@mengine/medeo-client 1.2.1-alpha.4 → 1.2.1-alpha.6

package/dist/{document-C-h5-rCG.js → document-49OdvviW.js}

@@ -111,7 +111,7 @@ const speechPart = schema.LoroMap({
 	volume: schema.Number(),
 	audio_storage_key: schema.String(),
 	origin_speech_id: schema.String(),
-	voice: schema.String({ required: false }).transform(jsonTransform()),
+	voice: schema.String().transform(jsonTransform()),
 	caption_ids: schema.String().transform(jsonTransform())
 }, { required: false });
 const captionStyle = schema.LoroMap({
@@ -322,71 +322,6 @@ function syncAggregatedClipsTimePosition(doc) {
 	}
 }
 /**
-* Resolve authoritative `time_position` edges without inventing parents from
-* geometric overlap. Roots stay absolute/sequential; every anchored descendant
-* follows its declared parent, regardless of whether that parent is video, BGM,
-* or speech.
-*/
-function resolveAnchoredClipsTimePosition(doc) {
-	const items = [
-		...doc.main_track,
-		...doc.bgm_track,
-		...doc.speech_track,
-		...doc.caption_track
-	];
-	const byPart = indexByPart(items);
-	const resolved = /* @__PURE__ */ new Set();
-	const visiting = /* @__PURE__ */ new Set();
-	const resolve = (item) => {
-		if (resolved.has(item.part_id)) return item.abs_time_position;
-		if (visiting.has(item.part_id)) throw new Error(`anchored position cycle at ${item.part_id}`);
-		visiting.add(item.part_id);
-		const position = item.time_position;
-		if (position.mode === "absolute") item.abs_time_position = Math.round(position.offsetMs);
-		else if (position.mode === "anchored") {
-			const parent = byPart.get(position.anchorPartId);
-			if (parent != null) item.abs_time_position = resolve(parent) + Math.round(position.offsetMs);
-		}
-		visiting.delete(item.part_id);
-		resolved.add(item.part_id);
-		return item.abs_time_position;
-	};
-	for (const item of items) resolve(item);
-}
-/** Preserve the historical orphan fallback without re-parenting valid roots. */
-function recoverOrphanedSpeechAnchorsByTime(doc) {
-	const placedPartIds = new Set([
-		...doc.main_track,
-		...doc.bgm_track,
-		...doc.speech_track,
-		...doc.caption_track
-	].map((item) => item.part_id));
-	const ranges = doc.main_track.map((item) => ({
-		part_id: item.part_id,
-		start: item.abs_time_position,
-		end: item.abs_time_position + partDurationMs(doc, item.part_id)
-	}));
-	if (ranges.length === 0) return;
-	for (const speech of doc.speech_track) {
-		const position = speech.time_position;
-		if (position.mode !== "anchored" || placedPartIds.has(position.anchorPartId)) continue;
-		const range = ranges.find((candidate) => candidate.start <= speech.abs_time_position && speech.abs_time_position < candidate.end) ?? (speech.abs_time_position < ranges[0].start ? ranges[0] : ranges[ranges.length - 1]);
-		const relative = Math.max(0, Math.round(speech.abs_time_position - range.start));
-		let aggregation = doc.aggregations.find((candidate) => candidate.body_part_id === range.part_id);
-		if (aggregation == null) {
-			aggregation = {
-				body_part_id: range.part_id,
-				attachments: []
-			};
-			doc.aggregations.push(aggregation);
-		}
-		aggregation.attachments.push({
-			part_id: speech.part_id,
-			relative_time_position: relative
-		});
-	}
-}
-/**
 * Reassign each speech to the video clip whose time range contains its start,
 * rebuilding `aggregations`. A speech before the first clip or after the last
 * falls back to the first / last clip respectively (FE: see §4 note — FE falls
@@ -589,14 +524,14 @@ function hostVideoOf(doc, speechPartId) {
 //#endregion
 //#region src/timeline-core/entrypoints.ts
 /**
-* The full solve pipeline: arrange → resolve declared parent edges →
-* resolve-overlap → fill-gaps → recalc. The single cascade the read-side projection runs; ops
+* The full solve pipeline: arrange → sync → reassign → resolve-overlap →
+* fill-gaps → recalc. The single cascade the read-side projection runs; ops
 * never call it (they write only facts — RFC 02 §7/§10).
 */
 function cascadeAfterVideoClipChanges(doc, makeEmptyPart) {
 	arrangeMainTrackSeamlessly(doc);
-	resolveAnchoredClipsTimePosition(doc);
-	recoverOrphanedSpeechAnchorsByTime(doc);
+	syncAggregatedClipsTimePosition(doc);
+	reassignSpeechesToVideoClipsByTime(doc);
 	resolveAllSpeechOverlaps(doc);
 	fillMainTrackTimeGaps(doc, makeEmptyPart);
 	recalculateTimelineDuration(doc);
@@ -670,7 +605,7 @@ function videoDocumentToTimelineDoc(document) {
 		caption_track: captionItems,
 		bgm_track: bgmItems,
 		part_library: partLibrary,
-		aggregations: seedAggregations([...mainItems, ...bgmItems], speechItems),
+		aggregations: seedAggregations(mainItems, speechItems),
 		timeline: {
 			duration_ms: 0,
 			unit_time_ms: document.timeline?.unit_time_ms ?? 0
@@ -694,18 +629,19 @@ function seedAbs(position, fallbackAbsMs) {
 	}
 }
 /**
-* Build the read-view `aggregations` parent-map from anchored speech items.
-* Video and BGM roots are both valid hosts; absolute speech roots have no entry.
-* Caption→speech ownership remains in `speech_part_id` / `caption_ids`.
+* Build the cascade `aggregations` parent-map from anchored items: an anchored
+* item's `anchorPartId` is the host's `part_id` (= `body_part_id`), `offsetMs`
+* is `relative_time_position`. Only main-track (video) hosts form aggregations;
+* caption→speech relations are recomputed by the cascade from `caption.start_ms`.
 */
-function seedAggregations(hostItems, speechItems) {
-	const hostPartIds = new Set(hostItems.map((item) => item.part_id));
+function seedAggregations(mainItems, speechItems) {
+	const mainPartIds = new Set(mainItems.map((item) => item.part_id));
 	const byHost = /* @__PURE__ */ new Map();
 	const order = [];
 	for (const speech of speechItems) {
 		if (speech.time_position.mode !== "anchored") continue;
 		const host = speech.time_position.anchorPartId;
-		if (!hostPartIds.has(host)) continue;
+		if (!mainPartIds.has(host)) continue;
 		let aggregation = byHost.get(host);
 		if (aggregation == null) {
 			aggregation = {
@@ -870,7 +806,7 @@ const speechPartSchema = z.object({
 	volume: finiteNumber,
 	audio_storage_key: z.string(),
 	origin_speech_id: z.string(),
-	voice: z.unknown().optional(),
+	voice: z.unknown(),
 	caption_ids: z.array(z.string())
 }).passthrough();
 const captionPartSchema = z.object({
@@ -1831,4 +1767,4 @@ function createPlainMemoryAdapter(document, options) {
 	return new PlainMemoryAdapter(document, options);
 }
 //#endregion
-export { resolveAllSpeechOverlaps as A, effectiveVideoClipDurationMs as B, solveVideoDocument as C, reassignSpeechesToVideoClipsByTime as D, fillMainTrackTimeGaps as E, isEmptyVideoClip as F, base64ToBytes as G, partUnionToDraft as H, partDurationMs as I, bytesToBase64 as K, safeDurationMs as L, resolveSpeechOverlapByShiftingVideos as M, syncAggregatedClipsTimePosition as N, recalculateTimelineDuration as O, TIMELINE_SKELETON_DURATION_MS as P, DEFAULT_UNIT_TIME_MS as R, laneTrackId as S, arrangeMainTrackSeamlessly as T, recordEntries as U, speedOf as V, videoDocumentMirrorSchema as W, partUnionSchema as _, createMirrorVideoDocument as a, ensureLaneTrack as b, readVideoDocumentFromDraft as c, derivePositionFromAbs as d, fromVideoDocument as f, validateVideoDocument as g, assertValidVideoDocument as h, MirrorVideoDocumentAdapter as i, resolveAnchoredClipsTimePosition as j, recoverOrphanedSpeechAnchorsByTime as k, buildInitialVideoDocument as l, VideoDocumentValidationError as m, createPlainMemoryAdapter as n, createMirrorVideoDocumentAdapter as o, toVideoDocument as p, generatePartId as r, writeVideoDocumentToDraft as s, PlainMemoryAdapter as t, buildSpeechHostMap as u, videoDocumentSchema as v, cascadeAfterVideoClipChanges as w, findLaneTrack as x, LANE_KINDS_IN_STACK_ORDER as y, VIDEO_DOCUMENT_SCHEMA_VERSION as z };
+export { resolveSpeechOverlapByShiftingVideos as A, partUnionToDraft as B, solveVideoDocument as C, reassignSpeechesToVideoClipsByTime as D, fillMainTrackTimeGaps as E, safeDurationMs as F, videoDocumentMirrorSchema as H, DEFAULT_UNIT_TIME_MS as I, VIDEO_DOCUMENT_SCHEMA_VERSION as L, TIMELINE_SKELETON_DURATION_MS as M, isEmptyVideoClip as N, recalculateTimelineDuration as O, partDurationMs as P, effectiveVideoClipDurationMs as R, laneTrackId as S, arrangeMainTrackSeamlessly as T, base64ToBytes as U, recordEntries as V, bytesToBase64 as W, partUnionSchema as _, createMirrorVideoDocument as a, ensureLaneTrack as b, readVideoDocumentFromDraft as c, derivePositionFromAbs as d, fromVideoDocument as f, validateVideoDocument as g, assertValidVideoDocument as h, MirrorVideoDocumentAdapter as i, syncAggregatedClipsTimePosition as j, resolveAllSpeechOverlaps as k, buildInitialVideoDocument as l, VideoDocumentValidationError as m, createPlainMemoryAdapter as n, createMirrorVideoDocumentAdapter as o, toVideoDocument as p, generatePartId as r, writeVideoDocumentToDraft as s, PlainMemoryAdapter as t, buildSpeechHostMap as u, videoDocumentSchema as v, cascadeAfterVideoClipChanges as w, findLaneTrack as x, LANE_KINDS_IN_STACK_ORDER as y, speedOf as z };

package/dist/{index-DkJ0JK2t.d.ts → index-BO2fSUKi.d.ts}

@@ -1,4 +1,4 @@
-import { A as MoveVideoClipsInput, B as MoveSpeechesInput, J as DeleteBgmInput, K as DeleteSpeechesInput, N as MoveVideoClipsByAnchorInput, O as ReplaceVideoClipContentInput, R as MoveBgmInput, S as SetBgmInput, T as ReplaceVideoClipSequenceInput, U as DeleteVideoClipsInput, X as ChangeSpeechScriptInput, Z as ChangeSpeechVoiceInput, b as SetCaptionStyleInput, ct as AddVideoClipsInput, et as AdjustVideoClipVolumeInput, g as SetVideoClipSpeedShiftInput, it as AdjustSpeechVolumeInput, nt as AdjustVideoClipDurationInput, ot as AdjustBgmVolumeInput, ut as AddSpeechesInput, v as SetCaptionVisibilityInput } from "./index-CiDSfSMi.js";
+import { A as MoveVideoClipsInput, J as ChangeSpeechScriptInput, K as DeleteBgmInput, N as MoveVideoClipsByAnchorInput, O as ReplaceVideoClipContentInput, Q as AdjustVideoClipVolumeInput, R as MoveSpeechesInput, S as SetBgmInput, T as ReplaceVideoClipSequenceInput, V as DeleteVideoClipsInput, W as DeleteSpeechesInput, Y as ChangeSpeechVoiceInput, b as SetCaptionStyleInput, ct as AddSpeechesInput, et as AdjustVideoClipDurationInput, g as SetVideoClipSpeedShiftInput, it as AdjustBgmVolumeInput, nt as AdjustSpeechVolumeInput, ot as AddVideoClipsInput, v as SetCaptionVisibilityInput } from "./index-InEWX9rk.js";
 import { InferInputType } from "loro-mirror";
 import { z } from "zod";
 import { LoroDoc, PeerID } from "loro-crdt";
@@ -1025,7 +1025,6 @@ declare class SchemaValidator {
   validateChangeSpeechVoice(input: ChangeSpeechVoiceInput, doc: SnapshotReadable): void;
   validateDeleteSpeeches(input: DeleteSpeechesInput, doc: SnapshotReadable): void;
   validateMoveSpeeches(input: MoveSpeechesInput, doc: SnapshotReadable): void;
-  validateMoveBgm(input: MoveBgmInput, doc: SnapshotReadable): void;
   validateSetBgm(input: SetBgmInput, _doc: SnapshotReadable): void;
   validateDeleteBgm(input: DeleteBgmInput, _doc: SnapshotReadable): void;
   validateSetCaptionVisibility(input: SetCaptionVisibilityInput, _doc: SnapshotReadable): void;
@@ -1037,8 +1036,12 @@ declare class SchemaValidator {
   /** Each regenerated speech (re-TTS) must already exist in the document. */
   private assertSpeechesExist;
   /**
-   * Anchored speeches may follow a video or BGM root. A null anchor is an
-   * explicit absolute speech root. Missing and wrong-kind parents are rejected.
+   * Each speech's `anchor_part_id` must point at a video clip that already
+   * exists. A relative speech anchored to a missing clip cannot be positioned
+   * by the projection and has no host to recover to (RFC 02 §4/§11.1), so the
+   * write must be rejected rather than landing a dangling reference. (Restores
+   * the dangling-reference guard the old `validateMaterializedPatch` carried;
+   * ADR 0009.)
    */
   private assertAnchorsExist;
   private assertPartKind;
@@ -1073,7 +1076,7 @@ declare class SchemaValidator {
  * side-effect payload contract in
  * `docs/projects/medeo-integration/results/phase-4-side-effect-payload-contract.md`.
  */
-type ImplementedSemanticOpKind = 'MoveVideoClips' | 'MoveVideoClipsByAnchor' | 'DeleteVideoClips' | 'AddVideoClips' | 'AdjustVideoClipVolume' | 'SetVideoClipSpeedShift' | 'ReplaceVideoClipContent' | 'ReplaceVideoClipSequence' | 'AdjustVideoClipDuration' | 'AddSpeeches' | 'DeleteSpeeches' | 'MoveSpeeches' | 'MoveBgm' | 'ChangeSpeechScript' | 'ChangeSpeechVoice' | 'AdjustSpeechVolume' | 'SetCaptionVisibility' | 'SetCaptionStyle' | 'SetBgm' | 'DeleteBgm' | 'AdjustBgmVolume';
+type ImplementedSemanticOpKind = 'MoveVideoClips' | 'MoveVideoClipsByAnchor' | 'DeleteVideoClips' | 'AddVideoClips' | 'AdjustVideoClipVolume' | 'SetVideoClipSpeedShift' | 'ReplaceVideoClipContent' | 'ReplaceVideoClipSequence' | 'AdjustVideoClipDuration' | 'AddSpeeches' | 'DeleteSpeeches' | 'MoveSpeeches' | 'ChangeSpeechScript' | 'ChangeSpeechVoice' | 'AdjustSpeechVolume' | 'SetCaptionVisibility' | 'SetCaptionStyle' | 'SetBgm' | 'DeleteBgm' | 'AdjustBgmVolume';
 /**
  * Operations on the roadmap but NOT yet implemented by `SemanticEditor`. Empty:
  * Phase 4 covers every op with a real entry point. Kept as a named type so the
@@ -1140,7 +1143,7 @@ interface CommitOptions {
   /** Author of this op; omitted for writes with no acting user (bootstrap, repair). */
   actor?: OpActor;
 }
-type SemanticOpInput = MoveVideoClipsInput | MoveVideoClipsByAnchorInput | DeleteVideoClipsInput | AddVideoClipsInput | AdjustVideoClipVolumeInput | AdjustSpeechVolumeInput | AdjustBgmVolumeInput | SetVideoClipSpeedShiftInput | ReplaceVideoClipContentInput | ReplaceVideoClipSequenceInput | AdjustVideoClipDurationInput | AddSpeechesInput | DeleteSpeechesInput | MoveSpeechesInput | MoveBgmInput | SetCaptionVisibilityInput | SetCaptionStyleInput | SetBgmInput | DeleteBgmInput;
+type SemanticOpInput = MoveVideoClipsInput | MoveVideoClipsByAnchorInput | DeleteVideoClipsInput | AddVideoClipsInput | AdjustVideoClipVolumeInput | AdjustSpeechVolumeInput | AdjustBgmVolumeInput | SetVideoClipSpeedShiftInput | ReplaceVideoClipContentInput | ReplaceVideoClipSequenceInput | AdjustVideoClipDurationInput | AddSpeechesInput | DeleteSpeechesInput | MoveSpeechesInput | SetCaptionVisibilityInput | SetCaptionStyleInput | SetBgmInput | DeleteBgmInput;
 type SemanticOpName = ImplementedSemanticOpKind;
 /** Audit metadata committed alongside an op's writes. */
 interface TransactAudit {
@@ -1257,8 +1260,6 @@ declare class SemanticEditor {
    * derives absolute positions on read.
    */
   moveSpeeches(input: MoveSpeechesInput, options?: CommitOptions): Promise<void>;
-  /** Move the BGM root; any speech/caption descendants follow through their anchors. */
-  moveBgm(input: MoveBgmInput, options?: CommitOptions): Promise<void>;
   /** Change a speech's script. Re-TTS runs upstream; the regenerated parts arrive materialized (no cascade). */
   changeSpeechScript(input: ChangeSpeechScriptInput, options?: CommitOptions): Promise<void>;
   /** Change a speech's voice. Re-TTS runs upstream; the regenerated parts arrive materialized (no cascade). */
@@ -1455,7 +1456,7 @@ declare const partUnionSchema: z.ZodUnion<readonly [z.ZodObject<{
     volume: z.ZodNumber;
     audio_storage_key: z.ZodString;
     origin_speech_id: z.ZodString;
-    voice: z.ZodOptional<z.ZodUnknown>;
+    voice: z.ZodUnknown;
     caption_ids: z.ZodArray<z.ZodString>;
   }, z.core.$loose>;
 }, z.core.$loose>, z.ZodObject<{
@@ -1555,7 +1556,7 @@ declare const videoDocumentSchema: z.ZodObject<{
       volume: z.ZodNumber;
       audio_storage_key: z.ZodString;
       origin_speech_id: z.ZodString;
-      voice: z.ZodOptional<z.ZodUnknown>;
+      voice: z.ZodUnknown;
       caption_ids: z.ZodArray<z.ZodString>;
     }, z.core.$loose>;
   }, z.core.$loose>, z.ZodObject<{

package/dist/{index-CiDSfSMi.d.ts → index-InEWX9rk.d.ts}

@@ -11,16 +11,16 @@ import { z } from "zod";
 declare const addSpeechesInputSchema: z.ZodObject<{
   speeches: z.ZodArray<z.ZodObject<{
     speech_id: z.ZodString;
-    anchor_part_id: z.ZodNullable<z.ZodString>;
+    anchor_part_id: z.ZodString;
     offset_ms: z.ZodNumber;
     audio_storage_key: z.ZodString;
     duration_ms: z.ZodNumber;
     audio_script: z.ZodString;
     volume: z.ZodNumber;
-    voice: z.ZodOptional<z.ZodObject<{
+    voice: z.ZodObject<{
       id: z.ZodString;
       name: z.ZodString;
-    }, z.core.$strip>>;
+    }, z.core.$strip>;
     origin_speech_id: z.ZodString;
     caption_ids: z.ZodArray<z.ZodString>;
   }, z.core.$strip>>;
@@ -116,16 +116,16 @@ type AdjustVideoClipVolumeInput = z.infer<typeof adjustVideoClipVolumeInputSchem
 declare const changeSpeechScriptInputSchema: z.ZodObject<{
   speeches: z.ZodArray<z.ZodObject<{
     speech_id: z.ZodString;
-    anchor_part_id: z.ZodNullable<z.ZodString>;
+    anchor_part_id: z.ZodString;
     offset_ms: z.ZodNumber;
     audio_storage_key: z.ZodString;
     duration_ms: z.ZodNumber;
     audio_script: z.ZodString;
     volume: z.ZodNumber;
-    voice: z.ZodOptional<z.ZodObject<{
+    voice: z.ZodObject<{
       id: z.ZodString;
       name: z.ZodString;
-    }, z.core.$strip>>;
+    }, z.core.$strip>;
     origin_speech_id: z.ZodString;
     caption_ids: z.ZodArray<z.ZodString>;
   }, z.core.$strip>>;
@@ -140,16 +140,16 @@ declare const changeSpeechScriptInputSchema: z.ZodObject<{
 declare const changeSpeechVoiceInputSchema: z.ZodObject<{
   speeches: z.ZodArray<z.ZodObject<{
     speech_id: z.ZodString;
-    anchor_part_id: z.ZodNullable<z.ZodString>;
+    anchor_part_id: z.ZodString;
     offset_ms: z.ZodNumber;
     audio_storage_key: z.ZodString;
     duration_ms: z.ZodNumber;
     audio_script: z.ZodString;
     volume: z.ZodNumber;
-    voice: z.ZodOptional<z.ZodObject<{
+    voice: z.ZodObject<{
       id: z.ZodString;
       name: z.ZodString;
-    }, z.core.$strip>>;
+    }, z.core.$strip>;
     origin_speech_id: z.ZodString;
     caption_ids: z.ZodArray<z.ZodString>;
   }, z.core.$strip>>;
@@ -219,12 +219,6 @@ declare const moveSpeechesInputSchema: z.ZodObject<{
 }, z.core.$strip>;
 type MoveSpeechesInput = z.infer<typeof moveSpeechesInputSchema>;
 //#endregion
-//#region src/editor/schemas/move-bgm.d.ts
-declare const moveBgmInputSchema: z.ZodObject<{
-  new_start_ms: z.ZodNumber;
-}, z.core.$strip>;
-type MoveBgmInput = z.infer<typeof moveBgmInputSchema>;
-//#endregion
 //#region src/editor/schemas/move-video-clips-by-anchor.d.ts
 /**
  * Where the moved block lands on the main track.
@@ -551,23 +545,25 @@ declare const voiceSchema: z.ZodObject<{
  * stable speech + caption parts and writes them as authoritative facts. No
  * cascade runs on write — the projection derives absolute positions on read.
  *
- * `anchor_part_id` states the root relationship explicitly: a video/BGM id
- * creates an anchored speech; `null` creates an absolute audio root.
- * `offset_ms` is relative to the anchor or absolute for a root. Captions always
- * anchor to their speech via `start_ms`.
+ * Each speech carries the anchoring fact directly (RFC 02 §4): the host video
+ * clip `anchor_part_id` and the `offset_ms` within it. The upstream caller
+ * already knows which clip a speech attaches to, so the op writes
+ * `{ mode:'anchored', anchorPartId, offsetMs }` verbatim — no write-time
+ * host-picking. Captions anchor to their speech via the caption part's
+ * `start_ms` (offset within the speech).
  */
 declare const speechAssetSchema: z.ZodObject<{
   speech_id: z.ZodString;
-  anchor_part_id: z.ZodNullable<z.ZodString>;
+  anchor_part_id: z.ZodString;
   offset_ms: z.ZodNumber;
   audio_storage_key: z.ZodString;
   duration_ms: z.ZodNumber;
   audio_script: z.ZodString;
   volume: z.ZodNumber;
-  voice: z.ZodOptional<z.ZodObject<{
+  voice: z.ZodObject<{
     id: z.ZodString;
     name: z.ZodString;
-  }, z.core.$strip>>;
+  }, z.core.$strip>;
   origin_speech_id: z.ZodString;
   caption_ids: z.ZodArray<z.ZodString>;
 }, z.core.$strip>;
@@ -582,16 +578,16 @@ declare const captionAssetSchema: z.ZodObject<{
 declare const speechAssetsSchema: z.ZodObject<{
   speeches: z.ZodArray<z.ZodObject<{
     speech_id: z.ZodString;
-    anchor_part_id: z.ZodNullable<z.ZodString>;
+    anchor_part_id: z.ZodString;
     offset_ms: z.ZodNumber;
     audio_storage_key: z.ZodString;
     duration_ms: z.ZodNumber;
     audio_script: z.ZodString;
     volume: z.ZodNumber;
-    voice: z.ZodOptional<z.ZodObject<{
+    voice: z.ZodObject<{
       id: z.ZodString;
       name: z.ZodString;
-    }, z.core.$strip>>;
+    }, z.core.$strip>;
     origin_speech_id: z.ZodString;
     caption_ids: z.ZodArray<z.ZodString>;
   }, z.core.$strip>>;
@@ -607,7 +603,7 @@ type SpeechAsset = z.infer<typeof speechAssetSchema>;
 type CaptionAsset = z.infer<typeof captionAssetSchema>;
 type SpeechAssets = z.infer<typeof speechAssetsSchema>;
 declare namespace index_d_exports {
-  export { AddSpeechesInput, AddVideoClipsInput, AdjustBgmVolumeInput, AdjustSpeechVolumeInput, AdjustVideoClipDurationInput, AdjustVideoClipVolumeInput, AnchoredDeletePolicy, AnchoredReplacePolicy, CaptionAsset, ChangeSpeechScriptInput, ChangeSpeechVoiceInput, DeleteBgmInput, DeleteSpeechesInput, DeleteVideoClipsInput, MoveAnchor, MoveBgmInput, MoveSpeechesInput, MoveVideoClipsByAnchorInput, MoveVideoClipsInput, MovedClipAnchoredPolicy, ReplaceVideoClipContentInput, ReplaceVideoClipSequenceInput, SetBgmInput, SetCaptionStyleInput, SetCaptionVisibilityInput, SetVideoClipSpeedShiftInput, SpeechAsset, SpeechAssets, addSpeechesInputSchema, addVideoClipsInputSchema, adjustBgmVolumeInputSchema, adjustSpeechVolumeInputSchema, adjustVideoClipDurationInputSchema, adjustVideoClipVolumeInputSchema, anchoredDeletePolicySchema, anchoredReplacePolicySchema, changeSpeechScriptInputSchema, changeSpeechVoiceInputSchema, clipIdSchema, clipIdsSchema, deleteBgmInputSchema, deleteSpeechesInputSchema, deleteVideoClipsInputSchema, mediaIdSchema, moveAnchorSchema, moveBgmInputSchema, moveSpeechesInputSchema, moveVideoClipsByAnchorInputSchema, moveVideoClipsInputSchema, movedClipAnchoredPolicySchema, positiveMsSchema, replaceVideoClipContentInputSchema, replaceVideoClipSequenceInputSchema, setBgmInputSchema, setCaptionStyleInputSchema, setCaptionVisibilityInputSchema, setVideoClipSpeedShiftInputSchema, speechAssetsSchema, speechIdSchema, speechIdsSchema, speedShiftSchema, timelineMsSchema, voiceSchema, volumeSchema };
+  export { AddSpeechesInput, AddVideoClipsInput, AdjustBgmVolumeInput, AdjustSpeechVolumeInput, AdjustVideoClipDurationInput, AdjustVideoClipVolumeInput, AnchoredDeletePolicy, AnchoredReplacePolicy, CaptionAsset, ChangeSpeechScriptInput, ChangeSpeechVoiceInput, DeleteBgmInput, DeleteSpeechesInput, DeleteVideoClipsInput, MoveAnchor, MoveSpeechesInput, MoveVideoClipsByAnchorInput, MoveVideoClipsInput, MovedClipAnchoredPolicy, ReplaceVideoClipContentInput, ReplaceVideoClipSequenceInput, SetBgmInput, SetCaptionStyleInput, SetCaptionVisibilityInput, SetVideoClipSpeedShiftInput, SpeechAsset, SpeechAssets, addSpeechesInputSchema, addVideoClipsInputSchema, adjustBgmVolumeInputSchema, adjustSpeechVolumeInputSchema, adjustVideoClipDurationInputSchema, adjustVideoClipVolumeInputSchema, anchoredDeletePolicySchema, anchoredReplacePolicySchema, changeSpeechScriptInputSchema, changeSpeechVoiceInputSchema, clipIdSchema, clipIdsSchema, deleteBgmInputSchema, deleteSpeechesInputSchema, deleteVideoClipsInputSchema, mediaIdSchema, moveAnchorSchema, moveSpeechesInputSchema, moveVideoClipsByAnchorInputSchema, moveVideoClipsInputSchema, movedClipAnchoredPolicySchema, positiveMsSchema, replaceVideoClipContentInputSchema, replaceVideoClipSequenceInputSchema, setBgmInputSchema, setCaptionStyleInputSchema, setCaptionVisibilityInputSchema, setVideoClipSpeedShiftInputSchema, speechAssetsSchema, speechIdSchema, speechIdsSchema, speedShiftSchema, timelineMsSchema, voiceSchema, volumeSchema };
 }
 //#endregion
-export { changeSpeechVoiceInputSchema as $, MoveVideoClipsInput as A, MoveSpeechesInput as B, setBgmInputSchema as C, replaceVideoClipSequenceInputSchema as D, anchoredReplacePolicySchema as E, moveAnchorSchema as F, deleteVideoClipsInputSchema as G, AnchoredDeletePolicy as H, moveVideoClipsByAnchorInputSchema as I, DeleteBgmInput as J, DeleteSpeechesInput as K, movedClipAnchoredPolicySchema as L, MoveAnchor as M, MoveVideoClipsByAnchorInput as N, ReplaceVideoClipContentInput as O, MovedClipAnchoredPolicy as P, changeSpeechScriptInputSchema as Q, MoveBgmInput as R, SetBgmInput as S, ReplaceVideoClipSequenceInput as T, DeleteVideoClipsInput as U, moveSpeechesInputSchema as V, anchoredDeletePolicySchema as W, ChangeSpeechScriptInput as X, deleteBgmInputSchema as Y, ChangeSpeechVoiceInput as Z, setVideoClipSpeedShiftInputSchema as _, speechAssetsSchema as a, adjustSpeechVolumeInputSchema as at, SetCaptionStyleInput as b, mediaIdSchema as c, AddVideoClipsInput as ct, speechIdsSchema as d, addSpeechesInputSchema as dt, AdjustVideoClipVolumeInput as et, speedShiftSchema as f, SetVideoClipSpeedShiftInput as g, volumeSchema as h, SpeechAssets as i, AdjustSpeechVolumeInput as it, moveVideoClipsInputSchema as j, replaceVideoClipContentInputSchema as k, positiveMsSchema as l, addVideoClipsInputSchema as lt, voiceSchema as m, CaptionAsset as n, AdjustVideoClipDurationInput as nt, clipIdSchema as o, AdjustBgmVolumeInput as ot, timelineMsSchema as p, deleteSpeechesInputSchema as q, SpeechAsset as r, adjustVideoClipDurationInputSchema as rt, clipIdsSchema as s, adjustBgmVolumeInputSchema as st, index_d_exports as t, adjustVideoClipVolumeInputSchema as tt, speechIdSchema as u, AddSpeechesInput as ut, SetCaptionVisibilityInput as v, AnchoredReplacePolicy as w, setCaptionStyleInputSchema as x, setCaptionVisibilityInputSchema as y, moveBgmInputSchema as z };
+export { adjustVideoClipVolumeInputSchema as $, MoveVideoClipsInput as A, AnchoredDeletePolicy as B, setBgmInputSchema as C, replaceVideoClipSequenceInputSchema as D, anchoredReplacePolicySchema as E, moveAnchorSchema as F, deleteSpeechesInputSchema as G, anchoredDeletePolicySchema as H, moveVideoClipsByAnchorInputSchema as I, ChangeSpeechScriptInput as J, DeleteBgmInput as K, movedClipAnchoredPolicySchema as L, MoveAnchor as M, MoveVideoClipsByAnchorInput as N, ReplaceVideoClipContentInput as O, MovedClipAnchoredPolicy as P, AdjustVideoClipVolumeInput as Q, MoveSpeechesInput as R, SetBgmInput as S, ReplaceVideoClipSequenceInput as T, deleteVideoClipsInputSchema as U, DeleteVideoClipsInput as V, DeleteSpeechesInput as W, changeSpeechScriptInputSchema as X, ChangeSpeechVoiceInput as Y, changeSpeechVoiceInputSchema as Z, setVideoClipSpeedShiftInputSchema as _, speechAssetsSchema as a, adjustBgmVolumeInputSchema as at, SetCaptionStyleInput as b, mediaIdSchema as c, AddSpeechesInput as ct, speechIdsSchema as d, AdjustVideoClipDurationInput as et, speedShiftSchema as f, SetVideoClipSpeedShiftInput as g, volumeSchema as h, SpeechAssets as i, AdjustBgmVolumeInput as it, moveVideoClipsInputSchema as j, replaceVideoClipContentInputSchema as k, positiveMsSchema as l, addSpeechesInputSchema as lt, voiceSchema as m, CaptionAsset as n, AdjustSpeechVolumeInput as nt, clipIdSchema as o, AddVideoClipsInput as ot, timelineMsSchema as p, deleteBgmInputSchema as q, SpeechAsset as r, adjustSpeechVolumeInputSchema as rt, clipIdsSchema as s, addVideoClipsInputSchema as st, index_d_exports as t, adjustVideoClipDurationInputSchema as tt, speechIdSchema as u, SetCaptionVisibilityInput as v, AnchoredReplacePolicy as w, setCaptionStyleInputSchema as x, setCaptionVisibilityInputSchema as y, moveSpeechesInputSchema as z };

package/dist/index.d.ts

@@ -1,5 +1,5 @@
-import { t as index_d_exports } from "./index-CiDSfSMi.js";
-import { $ as Track, A as TrackDraft, B as DerivedItemPosition, C as SemanticOpKind, D as ValidationError, E as SnapshotReadable, F as VideoDocumentValidationError, G as toVideoDocument, H as buildSpeechHostMap, I as assertValidVideoDocument, J as DEFAULT_UNIT_TIME_MS, K as BgmPart, L as validateVideoDocument, M as VideoDocumentDraft, N as VideoDocumentMirrorSchema, O as PartIdFactory, P as videoDocumentMirrorSchema, Q as Timeline, R as InitialDocumentFacts, S as PlannedSemanticOpKind, T as SchemaValidator, U as derivePositionFromAbs, V as SpeechHostMap, W as fromVideoDocument, X as PartUnion, Y as PartKind, Z as SpeechPart, _ as SemanticOpInput, _t as Timeline$1, a as PlainMemoryAdapter, at as VideoDocumentSchemaVersion, b as IMPLEMENTED_SEMANTIC_OP_KINDS, c as MirrorVideoDocumentAdapter, ct as VideoDraft, d as createMirrorVideoDocumentAdapter, dt as speedOf, et as TrackItem, f as writeVideoDocumentToDraft, ft as Attachment, g as SemanticEditor, gt as SpeedShift, h as SemanticDocumentAdapter, ht as PartAggregation, i as JournalEntry, it as VideoDocument, j as TrackItemDraft, k as generatePartId, l as MirrorVideoDocumentOptions, lt as VideoDraftPartUnion, m as OpActor, mt as CaptionStyle, n as videoDocumentSchema, nt as VIDEO_DOCUMENT_SCHEMA_VERSION, o as PlainMemoryAdapterOptions, ot as VideoDocumentValidationIssue, p as CommitOptions, pt as CaptionPart$1, q as CaptionPart, r as readVideoDocumentFromDraft, rt as VideoClipPart, s as createPlainMemoryAdapter, st as VideoDocumentValidationIssueCode, t as partUnionSchema, tt as TrackItemTimePosition, u as createMirrorVideoDocument, ut as effectiveVideoClipDurationMs, v as SemanticOpName, vt as Track$1, w as isImplementedSemanticOpKind, x as ImplementedSemanticOpKind, y as TransactAudit, yt as TrackItem$1, z as buildInitialVideoDocument } from "./index-DkJ0JK2t.js";
+import { t as index_d_exports } from "./index-InEWX9rk.js";
+import { $ as Track, A as TrackDraft, B as DerivedItemPosition, C as SemanticOpKind, D as ValidationError, E as SnapshotReadable, F as VideoDocumentValidationError, G as toVideoDocument, H as buildSpeechHostMap, I as assertValidVideoDocument, J as DEFAULT_UNIT_TIME_MS, K as BgmPart, L as validateVideoDocument, M as VideoDocumentDraft, N as VideoDocumentMirrorSchema, O as PartIdFactory, P as videoDocumentMirrorSchema, Q as Timeline, R as InitialDocumentFacts, S as PlannedSemanticOpKind, T as SchemaValidator, U as derivePositionFromAbs, V as SpeechHostMap, W as fromVideoDocument, X as PartUnion, Y as PartKind, Z as SpeechPart, _ as SemanticOpInput, _t as Timeline$1, a as PlainMemoryAdapter, at as VideoDocumentSchemaVersion, b as IMPLEMENTED_SEMANTIC_OP_KINDS, c as MirrorVideoDocumentAdapter, ct as VideoDraft, d as createMirrorVideoDocumentAdapter, dt as speedOf, et as TrackItem, f as writeVideoDocumentToDraft, ft as Attachment, g as SemanticEditor, gt as SpeedShift, h as SemanticDocumentAdapter, ht as PartAggregation, i as JournalEntry, it as VideoDocument, j as TrackItemDraft, k as generatePartId, l as MirrorVideoDocumentOptions, lt as VideoDraftPartUnion, m as OpActor, mt as CaptionStyle, n as videoDocumentSchema, nt as VIDEO_DOCUMENT_SCHEMA_VERSION, o as PlainMemoryAdapterOptions, ot as VideoDocumentValidationIssue, p as CommitOptions, pt as CaptionPart$1, q as CaptionPart, r as readVideoDocumentFromDraft, rt as VideoClipPart, s as createPlainMemoryAdapter, st as VideoDocumentValidationIssueCode, t as partUnionSchema, tt as TrackItemTimePosition, u as createMirrorVideoDocument, ut as effectiveVideoClipDurationMs, v as SemanticOpName, vt as Track$1, w as isImplementedSemanticOpKind, x as ImplementedSemanticOpKind, y as TransactAudit, yt as TrackItem$1, z as buildInitialVideoDocument } from "./index-BO2fSUKi.js";
 import { PeerID } from "loro-crdt";
 import { DocState } from "@mengine/sync";
 import { BaseDocStorage, Connection, DocDiff, DocPushReceipt, DocSnapshotRecord, DocStorage, DocStorageOptions, DocUpdate, DocUpdateRecord } from "@mengine/storage";
@@ -866,15 +866,6 @@ declare function arrangeMainTrackSeamlessly(doc: TimelineDoc): void;
  *   caption.abs  = speech.abs + caption.start_ms
  */
 declare function syncAggregatedClipsTimePosition(doc: TimelineDoc): void;
-/**
- * Resolve authoritative `time_position` edges without inventing parents from
- * geometric overlap. Roots stay absolute/sequential; every anchored descendant
- * follows its declared parent, regardless of whether that parent is video, BGM,
- * or speech.
- */
-declare function resolveAnchoredClipsTimePosition(doc: TimelineDoc): void;
-/** Preserve the historical orphan fallback without re-parenting valid roots. */
-declare function recoverOrphanedSpeechAnchorsByTime(doc: TimelineDoc): void;
 /**
  * Reassign each speech to the video clip whose time range contains its start,
  * rebuilding `aggregations`. A speech before the first clip or after the last
@@ -940,8 +931,8 @@ type MakeEmptyPart = (durationMs: number) => {
   partId: string;
 };
 /**
- * The full solve pipeline: arrange → resolve declared parent edges →
- * resolve-overlap → fill-gaps → recalc. The single cascade the read-side projection runs; ops
+ * The full solve pipeline: arrange → sync → reassign → resolve-overlap →
+ * fill-gaps → recalc. The single cascade the read-side projection runs; ops
  * never call it (they write only facts — RFC 02 §7/§10).
  */
 declare function cascadeAfterVideoClipChanges(doc: TimelineDoc, makeEmptyPart: MakeEmptyPart): void;
@@ -1086,4 +1077,4 @@ declare function hostForAbsMs(ranges: MainClipRange[], absMs: number): MainClipR
  */
 declare function relativePositionForAbs(ranges: MainClipRange[], absMs: number): TrackItemTimePosition;
 //#endregion
-export { type Aggregation, type Attachment, type BgmPart, type CaptionPart, type CaptionStyle, type CommitOptions, DEFAULT_UNIT_TIME_MS, type DerivedItemPosition, type DocStorageLike, type DocVersionMark, IMPLEMENTED_SEMANTIC_OP_KINDS, type ImplementedSemanticOpKind, type InitialDocumentFacts, type JournalEntry, LANE_KINDS_IN_STACK_ORDER, type LaneKind, type MainClipRange, type MakeEmptyPart, ManualSyncDoc, type ManualSyncDocOptions, MedeoHttpDocStorage, type MedeoHttpDocStorageOptions, MemoryDocStorage, MengineAckFailedError, type MengineAuditEntry, type MengineAuditResponse, MengineDocSession, type MengineDocSessionOptions, type MengineDocSessionUpdateEvent, type MengineDocumentVersion, type MengineEventStreamOptions, MengineHttpClient, type MengineHttpClientOptions, MengineHttpRequestError, MenginePushRejectedError, type MenginePushResponse, type MenginePushUpdateResponse, type MengineRejectedResponse, type MengineSnapshotResponse, type MengineSseUpdateEvent, type MengineSyncResponse, type MengineUpdateMeta, MirrorVideoDocumentAdapter, type MirrorVideoDocumentOptions, type OpActor, type PartAggregation, type PartIdFactory, type PartKind, type PartUnion, PlainMemoryAdapter, type PlainMemoryAdapterOptions, type PlannedSemanticOpKind, type PullFailureReason, type PullResult, type PushOutcome, type PushOutcomeKind, type PushResult, type PushResultKind, SchemaValidator, type SemanticDocumentAdapter, SemanticEditor, type SemanticOpInput, type SemanticOpKind, type SemanticOpName, 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, VIDEO_DOCUMENT_SCHEMA_VERSION, ValidationError, type VideoClipPart, type VideoDocument, type VideoDocumentDraft, type VideoDocumentMirrorSchema, type VideoDocumentSchemaVersion, VideoDocumentValidationError, type VideoDocumentValidationIssue, type VideoDocumentValidationIssueCode, type VideoDraft, type CaptionPart$1 as VideoDraftCaptionPart, type VideoDraftPartUnion, type Timeline$1 as VideoDraftTimeline, type Track$1 as VideoDraftTrack, type TrackItem$1 as VideoDraftTrackItem, type WaitForServerAckOptions, arrangeMainTrackSeamlessly, assertValidVideoDocument, base64ToBytes, buildInitialVideoDocument, buildSpeechHostMap, bytesToBase64, cascadeAfterVideoClipChanges, createEditSandbox, createMirrorVideoDocument, createMirrorVideoDocumentAdapter, createPlainMemoryAdapter, decodeDocVersionMark, derivePositionFromAbs, effectiveVideoClipDurationMs, encodeDocVersionMark, ensureLaneTrack, fillMainTrackTimeGaps, findLaneTrack, fromVideoDocument, generatePartId, getAt, hostForAbsMs, isEmptyVideoClip, isImplementedSemanticOpKind, isMap, laneTrackId, mainTrackRanges, partDurationMs, partUnionSchema, readMainTrackItems, readMengineEventStream, readPart, readPartDurationMs, readVideoDocumentFromDraft, reassignSpeechesToVideoClipsByTime, recalculateTimelineDuration, recoverOrphanedSpeechAnchorsByTime, relativePositionForAbs, replayJournal, resolveAllSpeechOverlaps, resolveAnchoredClipsTimePosition, resolveSpeechOverlapByShiftingVideos, safeDurationMs, index_d_exports as schemas, snapshotToPlain, solveVideoDocument, speedOf, syncAggregatedClipsTimePosition, toVideoDocument, validateVideoDocument, videoDocumentMirrorSchema, videoDocumentSchema, writeVideoDocumentToDraft };
+export { type Aggregation, type Attachment, type BgmPart, type CaptionPart, type CaptionStyle, type CommitOptions, DEFAULT_UNIT_TIME_MS, type DerivedItemPosition, type DocStorageLike, type DocVersionMark, IMPLEMENTED_SEMANTIC_OP_KINDS, type ImplementedSemanticOpKind, type InitialDocumentFacts, type JournalEntry, LANE_KINDS_IN_STACK_ORDER, type LaneKind, type MainClipRange, type MakeEmptyPart, ManualSyncDoc, type ManualSyncDocOptions, MedeoHttpDocStorage, type MedeoHttpDocStorageOptions, MemoryDocStorage, MengineAckFailedError, type MengineAuditEntry, type MengineAuditResponse, MengineDocSession, type MengineDocSessionOptions, type MengineDocSessionUpdateEvent, type MengineDocumentVersion, type MengineEventStreamOptions, MengineHttpClient, type MengineHttpClientOptions, MengineHttpRequestError, MenginePushRejectedError, type MenginePushResponse, type MenginePushUpdateResponse, type MengineRejectedResponse, type MengineSnapshotResponse, type MengineSseUpdateEvent, type MengineSyncResponse, type MengineUpdateMeta, MirrorVideoDocumentAdapter, type MirrorVideoDocumentOptions, type OpActor, type PartAggregation, type PartIdFactory, type PartKind, type PartUnion, PlainMemoryAdapter, type PlainMemoryAdapterOptions, type PlannedSemanticOpKind, type PullFailureReason, type PullResult, type PushOutcome, type PushOutcomeKind, type PushResult, type PushResultKind, SchemaValidator, type SemanticDocumentAdapter, SemanticEditor, type SemanticOpInput, type SemanticOpKind, type SemanticOpName, 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, VIDEO_DOCUMENT_SCHEMA_VERSION, ValidationError, type VideoClipPart, type VideoDocument, type VideoDocumentDraft, type VideoDocumentMirrorSchema, type VideoDocumentSchemaVersion, VideoDocumentValidationError, type VideoDocumentValidationIssue, type VideoDocumentValidationIssueCode, type VideoDraft, type CaptionPart$1 as VideoDraftCaptionPart, type VideoDraftPartUnion, type Timeline$1 as VideoDraftTimeline, type Track$1 as VideoDraftTrack, type TrackItem$1 as VideoDraftTrackItem, type WaitForServerAckOptions, arrangeMainTrackSeamlessly, assertValidVideoDocument, base64ToBytes, buildInitialVideoDocument, buildSpeechHostMap, bytesToBase64, cascadeAfterVideoClipChanges, createEditSandbox, createMirrorVideoDocument, createMirrorVideoDocumentAdapter, createPlainMemoryAdapter, decodeDocVersionMark, derivePositionFromAbs, effectiveVideoClipDurationMs, encodeDocVersionMark, ensureLaneTrack, fillMainTrackTimeGaps, findLaneTrack, fromVideoDocument, generatePartId, getAt, hostForAbsMs, isEmptyVideoClip, isImplementedSemanticOpKind, isMap, laneTrackId, mainTrackRanges, partDurationMs, partUnionSchema, readMainTrackItems, readMengineEventStream, readPart, readPartDurationMs, readVideoDocumentFromDraft, reassignSpeechesToVideoClipsByTime, recalculateTimelineDuration, relativePositionForAbs, replayJournal, resolveAllSpeechOverlaps, resolveSpeechOverlapByShiftingVideos, safeDurationMs, index_d_exports as schemas, snapshotToPlain, solveVideoDocument, speedOf, syncAggregatedClipsTimePosition, toVideoDocument, validateVideoDocument, videoDocumentMirrorSchema, videoDocumentSchema, writeVideoDocumentToDraft };

package/dist/index.js

@@ -1,5 +1,5 @@
-import { A as resolveAllSpeechOverlaps, B as effectiveVideoClipDurationMs, C as solveVideoDocument, D as reassignSpeechesToVideoClipsByTime, E as fillMainTrackTimeGaps, F as isEmptyVideoClip, G as base64ToBytes, H as partUnionToDraft, I as partDurationMs, K as bytesToBase64, L as safeDurationMs, M as resolveSpeechOverlapByShiftingVideos, N as syncAggregatedClipsTimePosition, O as recalculateTimelineDuration, P as TIMELINE_SKELETON_DURATION_MS, R as DEFAULT_UNIT_TIME_MS, S as laneTrackId, T as arrangeMainTrackSeamlessly, U as recordEntries, V as speedOf, W as videoDocumentMirrorSchema, _ as partUnionSchema, a as createMirrorVideoDocument, b as ensureLaneTrack, c as readVideoDocumentFromDraft, d as derivePositionFromAbs, f as fromVideoDocument, g as validateVideoDocument, h as assertValidVideoDocument, i as MirrorVideoDocumentAdapter, j as resolveAnchoredClipsTimePosition, k as recoverOrphanedSpeechAnchorsByTime, l as buildInitialVideoDocument, m as VideoDocumentValidationError, n as createPlainMemoryAdapter, o as createMirrorVideoDocumentAdapter, p as toVideoDocument, r as generatePartId, s as writeVideoDocumentToDraft, t as PlainMemoryAdapter, u as buildSpeechHostMap, v as videoDocumentSchema, w as cascadeAfterVideoClipChanges, x as findLaneTrack, y as LANE_KINDS_IN_STACK_ORDER, z as VIDEO_DOCUMENT_SCHEMA_VERSION } from "./document-C-h5-rCG.js";
-import { addSpeechesInputSchema, addVideoClipsInputSchema, adjustBgmVolumeInputSchema, adjustSpeechVolumeInputSchema, adjustVideoClipDurationInputSchema, adjustVideoClipVolumeInputSchema, changeSpeechScriptInputSchema, changeSpeechVoiceInputSchema, deleteBgmInputSchema, deleteSpeechesInputSchema, deleteVideoClipsInputSchema, moveBgmInputSchema, moveSpeechesInputSchema, moveVideoClipsByAnchorInputSchema, moveVideoClipsInputSchema, replaceVideoClipContentInputSchema, replaceVideoClipSequenceInputSchema, setBgmInputSchema, setCaptionStyleInputSchema, setCaptionVisibilityInputSchema, setVideoClipSpeedShiftInputSchema, t as schemas_exports } from "./schemas.js";
+import { A as resolveSpeechOverlapByShiftingVideos, B as partUnionToDraft, C as solveVideoDocument, D as reassignSpeechesToVideoClipsByTime, E as fillMainTrackTimeGaps, F as safeDurationMs, H as videoDocumentMirrorSchema, I as DEFAULT_UNIT_TIME_MS, L as VIDEO_DOCUMENT_SCHEMA_VERSION, M as TIMELINE_SKELETON_DURATION_MS, N as isEmptyVideoClip, O as recalculateTimelineDuration, P as partDurationMs, R as effectiveVideoClipDurationMs, S as laneTrackId, T as arrangeMainTrackSeamlessly, U as base64ToBytes, V as recordEntries, W as bytesToBase64, _ as partUnionSchema, a as createMirrorVideoDocument, b as ensureLaneTrack, c as readVideoDocumentFromDraft, d as derivePositionFromAbs, f as fromVideoDocument, g as validateVideoDocument, h as assertValidVideoDocument, i as MirrorVideoDocumentAdapter, j as syncAggregatedClipsTimePosition, k as resolveAllSpeechOverlaps, l as buildInitialVideoDocument, m as VideoDocumentValidationError, n as createPlainMemoryAdapter, o as createMirrorVideoDocumentAdapter, p as toVideoDocument, r as generatePartId, s as writeVideoDocumentToDraft, t as PlainMemoryAdapter, u as buildSpeechHostMap, v as videoDocumentSchema, w as cascadeAfterVideoClipChanges, x as findLaneTrack, y as LANE_KINDS_IN_STACK_ORDER, z as speedOf } from "./document-49OdvviW.js";
+import { addSpeechesInputSchema, addVideoClipsInputSchema, adjustBgmVolumeInputSchema, adjustSpeechVolumeInputSchema, adjustVideoClipDurationInputSchema, adjustVideoClipVolumeInputSchema, changeSpeechScriptInputSchema, changeSpeechVoiceInputSchema, deleteBgmInputSchema, deleteSpeechesInputSchema, deleteVideoClipsInputSchema, moveSpeechesInputSchema, moveVideoClipsByAnchorInputSchema, moveVideoClipsInputSchema, replaceVideoClipContentInputSchema, replaceVideoClipSequenceInputSchema, setBgmInputSchema, setCaptionStyleInputSchema, setCaptionVisibilityInputSchema, setVideoClipSpeedShiftInputSchema, t as schemas_exports } from "./schemas.js";
 import { LoroDoc, VersionVector } from "loro-crdt";
 import { ClientServerSynchronizer, DocManager } from "@mengine/sync";
 import { DisposableSet, EventBus, Task } from "@mengine/utils";
@@ -445,10 +445,6 @@ var SchemaValidator = class {
 		const snapshot = doc.snapshot();
 		for (const s of parsed.speeches) this.assertPartKind(snapshot, s.speech_id, "speech", "move_speeches");
 	}
-	validateMoveBgm(input, doc) {
-		this.parse(moveBgmInputSchema, input, "move_bgm_invalid_input");
-		if (doc.snapshot().tracks?.find((track) => track.parts_kind === "bgm")?.items?.[0] == null) throw new ValidationError("move_bgm_not_found", "document has no BGM item");
-	}
 	validateSetBgm(input, _doc) {
 		this.parse(setBgmInputSchema, input, "set_bgm_invalid_input");
 	}
@@ -484,20 +480,16 @@ var SchemaValidator = class {
 		for (const speech of assets.speeches) this.assertPartKind(snapshot, speech.speech_id, "speech", opName);
 	}
 	/**
-	* Anchored speeches may follow a video or BGM root. A null anchor is an
-	* explicit absolute speech root. Missing and wrong-kind parents are rejected.
+	* Each speech's `anchor_part_id` must point at a video clip that already
+	* exists. A relative speech anchored to a missing clip cannot be positioned
+	* by the projection and has no host to recover to (RFC 02 §4/§11.1), so the
+	* write must be rejected rather than landing a dangling reference. (Restores
+	* the dangling-reference guard the old `validateMaterializedPatch` carried;
+	* ADR 0009.)
 	*/
 	assertAnchorsExist(assets, doc, opName) {
 		const snapshot = doc.snapshot();
-		for (const speech of assets.speeches) {
-			if (speech.anchor_part_id == null) continue;
-			const part = readPart(snapshot, speech.anchor_part_id);
-			if (part == null) throw new ValidationError(`${opName}_anchor_part_not_found`, `part_library has no parent`, { part_id: speech.anchor_part_id });
-			if (part.part_kind !== "video_clip" && part.part_kind !== "bgm") throw new ValidationError(`${opName}_anchor_wrong_part_kind`, `speech parent must be video_clip or bgm`, {
-				part_id: speech.anchor_part_id,
-				part_kind: part.part_kind
-			});
-		}
+		for (const speech of assets.speeches) this.assertPartKind(snapshot, speech.anchor_part_id, "video_clip", `${opName}_anchor`);
 	}
 	assertPartKind(snapshot, partId, kind, opName) {
 		const part = readPart(snapshot, partId);
@@ -846,9 +838,8 @@ var SemanticEditor = class {
 	*/
 	async addSpeeches(input, options) {
 		this.validator.validateAddSpeeches(input, this.doc);
-		const absByPartId = solveVideoDocument(this.doc.snapshot()).absByPartId;
 		this.doc.transact((draft) => {
-			writeSpeechAssets(draft, input, absByPartId);
+			writeSpeechAssets(draft, input);
 		}, audit("AddSpeeches", input, options));
 	}
 	/** Delete speeches with their captions; the subtree is removed (§9.2). Surviving lanes keep their facts. */
@@ -866,60 +857,29 @@ var SemanticEditor = class {
 	*/
 	async moveSpeeches(input, options) {
 		this.validator.validateMoveSpeeches(input, this.doc);
-		const snapshot = this.doc.snapshot();
-		const absByPartId = solveVideoDocument(snapshot).absByPartId;
 		this.doc.transact((draft) => {
-			const track = findLaneTrack(draft, "speech");
 			const ranges = mainTrackRanges(draft);
+			const track = findLaneTrack(draft, "speech");
 			for (const move of input.speeches) {
 				const item = (track?.items ?? []).find((it) => it?.part_id === move.speech_id);
 				if (item == null) continue;
-				const current = item.time_position;
-				if (current?.mode === "anchored") {
-					const parentStartMs = absByPartId.get(current.anchorPartId);
-					item.time_position = readPart(snapshot, current.anchorPartId)?.part_kind === "video_clip" ? relativePositionForAbs(ranges, move.new_start_ms) : parentStartMs == null ? {
-						mode: "absolute",
-						offsetMs: move.new_start_ms
-					} : {
-						mode: "anchored",
-						anchorPartId: current.anchorPartId,
-						offsetMs: Math.max(0, move.new_start_ms - parentStartMs)
-					};
-				} else item.time_position = {
-					mode: "absolute",
-					offsetMs: move.new_start_ms
-				};
+				item.time_position = relativePositionForAbs(ranges, move.new_start_ms);
 				item.fallback_abs_ms = move.new_start_ms;
 			}
 		}, audit("MoveSpeeches", input, options));
 	}
-	/** Move the BGM root; any speech/caption descendants follow through their anchors. */
-	async moveBgm(input, options) {
-		this.validator.validateMoveBgm(input, this.doc);
-		this.doc.transact((draft) => {
-			const item = findLaneTrack(draft, "bgm")?.items?.[0];
-			if (item == null) return;
-			item.time_position = {
-				mode: "absolute",
-				offsetMs: input.new_start_ms
-			};
-			item.fallback_abs_ms = void 0;
-		}, audit("MoveBgm", input, options));
-	}
 	/** Change a speech's script. Re-TTS runs upstream; the regenerated parts arrive materialized (no cascade). */
 	async changeSpeechScript(input, options) {
 		this.validator.validateChangeSpeechScript(input, this.doc);
-		const absByPartId = solveVideoDocument(this.doc.snapshot()).absByPartId;
 		this.doc.transact((draft) => {
-			writeSpeechAssets(draft, input, absByPartId);
+			writeSpeechAssets(draft, input);
 		}, audit("ChangeSpeechScript", input, options));
 	}
 	/** Change a speech's voice. Re-TTS runs upstream; the regenerated parts arrive materialized (no cascade). */
 	async changeSpeechVoice(input, options) {
 		this.validator.validateChangeSpeechVoice(input, this.doc);
-		const absByPartId = solveVideoDocument(this.doc.snapshot()).absByPartId;
 		this.doc.transact((draft) => {
-			writeSpeechAssets(draft, input, absByPartId);
+			writeSpeechAssets(draft, input);
 		}, audit("ChangeSpeechVoice", input, options));
 	}
 	/**
@@ -1148,7 +1108,7 @@ function normalizeSpeedShift(input) {
 * projection derives it on read. On re-TTS the speech id is preserved, so an
 * existing item is re-positioned rather than duplicated.
 */
-function writeSpeechAssets(draft, assets, absByPartId) {
+function writeSpeechAssets(draft, assets) {
 	const captionById = new Map(assets.captions.map((c) => [c.caption_id, c]));
 	const speechTrack = ensureLaneTrack(draft, "speech");
 	speechTrack.items ??= [];
@@ -1161,6 +1121,8 @@ function writeSpeechAssets(draft, assets, absByPartId) {
 		}
 		return captionItems;
 	};
+	const ranges = mainTrackRanges(draft);
+	const hostStartMs = (anchorPartId) => ranges.find((r) => r.partId === anchorPartId)?.startMs;
 	for (const speech of assets.speeches) {
 		const priorCaptionIds = draft.part_library?.[speech.speech_id]?.speech?.caption_ids ?? [];
 		const nextCaptionIds = new Set(speech.caption_ids);
@@ -1184,10 +1146,9 @@ function writeSpeechAssets(draft, assets, absByPartId) {
 			voice: speech.voice,
 			caption_ids: speech.caption_ids
 		} });
-		const hostStart = speech.anchor_part_id == null ? void 0 : absByPartId.get(speech.anchor_part_id);
-		const speechAbs = speech.anchor_part_id == null ? speech.offset_ms : hostStart == null ? void 0 : hostStart + speech.offset_ms;
-		if (speech.anchor_part_id == null) placeAbsolute(speechTrack.items, speech.speech_id, speech.offset_ms);
-		else placeRelative(speechTrack.items, speech.speech_id, speech.anchor_part_id, speech.offset_ms, speechAbs);
+		const hostStart = hostStartMs(speech.anchor_part_id);
+		const speechAbs = hostStart == null ? void 0 : hostStart + speech.offset_ms;
+		placeRelative(speechTrack.items, speech.speech_id, speech.anchor_part_id, speech.offset_ms, speechAbs);
 		for (const captionId of speech.caption_ids) {
 			const caption = captionById.get(captionId);
 			if (caption == null) continue;
@@ -1232,23 +1193,6 @@ function placeRelative(items, partId, anchorPartId, offsetMs, fallbackAbsMs) {
 		fallback_abs_ms: fallbackAbsMs
 	});
 }
-function placeAbsolute(items, partId, startMs) {
-	const timePosition = {
-		mode: "absolute",
-		offsetMs: Math.max(0, Math.round(startMs))
-	};
-	const existing = items.find((item) => item?.part_id === partId);
-	if (existing != null) {
-		existing.time_position = timePosition;
-		existing.fallback_abs_ms = void 0;
-		return;
-	}
-	items.push({
-		part_id: partId,
-		time_position: timePosition,
-		fallback_abs_ms: void 0
-	});
-}
 /**
 * After the main-track flow order changes (move / reorder), re-parent each
 * anchored speech to the clip its *preserved* absolute landing now falls in
@@ -1446,9 +1390,6 @@ async function dispatchEntry(editor, entry) {
 		case "MoveSpeeches":
 			await editor.moveSpeeches(payload, options);
 			return;
-		case "MoveBgm":
-			await editor.moveBgm(payload, options);
-			return;
 		case "ChangeSpeechScript":
 			await editor.changeSpeechScript(payload, options);
 			return;
@@ -1495,7 +1436,6 @@ const IMPLEMENTED_SEMANTIC_OP_KINDS = [
 	"AddSpeeches",
 	"DeleteSpeeches",
 	"MoveSpeeches",
-	"MoveBgm",
 	"ChangeSpeechScript",
 	"ChangeSpeechVoice",
 	"AdjustSpeechVolume",
@@ -2414,4 +2354,4 @@ function serverCovers(serverVV, target) {
 	return covers(server, target);
 }
 //#endregion
-export { DEFAULT_UNIT_TIME_MS, IMPLEMENTED_SEMANTIC_OP_KINDS, LANE_KINDS_IN_STACK_ORDER, ManualSyncDoc, MedeoHttpDocStorage, MemoryDocStorage, MengineAckFailedError, MengineDocSession, MengineHttpClient, MengineHttpRequestError, MenginePushRejectedError, MirrorVideoDocumentAdapter, PlainMemoryAdapter, SchemaValidator, SemanticEditor, TIMELINE_SKELETON_DURATION_MS, VIDEO_DOCUMENT_SCHEMA_VERSION, ValidationError, VideoDocumentValidationError, arrangeMainTrackSeamlessly, assertValidVideoDocument, base64ToBytes, buildInitialVideoDocument, buildSpeechHostMap, bytesToBase64, cascadeAfterVideoClipChanges, createEditSandbox, createMirrorVideoDocument, createMirrorVideoDocumentAdapter, createPlainMemoryAdapter, decodeDocVersionMark, derivePositionFromAbs, effectiveVideoClipDurationMs, encodeDocVersionMark, ensureLaneTrack, fillMainTrackTimeGaps, findLaneTrack, fromVideoDocument, generatePartId, getAt, hostForAbsMs, isEmptyVideoClip, isImplementedSemanticOpKind, isMap, laneTrackId, mainTrackRanges, partDurationMs, partUnionSchema, readMainTrackItems, readMengineEventStream, readPart, readPartDurationMs, readVideoDocumentFromDraft, reassignSpeechesToVideoClipsByTime, recalculateTimelineDuration, recoverOrphanedSpeechAnchorsByTime, relativePositionForAbs, replayJournal, resolveAllSpeechOverlaps, resolveAnchoredClipsTimePosition, resolveSpeechOverlapByShiftingVideos, safeDurationMs, schemas_exports as schemas, snapshotToPlain, solveVideoDocument, speedOf, syncAggregatedClipsTimePosition, toVideoDocument, validateVideoDocument, videoDocumentMirrorSchema, videoDocumentSchema, writeVideoDocumentToDraft };
+export { DEFAULT_UNIT_TIME_MS, IMPLEMENTED_SEMANTIC_OP_KINDS, LANE_KINDS_IN_STACK_ORDER, ManualSyncDoc, MedeoHttpDocStorage, MemoryDocStorage, MengineAckFailedError, MengineDocSession, MengineHttpClient, MengineHttpRequestError, MenginePushRejectedError, MirrorVideoDocumentAdapter, PlainMemoryAdapter, SchemaValidator, SemanticEditor, TIMELINE_SKELETON_DURATION_MS, VIDEO_DOCUMENT_SCHEMA_VERSION, ValidationError, VideoDocumentValidationError, arrangeMainTrackSeamlessly, assertValidVideoDocument, base64ToBytes, buildInitialVideoDocument, buildSpeechHostMap, bytesToBase64, cascadeAfterVideoClipChanges, createEditSandbox, createMirrorVideoDocument, createMirrorVideoDocumentAdapter, createPlainMemoryAdapter, decodeDocVersionMark, derivePositionFromAbs, effectiveVideoClipDurationMs, encodeDocVersionMark, ensureLaneTrack, fillMainTrackTimeGaps, findLaneTrack, fromVideoDocument, generatePartId, getAt, hostForAbsMs, isEmptyVideoClip, isImplementedSemanticOpKind, isMap, laneTrackId, mainTrackRanges, partDurationMs, partUnionSchema, readMainTrackItems, readMengineEventStream, readPart, readPartDurationMs, readVideoDocumentFromDraft, reassignSpeechesToVideoClipsByTime, recalculateTimelineDuration, relativePositionForAbs, replayJournal, resolveAllSpeechOverlaps, resolveSpeechOverlapByShiftingVideos, safeDurationMs, schemas_exports as schemas, snapshotToPlain, solveVideoDocument, speedOf, syncAggregatedClipsTimePosition, toVideoDocument, validateVideoDocument, videoDocumentMirrorSchema, videoDocumentSchema, writeVideoDocumentToDraft };

package/dist/schemas.d.ts

@@ -1,2 +1,2 @@
-import { $ as changeSpeechVoiceInputSchema, A as MoveVideoClipsInput, B as MoveSpeechesInput, C as setBgmInputSchema, D as replaceVideoClipSequenceInputSchema, E as anchoredReplacePolicySchema, F as moveAnchorSchema, G as deleteVideoClipsInputSchema, H as AnchoredDeletePolicy, I as moveVideoClipsByAnchorInputSchema, J as DeleteBgmInput, K as DeleteSpeechesInput, L as movedClipAnchoredPolicySchema, M as MoveAnchor, N as MoveVideoClipsByAnchorInput, O as ReplaceVideoClipContentInput, P as MovedClipAnchoredPolicy, Q as changeSpeechScriptInputSchema, R as MoveBgmInput, S as SetBgmInput, T as ReplaceVideoClipSequenceInput, U as DeleteVideoClipsInput, V as moveSpeechesInputSchema, W as anchoredDeletePolicySchema, X as ChangeSpeechScriptInput, Y as deleteBgmInputSchema, Z as ChangeSpeechVoiceInput, _ as setVideoClipSpeedShiftInputSchema, a as speechAssetsSchema, at as adjustSpeechVolumeInputSchema, b as SetCaptionStyleInput, c as mediaIdSchema, ct as AddVideoClipsInput, d as speechIdsSchema, dt as addSpeechesInputSchema, et as AdjustVideoClipVolumeInput, f as speedShiftSchema, g as SetVideoClipSpeedShiftInput, h as volumeSchema, i as SpeechAssets, it as AdjustSpeechVolumeInput, j as moveVideoClipsInputSchema, k as replaceVideoClipContentInputSchema, l as positiveMsSchema, lt as addVideoClipsInputSchema, m as voiceSchema, n as CaptionAsset, nt as AdjustVideoClipDurationInput, o as clipIdSchema, ot as AdjustBgmVolumeInput, p as timelineMsSchema, q as deleteSpeechesInputSchema, r as SpeechAsset, rt as adjustVideoClipDurationInputSchema, s as clipIdsSchema, st as adjustBgmVolumeInputSchema, tt as adjustVideoClipVolumeInputSchema, u as speechIdSchema, ut as AddSpeechesInput, v as SetCaptionVisibilityInput, w as AnchoredReplacePolicy, x as setCaptionStyleInputSchema, y as setCaptionVisibilityInputSchema, z as moveBgmInputSchema } from "./index-CiDSfSMi.js";
-export { AddSpeechesInput, AddVideoClipsInput, AdjustBgmVolumeInput, AdjustSpeechVolumeInput, AdjustVideoClipDurationInput, AdjustVideoClipVolumeInput, AnchoredDeletePolicy, AnchoredReplacePolicy, CaptionAsset, ChangeSpeechScriptInput, ChangeSpeechVoiceInput, DeleteBgmInput, DeleteSpeechesInput, DeleteVideoClipsInput, MoveAnchor, MoveBgmInput, MoveSpeechesInput, MoveVideoClipsByAnchorInput, MoveVideoClipsInput, MovedClipAnchoredPolicy, ReplaceVideoClipContentInput, ReplaceVideoClipSequenceInput, SetBgmInput, SetCaptionStyleInput, SetCaptionVisibilityInput, SetVideoClipSpeedShiftInput, SpeechAsset, SpeechAssets, addSpeechesInputSchema, addVideoClipsInputSchema, adjustBgmVolumeInputSchema, adjustSpeechVolumeInputSchema, adjustVideoClipDurationInputSchema, adjustVideoClipVolumeInputSchema, anchoredDeletePolicySchema, anchoredReplacePolicySchema, changeSpeechScriptInputSchema, changeSpeechVoiceInputSchema, clipIdSchema, clipIdsSchema, deleteBgmInputSchema, deleteSpeechesInputSchema, deleteVideoClipsInputSchema, mediaIdSchema, moveAnchorSchema, moveBgmInputSchema, moveSpeechesInputSchema, moveVideoClipsByAnchorInputSchema, moveVideoClipsInputSchema, movedClipAnchoredPolicySchema, positiveMsSchema, replaceVideoClipContentInputSchema, replaceVideoClipSequenceInputSchema, setBgmInputSchema, setCaptionStyleInputSchema, setCaptionVisibilityInputSchema, setVideoClipSpeedShiftInputSchema, speechAssetsSchema, speechIdSchema, speechIdsSchema, speedShiftSchema, timelineMsSchema, voiceSchema, volumeSchema };
+import { $ as adjustVideoClipVolumeInputSchema, A as MoveVideoClipsInput, B as AnchoredDeletePolicy, C as setBgmInputSchema, D as replaceVideoClipSequenceInputSchema, E as anchoredReplacePolicySchema, F as moveAnchorSchema, G as deleteSpeechesInputSchema, H as anchoredDeletePolicySchema, I as moveVideoClipsByAnchorInputSchema, J as ChangeSpeechScriptInput, K as DeleteBgmInput, L as movedClipAnchoredPolicySchema, M as MoveAnchor, N as MoveVideoClipsByAnchorInput, O as ReplaceVideoClipContentInput, P as MovedClipAnchoredPolicy, Q as AdjustVideoClipVolumeInput, R as MoveSpeechesInput, S as SetBgmInput, T as ReplaceVideoClipSequenceInput, U as deleteVideoClipsInputSchema, V as DeleteVideoClipsInput, W as DeleteSpeechesInput, X as changeSpeechScriptInputSchema, Y as ChangeSpeechVoiceInput, Z as changeSpeechVoiceInputSchema, _ as setVideoClipSpeedShiftInputSchema, a as speechAssetsSchema, at as adjustBgmVolumeInputSchema, b as SetCaptionStyleInput, c as mediaIdSchema, ct as AddSpeechesInput, d as speechIdsSchema, et as AdjustVideoClipDurationInput, f as speedShiftSchema, g as SetVideoClipSpeedShiftInput, h as volumeSchema, i as SpeechAssets, it as AdjustBgmVolumeInput, j as moveVideoClipsInputSchema, k as replaceVideoClipContentInputSchema, l as positiveMsSchema, lt as addSpeechesInputSchema, m as voiceSchema, n as CaptionAsset, nt as AdjustSpeechVolumeInput, o as clipIdSchema, ot as AddVideoClipsInput, p as timelineMsSchema, q as deleteBgmInputSchema, r as SpeechAsset, rt as adjustSpeechVolumeInputSchema, s as clipIdsSchema, st as addVideoClipsInputSchema, tt as adjustVideoClipDurationInputSchema, u as speechIdSchema, v as SetCaptionVisibilityInput, w as AnchoredReplacePolicy, x as setCaptionStyleInputSchema, y as setCaptionVisibilityInputSchema, z as moveSpeechesInputSchema } from "./index-InEWX9rk.js";
+export { AddSpeechesInput, AddVideoClipsInput, AdjustBgmVolumeInput, AdjustSpeechVolumeInput, AdjustVideoClipDurationInput, AdjustVideoClipVolumeInput, AnchoredDeletePolicy, AnchoredReplacePolicy, CaptionAsset, ChangeSpeechScriptInput, ChangeSpeechVoiceInput, DeleteBgmInput, DeleteSpeechesInput, DeleteVideoClipsInput, MoveAnchor, MoveSpeechesInput, MoveVideoClipsByAnchorInput, MoveVideoClipsInput, MovedClipAnchoredPolicy, ReplaceVideoClipContentInput, ReplaceVideoClipSequenceInput, SetBgmInput, SetCaptionStyleInput, SetCaptionVisibilityInput, SetVideoClipSpeedShiftInput, SpeechAsset, SpeechAssets, addSpeechesInputSchema, addVideoClipsInputSchema, adjustBgmVolumeInputSchema, adjustSpeechVolumeInputSchema, adjustVideoClipDurationInputSchema, adjustVideoClipVolumeInputSchema, anchoredDeletePolicySchema, anchoredReplacePolicySchema, changeSpeechScriptInputSchema, changeSpeechVoiceInputSchema, clipIdSchema, clipIdsSchema, deleteBgmInputSchema, deleteSpeechesInputSchema, deleteVideoClipsInputSchema, mediaIdSchema, moveAnchorSchema, moveSpeechesInputSchema, moveVideoClipsByAnchorInputSchema, moveVideoClipsInputSchema, movedClipAnchoredPolicySchema, positiveMsSchema, replaceVideoClipContentInputSchema, replaceVideoClipSequenceInputSchema, setBgmInputSchema, setCaptionStyleInputSchema, setCaptionVisibilityInputSchema, setVideoClipSpeedShiftInputSchema, speechAssetsSchema, speechIdSchema, speechIdsSchema, speedShiftSchema, timelineMsSchema, voiceSchema, volumeSchema };

package/dist/schemas.js

@@ -45,28 +45,24 @@ const voiceSchema = z.object({
 * stable speech + caption parts and writes them as authoritative facts. No
 * cascade runs on write — the projection derives absolute positions on read.
 *
-* `anchor_part_id` states the root relationship explicitly: a video/BGM id
-* creates an anchored speech; `null` creates an absolute audio root.
-* `offset_ms` is relative to the anchor or absolute for a root. Captions always
-* anchor to their speech via `start_ms`.
+* Each speech carries the anchoring fact directly (RFC 02 §4): the host video
+* clip `anchor_part_id` and the `offset_ms` within it. The upstream caller
+* already knows which clip a speech attaches to, so the op writes
+* `{ mode:'anchored', anchorPartId, offsetMs }` verbatim — no write-time
+* host-picking. Captions anchor to their speech via the caption part's
+* `start_ms` (offset within the speech).
 */
 const speechAssetSchema = z.object({
 	speech_id: speechIdSchema.describe("The speech part ID (= side-effect speech_parts[].id)"),
-	anchor_part_id: z.string().min(1).nullable().describe("Video/BGM parent part ID, or null when the speech itself is the root"),
-	offset_ms: timelineMsSchema.describe("Relative offset from the parent, or absolute start for a root speech"),
+	anchor_part_id: clipIdSchema.describe("Host video clip part ID the speech anchors to (RFC 02 §4)"),
+	offset_ms: timelineMsSchema.describe("Offset within the host clip (speech.abs = host.abs + offset_ms)"),
 	audio_storage_key: z.string().min(1),
 	duration_ms: positiveMsSchema,
 	audio_script: z.string(),
 	volume: volumeSchema,
-	voice: voiceSchema.optional(),
+	voice: voiceSchema,
 	origin_speech_id: z.string().min(1),
 	caption_ids: z.array(z.string().min(1)).describe("Caption part IDs owned by this speech")
-}).superRefine((speech, ctx) => {
-	if (speech.audio_storage_key === speech.speech_id || speech.audio_storage_key === speech.origin_speech_id) ctx.addIssue({
-		code: z.ZodIssueCode.custom,
-		path: ["audio_storage_key"],
-		message: "audio_storage_key must be the materialized storage key, not a speech id"
-	});
 });
 const captionAssetSchema = z.object({
 	caption_id: z.string().min(1).describe("The caption part ID (= side-effect created_caption_parts[].id)"),
@@ -79,21 +75,6 @@ const captionAssetSchema = z.object({
 const speechAssetsSchema = z.object({
 	speeches: z.array(speechAssetSchema).min(1).describe("Materialized speech parts to write"),
 	captions: z.array(captionAssetSchema).describe("Materialized caption parts owned by the speeches")
-}).superRefine((assets, ctx) => {
-	const ownerByStorageKey = /* @__PURE__ */ new Map();
-	for (const [index, speech] of assets.speeches.entries()) {
-		const owner = ownerByStorageKey.get(speech.audio_storage_key);
-		if (owner != null && owner !== speech.speech_id) ctx.addIssue({
-			code: z.ZodIssueCode.custom,
-			path: [
-				"speeches",
-				index,
-				"audio_storage_key"
-			],
-			message: "one materialized audio_storage_key must belong to one speech part"
-		});
-		ownerByStorageKey.set(speech.audio_storage_key, speech.speech_id);
-	}
 });
 //#endregion
 //#region src/editor/schemas/add-speeches.ts
@@ -265,9 +246,6 @@ const moveSpeechesInputSchema = z.object({ speeches: z.array(z.object({
 	new_start_ms: timelineMsSchema.describe("New absolute start time on the timeline")
 })).min(1).describe("Speeches to move to new positions") });
 //#endregion
-//#region src/editor/schemas/move-bgm.ts
-const moveBgmInputSchema = z.object({ new_start_ms: timelineMsSchema.describe("New absolute BGM start on the timeline") });
-//#endregion
 //#region src/editor/schemas/move-video-clips-by-anchor.ts
 /**
 * Where the moved block lands on the main track.
@@ -517,7 +495,6 @@ var schemas_exports = /* @__PURE__ */ __exportAll({
 	deleteVideoClipsInputSchema: () => deleteVideoClipsInputSchema,
 	mediaIdSchema: () => mediaIdSchema,
 	moveAnchorSchema: () => moveAnchorSchema,
-	moveBgmInputSchema: () => moveBgmInputSchema,
 	moveSpeechesInputSchema: () => moveSpeechesInputSchema,
 	moveVideoClipsByAnchorInputSchema: () => moveVideoClipsByAnchorInputSchema,
 	moveVideoClipsInputSchema: () => moveVideoClipsInputSchema,
@@ -538,4 +515,4 @@ var schemas_exports = /* @__PURE__ */ __exportAll({
 	volumeSchema: () => volumeSchema
 });
 //#endregion
-export { addSpeechesInputSchema, addVideoClipsInputSchema, adjustBgmVolumeInputSchema, adjustSpeechVolumeInputSchema, adjustVideoClipDurationInputSchema, adjustVideoClipVolumeInputSchema, anchoredDeletePolicySchema, anchoredReplacePolicySchema, changeSpeechScriptInputSchema, changeSpeechVoiceInputSchema, clipIdSchema, clipIdsSchema, deleteBgmInputSchema, deleteSpeechesInputSchema, deleteVideoClipsInputSchema, mediaIdSchema, moveAnchorSchema, moveBgmInputSchema, moveSpeechesInputSchema, moveVideoClipsByAnchorInputSchema, moveVideoClipsInputSchema, movedClipAnchoredPolicySchema, positiveMsSchema, replaceVideoClipContentInputSchema, replaceVideoClipSequenceInputSchema, setBgmInputSchema, setCaptionStyleInputSchema, setCaptionVisibilityInputSchema, setVideoClipSpeedShiftInputSchema, speechAssetsSchema, speechIdSchema, speechIdsSchema, speedShiftSchema, schemas_exports as t, timelineMsSchema, voiceSchema, volumeSchema };
+export { addSpeechesInputSchema, addVideoClipsInputSchema, adjustBgmVolumeInputSchema, adjustSpeechVolumeInputSchema, adjustVideoClipDurationInputSchema, adjustVideoClipVolumeInputSchema, anchoredDeletePolicySchema, anchoredReplacePolicySchema, changeSpeechScriptInputSchema, changeSpeechVoiceInputSchema, clipIdSchema, clipIdsSchema, deleteBgmInputSchema, deleteSpeechesInputSchema, deleteVideoClipsInputSchema, mediaIdSchema, moveAnchorSchema, moveSpeechesInputSchema, moveVideoClipsByAnchorInputSchema, moveVideoClipsInputSchema, movedClipAnchoredPolicySchema, positiveMsSchema, replaceVideoClipContentInputSchema, replaceVideoClipSequenceInputSchema, setBgmInputSchema, setCaptionStyleInputSchema, setCaptionVisibilityInputSchema, setVideoClipSpeedShiftInputSchema, speechAssetsSchema, speechIdSchema, speechIdsSchema, speedShiftSchema, schemas_exports as t, timelineMsSchema, voiceSchema, volumeSchema };

package/dist/testing.d.ts

@@ -1,4 +1,4 @@
-import { c as MirrorVideoDocumentAdapter, ct as VideoDraft } from "./index-DkJ0JK2t.js";
+import { c as MirrorVideoDocumentAdapter, ct as VideoDraft } from "./index-BO2fSUKi.js";
 
 //#region src/testing/in-memory-mengine-server.d.ts
 /**

package/dist/testing.js

@@ -1,4 +1,4 @@
-import { G as base64ToBytes, K as bytesToBase64, o as createMirrorVideoDocumentAdapter, z as VIDEO_DOCUMENT_SCHEMA_VERSION } from "./document-C-h5-rCG.js";
+import { L as VIDEO_DOCUMENT_SCHEMA_VERSION, U as base64ToBytes, W as bytesToBase64, o as createMirrorVideoDocumentAdapter } from "./document-49OdvviW.js";
 import { t as classifyUpdate } from "./loro-relay-doc-ssdYpuef.js";
 import { LoroDoc, VersionVector, encodeFrontiers } from "loro-crdt";
 //#region src/testing/in-memory-mengine-server.ts

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mengine/medeo-client",
-  "version": "1.2.1-alpha.4",
+  "version": "1.2.1-alpha.6",
   "license": "UNLICENSED",
   "repository": {
     "type": "git",
@@ -27,9 +27,9 @@
     "immer": "^10.2.0",
     "loro-mirror": "^2.2.0",
     "zod": "^4.4.3",
-    "@mengine/storage": "1.2.1-alpha.4",
-    "@mengine/utils": "1.2.1-alpha.4",
-    "@mengine/sync": "1.2.1-alpha.4"
+    "@mengine/sync": "1.2.1-alpha.6",
+    "@mengine/utils": "1.2.1-alpha.6",
+    "@mengine/storage": "1.2.1-alpha.6"
   },
   "devDependencies": {
     "@types/node": "^25.9.1",