@mengine/medeo-client 2.0.1-alpha.20 → 2.0.1-alpha.21

package/dist/index.d.ts

@@ -1,5 +1,5 @@
 import { t as index_d_exports } from "./index-CMa6ZCtX.js";
-import { $ as SpeechHostMap, A as assertValidVideoDocument, At as VideoDocumentValidationIssueCode, B as EntityRow, Bt as SpeedShift, C as isImplementedSemanticOpKind, Ct as Timeline, D as PartIdFactory, Dt as VideoClipPart, E as ValidationError, Et as TrackItemTimePosition, F as ResolvedEntityTrack, Ft as speedOf, G as SequenceDuration, H as CaptionSegmentSelection, Ht as Track$1, I as resolveEntityTimelineLayout, It as Attachment, J as JsonObject, K as SequenceRange, L as projectEntityTimeline, Lt as CaptionPart$1, M as EntityTimelineTrackRole, Mt as VideoDraftContent, N as ResolvedEntityClipPlacement, Nt as VideoDraftPartUnion, O as generatePartId, Ot as VideoDocument, P as ResolvedEntityTimelineLayout, Pt as effectiveVideoClipDurationMs, Q as DerivedItemPosition, R as EntityTimelineProjectionError, Rt as CaptionStyle, S as SemanticOpKind, St as SpeechPart, T as SnapshotReadable, Tt as TrackItem, U as CaptionStyleFields, Ut as TrackItem$1, V as RelationRow, Vt as Timeline$1, W as ScriptTextSegment, X as EntityId, Y as JsonValue, Z as buildInitialVideoDocument, _ as SemanticOpName, _t as CaptionDisplayCue, a as PlainMemoryAdapter, at as LaneKind, b as ImplementedSemanticOpKind, bt as PartKind, c as MirrorVideoDocumentAdapter, ct as findLaneTrack, d as createMirrorVideoDocumentAdapter, dt as TrackDraft, et as buildSpeechHostMap, f as CommitOptions, ft as TrackItemDraft, g as SemanticOpInput, gt as BgmPart, h as SemanticEditor, ht as videoDocumentMirrorSchema, i as JournalEntry, it as LANE_KINDS_IN_STACK_ORDER, j as validateVideoDocument, jt as VideoDraft, k as VideoDocumentValidationError, kt as VideoDocumentValidationIssue, l as MirrorVideoDocumentOptions, lt as laneTrackId, m as SemanticDocumentAdapter, mt as VideoDocumentMirrorSchema, n as videoDocumentSchema, nt as fromVideoDocument, o as PlainMemoryAdapterOptions, ot as SolvedVideoDocument, p as OpActor, pt as VideoDocumentDraft, q as VoiceDescriptor, r as readVideoDocumentFromDraft, rt as toVideoDocument, s as createPlainMemoryAdapter, st as ensureLaneTrack, t as partUnionSchema, tt as derivePositionFromAbs, u as createMirrorVideoDocument, ut as solveVideoDocument, v as TransactAudit, vt as CaptionPart, w as SchemaValidator, wt as Track, x as PlannedSemanticOpKind, xt as PartUnion, y as IMPLEMENTED_SEMANTIC_OP_KINDS, yt as DEFAULT_UNIT_TIME_MS, z as EntityRelationRows, zt as PartAggregation } from "./index-C0uUWqkz.js";
+import { $ as DerivedItemPosition, A as assertValidVideoDocument, At as VideoDocument, B as EntityRelationRows, Bt as CaptionStyle, C as isImplementedSemanticOpKind, Ct as SpeechOriginAssetSystem, D as PartIdFactory, Dt as TrackItem, E as ValidationError, Et as Track, F as ResolvedEntityTrack, Ft as VideoDraftPartUnion, G as ScriptTextSegment, Gt as TrackItem$1, H as RelationRow, Ht as SpeedShift, I as resolveEntityTimelineLayout, It as effectiveVideoClipDurationMs, J as VoiceDescriptor, K as SequenceDuration, L as EntityTimelineProjectionDocument, Lt as speedOf, M as EntityTimelineTrackRole, Mt as VideoDocumentValidationIssueCode, N as ResolvedEntityClipPlacement, Nt as VideoDraft, O as generatePartId, Ot as TrackItemTimePosition, P as ResolvedEntityTimelineLayout, Pt as VideoDraftContent, Q as buildInitialVideoDocument, R as projectEntityTimeline, Rt as Attachment, S as SemanticOpKind, St as PartUnion, T as SnapshotReadable, Tt as Timeline, U as CaptionSegmentSelection, Ut as Timeline$1, V as EntityRow, Vt as PartAggregation, W as CaptionStyleFields, Wt as Track$1, X as JsonValue, Y as JsonObject, Z as EntityId, _ as SemanticOpName, _t as BgmPart, a as PlainMemoryAdapter, at as LANE_KINDS_IN_STACK_ORDER, b as ImplementedSemanticOpKind, bt as DEFAULT_UNIT_TIME_MS, c as MirrorVideoDocumentAdapter, ct as ensureLaneTrack, d as createMirrorVideoDocumentAdapter, dt as solveVideoDocument, et as SpeechHostMap, f as CommitOptions, ft as TrackDraft, g as SemanticOpInput, gt as videoDocumentMirrorSchema, h as SemanticEditor, ht as VideoDocumentMirrorSchema, i as JournalEntry, it as toVideoDocument, j as validateVideoDocument, jt as VideoDocumentValidationIssue, k as VideoDocumentValidationError, kt as VideoClipPart, l as MirrorVideoDocumentOptions, lt as findLaneTrack, m as SemanticDocumentAdapter, mt as VideoDocumentDraft, n as videoDocumentSchema, nt as derivePositionFromAbs, o as PlainMemoryAdapterOptions, ot as LaneKind, p as OpActor, pt as TrackItemDraft, q as SequenceRange, r as readVideoDocumentFromDraft, rt as fromVideoDocument, s as createPlainMemoryAdapter, st as SolvedVideoDocument, t as partUnionSchema, tt as buildSpeechHostMap, u as createMirrorVideoDocument, ut as laneTrackId, v as TransactAudit, vt as CaptionDisplayCue, w as SchemaValidator, wt as SpeechPart, x as PlannedSemanticOpKind, xt as PartKind, y as IMPLEMENTED_SEMANTIC_OP_KINDS, yt as CaptionPart, z as EntityTimelineProjectionError, zt as CaptionPart$1 } from "./index-CzDNr1Tb.js";
 import { LoroDoc, PeerID } from "loro-crdt";
 import { DocState } from "@mengine/sync";
 import { BaseDocStorage, Connection, DocDiff, DocPushReceipt, DocSnapshotRecord, DocStorage, DocStorageOptions, DocUpdate, DocUpdateRecord } from "@mengine/storage";
@@ -262,6 +262,171 @@ declare class EntityGraphHttpClient {
   private request;
 }
 //#endregion
+//#region src/entity-crdt/field-changes.d.ts
+/** List members are addressed by identity, never a transient array offset. */
+type FieldPath = readonly (string | {
+  elementId: string;
+})[];
+type FieldChange = {
+  op: 'set';
+  path: FieldPath;
+  value: JsonValue;
+} | {
+  op: 'unset';
+  path: FieldPath;
+} | {
+  op: 'text.splice';
+  path: FieldPath;
+  index: number;
+  deleteCount: number;
+  text: string;
+} | {
+  op: 'list.insert';
+  path: FieldPath;
+  value: JsonValue;
+  beforeElementId: string | null;
+} | {
+  op: 'list.move';
+  path: FieldPath;
+  elementId: string;
+  beforeElementId: string | null;
+} | {
+  op: 'list.remove';
+  path: FieldPath;
+  elementId: string;
+};
+declare function assertFieldChanges(changes: readonly FieldChange[]): void;
+/** Resolve against the sequential causal view; the numeric path never crosses the API boundary. */
+declare function resolveFieldPath(payload: JsonObject, path: FieldPath): (string | number)[];
+/** Detached plan evaluation. Native compilation replays the original operations, not this snapshot. */
+declare function applyFieldChanges(payload: JsonObject, changes: readonly FieldChange[]): JsonObject;
+//#endregion
+//#region src/entity-crdt/entity-document.d.ts
+declare const LORO_ENTITY_SCHEMA = "medeo.entities.loro.v1";
+interface EntityDocumentState {
+  readonly rows: EntityRelationRows;
+  readonly project: Readonly<Record<string, string>>;
+}
+type Command = {
+  type: 'create';
+  row: EntityRow;
+} | {
+  type: 'selection-baseline';
+  root: string;
+  fingerprint: string;
+} | {
+  type: 'order';
+  group: string;
+  roots: string[];
+  token: string;
+} | {
+  type: 'write';
+  root: string;
+  payload: JsonObject;
+  token: string;
+} | {
+  type: 'text';
+  root: string;
+  path: readonly (string | number)[];
+  start: number;
+  deleteCount: number;
+  text: string;
+  token: string;
+} | {
+  type: 'field';
+  root: string;
+  path: (string | number)[];
+  change: FieldChange;
+  token: string;
+} | {
+  type: 'relation-field';
+  root: string;
+  path: (string | number)[];
+  change: FieldChange;
+} | {
+  type: 'delete';
+  root: string;
+} | {
+  type: 'relation';
+  row: RelationRow;
+} | {
+  type: 'relation-fields';
+  root: string;
+  metadata: JsonObject;
+  trace: JsonObject;
+} | {
+  type: 'unlink';
+  root: string;
+} | {
+  type: 'attach';
+  slot: string;
+  entityId: string;
+};
+/** Plain domain objects assembled from native Loro state, never a second writable authority. */
+declare function readEntityDocument(doc: LoroDoc, allowSelectionConflicts?: boolean): EntityDocumentState;
+/** Native editing containers survive domain version changes and merge independently. */
+declare class LoroEntityDocument {
+  readonly doc: LoroDoc;
+  private readonly validate?;
+  private readonly undoManager;
+  constructor(doc: LoroDoc, validate?: ((state: EntityDocumentState) => void) | undefined);
+  /** Open a causal snapshot using the same Loro runtime as the compiler. */
+  static fromSnapshot(snapshot: Uint8Array, validate?: (state: EntityDocumentState) => void): LoroEntityDocument;
+  static create(rows: EntityRelationRows, project?: Readonly<Record<string, string>>, peerId?: PeerID): LoroEntityDocument;
+  snapshot(): EntityDocumentState;
+  relationsOf(entityId: string): readonly RelationRow[];
+  /** Read unlinked Relation records for durable effect receipts; never part of the current semantic graph. */
+  tombstonedRelations(): readonly RelationRow[];
+  /** Resolve a queued edit target through its native identity without changing history. */
+  resolveCurrentEntityId(id: string): string;
+  getVersion(id: string): EntityRow | undefined;
+  /** The callback and every domain validation execute on a causally complete fork first. */
+  transact(edit: (draft: LoroEntityDraft) => void): Uint8Array;
+  /** Import preserves native operation identities; projection errors never discard remote operations. */
+  import(update: Uint8Array): void;
+  /**
+   * Bind a saved snapshot without rewriting text operations or resolving
+   * selection conflicts. The script names the bytes the same way everything
+   * else does — an Asset entity plus a Relation — and keeps only the
+   * fingerprint that tells a stale snapshot from a current one. The Relation is
+   * `to-asset`: these bytes are the script's own content written out, not a
+   * source it was made from.
+   */
+  bindAudioScriptAsset(entityId: string, assetId: string, expectedContentHash: string): Uint8Array;
+  canUndo(): boolean;
+  canRedo(): boolean;
+  undo(): boolean;
+  redo(): boolean;
+  private historyEdit;
+}
+declare class LoroEntityDraft {
+  private readonly doc;
+  readonly commands: Command[];
+  private readonly selectedRoots;
+  private readonly relationAliases;
+  constructor(doc: LoroDoc);
+  snapshot(): EntityDocumentState;
+  order(groupId: string, entityIds: readonly string[]): void;
+  /** Project the final scalar placement intents onto the existing native ordering containers. */
+  reconcileOrder(rows: EntityRelationRows): void;
+  create(row: EntityRow): void;
+  delete(entityId: string): void;
+  attach(slot: string, entityId: string): void;
+  link(row: RelationRow): void;
+  patchRelation(relationId: string, metadata: JsonObject, trace: JsonObject): void;
+  unlink(relationId: string): void;
+  private relationRoot;
+  update(entityId: string, fields: JsonObject): void;
+  declare(entityId: string, fields: JsonObject): void;
+  replaceOwned(entityId: string, payload: JsonObject): void;
+  editText(entityId: string, path: readonly (string | number)[], start: number, deleteCount: number, text: string): void;
+  /** Generic assembled-field edit compiled at the author's causal position. */
+  change(entityId: string, changes: readonly FieldChange[]): void;
+  changeRelation(relationId: string, changes: readonly FieldChange[]): void;
+  captureSelectionBaselines(): void;
+  private record;
+}
+//#endregion
 //#region src/entity-editor/types.d.ts
 /** Injectable domain-id mint shared by Entity and Relation authoring surfaces. */
 type EntityTimelineIdFactory = (prefix: 'entity' | 'relation') => string;
@@ -305,6 +470,71 @@ interface VoiceoverMediaAssetFact {
 }
 /** Facts resolved from media storage. A trim window never substitutes for intrinsic duration. */
 type MediaAssetFact = ImageMediaAssetFact | VideoMediaAssetFact | AudioMediaAssetFact | VoiceoverMediaAssetFact;
+/** Host-resolved ASR artifact. The SDK treats its storage fields as opaque facts. */
+interface CaptionAssetFact {
+  readonly assetId: string;
+  readonly storageKey?: string;
+}
+interface AsrCaptionSegment extends ScriptTextSegment {
+  readonly startMs: number;
+  readonly endMs: number;
+}
+/** One closed host command for materializing an ASR result into the Entity graph. */
+interface ApplyCaptionAssetsInput {
+  readonly effectId: string;
+  readonly timelineEntityId: string;
+  readonly trackEntityId: string;
+  readonly source: {
+    readonly entityId: string;
+    readonly entityKind: 'audio' | 'video';
+  };
+  /** The exact placed use whose source-relative ASR timing is being materialized. */
+  readonly sourceOccurrence: {
+    readonly clipEntityId: string;
+    readonly markerEntityId: string;
+  };
+  readonly captionAsset: CaptionAssetFact;
+  readonly segments: readonly AsrCaptionSegment[];
+}
+interface ApplyCaptionAssetsResult {
+  readonly audioScriptEntityId: string;
+  readonly audioScriptMarkerEntityId: string;
+  readonly captionAssetEntityId: string;
+  readonly captionEntityIds: readonly string[];
+  readonly captionClipEntityIds: readonly string[];
+}
+interface CaptionAssetsEffectEnvelope {
+  readonly version: 1;
+  readonly baselineFingerprint: string;
+  readonly effectId: string;
+  readonly signature: string;
+  readonly input: ApplyCaptionAssetsInput;
+  readonly decisions: {
+    readonly sourceOccurrence: {
+      readonly sourceStartMs: number;
+      readonly sourceEndMs: number;
+      readonly timelineStartMs: number;
+      readonly rate: number;
+    };
+    readonly targetRanges: readonly {
+      readonly segmentId: string;
+      readonly start: number;
+      readonly end: number;
+    }[];
+    readonly captionAsset: {
+      readonly mode: 'create' | 'reuse' | 'replace';
+      readonly priorAssetEntityId?: string;
+      readonly resultAssetEntityId: string;
+      readonly reboundRelationIds: readonly string[];
+    };
+  };
+}
+interface CompiledCaptionAssetsEffect {
+  readonly result: ApplyCaptionAssetsResult;
+  readonly replayed: boolean;
+  readonly update?: Uint8Array;
+  readonly rows: EntityRelationRows;
+}
 type ClipPlacement = {
   readonly kind: 'sequential';
   readonly order: number;
@@ -437,6 +667,157 @@ interface SetClipSpeedInput {
   readonly clipEntityId: string;
   readonly timeRemapping: LinearClipSpeed | null;
 }
+type VideoClipSpeedMode = '0.5x' | '0.75x' | '1.0x' | '1.5x' | '2.0x';
+interface SetVideoClipSpeedEffectInput {
+  readonly clips: readonly {
+    readonly videoClipEntityId: string;
+    readonly mode: VideoClipSpeedMode;
+  }[];
+}
+interface GeneratedVisualClipFact {
+  readonly media: VisualMediaAssetFact;
+  /** Finite timeline use duration. Required for still images and bounded for video. */
+  readonly durationMs: number;
+  readonly generationJobId: string;
+  readonly sourceMediaEntityIds: readonly string[];
+  readonly volume?: number;
+}
+interface GeneratedVisualsInputBase {
+  readonly effectId: string;
+  readonly timelineEntityId: string;
+  readonly clips: readonly GeneratedVisualClipFact[];
+}
+/** An omitted occurrence anchor means the start of the main visual Track. */
+type AddGeneratedVisualsInput = GeneratedVisualsInputBase & ({
+  readonly beforeClipEntityId: string;
+  readonly afterClipEntityId?: never;
+} | {
+  readonly beforeClipEntityId?: never;
+  readonly afterClipEntityId: string;
+} | {
+  readonly beforeClipEntityId?: never;
+  readonly afterClipEntityId?: never;
+});
+interface ContinueVideoInput {
+  readonly effectId: string;
+  readonly timelineEntityId: string;
+  readonly afterClipEntityId: string;
+  readonly tailFrameEntityId: string;
+  readonly sourceMediaEntityId: string;
+  readonly media: VideoMediaAssetFact;
+}
+interface GeneratedVisualEffectEnvelope {
+  readonly version: 1;
+  readonly baselineFingerprint: string;
+  readonly effectId: string;
+  readonly signature: string;
+  readonly input: AddGeneratedVisualsInput;
+  readonly decisions: {
+    readonly trackEntityId: string;
+    readonly anchorOrder: number | null;
+    readonly anchorPosition: 'before' | 'after' | 'track_start';
+    readonly insertionOrders: readonly number[];
+  };
+}
+interface GeneratedVisualEffectResult {
+  readonly clipEntityIds: readonly string[];
+  readonly mediaEntityIds: readonly string[];
+}
+interface CompiledGeneratedVisualEffect {
+  readonly result: GeneratedVisualEffectResult;
+  readonly replayed: boolean;
+  readonly update?: Uint8Array;
+  readonly rows: EntityRelationRows;
+}
+interface GeneratedSpeechCaptionFact {
+  readonly startMs: number;
+  readonly endMs: number;
+  readonly selection: CaptionSegmentSelection;
+}
+/** Host-resolved current Loro segment; arbitrary immutable JSON fields remain owned by the script. */
+type ResolvedSpeechScriptSegment = ScriptTextSegment & JsonObject;
+type ResolvedAsrSpeechSegment = AsrCaptionSegment & JsonObject;
+type GeneratedSpeechSourceFact = {
+  readonly kind: 'phonetic-script';
+  readonly audioScriptEntityId: string;
+  readonly phoneticScriptEntityId: string;
+  readonly segments: readonly ResolvedSpeechScriptSegment[];
+  readonly phonemeScript?: string;
+} | {
+  readonly kind: 'audio-script';
+  readonly audioScriptEntityId?: string;
+  readonly segments: readonly ResolvedAsrSpeechSegment[];
+  readonly sourceMediaEntityIds: readonly string[];
+  readonly sourceStartMs: number;
+};
+interface GeneratedSpeechFactBase {
+  readonly source: GeneratedSpeechSourceFact;
+  readonly captions: readonly GeneratedSpeechCaptionFact[];
+  readonly volume?: number;
+}
+type GeneratedSpeechFact = GeneratedSpeechFactBase & ({
+  readonly renderMode: 'recorded';
+  readonly media: AudioMediaAssetFact;
+  readonly source: Extract<GeneratedSpeechSourceFact, {
+    kind: 'audio-script';
+  }>;
+} | {
+  readonly renderMode: 'synthesized';
+  readonly media: Omit<VoiceoverMediaAssetFact, 'voice'> & {
+    readonly voice: VoiceDescriptor;
+  };
+});
+interface AddSpeechAssetsInput {
+  readonly effectId: string;
+  readonly timelineEntityId: string;
+  readonly items: readonly (GeneratedSpeechFact & {
+    readonly placementStartMs: number;
+  })[];
+}
+interface ReplaceVoiceSpeechAssetsInput {
+  readonly effectId: string;
+  readonly timelineEntityId: string;
+  readonly items: readonly (Extract<GeneratedSpeechFact, {
+    renderMode: 'synthesized';
+  }> & {
+    readonly targetSpeechClipEntityId: string; /** Explicit current Caption uses replaced with this selected speech use. */
+    readonly targetCaptionClipEntityIds: readonly string[];
+  })[];
+}
+type SpeechAssetsEffectInput = {
+  readonly mode: 'add';
+  readonly value: AddSpeechAssetsInput;
+} | {
+  readonly mode: 'replace-voice';
+  readonly value: ReplaceVoiceSpeechAssetsInput;
+};
+interface SpeechAssetsEffectEnvelope {
+  readonly version: 1;
+  readonly baselineFingerprint: string;
+  readonly effectId: string;
+  readonly signature: string;
+  readonly input: SpeechAssetsEffectInput;
+  readonly decisions: {
+    readonly speechTrackEntityId: string;
+    readonly captionTrackEntityId?: string;
+    readonly targetStartsMs: readonly number[]; /** Existing raw-Audio occurrences reused by recorded items; positional with input items. */
+    readonly recordedSpeechClipEntityIds: readonly (string | null)[];
+  };
+}
+interface SpeechAssetsEffectResult {
+  readonly speechClipEntityIds: readonly string[];
+  readonly speechAudioEntityIds: readonly string[];
+  readonly audioScriptEntityIds: readonly string[];
+  /** Positional per input item; recorded audio has no pronunciation variant. */
+  readonly phoneticScriptEntityIds: readonly (string | null)[];
+  readonly captionClipEntityIds: readonly (readonly string[])[];
+}
+interface CompiledSpeechAssetsEffect {
+  readonly result: SpeechAssetsEffectResult;
+  readonly replayed: boolean;
+  readonly update?: Uint8Array;
+  readonly rows: EntityRelationRows;
+}
 interface TrimClipInput {
   readonly clipEntityId: string;
   readonly sourceRange: SequenceRange<number>;
@@ -542,6 +923,7 @@ type ClipEntityId = EntityId;
  * and committed only after validation succeeds.
  */
 declare class EntityTimelineEditor {
+  #private;
   private readonly idFactory;
   private state;
   constructor(initialRows: EntityRelationRows, idFactory?: EntityTimelineIdFactory);
@@ -559,6 +941,9 @@ declare class EntityTimelineEditor {
   replaceMediaClip(input: ReplaceMediaClipInput): void;
   setClipVolume(input: SetClipVolumeInput): void;
   setClipSpeed(input: SetClipSpeedInput): void;
+  /** Apply one typed Director speed effect as a single graph transaction. */
+  setVideoClipSpeeds(input: SetVideoClipSpeedEffectInput): void;
+  private setClipSpeedIntoDraft;
   trimClip(input: TrimClipInput): void;
   replaceSequentialClips(input: ReplaceSequentialClipsInput): readonly ClipEntityId[];
   deleteClip(input: DeleteClipInput): void;
@@ -579,6 +964,9 @@ declare class EntityTimelineEditor {
    * never the Caption entity or the AudioScript.
    */
   insertCaptionClip(input: InsertCaptionClipInput): ClipEntityId;
+  static compileCaptionAssetsEffect(base: LoroEntityDocument, envelope: CaptionAssetsEffectEnvelope): CompiledCaptionAssetsEffect;
+  static compileGeneratedVisualEffect(base: LoroEntityDocument, envelope: GeneratedVisualEffectEnvelope): CompiledGeneratedVisualEffect;
+  static compileSpeechAssetsEffect(base: LoroEntityDocument, envelope: SpeechAssetsEffectEnvelope): CompiledSpeechAssetsEffect;
   private insertPlacedClipIntoDraft;
   private ensureRoleTrack;
   /** The existing pronunciation variant a voiceover consumes; composition is required. */
@@ -607,6 +995,21 @@ declare class EntityTimelineEditor {
   private relinkRelation;
   private mintRelationId;
 }
+/** Freeze every state-dependent choice once in the sole FE tool-effect handler. */
+declare function prepareCaptionAssetsEffect(base: LoroEntityDocument, rawInput: ApplyCaptionAssetsInput): CaptionAssetsEffectEnvelope;
+/**
+ * Compile only a frozen envelope on its exact causal baseline. Collaborators
+ * receive the resulting Loro update, never the raw effect.
+ */
+declare function compileCaptionAssetsEffect(base: LoroEntityDocument, envelope: CaptionAssetsEffectEnvelope): CompiledCaptionAssetsEffect;
+/** Freeze the insertion position and lineage on one causal baseline. */
+declare function prepareGeneratedVisualEffect(base: LoroEntityDocument, rawInput: AddGeneratedVisualsInput): GeneratedVisualEffectEnvelope;
+declare function prepareContinueVideoEffect(base: LoroEntityDocument, input: ContinueVideoInput): GeneratedVisualEffectEnvelope;
+/** Compile once at the FE authority; retain the returned update bytes for retry and collaborators. */
+declare function compileGeneratedVisualEffect(base: LoroEntityDocument, envelope: GeneratedVisualEffectEnvelope): CompiledGeneratedVisualEffect;
+declare function prepareSpeechAssetsEffect(base: LoroEntityDocument, rawInput: SpeechAssetsEffectInput): SpeechAssetsEffectEnvelope;
+/** Compile once at the FE authority; retain the returned update bytes for retry and collaborators. */
+declare function compileSpeechAssetsEffect(base: LoroEntityDocument, envelope: SpeechAssetsEffectEnvelope): CompiledSpeechAssetsEffect;
 //#endregion
 //#region src/entity-editor/editor-resources.d.ts
 declare const roleOrder: {
@@ -659,6 +1062,12 @@ interface ImportedMediaAsset {
   /** The timbre identity of a synthesized voiceover; absent for recorded media. */
   readonly voiceEntityId?: string;
 }
+interface ImportedCaptionAsset {
+  readonly rows: EntityRelationRows;
+  readonly assetEntityId: string;
+}
+/** Import only the fixed ASR artifact; Caption entities are authored separately. */
+declare function importCaptionAsset(initial: EntityRelationRows, fact: CaptionAssetFact, idFactory?: EntityTimelineIdFactory): ImportedCaptionAsset;
 /**
  * Import at the editor boundary, not at asset generation time. Existing
  * provenance is retained.
@@ -701,169 +1110,6 @@ interface DocumentAudioScriptState {
 /** Read only the attached version and preserve every segment; never aggregate other scripts. */
 declare function readDocumentAudioScript(state: DocumentAudioScriptState): DocumentAudioScript;
 //#endregion
-//#region src/entity-crdt/field-changes.d.ts
-/** List members are addressed by identity, never a transient array offset. */
-type FieldPath = readonly (string | {
-  elementId: string;
-})[];
-type FieldChange = {
-  op: 'set';
-  path: FieldPath;
-  value: JsonValue;
-} | {
-  op: 'unset';
-  path: FieldPath;
-} | {
-  op: 'text.splice';
-  path: FieldPath;
-  index: number;
-  deleteCount: number;
-  text: string;
-} | {
-  op: 'list.insert';
-  path: FieldPath;
-  value: JsonValue;
-  beforeElementId: string | null;
-} | {
-  op: 'list.move';
-  path: FieldPath;
-  elementId: string;
-  beforeElementId: string | null;
-} | {
-  op: 'list.remove';
-  path: FieldPath;
-  elementId: string;
-};
-declare function assertFieldChanges(changes: readonly FieldChange[]): void;
-/** Resolve against the sequential causal view; the numeric path never crosses the API boundary. */
-declare function resolveFieldPath(payload: JsonObject, path: FieldPath): (string | number)[];
-/** Detached plan evaluation. Native compilation replays the original operations, not this snapshot. */
-declare function applyFieldChanges(payload: JsonObject, changes: readonly FieldChange[]): JsonObject;
-//#endregion
-//#region src/entity-crdt/entity-document.d.ts
-declare const LORO_ENTITY_SCHEMA = "medeo.entities.loro.v1";
-interface EntityDocumentState {
-  readonly rows: EntityRelationRows;
-  readonly project: Readonly<Record<string, string>>;
-}
-type Command = {
-  type: 'create';
-  row: EntityRow;
-} | {
-  type: 'selection-baseline';
-  root: string;
-  fingerprint: string;
-} | {
-  type: 'order';
-  group: string;
-  roots: string[];
-  token: string;
-} | {
-  type: 'write';
-  root: string;
-  payload: JsonObject;
-  token: string;
-} | {
-  type: 'text';
-  root: string;
-  path: readonly (string | number)[];
-  start: number;
-  deleteCount: number;
-  text: string;
-  token: string;
-} | {
-  type: 'field';
-  root: string;
-  path: (string | number)[];
-  change: FieldChange;
-  token: string;
-} | {
-  type: 'relation-field';
-  root: string;
-  path: (string | number)[];
-  change: FieldChange;
-} | {
-  type: 'delete';
-  root: string;
-} | {
-  type: 'relation';
-  row: RelationRow;
-} | {
-  type: 'relation-fields';
-  root: string;
-  metadata: JsonObject;
-  trace: JsonObject;
-} | {
-  type: 'unlink';
-  root: string;
-} | {
-  type: 'attach';
-  slot: string;
-  entityId: string;
-};
-/** Plain domain objects assembled from native Loro state, never a second writable authority. */
-declare function readEntityDocument(doc: LoroDoc, allowSelectionConflicts?: boolean): EntityDocumentState;
-/** Native editing containers survive domain version changes and merge independently. */
-declare class LoroEntityDocument {
-  readonly doc: LoroDoc;
-  private readonly validate?;
-  private readonly undoManager;
-  constructor(doc: LoroDoc, validate?: ((state: EntityDocumentState) => void) | undefined);
-  /** Open a causal snapshot using the same Loro runtime as the compiler. */
-  static fromSnapshot(snapshot: Uint8Array, validate?: (state: EntityDocumentState) => void): LoroEntityDocument;
-  static create(rows: EntityRelationRows, project?: Readonly<Record<string, string>>, peerId?: PeerID): LoroEntityDocument;
-  snapshot(): EntityDocumentState;
-  relationsOf(entityId: string): readonly RelationRow[];
-  /** Resolve a queued edit target through its native identity without changing history. */
-  resolveCurrentEntityId(id: string): string;
-  getVersion(id: string): EntityRow | undefined;
-  /** The callback and every domain validation execute on a causally complete fork first. */
-  transact(edit: (draft: LoroEntityDraft) => void): Uint8Array;
-  /** Import preserves native operation identities; projection errors never discard remote operations. */
-  import(update: Uint8Array): void;
-  /**
-   * Bind a saved snapshot without rewriting text operations or resolving
-   * selection conflicts. The script names the bytes the same way everything
-   * else does — an Asset entity plus a Relation — and keeps only the
-   * fingerprint that tells a stale snapshot from a current one. The Relation is
-   * `to-asset`: these bytes are the script's own content written out, not a
-   * source it was made from.
-   */
-  bindAudioScriptAsset(entityId: string, assetId: string, expectedContentHash: string): Uint8Array;
-  canUndo(): boolean;
-  canRedo(): boolean;
-  undo(): boolean;
-  redo(): boolean;
-  private historyEdit;
-}
-declare class LoroEntityDraft {
-  private readonly doc;
-  readonly commands: Command[];
-  private readonly selectedRoots;
-  private readonly relationAliases;
-  constructor(doc: LoroDoc);
-  snapshot(): EntityDocumentState;
-  order(groupId: string, entityIds: readonly string[]): void;
-  /** Project the final scalar placement intents onto the existing native ordering containers. */
-  reconcileOrder(rows: EntityRelationRows): void;
-  create(row: EntityRow): void;
-  delete(entityId: string): void;
-  attach(slot: string, entityId: string): void;
-  link(row: RelationRow): void;
-  patchRelation(relationId: string, metadata: JsonObject, trace: JsonObject): void;
-  unlink(relationId: string): void;
-  private relationRoot;
-  update(entityId: string, fields: JsonObject): void;
-  declare(entityId: string, fields: JsonObject): void;
-  replaceOwned(entityId: string, payload: JsonObject): void;
-  editText(entityId: string, path: readonly (string | number)[], start: number, deleteCount: number, text: string): void;
-  /** Generic assembled-field edit compiled at the author's causal position. */
-  change(entityId: string, changes: readonly FieldChange[]): void;
-  changeRelation(relationId: string, changes: readonly FieldChange[]): void;
-  captureSelectionBaselines(): void;
-  private record;
-}
-//#endregion
 //#region src/entity-crdt/compile-rows.d.ts
 /** Lower a detached ordinary-object edit against its actual causal Loro baseline. */
 declare function compileEntityRows(snapshot: Uint8Array, rows: EntityRelationRows, validate?: (state: EntityDocumentState) => void): {
@@ -1772,4 +2018,4 @@ declare function phoneticScriptAssetHash(rows: EntityRelationRows, entityId: str
 /** The Asset a PhoneticScript was written out to, read from the graph. */
 declare function phoneticScriptAssetOf(rows: EntityRelationRows, entityId: string): EntityRow | undefined;
 //#endregion
-export { type Aggregation, type AssembledCaptionContent, type AssembledPhoneticScriptContent, type Attachment, type AudioMediaAssetFact, type BgmPart, type CaptionContentInput, type CaptionDisplayCue, type CaptionPart, type CaptionSegmentSelection, type CaptionStyle, type ClipEntityId, type ClipPlacement, type CommitOptions, DEFAULT_UNIT_TIME_MS, type DeleteBgmInput, type DeleteClipInput, type DeleteClipTreeInput, type DeleteVoiceoverInput, type DerivedItemPosition, type DocStorageLike, type DocVersionMark, type DocumentAudioScript, type DocumentAudioScriptState, type EditorFoundation, EntityDocumentState, type EntityGraphCommitOptions, EntityGraphHttpClient, type EntityGraphState, EntityTimelineEditor, type EntityTimelineIdFactory, EntityTimelineProjectionError, type EntityTimelineTrackRole, type FieldChange, type FieldPath, IMPLEMENTED_SEMANTIC_OP_KINDS, type ImageMediaAssetFact, type ImplementedSemanticOpKind, type ImportedMediaAsset, type InsertCaptionClipInput, type InsertClipInput, type InsertMediaClipInput, type InsertMediaClipsInput, type InsertPlacedClipInput, type JournalEntry, LANE_KINDS_IN_STACK_ORDER, LORO_ENTITY_SCHEMA, type LaneKind, type LinearClipSpeed, LoroEntityDocument, LoroEntityDraft, type MainClipRange, type MakeEmptyPart, ManualSyncDoc, type ManualSyncDocOptions, MedeoHttpDocStorage, type MedeoHttpDocStorageOptions, type MediaAssetFact, type MediaClipInsertion, MemoryDocStorage, MengineAckFailedError, type MengineAuditEntry, type MengineAuditResponse, MengineDocSession, type MengineDocSessionOptions, type MengineDocSessionUpdateEvent, type MengineDocSyncState, type MengineDocumentVersion, type MengineEventStreamOptions, MengineHttpClient, type MengineHttpClientOptions, MengineHttpRequestError, MenginePushRejectedError, type MenginePushResponse, type MenginePushUpdateResponse, type MengineRejectedResponse, type MengineSnapshotResponse, type MengineSseUpdateEvent, type MengineSyncResponse, type MengineUndoState, type MengineUpdateMeta, MirrorVideoDocumentAdapter, type MirrorVideoDocumentOptions, type MoveClipInput, type MoveClipsToStartsInput, type MoveSequentialClipsInput, type MoveVoiceoverInput, type OpActor, type PartAggregation, type PartIdFactory, type PartKind, type PartUnion, type PatchCaptionStyleInput, PlainMemoryAdapter, type PlainMemoryAdapterOptions, type PlannedSemanticOpKind, type PullFailureReason, type PullResult, type PushOutcome, type PushOutcomeKind, type PushResult, type PushResultKind, type ReplaceClipContentInput, type ReplaceMediaClipInput, type ReplaceSequentialClipsInput, type ReplacementMediaClipInput, type ResolvedEntityClipPlacement, type ResolvedEntityTimelineLayout, type ResolvedEntityTrack, SchemaValidator, ScriptCompositionError, type SemanticDocumentAdapter, SemanticEditor, type SemanticOpInput, type SemanticOpKind, type SemanticOpName, type SequentialClipAnchor, type SetBgmInput, type SetCaptionVisibilityInput, type SetClipPlacementInput, type SetClipSpeedInput, type SetClipVolumeInput, type SnapshotReadable, type SolvedVideoDocument, type SpeechHostMap, type SpeechPart, type SpeedShift, TIMELINE_SKELETON_DURATION_MS, type Timeline, type TimelineDoc, type TimelineItem, type Track, type TrackDraft, type TrackItem, type TrackItemDraft, type TrackItemTimePosition, type TransactAudit, type TrimClipInput, type UpdateClipInput, type UpdateClipMarkerInput, ValidationError, type VideoClipPart, type VideoDocument, type VideoDocumentDraft, type VideoDocumentMirrorSchema, VideoDocumentValidationError, type VideoDocumentValidationIssue, type VideoDocumentValidationIssueCode, type VideoDraft, type CaptionPart$1 as VideoDraftCaptionPart, type VideoDraftContent, type VideoDraftPartUnion, type Timeline$1 as VideoDraftTimeline, type Track$1 as VideoDraftTrack, type TrackItem$1 as VideoDraftTrackItem, type VideoMediaAssetFact, type VisualMediaAssetFact, type VoiceoverCaptionFact, type VoiceoverMediaAssetFact, type VoiceoverTakeInput, type VoiceoverTakeResult, type WaitForServerAckOptions, applyEntityRows, applyFieldChanges, arrangeMainTrackSeamlessly, assembleCaptionContent, assembleEntityContent, assemblePhoneticScriptContent, assembleScriptText, assertCanonicalEditorResources, assertEntityProjectionPreservesLegacy, assertFieldChanges, assertMediaAssetWritePolicy, assertValidVideoDocument, audioScriptAssetContent, audioScriptAssetFields, audioScriptAssetHash, audioScriptAssetOf, base64ToBytes, buildInitialVideoDocument, buildSpeechHostMap, bytesToBase64, cascadeAfterVideoClipChanges, compileEntityRows, createEditSandbox, createMirrorVideoDocument, createMirrorVideoDocumentAdapter, createPlainMemoryAdapter, decodeDocVersionMark, derivePositionFromAbs, effectiveVideoClipDurationMs, encodeDocVersionMark, ensureEditorFoundation, ensureLaneTrack, fillMainTrackTimeGaps, findComposedAudioScript, findLaneTrack, fromVideoDocument, generatePartId, getAt, hostForAbsMs, importMediaAsset, isEmptyVideoClip, isImplementedSemanticOpKind, isMap, laneTrackId, mainTrackRanges, migrateLegacyTimelineToEntities, partDurationMs, partUnionSchema, phoneticScriptAssetContent, phoneticScriptAssetHash, phoneticScriptAssetOf, prepareCaptionContent, projectEntityTimeline, readDocumentAudioScript, readEntityDocument, readMainTrackItems, readMengineEventStream, readPart, readPartDurationMs, readVideoDocumentFromDraft, reassignSpeechesToVideoClipsByTime, recalculateTimelineDuration, relativePositionForAbs, replayJournal, resolveAllSpeechOverlaps, resolveEntityTimelineLayout, resolveFieldPath, resolveSpeechOverlapByShiftingVideos, safeDurationMs, index_d_exports as schemas, scriptSourceAssetOf, selectAudioScriptSegment, snapshotToPlain, solveVideoDocument, speedOf, syncAggregatedClipsTimePosition, toVideoDocument, validateVideoDocument, videoDocumentMirrorSchema, videoDocumentSchema };
+export { type AddGeneratedVisualsInput, type AddSpeechAssetsInput, type Aggregation, type ApplyCaptionAssetsInput, type ApplyCaptionAssetsResult, type AsrCaptionSegment, type AssembledCaptionContent, type AssembledPhoneticScriptContent, type Attachment, type AudioMediaAssetFact, type BgmPart, type CaptionAssetFact, type CaptionAssetsEffectEnvelope, type CaptionContentInput, type CaptionDisplayCue, type CaptionPart, type CaptionSegmentSelection, type CaptionStyle, type ClipEntityId, type ClipPlacement, type CommitOptions, type CompiledCaptionAssetsEffect, type CompiledGeneratedVisualEffect, type CompiledSpeechAssetsEffect, type ContinueVideoInput, DEFAULT_UNIT_TIME_MS, type DeleteBgmInput, type DeleteClipInput, type DeleteClipTreeInput, type DeleteVoiceoverInput, type DerivedItemPosition, type DocStorageLike, type DocVersionMark, type DocumentAudioScript, type DocumentAudioScriptState, type EditorFoundation, EntityDocumentState, type EntityGraphCommitOptions, EntityGraphHttpClient, type EntityGraphState, EntityTimelineEditor, type EntityTimelineIdFactory, type EntityTimelineProjectionDocument, EntityTimelineProjectionError, type EntityTimelineTrackRole, type FieldChange, type FieldPath, type GeneratedSpeechCaptionFact, type GeneratedSpeechFact, type GeneratedSpeechSourceFact, type GeneratedVisualClipFact, type GeneratedVisualEffectEnvelope, type GeneratedVisualEffectResult, IMPLEMENTED_SEMANTIC_OP_KINDS, type ImageMediaAssetFact, type ImplementedSemanticOpKind, type ImportedCaptionAsset, type ImportedMediaAsset, type InsertCaptionClipInput, type InsertClipInput, type InsertMediaClipInput, type InsertMediaClipsInput, type InsertPlacedClipInput, type JournalEntry, LANE_KINDS_IN_STACK_ORDER, LORO_ENTITY_SCHEMA, type LaneKind, type LinearClipSpeed, LoroEntityDocument, LoroEntityDraft, type MainClipRange, type MakeEmptyPart, ManualSyncDoc, type ManualSyncDocOptions, MedeoHttpDocStorage, type MedeoHttpDocStorageOptions, type MediaAssetFact, type MediaClipInsertion, MemoryDocStorage, MengineAckFailedError, type MengineAuditEntry, type MengineAuditResponse, MengineDocSession, type MengineDocSessionOptions, type MengineDocSessionUpdateEvent, type MengineDocSyncState, type MengineDocumentVersion, type MengineEventStreamOptions, MengineHttpClient, type MengineHttpClientOptions, MengineHttpRequestError, MenginePushRejectedError, type MenginePushResponse, type MenginePushUpdateResponse, type MengineRejectedResponse, type MengineSnapshotResponse, type MengineSseUpdateEvent, type MengineSyncResponse, type MengineUndoState, type MengineUpdateMeta, MirrorVideoDocumentAdapter, type MirrorVideoDocumentOptions, type MoveClipInput, type MoveClipsToStartsInput, type MoveSequentialClipsInput, type MoveVoiceoverInput, type OpActor, type PartAggregation, type PartIdFactory, type PartKind, type PartUnion, type PatchCaptionStyleInput, PlainMemoryAdapter, type PlainMemoryAdapterOptions, type PlannedSemanticOpKind, type PullFailureReason, type PullResult, type PushOutcome, type PushOutcomeKind, type PushResult, type PushResultKind, type ReplaceClipContentInput, type ReplaceMediaClipInput, type ReplaceSequentialClipsInput, type ReplaceVoiceSpeechAssetsInput, type ReplacementMediaClipInput, type ResolvedAsrSpeechSegment, type ResolvedEntityClipPlacement, type ResolvedEntityTimelineLayout, type ResolvedEntityTrack, type ResolvedSpeechScriptSegment, SchemaValidator, ScriptCompositionError, type SemanticDocumentAdapter, SemanticEditor, type SemanticOpInput, type SemanticOpKind, type SemanticOpName, type SequentialClipAnchor, type SetBgmInput, type SetCaptionVisibilityInput, type SetClipPlacementInput, type SetClipSpeedInput, type SetClipVolumeInput, type SetVideoClipSpeedEffectInput, type SnapshotReadable, type SolvedVideoDocument, type SpeechAssetsEffectEnvelope, type SpeechAssetsEffectInput, type SpeechAssetsEffectResult, type SpeechHostMap, type SpeechOriginAssetSystem, type SpeechPart, type SpeedShift, TIMELINE_SKELETON_DURATION_MS, type Timeline, type TimelineDoc, type TimelineItem, type Track, type TrackDraft, type TrackItem, type TrackItemDraft, type TrackItemTimePosition, type TransactAudit, type TrimClipInput, type UpdateClipInput, type UpdateClipMarkerInput, ValidationError, type VideoClipPart, type VideoClipSpeedMode, type VideoDocument, type VideoDocumentDraft, type VideoDocumentMirrorSchema, VideoDocumentValidationError, type VideoDocumentValidationIssue, type VideoDocumentValidationIssueCode, type VideoDraft, type CaptionPart$1 as VideoDraftCaptionPart, type VideoDraftContent, type VideoDraftPartUnion, type Timeline$1 as VideoDraftTimeline, type Track$1 as VideoDraftTrack, type TrackItem$1 as VideoDraftTrackItem, type VideoMediaAssetFact, type VisualMediaAssetFact, type VoiceoverCaptionFact, type VoiceoverMediaAssetFact, type VoiceoverTakeInput, type VoiceoverTakeResult, type WaitForServerAckOptions, applyEntityRows, applyFieldChanges, arrangeMainTrackSeamlessly, assembleCaptionContent, assembleEntityContent, assemblePhoneticScriptContent, assembleScriptText, assertCanonicalEditorResources, assertEntityProjectionPreservesLegacy, assertFieldChanges, assertMediaAssetWritePolicy, assertValidVideoDocument, audioScriptAssetContent, audioScriptAssetFields, audioScriptAssetHash, audioScriptAssetOf, base64ToBytes, buildInitialVideoDocument, buildSpeechHostMap, bytesToBase64, cascadeAfterVideoClipChanges, compileCaptionAssetsEffect, compileEntityRows, compileGeneratedVisualEffect, compileSpeechAssetsEffect, createEditSandbox, createMirrorVideoDocument, createMirrorVideoDocumentAdapter, createPlainMemoryAdapter, decodeDocVersionMark, derivePositionFromAbs, effectiveVideoClipDurationMs, encodeDocVersionMark, ensureEditorFoundation, ensureLaneTrack, fillMainTrackTimeGaps, findComposedAudioScript, findLaneTrack, fromVideoDocument, generatePartId, getAt, hostForAbsMs, importCaptionAsset, importMediaAsset, isEmptyVideoClip, isImplementedSemanticOpKind, isMap, laneTrackId, mainTrackRanges, migrateLegacyTimelineToEntities, partDurationMs, partUnionSchema, phoneticScriptAssetContent, phoneticScriptAssetHash, phoneticScriptAssetOf, prepareCaptionAssetsEffect, prepareCaptionContent, prepareContinueVideoEffect, prepareGeneratedVisualEffect, prepareSpeechAssetsEffect, projectEntityTimeline, readDocumentAudioScript, readEntityDocument, readMainTrackItems, readMengineEventStream, readPart, readPartDurationMs, readVideoDocumentFromDraft, reassignSpeechesToVideoClipsByTime, recalculateTimelineDuration, relativePositionForAbs, replayJournal, resolveAllSpeechOverlaps, resolveEntityTimelineLayout, resolveFieldPath, resolveSpeechOverlapByShiftingVideos, safeDurationMs, index_d_exports as schemas, scriptSourceAssetOf, selectAudioScriptSegment, snapshotToPlain, solveVideoDocument, speedOf, syncAggregatedClipsTimePosition, toVideoDocument, validateVideoDocument, videoDocumentMirrorSchema, videoDocumentSchema };