named paths are shadowed any more.

local reverted = vfs.revertPlayShadow("/zero/source/Foo.component/init.luau")
local dropped = vfs.revertPlayShadow({ "/zero/source/a.md", "/zero/source/b.md" })

globals/vfs/sessionLost

vfs.sessionLost() -> boolean

Whether the live world session this engine established is lost and reconnecting. While it is, every write this engine makes to authored source (/source outside /source/tmp/) is refused, because the backend of a dropped session can never receive it; the refusal names the reconnect. It answers false again from the frame the session re-establishes, and is false on an engine whose session never established (an offline boot, or a world still binding) unless vfs.simulateSessionLoss put it in that state. Ask it before, or after a failure of, work that writes source in steps: the importers do, and owe such an import until the session is back.

Returns boolean — True while writes to authored source are refused because the live session is lost.

if vfs.sessionLost() then log.warn("writes wait for the live session") end

globals/vfs/simulateSessionLoss

vfs.simulateSessionLoss(lost: boolean)

Put the VFS in the state a lost live session puts it in (true), or take it out of it (false), without touching the connection: vfs.sessionLost() answers it, and every write to authored source is refused exactly as it is while a real session is reconnecting. Once a live session has established, the engine's connection observer states the real session every frame, so the simulation holds on an engine whose live session never established (an offline boot), which is where a test drives the paths that answer a lost session.

Parameters

lost boolean — True to refuse writes to authored source as a lost session does, false to accept them again.

vfs.simulateSessionLoss(true)
vfs.simulateSessionLoss(false)

globals/vfs/sink

vfs.sink(path: string, opts: SinkOpts) -> PluginHandle

A sink for the file at path, to pass a plugin in a call. The plugin writes it from empty, up to opts.maxBytes, and the file lands when the plugin closes the sink or drops it. Only the world's own files take a sink.

Parameters

path string — The file to write.
opts SinkOpts — { maxBytes = <bytes> }, required.

Returns PluginHandle — The sink.

local out = vfs.sink("/source/imported/sphere.zvol", { maxBytes = 64 * 1024 * 1024 })

globals/vfs/stat

vfs.stat(path: string, opts: VfsOpts?) -> (VfsStat?, string?)

What is at path, how big it is and when it last changed, without reading it. size is how many bytes vfs.read(path) returns: the length the layer holding the file states (a file held in memory, a synced or staged file), so a large file costs nothing to measure, and where no layer states one (a runtime node, a built-in library file) the length of the bytes this engine holds. It is absent for a directory and for a file whose length nothing states and whose bytes are not on this engine. modified is when this engine's copy last changed, in seconds since the Unix epoch (the scale os.time() reads): the write or arrival that put the current content there, which a move carries along with the content. The memory layer records it for every file whose bytes it holds: text up to the inline size threshold, and everything on an engine with no world bound. It is absent where no time is recorded: a directory nothing was written into as such, a runtime node, and a file whose bytes the blob store holds (binary content, and text past the inline size threshold, on a world-bound engine). For a file, the shell's stat and ls -l print the same size and time. A path nothing is at answers nil and the reason, the sentence a vfs.read of it gives.

To tell whether a file changed without reading it: vfs.contentAddress(path) names the bytes the blob store holds, and vfs.contentVersion(path) moves on every change this session to a file the memory layer holds.

Parameters

path string — VFS path.
opts VfsOpts (optional) — { root = "/source/" }.

Returns (VfsStat?, string?) — The entry's description, or nil and the reason nothing is at the path.

local s = vfs.stat("/zero/source/scenes/main.scene/scene.json"); if s then print(s.size, s.modified) end

globals/vfs/unmarkPlayShadow

vfs.unmarkPlayShadow(path: string) -> boolean

Forget a single play-shadow path after it has been promoted (saved over source) or discarded. Tracking only — never touches the bytes.

Parameters

path string — VFS path to unmark.

Returns boolean — Always true.

vfs.unmarkPlayShadow("/zero/source/Foo.component/init.luau")

globals/vfs/unwatch

vfs.unwatch(watcherId: number) -> boolean

Remove a previously registered VFS watcher.

Parameters

watcherId number — Watcher id returned by vfs.watch.

Returns boolean — True if the watcher was found and removed.

vfs.unwatch(id)

globals/vfs/watch

vfs.watch(path: string, callback: (string, string, string) -> ()) -> number

Register a callback that fires when a VFS path is written or removed. Two match modes: exact, or folder/prefix (key ends with /, and fires for any descendant). The callback runs in the VM that registered it. Returns a watcher id for vfs.unwatch.

Parameters

path string — Exact path, or folder path ending in /.
callback (string, string, string) -> () — (mutated_path, kind, origin) -> (), kind "write" or "remove"; origin "local" for a change this engine made (through vfs.*, the shell or any other surface) and "remote" for one that arrived from another participant in the world.

Returns number — Watcher id.

local id = vfs.watch("/zero/source/", function(path, kind, origin) print(kind, origin, path) end)

globals/vfs/write

vfs.write(path: string, content: string, opts: VfsOpts?) -> (boolean, (WriteReport | string)?)

Write content to a file. Binary-safe. Overwrites existing files by default — pass opts.overwrite = false to refuse to clobber. While play is running a /source write lands on the play shadow: it succeeds and reads back, live in the session, and is undone by a guarded play-exit unless accepted; the report's survivesRestart says whether a restart keeps it. Scratch under /source/tmp/ writes through untouched, in edit mode and in play, and is gone when the engine stops or restarts, so its report says durable = false, survivesRestart = false and a warning opening SCRATCH. Pass opts.durable = true to say these bytes ARE the source: the write reaches canonical /source with play still running and the session still in play, hot-reloading the modules and components that read it, so the edit is observed running in the same play session with nothing left to promote. A durable write RAISES with the reason when the bytes cannot become canonical source. One write carries at most 256 MiB: longer content RAISES, naming its size and that limit, and the path keeps what it held.

Parameters

path string — VFS path to write to.
content string — File content (binary-safe).
opts VfsOpts (optional) — { root = "/source/", overwrite = true, quiet = false, durable = false }.

Returns (boolean, (WriteReport | string)?) — True and a WriteReport on success: durable says where the bytes went, with survivesRestart / warning / playShadow / shadowed beside it when the play shadow took them and survivesRestart / warning / scratch when they landed in /source/tmp/ scratch, diagnostics carries what the static passes read in the Luau it landed, and contentRequirements names what the asset it landed in must still say before the world publishes. It is the answer the MCP write_file tool gives for the same bytes at the same path. On failure returns false + error message; a durable write, and content past 256 MiB, raise instead of returning false.

vfs.write("/zero/source/notes.md", body)
local ok, report = vfs.write("/zero/source/game/Vent.component/init.luau", src)
if report.diagnostics then for _, d in report.diagnostics do print(d.severity, d.path, d.line, d.message) end end
if report.contentRequirements then for _, r in report.contentRequirements do print(r.path, r.detail) end end
if not report.durable then log.warn(report.warning) end
vfs.write("/zero/source/game/Vent.component/init.luau", src, { durable = true })

globals/video/create

video.create(url: string, options: VideoOptions?) -> string

Create a video player. Returns a texture handle (e.g. "video_0") usable directly in material.setTexture() — its frames sample like any other texture.

Parameters

url string — URL or asset path to an MP4 video file.
options VideoOptions (optional) — Playback options: loop (default false), autoplay (default false), rate (default 1.0).

Returns string — Texture handle.

local tex = video.create("http://example.com/clip.mp4", { autoplay = true })

globals/video/destroy

video.destroy(handle: string) -> boolean

Destroy a video player and free the render target and all resources.

Parameters

handle string — Video handle from video.create.

Returns boolean — True if the player was found and destroyed.

video.destroy(rt)

globals/video/getInfo

video.getInfo(handle: string) -> VideoInfo?

Get video information and current playback state. codec is the configuration the stream declares, such as "avc1.64002a", and codecSupported is this host's answer about decoding it: true where it has a decoder, false where it has none, and nil while the answer is still being fetched. state reads "unsupported" for a stream this host has no decoder for, with error naming the codec and what refused it, and no frame is produced for it. The frame counts describe playback since create: framesDecoded is how many frames the decoder produced, framesPresented how many the player handed to its texture, framesExpected how many frames the playback clock reached (one per frame interval a tick landed in), and frameThreadDecodes how many samples were decoded on the engine's frame thread rather than on the decoder's own.

Parameters

handle string — Video handle.

Returns VideoInfo? — { width, height, duration, currentTime, state, rate, loop, codec, codecSupported, error, framesDecoded, framesPresented, framesExpected, frameThreadDecodes } or nil if the handle is invalid.

local i = video.getInfo(rt); print(i.currentTime, "/", i.duration)

globals/video/pause

video.pause(handle: string) -> boolean

Pause video playback. Can be resumed with video.play.

Parameters

handle string — Video handle.

Returns boolean — True if the video was playing and is now paused, or was still loading with a play carried and that play is withdrawn. False for a player that was not playing (ready, paused, ended, error, unsupported) and for a handle that names no player.

video.pause(rt)

globals/video/play

video.play(handle: string) -> boolean

Start or resume video playback.

Parameters

handle string — Video handle from video.create.

Returns boolean — True if the command was accepted: the player is loading (the play is carried until the load lands), ready, playing, paused or ended. False for a handle that names no player and for one in error or unsupported, which never plays.

video.play(rt)

globals/video/seek

video.seek(handle: string, time: number) -> boolean

Seek to a specific time (seconds) in the video.

Parameters

handle string — Video handle.
time number — Target time in seconds.

Returns boolean — True if the seek was accepted; false for a handle that names no player and for one in error or unsupported.

video.seek(rt, 30.5)

globals/video/setLoop

video.setLoop(handle: string, loop: boolean) -> boolean

Enable or disable looping.

Parameters

handle string — Video handle.
loop boolean — Whether to loop playback.

Returns boolean — True if the setting was applied.

video.setLoop(rt, true)

globals/video/setRate

video.setRate(handle: string, rate: number) -> boolean

Set the playback speed multiplier. 1.0 = normal, 2.0 = double speed, 0.5 = half speed.

Parameters

handle string — Video handle.
rate number — Playback rate.

Returns boolean — True if the rate was set.

video.setRate(rt, 2.0)

globals/video/stop

video.stop(handle: string) -> boolean

Stop video playback and reset to the beginning.

Parameters

handle string — Video handle.

Returns boolean — True if the command was accepted; false for a handle that names no player and for one in error or unsupported.

video.stop(rt)

globals/warn

warn(...)

Emit a warning message at the warn log level. Roblox/Luau-style counterpart to print.

Returns nil

globals/world/add

world.add(paths: string | { string }, opts: AddOpts?) -> AddReport

Stage one path's manifest row, or each of a list of paths, expanding to the full asset family if a path lives inside a composite asset. Idempotent at (stage, manifest_row). A list stages every row its paths select once, so naming several files of one asset stages that asset's family once, as git add a b c does. Pass { force = true } to bypass the .zmignore / .gitignore gate — same intent as git add -f. Without force, attempts to stage an ignored path (or a path whose .refs points at an ignored dep) error. { stage = "<name>" } stages into the named area, shared with whoever else names it, instead of the caller's own, so a commit naming that area freezes these paths and leaves every other area's staged. Staging is advisory, and a path is staged in one area at a time. Naming a path another staging area holds takes it from that area, whichever account opened it and whether or not its session is open, and the report names the area, its account and whether that session is working now in released. A take from an area of another account warns, naming that account, the area and its session; a take from another agent or engine of the caller's own account does not. A staged path a caller of another account is writing now is named in written and warned about the same way. A staged source that references an asset whose current version no commit holds needs that version committed with it. The asset is staged beside the source, taken from any other area holding it, and named in carried. A .scene stages as its files on disk hold it, in edit and in play alike; live entity edits nobody saved stay in the session, and a notice names them with the calls that keep or drop them. An installed asset is staged by its root path: every row of the install (its files, its dependencies under /zero/source/deps/) is staged with it, as world.vcsStatus().installs lists it. A path that is one of the other rows of an install raises, naming the install and its root. A file world.vcsStatus lists with the reason references moved stages like any change: its row carries its live reference table and its bytes are the ones the branch head already holds.

Parameters

paths string | { string } — The path to stage, or a non-empty list of paths. Each must be a non-empty string; a gate refusing any of them refuses the call before anything is staged. { renormalize = true } first restates every folder asset's row at or beneath the named paths from the files beneath it, as git add --renormalize does: a folder whose stated version is not the folder hash of its files is staged at the hash its files hold, and a folder that already states it is left out.
opts AddOpts (optional) — Optional { force: boolean?, renormalize: boolean?, stage: string? }. Defaults to { force = false } on the calling caller's own staging area.

Returns AddReport — What the call staged: staged lists the rows the paths name (the asset family of a path inside an asset; otherwise the path and every changed row beneath it, deletions included, so a deleted asset folder stages the deletion of each of its files), leaving out with a warning naming it each path no row at or beneath which holds a change, carried the referenced assets staged beside them because no commit holds their current version, released each path taken from another staging area (stage, its account, same_account, its session and whether that session is live), and written each staged path a caller of another account is writing now (by, account). Each carried row, and each released and written row of another account, is also written to the caller as a warning.

world.add("/source/foo.luau")
world.add({ "/source/a.luau", "/source/b.luau" })
world.add("/source/scene_dirty/entities/42.json", { force = true })
world.add("/source/fauna.module", { stage = "fauna" })
world.add("/source/fauna.module", { renormalize = true })

globals/world/add_all

world.add_all(opts: AddAllOpts?) -> ({ string }, AddAllReport)

Stage every dirty manifest row, whoever wrote it, skipping any path that matches .zmignore / .gitignore. Paths that match an ignore pattern are skipped; world.add(path, { force = true }) is the explicit way to override the gate for an individual path. Rows still flagged conflicted by world.pullAsset are held back too: resolve them (edit + world.add(path), or world.resolvePullConflict) and re-run. Staging is advisory. The working tree is one per branch and staging areas are not, and a path is staged in one area at a time: a path another staging area holds is taken from that area, whichever account opened it and whether or not its session is open, and named in released with the area, its account, its session and whether that session is live. A row another caller is writing is staged like any other. The calling account's own agents and engines never warn each other: their areas and their writes are the account's own. A take from another account's area warns once per area, naming the account, the area and its session, and saying that account is working on the paths now when its session is open; a staged path a caller of another account is writing now warns once per caller and is named in written. Each staged path brings its asset family. A file world.vcsStatus lists with the reason references moved is staged too. The call reads the working tree a fixed number of times and stages and releases in batches, so its cost follows the size of the dirty set. A scene stages as its files on disk hold it, in edit and in play alike: live entity edits nobody saved stay in the session, and a notice names each scene holding some, with the calls that write them into scene.json (sceneAuthoring acceptChanges, layers.active:save()) or drop them.

Parameters

opts AddAllOpts (optional) — Optional { stage: string? }. stage names the staging area to stage into; omitted, the call stages into the calling caller's own area.

Returns ({ string }, AddAllReport) — The paths that were held back: the conflicted ones. Empty when none were. The second value reports the call in full: staged lists the paths it added to the staging area, held lists every held-back path with its reason ("conflicted"), released lists each path taken from another staging area (stage, account, same_account, session, live), and written each staged path a caller of another account is writing now (by, account).

world.add_all()
world.add_all({ stage = "fauna" })
local held, report = world.add_all()

globals/world/affirm

world.affirm(token: string)

Consume an XXX-XXX-XXX-style affirmation token returned by a destructive op surface (e.g. vfs.remove). The destruction commits atomically with the pending-row delete; the token is one-shot. Errors verbatim on expiry / wrong-user.

Parameters

token string — The affirmation token.

world.affirm("ABC-DEF-GHI")

globals/world/args

world.args -> any

Returns any

globals/world/assetInstallable

world.assetInstallable(opts: AssetInstallableOpts) -> AssetInstallableReport

Report whether one published asset can be installed on its own, without raising. world.previewInstall plans a real install and raises when the closure will not resolve; this answers the prior question — does this guid name something ZeroMind will hand over by itself — as a verdict a caller can branch on. Content that ships inside a larger library (a module inside a package, a material inside a system) carries its own published identity, so the answer is per asset rather than per library. Reads only.

Parameters

opts AssetInstallableOpts — { guid } names the asset; ref pins a commit-id instead of the latest.

Returns AssetInstallableReport — An AssetInstallableReport — installable, a verdict, a one-line detail, and the closure's nodes / tree_children / deps shape when one resolved.

world.assetInstallable({ guid = asset.guid("@builtin::materials.neon") })

globals/world/avatar_default_edit

world.avatar_default_edit -> any

Returns any

globals/world/avatar_default_play

world.avatar_default_play -> any

Returns any

globals/world/awaitOutgoingSync

world.awaitOutgoingSync()

Wait until every /source write this session made has reached the branch it was written against. Raises naming the paths that did not land. add / commit / push / checkout / merge / pull already wait on their own; call this before rebinding after a burst of writes, which world.checkout and world.swap refuse over.

Returns Nothing. Raises when a write did not land.

world.awaitOutgoingSync() ; world.checkout("main")

globals/world/beforeCommit

world.beforeCommit(name: string, fn: ((any) -> ())?)

Open-or-resume a staging area, set the message, and materialise the commit. Commits ONLY what's already staged via world.add / world.add_all — git semantics, not git commit -a. The reducer auto-deletes the stage row on success so a subsequent world.commit opens a fresh one.

Where derive hooks compute outputs for the staged content, the commit is sealed under its id first and lands once every required output is computed. The id this returns can name a commit that has not landed yet: world.vcsStatus().sealed lists the sealed commits with what each waits on, and world.unseal returns one to the staging area. When an older sealed commit holds a refusal, this commit queues behind it and the call says so.

{ stage = "<name>" } materialises one of the caller's own staging areas, so the commit carries the paths staged under that name and leaves every other caller's staged. When a commit from another caller has landed since the area was opened, this brings the area onto the branch head there is now and commits it there.

A staged scene is frozen as staged. Live entity edits its overlay holds that nobody saved stay in the session, and a notice names them once the commit lands.

Pre-flight .zmignore refs gate: every staged source's aggregated deps (via asset.deps, which recurses composite asset folders) are checked against the live ignore set. If any dep target's path is currently ignored AND the dep target is not itself in the stage, the commit is refused. This mirrors the closure invariant — a commit whose deps can't resolve cleanly shouldn't land. Force-staging the dep alongside (world.add(dep, { force = true })) makes the ignored dep satisfy the gate. Run fn at the start of every world.commit, before the commit reads what is staged — the moment to write content that belongs in the commit being made, and stage it. fn receives { stage }, the staging area the commit materialises (nil for the caller's own), to pass to world.add. Registering a name again replaces what it held, and fn = nil removes it. A hook that raises refuses the commit, naming the hook.

Parameters

name string — What identifies the hook.
fn ((any) -> ()) (optional) — The hook, or nil to remove it.

Returns The newly-allocated commit id (ULID string).

local id = world.commit("feat: ship widget")
local id = world.commit("fauna: the swallow colony", { stage = "fauna" })
world.beforeCommit("procgen.diskcache", function(ctx) ... end)

globals/world/branches

world.branches() -> { { branch: string, commit_id: string, current: boolean } }

Every branch this world has, with the commit each one's head names and which one this session is on — git branch --list. Sorted by name. A branch exists for everyone in the world; which one you are on is yours alone, so current is true for at most one row here and says nothing about where anybody else is.

Returns { { branch: string, commit_id: string, current: boolean } } — Array of { branch, commit_id, current }.

for _, b in ipairs(world.branches()) do print(b.branch, b.commit_id) end

globals/world/camera_default_edit

world.camera_default_edit -> any

Returns any

globals/world/camera_default_play

world.camera_default_play -> any

Returns any

globals/world/checkUpdates

world.checkUpdates() -> ({ UpdateReport }, { UpdateSkip })

Discover which units of this world have an update available. Read-only — makes no local mutation and no VCS write. A unit is one of:

a library: a marker (/source/libs/@<name>, @builtin included) whose requested_ref (default main) now resolves in the library's world to a commit other than the marker's resolved_commit. Apply with world.updateLibrary(name).
an installed asset: an install root (the asset that was installed; its reference table marks it) whose recorded version differs from the newest version of that asset at its origin, one lookup per root. A dependency of an install is never compared or reported: only the root is. Apply with world.update(<root path>) (zeromind.update), which installs the root's new closure with each dependency at the version the new root records.

Installed assets whose origin world is a library's world are part of that library's unit. A pulled row that no table marks as an install root (an install older than the marks, content forked from another world) is not reported.

Returns ({ UpdateReport }, { UpdateSkip }) — Two values: an array of UpdateReport (libraries first), each { kind = "library", library, from, to } or { kind = "asset", root_path, root_guid, display_name, from, to }, empty when everything is current; and an array of { name, reason } for the units whose origin could not be reached, which never fail the whole call.

local reports, skipped = world.checkUpdates()

globals/world/checkout

world.checkout(branch: string) -> string

Switch this session to another branch — git checkout <branch>. The branch must already exist (create one with world.createBranch). The tree is replaced by the branch's own content. Which branch this session is on is this session's alone; the branch itself is shared, so others may be on the one you move onto. Your uncommitted changes and staged entries come with you, as git keeps them across a switch: each lands on branch as it read before. When the switch would replace a path you changed with a different version, the call refuses with git's message naming the paths and nothing moves — commit or stash them first. Another caller's uncommitted changes stay on the branch they were made on. The switch is all or nothing: a write of yours still on its way to the branch you leave is waited for, and one that does not arrive refuses the switch before anything moves; a switch refused after your changes moved puts them back, and names each path that does not read as it did. Returns only once the branch's content has landed, so world.head, world.log, world.commit and the VFS all target the new branch immediately afterwards.

Parameters

branch string — The branch to switch to.

Returns string — The branch now checked out.

world.checkout("feature")

globals/world/checkoutPaths

world.checkoutPaths(rev: string, paths: { string }, opts: CheckoutPathsOpts?) -> { string }

Write paths as a revision holds them, and stage them: git checkout <rev> -- <path>.... rev is any revision world.revParse reads. Each path comes back as the asset the revision records: its asset guid, its bytes and its kind, so what refers to it by guid keeps resolving, and restoring it again gives the same asset again. A folder, a composite asset's or a plain directory, brings back every file the revision holds beneath it, so a deleted component, module or package comes back whole; a file beneath it the revision does not hold stays. A file whose folder asset was deleted brings the folder back with it. Binary files come back by their content. A path whose version is the branch head's reads clean afterwards and leaves the staging area; any other is staged. The paths that name nothing the revision holds are refused together, naming each, and nothing is written. Returns once the working copy holds every written path as the revision records it.

Parameters

rev string — The revision to read.
paths { string } — The paths to write: files, composite asset folders or directories.
opts CheckoutPathsOpts (optional) — Optional CheckoutPathsOpts: stage = false writes the working tree and leaves the staging area as it was (git restore --source <rev> -- <path>).

Returns { string } — The paths written, as the revision names them.

world.checkoutPaths("HEAD~1", { "/source/foo.luau" })
world.checkoutPaths("HEAD", { "/source/Doom.component", "/source/notes.md" })
world.checkoutPaths("HEAD~2", { "/source/levels" }, { stage = false })

globals/world/checkoutSide

world.checkoutSide(path: string, side: string)

Take one side of an unmerged path — git checkout --ours|--theirs <path>. The path holds the side's whole content afterwards; a merge conflict's side is the session branch (ours) or the branch merged in (theirs), an installed asset's update conflict's side is your local version (ours) or the upstream one (theirs). Record the resolution with world.add and commit.

A merge conflict's side is that branch's committed version: its bytes, its asset guid, its kind and its references, so what refers to the asset by guid keeps resolving. Text, binary and composite assets alike: a composite asset (a .scene, .package, .component … folder) is taken whole — name its folder and the folder and every file in it come from that side. A side that deleted the path leaves it a pending delete. Taking a side again replaces the earlier choice.

Parameters

path string — The unmerged path, or the composite asset folder holding one.
side string — ours or theirs.

world.checkoutSide("/source/foo.luau", "theirs")

globals/world/commit

world.commit(message: string, opts: CommitOpts?) -> string

Parameters

message string
opts CommitOpts (optional)

Returns string

globals/world/conflicts

world.conflicts() -> { ConflictEntry }

List every locally-pulled row currently flagged conflicted — the findable surface world.pullAsset leaves behind on an unresolved merge. Read-only.

Returns { ConflictEntry } — An array of ConflictEntry, one per conflicted row. Empty when nothing is conflicted.

local list = world.conflicts()

globals/world/connectedUsers

world.connectedUsers -> any

Returns any

globals/world/contentRequirements

world.contentRequirements(scope: { string }?) -> { ContentRequirement }

List the world's unmet content requirements: user-authored assets whose type-declared content constraints are not yet satisfied (an empty README, a .metadata with no description or tags — the empty-skeleton state a fresh create emits for the author to fill, and a description the type asks for that still starts with the [[describe]] placeholder, which describes = true marks on the row; world.push asks for a description only of the assets the push changes). The same walk world.push gates on, over the whole world: the push refuses over the rows whose asset the calling staging area holds or committed, and publishes the rest with the shortfall recorded on the asset that carries it. world.publishBlockers reports the caller's own share as the content class beside the other two. Empty list = every checked asset meets its type's contract. A write into an asset reports the rows its type's type.yaml requirements give for that asset as it lands: contentRequirements on the vfs.write report and on the MCP write_file / edit_file result.

Parameters

scope { string } (optional) — Asset paths to restrict the check to — pass a status read's dirty + staged paths to check only content that would actually publish (a per-file path matches its containing asset; each path resolves directly, with no world enumeration). Omit for the full-world walk the push gate performs.

Returns { ContentRequirement } — Array of ContentRequirement rows — { asset, typeName, detail, path, describes? }, where path is the asset the row is about.

for _, r in ipairs(world.contentRequirements()) do print(r.asset, r.detail) end

globals/world/contribute

world.contribute(opts: ContributeOpts?) -> { ContributeOutcome }

Send improvements to installed content back upstream — git subtree push ending in a pull request. For each targeted origin world: the diverging subtree is remapped to the origin's canonical paths, three-way merged against the origin's CURRENT content (regions the origin also changed become local conflicts to resolve first), pushed as a contrib/<id> branch in the origin world, and opened as a pull request there. With merge (the default) the pull request is merged immediately when authorized — a refusal leaves it open and reported, never a failure. After a merge, the local fork re-pulls so its origin pins advance and the asset no longer reads as ahead.

Parameters

opts ContributeOpts (optional) — Optional ContributeOpts — targets (origin world guids; default all ahead), roots (install roots; default every installed asset ahead), merge (default true), title, description, dryRun.

Returns { ContributeOutcome } — Array of ContributeOutcome, one per targeted origin.

local r = world.contribute({})

globals/world/convertInstalls

world.convertInstalls(opts: { only: string?, dryRun: boolean? }?) -> { { root_path: string, record: string, rows_removed: number, overrides_kept: number, files_added: number, completion: string? } }

Bring installs that stand as pulled rows (an earlier engine landed each file as a row of the world) into one record row each, the form world.installAsset writes now. The record is built from the rows themselves, written, and once this engine has materialized the install from it the unchanged rows are taken out of the world; a file changed since the install stays a row of the world. Nothing is committed: rows that were committed leave as deletions in the next commit. Returns one entry per install, and with dryRun writes nothing. Once the rows are taken out the record is stated again from the published version it names, so an install the old engine cut short holds every file of the version: files_added counts the files that added, and completion says why the record stayed as the rows stated it (the published version unreachable, or an asset of it that does not load as its type). The files stay where they stand and the call yields between its steps, so it never holds a frame: a run that outlasts the execute budget continues as a task.

Parameters

opts { only: string?, dryRun: boolean? } (optional) — { only?: string, dryRun?: boolean }: only names one install by its root path.

Returns { { root_path: string, record: string, rows_removed: number, overrides_kept: number, files_added: number, completion: string? } } — { { root_path, record, rows_removed, overrides_kept, files_added, completion? } }.

world.convertInstalls({ dryRun = true })

globals/world/createBranch

world.createBranch(name: string, fromCommit: string?, opts: { force: boolean? }?)

Create a branch — git branch <name> [<start>]. The branch starts at fromCommit (defaults to the session branch's HEAD) and gets its own working tree, materialized from that commit. The session stays on its current branch; move onto it with world.checkout("<branch>") (git checkout). With { force = true } (git branch -f) an existing branch is pointed at fromCommit and its working tree rewritten to it, in one transaction; that refuses only when the branch's working tree holds uncommitted changes, which the rewrite would lose.

Parameters

name string — The new branch name.
fromCommit string (optional) — Commit id to start at. Defaults to world.head().
opts { force: boolean? } (optional) — Optional { force: boolean? }.

world.createBranch("feature")
world.createBranch("feature", world.revParse("HEAD~1"), { force = true })

globals/world/deleteBranch

world.deleteBranch(branch: string, opts: DeleteBranchOpts?)

Delete a branch — git branch -d|-D <name>. Drops the branch and the working tree it owns; its commits are left alone, since deleting a branch is dropping the name and the tree under it, not rewriting history. Without opts.force (-d) it refuses a branch whose commits the session's branch does not hold, and one whose working tree holds uncommitted changes, naming them, since those go with the tree; { force = true } (-D) deletes it regardless. Refuses the branch this session is on (check out another first) and the world's last branch.

Parameters

branch string — The branch to delete.
opts DeleteBranchOpts (optional) — Optional DeleteBranchOpts — force.

world.deleteBranch("feature")
world.deleteBranch("experiment", { force = true })

globals/world/derive

world.derive(name: string, spec: any?)

Register a computation whose output lands in the commit with the content it is computed from. The hook (scope = "asset") computes output inside each staged asset matches accepts, from the asset as staged. Its fingerprint and compute are entries of module, run on a worker. Staging starts the work, and a required output that is not computed holds the commit that carries it: world.commit seals the commit and returns its id at once, and the commit lands with the outputs once they are computed. When every required output is computed already, world.commit returns after the commit landed. spec = nil removes it.

Parameters

name string — What identifies the hook.
spec any (optional) — The hook, or nil to remove it.

world.derive("thumbs", { scope = "asset", output = "thumb.png", matches = function(ref) return ref.type == "texture" end, module = "@root::thumbs", fingerprint = "fingerprint", compute = "compute" })

globals/world/diff

world.diff(...: string) -> any

Mirror git diff's CLI arg shape. Returns per-file diffs by default; pass --stat for summary stats, --name-only for just paths. Positional commit ids drive the two sources; --staged pivots to staged-vs-HEAD. ---separated args scope the diff to specific paths. --stage=<name> reads the named staging area in place of the calling caller's own.

Parameters

... string — Variadic string args: flags, commit ids, --, path filters.

Returns any — Array of DiffFile tables (or string-list for --name-only).

local files = world.diff()
local files = world.diff("--staged")
local files = world.diff("abc", "def")
local names = world.diff("--name-only")
local files = world.diff("--staged", "--stage=fauna")

globals/world/diffUpstream

world.diffUpstream(root: string) -> { UpdateFile }

What your copy of an installed asset changed against the version it was installed from — the diff zm diff <root> upstream prints. One UpdateFile per file under root whose content differs from its origin version, with the unified diff from the origin version to yours for a text file (diff and hunks are nil for a binary one). Empty when the installed asset holds no change of yours.

Parameters

root string — The installed asset's root path (or any path inside it).

Returns { UpdateFile } — Array of UpdateFile, sorted by path.

for _, f in ipairs(world.diffUpstream("/source/Doom.component")) do print(f.diff) end

globals/world/discard

world.discard(opts: StageOpts?)

Drop a staging area without committing. Live manifest dirty flags are preserved so the user can re-stage later. No-op if the area doesn't exist.

Parameters

opts StageOpts (optional) — Optional { stage: string? } naming the staging area to drop. Omitted, the call drops the calling caller's own area.

world.discard()
world.discard({ stage = "fauna" })

globals/world/discardAll

world.discardAll(scope: string?) -> DiscardAllResult

Discard every change under a path in one operation — the whole working tree by default. Every file edited since its commit goes back to the committed version, every deletion comes back, and every file no commit holds, an uncommitted install included, leaves the tree. Each goes to trash first, where world.restore(<handle>) brings it back. The paths a staging area holds stay, a staged file edited since it was staged goes back to its staged copy, and a change another account is making and a dependency another install records stay as well: they are named in kept. One backend operation takes the whole scope back, and the call returns once the working copy reads as the committed content.

Parameters

scope string (optional) — The VFS path to discard beneath; omitted, every path.

Returns DiscardAllResult — { discarded, paths, kept } — how many rows went back, which, and which rows under the scope stayed because another account or another install holds them.

world.discardAll()
world.discardAll("/source/hud")

globals/world/discardFile

world.discardFile(path: string)

Discard a path's unstaged working edits, taking it back to what it was staged or committed as — the git restore <path> shape. A folder takes every file beneath it back with it, and a deleted path comes back as its committed version. The stage is the baseline when the path is staged, the last commit when it is not, and where it is neither there is nothing to come back to, so the path goes away. Staging is left exactly as it was; world.unstage is the verb that changes it. One shot: edited bytes that revert to a committed version are snapshotted to trash first, so that case is recoverable via world.restore(<handle>); a deletion taken back throws nothing away and leaves no trash entry. Returns once the working copy reads as the reverted content: text reads back as committed, and binary bytes are served at the committed version on the next read. Errors when nothing at or beneath the path is dirty (nothing to discard). An installed asset is discarded by its root path, which puts the whole install back at the version the branch head holds as one unit: its rows leave the staging area and then return to their committed version, or go away when no commit holds them. A dependency another install records too stays. Another row of an install raises, naming the install and its root. A file whose only difference from the head is its reference table (listed with the reason references moved) raises too, naming what moved its references and the undo: moving the library pin back (world.updateLibrary, or discarding the pin marker) or reverting the moved asset.

Parameters

path string — The VFS path whose unstaged edits to discard.

world.discardFile("/source/foo.luau")
world.discardFile("/source/Doom.component")

globals/world/dropDerived

world.dropDerived(outputPath: string, opts: { stage: string? }?) -> number

Leave the derived output at outputPath out of the commit. Only an output whose hook is best-effort can be dropped; a required one is refused, naming the other ways out (retry, provide, unseal).

Parameters

outputPath string — The output's VFS path.
opts { stage: string? } (optional) — Optional { stage: string? } naming the staging area.

Returns number — How many reservations were dropped.

world.dropDerived("/zero/source/gold.material/extra.json")

globals/world/dropOwedWrites

world.dropOwedWrites(path: string?) -> { string }

Give up the changes this session still owes the world at path and beneath it: the writes and removals vfs.pendingWrites() lists there (a write the world never took included), the ones world.syncStatus() names as conflicts or losses, and every stored record of them on this machine, so no later boot re-applies one. Each path then reads as the world's row holds it: the row's content, or nothing where the world has no row. Returns the paths given up. The bytes of a given-up change are not kept.

Parameters

path string (optional) — Canonical /source path; defaults to the whole source tree.

Returns { string } — The paths given up, sorted.

world.dropOwedWrites()
world.dropOwedWrites("/zero/source/hud")

globals/world/fetch

world.fetch(branch: string?) -> FetchResult

Update the origin/<branch> remote-tracking ref — git fetch. Mirrors the world's ZeroMind branch head into the local commit history (no working-tree change) and reports how the session branch relates to it: behind origin commits to pull, ahead local commits to push, diverged when both. A stale installed pin or out-of-band ZeroMind change shows up here as behind — reconcile with world.pull().

Parameters

branch string (optional) — Remote branch to fetch. Defaults to the session branch.

Returns FetchResult — A FetchResult table.

local f = world.fetch()

globals/world/forkLive

world.forkLive(opts: { source: string, sourceBranch: string?, maxBatches: number? }) -> number

Seed THIS (empty) world's live content from another world by copying its whole manifest as clean Pulled rows — the in-engine half of "fork a world". Provenance is preserved: each row points at the content's ORIGINAL owner (a fork of a fork-of-A still points at A), so the fork never claims to have authored what it pulled. The copy runs server-side in bounded batches (idempotent + resumable), looping until the source is fully mirrored. Pair with world.add_all() + world.commit() + world.push() to publish the fork.

Parameters

opts { source: string, sourceBranch: string?, maxBatches: number? } — { source, sourceBranch?, maxBatches? } — source is the source world GUID; branches default to "main"; maxBatches caps the batch loop (default 60 ⇒ up to ~120k entries).

Returns number — The number of pulled (dirty) rows now staged-pending on the fork.

world.forkLive({ source = "e89aa92e-4c1f-460e-acd8-73859dd3a346" })

globals/world/forkStatus

world.forkStatus() -> { ForkStatus }

Per-asset "ahead of origin" report — the fork analogue of git status against an upstream. Every installed (pulled) row whose content diverges from its pinned origin is listed, partitioned by the TRUE origin world it was pulled from (nested dependencies carry the world that authored them, not the intermediary they arrived through). This is information for judgment: decide whether a change belongs upstream, then world.contribute.

Returns { ForkStatus } — Array of ForkStatus partitions.

for _, f in ipairs(world.forkStatus()) do print(f.origin_world, #f.entries) end

globals/world/head

world.head() -> string?

Return the current branch HEAD commit id, or nil if the branch has no commits yet.

Returns string? — The commit id string, or nil.

local id = world.head()

globals/world/installAsset

world.installAsset(opts: InstallAssetOpts) -> InstallAssetResult

Install a published asset into this world, pulling the asset and every dependency it closes over and writing them into the source tree. The world holds ONE row for the install, its record <dest>.pull, which lists every file of the version and every dependency placed beside it; every engine that holds the row materializes the install from it, so a file of the install is a row of the world only once it is changed. Returns once this engine has materialized the install and it runs. Installing an asset where the world already holds it is safe to repeat: rows at this version are left as they are, and rows of an earlier version are moved to this one. The root lands at at, else where the world already installed it, else at /source/<name>; a copy of it the world holds at this version anywhere else (another install's dependency under /source/deps/) moves there, so the world holds the asset once and everything that depends on it resolves to the moved copy. A path holding another asset, or local edits to this one, refuses the install. An install that does not materialize takes its record out again, so the world holds nothing of it and the same install can simply be run again.

Parameters

opts InstallAssetOpts — { guid } names the root asset to install; ref pins a specific commit-id instead of the latest.

Returns InstallAssetResult — { assets_written, blobs_downloaded, root_guid, root_path, root_version, deps, pending, problems, outcome, moved_from, record, files, timings }. record is the path of the install's row, files the number of files and folders it states, timings the seconds each step took. outcome is installed, updated or already_present; moved_from names the copy the root moved from, nil when it moved from nowhere. blobs_downloaded counts the blobs the install fetched (bytes already cached cost none); pending lists the written paths still unreadable when the wait for them stalled, and is empty when every one reads back; problems names each reference site in the installed content that resolved to nothing in the world it was published from, and each file a folder version the install wrote holds at its version that ZeroMind gave no entry for, and is empty when every site resolves and every written folder holds all of its files. A folder version lands holding every file its version states: a file the closure leaves out of it is read from that version of the folder.

local r = world.installAsset({ guid = assetGuid })

globals/world/installLibrary

world.installLibrary(opts: InstallLibraryOpts) -> InstallLibraryResult

Declarative cross-world dependency. Writes a single marker file at /source/libs/@<name> whose body is the zero/world-import/v1 JSON. The next commit ships it as one regular manifest entry; ZM's import-derivation pass at finalize-time decodes the marker and stamps the new commit's imports[]. Unmodified library content never ships in the importing world's tree: the engine fetches the pinned commit from ZeroMind and mounts it as @<name>::<path>. The call returns once the library is mounted (require("@<name>::…") resolves and library.list() lists it), and raises naming why when it does not mount, leaving the marker as it was before the call. A marker already there whose library is mounted is refused (move it with world.updateLibrary); one whose library never mounted is replaced, so running the install again repairs it.

Parameters

opts InstallLibraryOpts — See InstallLibraryOpts. opts.world is the upstream world's guid (required). opts.commit is the upstream commit_id to pin (optional; resolves opts.ref or main if omitted). opts.as is the local library name (defaults to the upstream world's slug). opts.ref is the human-meaningful ref recorded in the marker.

Returns InstallLibraryResult summarising the install.

world.installLibrary({ world = "guid", as = "combat" })

globals/world/installedAssets

world.installedAssets() -> { InstalledAsset }

Every asset this world carries from ZeroMind, keyed by the published guid it was pulled from. The read that answers "what is actually in this world" by identity rather than by path — an installed asset's local name can be chosen by the installer, so a path is not the thing to check a pull against.

Returns { InstalledAsset } — An array of InstalledAsset sorted by local path, one per pulled row carrying a published guid.

for _, a in ipairs(world.installedAssets()) do print(a.asset_guid, a.path) end

globals/world/installedLibraries

world.installedLibraries() -> { LibraryMarker }

The libraries this world has installed: one row per library marker under /zero/source/libs, @builtin among them, sorted by name.

Returns { LibraryMarker } — { name, world_guid, requested_ref?, resolved_commit } per library: its name with the @, the world it comes from, the ref that was asked for and the commit it is pinned at.

for _, lib in ipairs(world.installedLibraries()) do print(lib.name, lib.resolved_commit) end

globals/world/installs

world.installs(opts: { rows: boolean? }?) -> { InstallInfo }

The installs this world holds, one entry each, whichever form each is stored in: record (one row, every engine materializes the files from it) or rows (an earlier engine landed one row per file, and world.convertInstalls brings it into one record). An asset installed at more than one place names the other places in also_at.

Parameters

opts { rows: boolean? } (optional) — { rows: boolean? }: rows = false leaves the row installs out, which reads no row of the world.

Returns { InstallInfo } — An array of InstallInfo sorted by path.

for _, i in ipairs(world.installs()) do print(i.root_path, i.form, i.files) end

globals/world/list

world.list() -> { WorldEntry }

List every world the authenticated user has access to. Calls the spacetime list_my_worlds procedure which wraps ZeroMind's GET /v1/me/worlds. Flattens each entry to one record per world with the role promoted to a top-level field.

Returns { WorldEntry } — Array of WorldEntry records.

local worlds = world.list()

globals/world/loading_screen

world.loading_screen -> any

Returns any

globals/world/log

world.log(opts: LogOpts?) -> { CommitRow }

The history of the session branch, newest first — git log: every commit reachable from HEAD through both parent edges, so a merged branch's commits and a pulled commit appear along with the commits recorded on this branch. opts.rev starts the history at another revision (git log <rev>). Pass opts.path to get the per-path history (git log -- <path>): only the commits that touched that file, newest-first.

Parameters

opts LogOpts (optional) — Optional. opts.limit caps the number of commits (default 50, 0 = all). opts.rev names where the history starts. opts.path scopes the log to one file.

Returns { CommitRow } — Array of CommitRow tables.

local commits = world.log({ limit = 20 })
local merged = world.log({ rev = "feature" })
local touched = world.log({ path = "/source/foo.luau" })

globals/world/merge

world.merge(sourceBranch: string, opts: MergeOpts?) -> MergeResult

Merge another branch into the session branch — git merge <source>. The merge runs locally in the world's SpacetimeDB clone and is abortable with world.mergeAbort; nothing reaches ZeroMind until the result is pushed. When the session branch's head is in the source's history the head fast-forwards onto it (unless opts.noFf); otherwise a two-parent merge commit lands (opts.ffOnly refuses that). Local changes follow git's rule: one on a path the merge does not rewrite stays in the working tree, and one on a path it does rewrite refuses the merge with git's message naming the paths. A scene's unsaved live edits are a local change to its scene.json: a fast-forward refuses over a scene whose scene.json it rewrites, naming the scene and how to save or discard the edits, and a merge commit refuses over any scene holding them. A staged entry refuses a merge commit. Conflicts → git-style markers are projected into each conflicting text file, the cleanly-merged remainder is applied as working-tree changes, and world.vcsStatus().unmerged lists what needs attention: resolve each path (edit out the markers, or take one side whole), then world.add + world.commit — that commit records the merge (second parent = the source head) and clears the unmerged set.

Parameters

sourceBranch string — The branch to merge in.
opts MergeOpts (optional) — Optional MergeOpts — ffOnly, noFf.

Returns MergeResult — A MergeResult — status is fast_forward or clean (with commit), conflicts (with conflicts), or up_to_date.

local r = world.merge("feature")
local r = world.merge("feature", { noFf = true })

globals/world/mergeAbort

world.mergeAbort()

Abort the in-progress merge — git merge --abort. Clears the unmerged set and restores the working tree to the pre-merge state (the target head's committed content; the branch head never moved during a conflicted merge). Errors when no merge is in progress.

world.mergeAbort()

globals/world/offLoaded

world.offLoaded(handle: number) -> boolean

Stop a callback registered with world.onLoaded from running.

Parameters

handle number — The handle world.onLoaded returned.

Returns boolean — true when the handle matched a registered callback.

world.offLoaded(h)

globals/world/offSaved

world.offSaved(handle: number) -> boolean

Stop a callback registered with world.onSaved from running.

Parameters

handle number — The handle world.onSaved returned.

Returns boolean — true when the handle matched a registered callback.

world.offSaved(h)

globals/world/offUnloaded

world.offUnloaded(handle: number) -> boolean

Stop a callback registered with world.onUnloaded from running.

Parameters

handle number — The handle world.onUnloaded returned.

Returns boolean — true when the handle matched a registered callback.

world.offUnloaded(h)

globals/world/onLoaded

world.onLoaded(cb: (...any) -> ()) -> number

Parameters

cb (...any) -> () — Called when the event fires, with whatever the event supplies.

Returns number — Handle for world.offLoaded.

local h = world.onLoaded(function() log.info("loaded") end)

globals/world/onSaved

world.onSaved(cb: (...any) -> ()) -> number

Parameters

cb (...any) -> () — Called when the event fires, with whatever the event supplies.

Returns number — Handle for world.offSaved.

local h = world.onSaved(function() log.info("saved") end)

globals/world/onUnloaded

world.onUnloaded(cb: (...any) -> ()) -> number

Parameters

cb (...any) -> () — Called when the event fires, with whatever the event supplies.

Returns number — Handle for world.offUnloaded.

local h = world.onUnloaded(function() log.info("unloaded") end)

globals/world/orientation

world.orientation -> string

Returns string

globals/world/prConflicts

world.prConflicts(worldGuid: string?, number: number) -> any

Read a pull request's conflicts — what stands between it and a merge. Returns the mergeability verdict, the merge base, both heads, and one entry per conflicting path. A conflicting TEXT path carries marked_text: the same <<<<<<< / ======= / >>>>>>> rendering a merge leaves in the working tree, with the source and target sides laid against their common ancestor. Resolve a path by writing the settled bytes back to it and committing on the source branch; the pull request re-analyses on the next read. A binary path carries the two sides' hashes and no text — pick a side. A mergeable pull request returns an empty conflict list.

Parameters

worldGuid string (optional) — The world the pull request lives in. Defaults to the bound world.
number number — The pull request number.

Returns any — Decoded ZeroMind conflicts response.

local c = world.prConflicts(nil, 3)
for _, m in ipairs(world.prConflicts(originGuid, 3).markers) do print(m.path, m.marked_text) end

globals/world/prList

world.prList(worldGuid: string?, number: number?) -> any

Parameters

worldGuid string (optional)
number number (optional)

Returns any

globals/world/prMerge

world.prMerge(worldGuid: string, number: number, strategy: string?) -> any

Merge a pull request — the agent-side merge button.

Parameters

worldGuid string — The world the pull request lives in.
number number — The pull request number.
strategy string (optional) — merge (default), squash, or fast_forward.

Returns any — Decoded ZeroMind merge response.

world.prMerge(originGuid, 3)

globals/world/prOpen

world.prOpen(opts: PrOpenOpts) -> any

List a world's pull requests, or fetch one. Open a pull request — gh pr create. Proposes the work on one (world, branch) pair to another. Defaults make the common cases one argument: from a fork, the target is the world it was forked from, so world.prOpen({ title = "..." }) proposes your work upstream. In an ordinary world the target is the same world, so you get a branch → main pull request. The PR lives in — and is numbered by — the world it targets, exactly as a forge numbers pull requests on the upstream repository. That is also where world.prList finds it.

Parameters

opts PrOpenOpts — title (required), plus description, sourceWorld, sourceBranch, targetWorld, targetBranch to address any leg explicitly.

Returns any — Decoded ZeroMind response. The decoded ZeroMind pull request.

local prs = world.prList()
world.prOpen({ title = "fix the door hinge" })
world.prOpen({ title = "port the fix", targetWorld = otherGuid })

globals/world/prView

world.prView(worldGuid: string?, number: number) -> any

Read one pull request in full — gh pr view. Returns the record plus a LIVE re-analysis against the current branch heads: mergeability (clean / conflicts / fast_forwardable / up_to_date / unrelated), conflict_count, and diff — every path the request adds, modifies or deletes with its checksums. Read this before merging: it is what tells you WHAT the request changes.

Parameters

worldGuid string (optional) — The world the pull request lives in (its target world). Defaults to the session world.
number number — The pull request number.

Returns any — The decoded pull request view.

world.prView(nil, 1)

globals/world/previewInstall

world.previewInstall(opts: InstallAssetOpts) -> PreviewResult

Preview what installing an asset WOULD write, without writing anything. Fetches + decodes the closure and plans placement (the same helpers world.installAsset uses), returning a flat node list plus rollup totals. A truncated closure is reported (not raised) so a caller can surface it and block import.

Parameters

opts InstallAssetOpts — { guid, at?, ref? } — same shape as installAsset.

Returns PreviewResult — nodes + totals + a truncated flag.

world.previewInstall({ guid = "..." })

globals/world/provideDerived

world.provideDerived(outputPath: string, opts: { stage: string? }?) -> number

Take the file at outputPath as the derived output: the file you wrote is staged as it stands, and the reservation is settled done with the fingerprint authored, which no recomputation replaces.

Parameters

outputPath string — The output's VFS path; the file must exist.
opts { stage: string? } (optional) — Optional { stage: string? } naming the staging area.

Returns number — How many reservations the file filled.

world.provideDerived("/zero/source/gold.material/preview.png")

globals/world/publishBlockers

world.publishBlockers(opts: StageOpts?) -> { PublishBlockerClass }

List every reason world.push would refuse to publish this world, as one entry per blocker class: script errors in user content (script), assets that don't meet their type's content requirements (content), computed asset references a component or scene entrypoint reaches (staticPin), and the conditions the publish gate refuses on: a component or scene entrypoint reaching a dynamic resolve over require/tools.use (gameplayPath), a reference pinned to an asset the world no longer holds (references), a reference of the world's own file recorded before the world pinned the library it names (pinPending, which the commit that takes the file clears by deriving its table at the pin), and an error-severity problem recorded on published content (problems). Each class carries the same title the refusal prints, one items entry per offending subject (an asset identity, or a <path>:<line> site), and the single remedy covering that class. This is the account world.push composes its refusal from, so it names the same blockers with no push attempted — and in full, where a refusal bounds how many of a class it prints. zm status prints this list. One more class blocks the next commit rather than the publish, and carries blocks = "commit": staleReferences, the files the caller's staged set holds or leaves out whose references do not resolve in it (a file whose references moved with a staged change and is left out, or a staged file stating a version the staged set leaves out). It is the class world.commit refuses with; world.push never reads it, since a push publishes what is committed. Empty list = the caller commits and publishes.

Parameters

opts StageOpts (optional) — Optional { stage: string? } naming a staging area, so the content class reports the assets that area answers for. Omitted, the read acts on the calling caller's own area.

Returns { PublishBlockerClass } — Array of PublishBlockerClass entries, empty when nothing blocks.

for _, c in ipairs(world.publishBlockers()) do
for _, i in ipairs(c.items) do print(c.kind, i.subject, i.detail) end
end

globals/world/pull

world.pull(branch: string?, opts: PullOpts?) -> PullResult

git pull: fetch origin/<branch> (default the session branch) and merge it into the session branch. When the session branch is strictly behind, its head fast-forwards to the fetched commit; when both sides moved, a two-parent merge commit lands with the same conflict/marker/resolve flow as world.merge (resolve the unmerged paths, then world.add + world.commit; abortable with world.mergeAbort); when the fetched commit is already in the session branch's history nothing changes. opts.ffOnly refuses anything but a fast-forward; opts.noFf always makes a merge commit. Local changes follow git's rule, as world.merge's do, unsaved scene edits included. A branch origin does not hold raises.

Parameters

branch string (optional) — The origin branch to pull. Defaults to the session branch.
opts PullOpts (optional) — { ffOnly?, noFf? }.

Returns PullResult — A PullResult table: status is up_to_date, fast_forward (commit the new head), clean (commit the merge commit) or conflicts (conflicts).

local r = world.pull()
local r = world.pull("feature")
local r = world.pull("main", { ffOnly = true })

globals/world/pullAsset

world.pullAsset(opts: PullAssetOpts?) -> PullAssetResult

Pull upstream changes into a previously-installed asset. Re-resolves the root's transitive closure at opts.ref (default latest) and lands only what differs from what the world holds (M.__updateRowPlan):

A file the world never edited (its bytes equal the version it installed) lands whole when its version, its reference table or a stale conflict flag differs from the closure's: bytes and table are written to the pulled row in one step, text and binary alike, and the write reaches source whatever the engine's mode.
A file the world edited is reconciled from (row.origin_checksum, row.checksum, entry.checksum) (M.__reconcileDecision): noop (upstream unmoved) leaves it alone, converged advances its origin pointer, and merge runs a three-way vcs.merge3 over (base, local, theirs): a clean result is written and the origin pointer advances, a conflicted result is written WITH markers and the row is flagged conflicted (base + theirs checksums recorded for world.resolvePullConflict). A binary file the world edited that upstream changed too is flagged conflicted with the structured base/theirs checksums (no marker write).
A folder moves to the closure's version and takes its table.
A row flagged conflicted that holds the bytes it installed carries no local content: it does not block the pull, and the pull clears it.

Closure drift: an entry the original install never landed is pulled fresh (added). A locally-pulled row nested under the root's own directory whose origin entry disappeared from the closure is removed when it holds the bytes it installed (pruned), or flagged conflicted when it carries edits.

Parameters

opts PullAssetOpts (optional) — See PullAssetOpts. opts.guid or opts.path is required; opts.ref pins the re-resolve to a specific upstream commit (defaults to latest finalized).

Returns PullAssetResult summarising what merged, conflicted, was pruned, and was newly added, the version of the root it landed (root_version), and the reference sites in the pulled content that resolved to nothing where it was published (problems).

world.pullAsset({ guid = "..." })

globals/world/push

world.push(commitId: string?, opts: PushOpts?) -> (string? | PushQueued, string)

Publish the current branch to ZeroMind. The no-argument form is a git merge --squash push: EVERY unpushed commit on the branch collapses into a SINGLE ZeroMind commit (latest content per path, parented on the branch's current remote HEAD). Because only the merged final state's bytes are uploaded, a superseded or lost intermediate-commit blob can never break the push — this is what makes a churn-heavy world publishable. The local commit history is preserved as the editing journal; on success every squashed commit shares the one remote commit id. The explicit commitId form still pushes that single commit verbatim via publish_commit (advanced / chain-replay use; its parent must already be on the remote).

Parameters

commitId string (optional) — Optional. A single commit to push verbatim. Omit for the squash push of the whole unpushed stack (the normal path). A push only moves the remote branch forward, as git's does: a remote head no local commit holds raises git's ! [rejected] <branch> -> <branch> (fetch first), and one the local head does not descend from raises (non-fast-forward); world.pull() first. opts.force is git push --force: a branch whose head is already published moves the remote branch there regardless (a Maintainer's call). A branch with commits not yet published is refused with force, nothing sent (pull first), and so is force with commitId: force moves the session branch.
opts PushOpts (optional) — Optional PushOpts: stage names a staging area, the way every other staging verb takes it, and force forces the move. The pre-push content gate refuses over the assets that area holds or committed; an asset another area answers for publishes with its shortfall recorded on it and is named to the caller with the area that holds it. Omitted, the call acts on the calling caller's own area, so the gate refuses over what that caller staged and committed and names the rest. What an area holds is read from the staged rows a status call reports and from the world's history, where every commit records the area it came through, so the answer is the same from every engine and every session bound to the world. An area named after a caller holds what it committed while that caller's session is open; an area a caller named holds it until another area commits inside the asset.

Returns (string? | PushQueued, string) — Two values: the ZeroMind-allocated commit id, and the verdict. "published" with the new commit id when this call published; "already-published" when ZeroMind already carries what this call would have published — the state a push asks for, so it returns rather than raising. That verdict carries the commit's existing ZeroMind id when the publish names one (the single-commit form), and a nil id when it names none (the whole-stack form, which reports a chain). A squash whose merged final state carries dep.unresolved problems takes the slow path inside the same call: the engine re-resolves each pending literal against its live asset index and submits the resolutions to the publish procedure, which completes the push. A reference literal that still resolves to nothing is published with the asset holding it and stays a problem recorded on that asset, while a dep pin the squash severed raises instead; either way the engine names each one with its path, line, reference and reason. A publish ZeroMind refuses raises naming the condition and the command that clears it — a branch that moved under this push names world.pull().

local zmId = world.push()
local zmId, verdict = world.push()
local queue, verdict = world.push() -- verdict "queued" while commits are sealed
world.push("01HABC...")

globals/world/reload

world.reload() -> { started: boolean?, guid: string?, branch: string?, error: string? }

Run the bound world again from the top, the way a fresh boot into it would: the world unloads, its content is taken again, its entrypoint and startup scene run. The load is asynchronous; the answer says whether it started. An engine with no world bound answers { error }.

Returns { started: boolean?, guid: string?, branch: string?, error: string? } — { started = true, guid, branch } or { error = string }.

world.reload()

globals/world/remoteBranches

world.remoteBranches() -> { { branch: string, commit_id: string, origin_head: string } }

Origin's branches as the last fetch of each recorded them — git branch -r. One row per branch of this world that world.fetch has fetched: branch is origin/<name>, commit_id the local commit that names origin's head, origin_head that head's ZeroMind id.

Returns { { branch: string, commit_id: string, origin_head: string } } — Array of { branch, commit_id, origin_head }, sorted by branch.

for _, r in ipairs(world.remoteBranches()) do print(r.branch) end

globals/world/renameBranch

world.renameBranch(oldName: string, newName: string, opts: { force: boolean? }?)

Rename a branch — git branch -m <old> <new>. The branch's head and history, and your uncommitted changes on it, go to newName; when oldName is the branch this session is on, the session moves with it. Another caller's uncommitted changes on oldName keep that branch in place beside the renamed one.

Parameters

oldName string — The branch to rename.
newName string — Its new name.
opts { force: boolean? } (optional)

world.renameBranch("feature", "feature-2")

globals/world/reset

world.reset(target: string, opts: ResetOpts?) -> ResetResult

Move the session branch's head to target — git reset [--soft|--mixed|--hard] <commit>. target is any revision world.revParse reads (a commit id, HEAD~1, a branch, ...). opts.mode: soft keeps the staged entries and the working tree and stages every path whose committed version the move changed; mixed (the default) empties your staging area and keeps the working tree, so a path the move changed reads as a local change; hard empties your staging area and brings the working tree to the new head: every path the move changed, every path your account changed, and every file of your account's that no commit holds (an uncommitted install and a scene's autosave overlay included), each to trash first. A write this session owes the world and the world is not taking is given up (world.dropOwedWrites). The local changes of another account stay and are listed in kept. A target equal to HEAD moves nothing, so world.reset("HEAD") unstages everything and world.reset("HEAD", { mode = "hard" }) discards your local changes. Non-destructive for history: every commit the head leaves behind becomes a trash entry that re-attaches it (in chain order).

Parameters

target string — The revision to move the head to.
opts ResetOpts (optional) — Optional ResetOpts — mode.

Returns ResetResult — A ResetResult — the new head, its subject line, and the commits left behind.

world.reset("HEAD~1")
world.reset("01HABC...", { mode = "hard" })

globals/world/resolveConflict

world.resolveConflict(path: string, mode: string, content: string?) -> ResolveResult

Resolve one /source record this session owes the world, one path at a time: a conflicted one, or one whose write or removal the server refused or the queue gave up on (unsaved_writes in world.syncStatus()), which waiting never carries and which holds a world.swap or world.checkout away from the world until it is settled here. mode is merge | apply | take-local | take-backend | discard. merge returns { merged, clean } and mutates nothing — a clean merge can be finalized with apply, and a conflicted one carries <<<<<<< / ======= / >>>>>>> markers to edit first. apply writes the finalized content to /source; take-local writes the retained local bytes; take-backend / discard keep the backend head. For a removal (deleted in world.syncStatus().conflicts), take-local removes the path again and merge errors, since a removal holds no content to merge. Errors when the record is absent, a merge has no common ancestor or hits binary content, or a write fails.

Parameters

path string — Canonical /source path of the conflicted record.
mode string — One of merge | apply | take-local | take-backend | discard.
content string (optional) — Finalized bytes for apply mode.

Returns ResolveResult

world.resolveConflict("/zero/source/foo.luau", "take-local")
local r = world.resolveConflict(p, "merge"); if r.clean then world.resolveConflict(p, "apply", r.merged) end

globals/world/resolvePullConflict

world.resolvePullConflict(path: string, choice: string)

Resolve a conflicted row. A TEXT conflict is one where the file currently contains conflict markers (world.pullAsset writes markers only for a text three-way merge that didn't resolve cleanly); a BINARY/composite conflict has no markers — the local bytes were left untouched.

choice = "theirs" fetches clean upstream text by content hash (conflict_theirs_blob_sha256 — path/rename-independent, since blobs are content-addressed) and overwrites the local file with it before clearing the flag. It errors, refusing to guess, when the row records no theirs blob (a binary/composite conflict — a text blob read can't address theirs for those; keep "ours" or re-install the asset instead).

choice = "ours" keeps the local side. When the file carries conflict markers, the local side is reconstructed from them (the marker writers put ours first, so dropping each block's base and theirs sections restores your bytes exactly) and written back; a marker-free file is kept as-is. Either way the flag clears.

Either way, staging the resolved file (world.add) is the normal git-add path once this returns — this call only clears the manifest-level flag and (for "theirs") the file content.

Parameters

path string — The conflicted row's local VFS path.
choice string — "ours" or "theirs".

world.resolvePullConflict("/source/combat/rules.luau", "theirs")

globals/world/restore

world.restore(handle: any?)

Restore one trash entry by row_id. Errors verbatim on handler-not-yet-implemented / world-mismatch.

Parameters

handle any (optional) — The trash row id. May be a number or a numeric string.

world.restore(42)

globals/world/retryDerived

world.retryDerived(outputPath: string, opts: { stage: string? }?) -> number

Compute the derived output at outputPath again, on the staging area or on a sealed commit that reserves it. A reservation that failed, was dropped or is done is reserved anew and computed from the inputs as they stand; a sealed commit waiting on it lands when it is done.

Parameters

outputPath string — The output's VFS path, as zm status names it.
opts { stage: string? } (optional) — Optional { stage: string? } naming the staging area.

Returns number — How many reservations were computed again.

world.retryDerived("/zero/source/gold.material/preview.png")

globals/world/revParse

world.revParse(rev: string) -> string

The commit a revision names — git rev-parse <rev>. A revision is HEAD, a branch, origin/<branch> (as the last fetch recorded it), a commit id or a unique prefix of one, each optionally followed by ~N (N first parents back), ^ (the first parent) or ^2 (a merge commit's second parent). Raises git's message for a revision that names nothing, HEAD on a branch with no commit included.

Parameters

rev string — The revision.

Returns string — The commit id.

local id = world.revParse("HEAD~1")

globals/world/reviewUpdate

world.reviewUpdate(unit: string, opts: ReviewUpdateOpts?) -> UpdateReview

Review what updating one unit changes, writing nothing: the update itself, never a preview of the asset. Answers the versions the unit moves from (what the world holds) and to (what its origin holds at opts.ref, default the ref world.checkUpdates compares it with), then commits: the commits of the origin world in between, newest first, each { id, message, author, date } (more is true past 200), then the files. modified names up front each file the world holds uncommitted changes to (a dirty or staged row) that the update touches, { path, outcome } with outcome merges or conflicts.

An installed asset (unit is its root path, as world.checkUpdates reports it; a path inside the install stands for its root): the commits that change the install's files, back to the one that moved the root from the version the world holds; files lists each file the update adds, changes or removes, { path, change, diff, hunks, local_edit }, where local_edit is none, or merges / conflicts for a file the world edited since it was installed.
A library (@builtin, @combat, with or without the @): the commits from the pinned commit to the new one and summary counts of the library files that change and of this world's files whose references move with the pin (they read references moved in status and are committed with the marker). opts.files = true adds files (each changed library file with its diff) and moving (those paths).

opts.commits = false reads no commits, and opts.files = false for an install leaves out files and its diffs: { commits = false, files = false } is the check before an update, answering modified and the counts alone. When a version of the install's root cannot be read the commit list stops there, more is true and commits_error says why.

Parameters

unit string — A library name or an install root path.
opts ReviewUpdateOpts (optional) — { ref?, files?, commits? }.

Returns UpdateReview: { kind, unit, from, to, commits, more, commits_error?, files?, moving?, modified, summary }; modified is one { path, outcome = "merges" | "conflicts" } per file holding uncommitted changes of its own that the update touches; summary counts added, changed, removed, modified, and merges / conflicts for an install or moving for a library.

local r = world.reviewUpdate("@builtin")
local r = world.reviewUpdate("/zero/source/Fire.module")
local r = world.reviewUpdate("/zero/source/Fire.module", { commits = false, files = false })

globals/world/show

world.show(...: string) -> any

Mirror git show's CLI arg shape. Default returns commit metadata + full diff vs parent. world.show("commit:/path") returns just the bytes. Flags: --stat, --name-only.

Parameters

... string — Variadic string args: commit id, optional path, optional flags.

Returns any — Either a ShowResult table or a string (for commit:/path).

local r = world.show("abc123")
local r = world.show("--stat", "abc123")
local bytes = world.show("abc123:/foo.luau")

globals/world/stageVersion

world.stageVersion(rev: string, path: string, opts: StageOpts?) -> boolean

Make the staging area's entry for path the version revision rev holds — git reset <rev> -- <path> (zm restore --staged --source <rev>). When that version is the one the branch head holds (both lacking the path included), the path is unstaged, as the index then matches HEAD there; otherwise a revision that lacks the path stages its deletion. The working tree is untouched.

Parameters

rev string — Any revision world.revParse reads.
path string — The path to stage.
opts StageOpts (optional) — Optional StageOpts — stage, the staging area.

Returns boolean — true when the staging area holds an entry for the path afterwards, false when the path ended unstaged.

world.stageVersion("HEAD~1", "/source/foo.luau")

globals/world/startup_scene

world.startup_scene -> any

Returns any

globals/world/stash

world.stash(label: string?) -> StashPushResult

Stash the caller's uncommitted changes — git stash push. The caller's dirty + staged state on the session branch is saved into a stash row (label names it), and each stashed path goes back to its committed version, unstaged. Binary and large content is saved by its stored blob, so applying the stash brings it back byte-identical. A path no commit holds keeps its working content in the tree, as git leaves an untracked file, and is listed under kept. A path that cannot be taken back raises, naming it; the stash is saved either way.

Parameters

label string (optional) — Optional. Free-form text label for the stash.

Returns StashPushResult — A StashPushResult.

local r = world.stash("wip widget refactor")

globals/world/stashApply

world.stashApply(ref: any?) -> StashApplyResult

Apply a stash and keep it — git stash apply [<stash>]. Author-only. ref is a row id, stash@{n}, or nil for the caller's newest stash. Text content is written back and staged entries are staged again; content recorded by checksum alone is listed under skipped. Refuses, writing nothing, when a path the stash restores holds a different local change.

Parameters

ref any (optional) — Optional stash reference.

Returns StashApplyResult — A StashApplyResult.

local r = world.stashApply()

globals/world/stashDrop

world.stashDrop(ref: any?) -> number

Drop a stash — git stash drop [<stash>]. Author-only. ref is a row id, stash@{n}, or nil for the caller's newest stash. The dropped stash goes to trash, where world.restore brings it back under a new row id.

Parameters

ref any (optional) — Optional stash reference.

Returns number — The row id of the dropped stash.

world.stashDrop()

globals/world/stashPop

world.stashPop(ref: any?) -> StashApplyResult

Apply a stash and drop it — git stash pop [<stash>]. Author-only. ref is a row id, stash@{n}, or nil for the caller's newest stash. Applies as world.stashApply does; a refused apply keeps the stash. The dropped stash goes to trash.

Parameters

ref any (optional) — Optional stash reference.

Returns StashApplyResult — A StashApplyResult.

local r = world.stashPop()

globals/world/stashes

world.stashes() -> { StashRow }

List every stash row in the world. Anyone with read access sees every stash; the per-row author_hex makes it clear which entries the caller can pop / drop themselves.

Returns { StashRow } — Array of StashRow tables.

local rows = world.stashes()

globals/world/status

world.status -> any

Returns any

globals/world/status_text

world.status_text -> any

Returns any

globals/world/syncStatus

world.syncStatus() -> SyncStatus

Read the durable-sync status: { subscribed, content_synced, content_sync_held, world_entry_held, outgoing_synced, uploads_outstanding, uploads_in_flight, progress, pending_writes, unrecorded_writes, owed_paths, unsaved_writes, held_uploads, uploads_abandoned, conflicts, binding }. Content sync runs in two directions and each has its own answer: content_synced is incoming, true once the world's content has been applied to this client, and it says nothing about this session's own changes. outgoing_synced is outgoing, true once every /source change this session made has reached the world: nothing queued, uploading, held for another attempt, unreflected by the manifest, kept in a durable record or a conflict, or refused. Read outgoing_synced to know whether your work is saved; uploads_outstanding counts the upload jobs it still waits on, and uploads_in_flight the ones among them a flush has taken and is carrying to the world. Both booleans read false with no world subscribed. content_sync_held is true while vfs.holdContentSync(true) holds the content sync, which keeps content_synced false until released. world_entry_held is true while world.holdEntry(true) holds the world entry: a world load whose content is ready waits before the world's entrypoint runs and its startup scene loads, with binding.step reading entryHeld, until it is released. owed_paths lists the /source paths this session wrote or removed that the server has not taken yet, the same list vfs.pendingWrites() returns. conflicts maps each conflicted /source path to { base_sha, local_sha, backend_sha, isBinary, deleted }; deleted is true when the author's side is a removal of the path. unsaved_writes lists the /source paths this session wrote or removed whose change the server never took, and uploads_abandoned counts the uploads the queue stopped carrying — read those two to tell a queue working through a backlog from one that gave content up, which pending_writes alone reads the same for. held_uploads lists the owed changes whose upload failed and is being tried again, as { path, reason }: a stage or commit of other content goes ahead past them, one naming such a path waits on it and reports the reason, and removing the path ends the hold. unrecorded_writes lists the owed paths whose durable record this machine could not store, which a restart loses; pending_writes counts the owed records it did store. binding names which world holds /source — bound, unbound, binding, session_only, or unclassified before the boot has decided — and, when a load attempt is on record, which attempt is running, the step it is held at and for how long, what the last one answered, and whether the load stopped. Read binding to tell a world that is still coming from one that was never asked for: subscribed answers false for both. Synchronous.

Returns SyncStatus

local s = world.syncStatus(); print(s.outgoing_synced, s.uploads_outstanding)
local s = world.syncStatus(); print(s.pending_writes)
local s = world.syncStatus(); for _, p in ipairs(s.unsaved_writes) do print(p) end
local s = world.syncStatus(); if s.binding.awaiting_world then print(s.binding.state, s.binding.attempt) end

globals/world/trash

world.trash() -> { TrashRow }

List trash entries for the world. Anyone with read access to the world can list trash — recovery is a shared safety net, not a privacy boundary. The handle (row_id) feeds back into world.restore.

Returns { TrashRow } — Array of TrashRow tables.

local rows = world.trash()

globals/world/trashEmpty

world.trashEmpty() -> number

Permanently delete the trash entries your own operations put there — zm trash empty — in one transaction. Entries other callers' operations trashed stay restorable. An emptied entry cannot be restored.

Returns number — The number of entries deleted.

local n = world.trashEmpty()

globals/world/uninstallLibrary

world.uninstallLibrary(name: string) -> string

Delete the library marker file. name accepts "@combat" or "combat" (the leading @ is the convention carried by the on-disk path).

Parameters

name string — The library name, with or without the leading @.

Returns string — The marker path that was removed.

world.uninstallLibrary("@combat")

globals/world/unseal

world.unseal(commitId: string?) -> string

Return a sealed commit to the staging area it was sealed from, with its message kept, so its paths can be changed and committed again. A commit world.commit sealed lands by itself once its required derived outputs are computed; this is the way out of one that cannot land.

Parameters

commitId string (optional) — The sealed commit's id. Omitted, the oldest sealed commit holding a failure, else the newest.

Returns string — The id of the commit returned.

world.unseal()

globals/world/unstage

world.unstage(path: string, opts: StageOpts?) -> UnstageReport

Remove a path from the staging area. Live manifest dirty state is untouched. A path is staged in one area at a time: a path the calling area does not hold that another staging area does is taken out of that area, whichever account opened it, and named in the report; a take from an area of another account warns, naming that account, the area and its session. An installed asset is unstaged by its root path, which takes every row of the install out of the area; another row of an install raises, naming the install and its root.

Parameters

path string — The path to unstage.
opts StageOpts (optional) — Optional { stage: string? } naming the staging area the path was staged into. Omitted, the call acts on the calling caller's own area.

Returns UnstageReport — { released }: each path taken out of another staging area (stage, account, same_account, session, live). Empty when the calling area held the path.

world.unstage("/source/foo.luau")
world.unstage("/source/foo.luau", { stage = "fauna" })

globals/world/update

world.update(units: (string | { string })?, opts: UpdateOpts?) -> UpdateResult

Apply updates. With no units, every unit world.checkUpdates lists; a name or a list of names (library names and install root paths, mixed) updates those. A library moves its pin (world.updateLibrary), an installed asset pulls its new closure (world.pullAsset), one unit at a time: a unit that fails is reported in its entry and the rest still run, and an install whose local edits conflict lists the conflicted paths. Nothing is staged, committed or pushed: the updated files wait in world.status(), to_commit lists every path the updates changed or moved, and warnings names each file holding uncommitted changes that an update merged into or conflicted.

Parameters

units (string | { string }) (optional) — Nil, a unit name, or a list of names.
opts UpdateOpts (optional) — { ref? }: a branch name or commit id every unit is updated to.

Returns UpdateResult — { results, to_commit, skipped, warnings, note }: results holds one { unit, kind, from, to, ok, error?, merged?, conflicted?, added?, pruned?, moving? } per unit, skipped the units the check could not read, warnings one { unit, path, outcome = "merged" | "conflicted" } per modified file an update touched, and note says nothing was committed.

local r = world.update()
local r = world.update("@combat")
local r = world.update({ "@builtin", "/zero/source/Fire.module" })

globals/world/updateLibrary

world.updateLibrary(name: string, ref: string?) -> UpdateLibraryResult

Move an installed library to the newest commit of a ref. Resolves ref (default main) in the world the library's marker at /source/libs/@<name> names; when that is the commit the marker already holds nothing is written, even when requested_ref differs; an imported library the engine does not hold at that commit is mounted and waited for, and a mount that fails raises naming why. Otherwise the marker is rewritten with the new commit (other marker fields are kept), the call waits until the engine holds the library at that commit (@builtin: its pin table is read; another library: its content is mounted), and the reference tables of the world's own sources that name the library are derived again. Each source whose table moved is listed by world.vcsStatus with the reason references moved and is committed with the marker: a commit holding the marker and leaving one of them out is refused, naming each. Discarding the marker puts the pin and its tables back. Works for @builtin and for every library world.installLibrary wrote. A failed wait puts the old marker back, waits for the library at the old commit and raises, saying where it ended. One move per library runs at a time, and none runs while play is running or in a runtime-profile session.

Parameters

name string — The library name, with or without the leading @.
ref string (optional) — A branch name or commit id of the library's world; defaults to main.

Returns UpdateLibraryResult — { name, from, to, changed, rederived }: the commits the pin moved between, whether it moved, and the source paths whose reference table was derived again, each one to commit with the marker.

local r = world.updateLibrary("@builtin")

globals/world/vcsStatus

world.vcsStatus(opts: StageOpts?) -> StatusResult

Return the working-tree VCS status: dirty paths, staged entries, ignored paths, untracked paths, and any unmerged ones. A file whose bytes are the ones the branch head holds and whose reference table is not (a library pin moved, or an asset it references was renamed, moved or got new bytes) is listed like any change, with reason = "references moved": in staged when the area holds it, in dirty otherwise. world.add / world.add_all stage it and it is committed with the change that moved it. A table that differs only in the install record the engine writes (an installed asset marked as the root of its install) is listed the same way with reason = "install recorded". An untracked path is one with no committed version behind it, and it appears in dirty as well — git add . picks up new files too. Named vcsStatus (not status) because world.status() is the runtime-snapshot accessor owned by world_status.module; the source-control surface keeps its own VCS-specific name so the two never shadow each other. Each dirty[i].dirtied_by is the identity the row was written under, which every session of one account shares; dirtied_by_caller is the stable caller key of the call that wrote it, dirtied_by_name the name that caller goes by on the world's roster now, and writer says what the row is to the calling caller: "you", "other" (a caller whose session is open), "ended" (a caller whose session has ended) or "unattributed". dirty_since_micros is the microsecond timestamp of the first write of the current dirty run. dirtied_by_account is the account the row's writer acts for and same_account whether that is the calling caller's account. local_identity is this connection's identity and local_caller the calling caller's own stable key, local_name the name it goes by now, local_account the account it acts for. claimed_by_other_stages names the paths some other staging area holds and who holds each: the area, its owner_account, whether that is the caller's (same_account), the session holding it and whether that session is open now (live). Staging is advisory: a staging call that stages one of them takes it out of that area. writer_accounts_refusal is set when the world's backend module predates the record of writers' accounts. stage names the area staged is the set of, so a caller reads which name reaches its own staged rows.

Parameters

opts StageOpts (optional) — Optional { stage: string? } naming the staging area whose staged set to report. Omitted, the read reports the calling caller's own area. The dirty, ignored and untracked sets are the world's working tree and read the same whichever area is named.

Returns StatusResult — A StatusResult table.

local s = world.vcsStatus()
local s = world.vcsStatus({ stage = "fauna" })

globals/worldToScreen

worldToScreen(x, y, z) -> { x, y, visible }

Project a world-space point to screen-space coordinates. Returns one table, not two numbers. visible reports whether the point is IN FRONT OF the camera, which is what makes x/y meaningful — a point in front but outside the viewport is still visible = true, so test the coordinates against getViewportSize() to decide whether it is on screen.

Parameters

x number — World X
y number — World Y
z number — World Z

Returns table — { x = number, y = number, visible = boolean } — x/y in pixels

globals/xpcall

xpcall(fn, handler, ...) -> ok, result

Protected call with custom error handler.

api
reference