@elizaos/plugin-local-inference 2.0.0-beta.1 → 2.0.3-beta.3

package/dist/services/voice/transcriber.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"transcriber.d.ts","sourceRoot":"","sources":["../../../src/services/voice/transcriber.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA+BG;AAEH,OAAO,KAAK,EACX,2BAA2B,EAC3B,iBAAiB,EACjB,MAAM,gBAAgB,CAAC;AACxB,OAAO,KAAK,EACX,QAAQ,EACR,oBAAoB,EAEpB,wBAAwB,EACxB,gBAAgB,EAEhB,cAAc,EACd,gBAAgB,EAChB,YAAY,EACZ,iBAAiB,EACjB,MAAM,SAAS,CAAC;AAEjB,0EAA0E;AAC1E,eAAO,MAAM,eAAe,QAAS,CAAC;AAEtC;;;;;GAKG;AACH,qBAAa,mBAAoB,SAAQ,KAAK;gBACjC,OAAO,EAAE,MAAM;CAI3B;AAED,MAAM,MAAM,oBAAoB,GAAG,MAAM,GAAG,OAAO,GAAG,WAAW,CAAC;AAElE,wBAAgB,6BAA6B,CAC5C,KAAK,EAAE,MAAM,GAAG,IAAI,GAAG,SAAS,GAC9B,oBAAoB,GAAG,IAAI,CAiB7B;AAED,wBAAgB,+BAA+B,CAC9C,GAAG,GAAE,MAAM,CAAC,UAAwB,GAClC,oBAAoB,GAAG,IAAI,CAE7B;AAcD;;;;;GAKG;AACH,wBAAgB,cAAc,CAC7B,GAAG,EAAE,YAAY,EACjB,QAAQ,EAAE,MAAM,EAChB,MAAM,EAAE,MAAM,GACZ,YAAY,CAad;AAED;;;;;GAKG;AACH,8BAAsB,wBAAyB,YAAW,oBAAoB;IAC7E,OAAO,CAAC,QAAQ,CAAC,SAAS,CAAuC;IACjE,OAAO,CAAC,QAAQ,CAA6B;IAC7C,sEAAsE;IACtE,SAAS,CAAC,WAAW,UAAS;IAC9B,+DAA+D;IAC/D,OAAO,CAAC,YAAY,CAAS;IAC7B,uFAAuF;IACvF,OAAO,CAAC,SAAS,CAAwB;IACzC,OAAO,CAAC,QAAQ,CAAC,gBAAgB,CAAkB;IACnD,OAAO,CAAC,QAAQ,CAA6B;IAC7C,OAAO,CAAC,QAAQ,CAAS;gBAEb,GAAG,CAAC,EAAE,cAAc,EAAE,QAAQ,GAAE,0BAA+B;IAQ3E,EAAE,CAAC,QAAQ,EAAE,wBAAwB,GAAG,MAAM,IAAI;IAKlD;;;;;;OAMG;IACH,mBAAmB,CAAC,QAAQ,EAAE,0BAA0B,GAAG,IAAI;IAI/D,IAAI,CAAC,KAAK,EAAE,QAAQ,GAAG,IAAI;IAkBrB,KAAK,IAAI,OAAO,CAAC,gBAAgB,CAAC;IAWxC,OAAO,IAAI,IAAI;IASf,6EAA6E;IAC7E,SAAS,CAAC,QAAQ,CAAC,OAAO,CAAC,KAAK,EAAE,QAAQ,GAAG,IAAI;IACjD,4FAA4F;IAC5F,SAAS,CAAC,QAAQ,CAAC,OAAO,IAAI,OAAO,CAAC,gBAAgB,CAAC;IACvD,+CAA+C;IAC/C,SAAS,CAAC,QAAQ,CAAC,SAAS,IAAI,IAAI;IAEpC,OAAO,CAAC,kBAAkB;IAU1B,OAAO,CAAC,eAAe;IAUvB,sFAAsF;IACtF,SAAS,CAAC,WAAW,CAAC,MAAM,EAAE,gBAAgB,GAAG,IAAI;IAYrD,OAAO,CAAC,YAAY;IAsCpB,OAAO,CAAC,IAAI;IAIZ,OAAO,CAAC,UAAU;CAsBlB;AAED,MAAM,WAAW,0BAA0B;IAC1C,MAAM,CAAC,EAAE,gBAAgB,CAAC;IAC1B,OAAO,CAAC,EAAE,YAAY,CAAC;IACvB,IAAI,CAAC,EAAE,iBAAiB,CAAC;CACzB;AAMD;;;;;GAKG;AACH,wBAAgB,uBAAuB,CACtC,GAAG,EAAE,iBAAiB,GAAG,IAAI,GAAG,SAAS,GACvC,OAAO,CAGT;AAED;;;;;;;;;;;;GAYG;AACH,qBAAa,uBAAwB,SAAQ,wBAAwB;IACpE,OAAO,CAAC,QAAQ,CAAC,GAAG,CAAoB;IACxC,OAAO,CAAC,QAAQ,CAAC,UAAU,CAAoC;IAC/D,gFAAgF;IAChF,OAAO,CAAC,QAAQ,CAAC,SAAS,CAAS;IACnC,OAAO,CAAC,MAAM,CAAuB;gBAEzB,IAAI,EAAE;QACjB,GAAG,EAAE,iBAAiB,CAAC;QACvB,UAAU,EAAE,MAAM,2BAA2B,CAAC;QAC9C,GAAG,CAAC,EAAE,cAAc,CAAC;QACrB,QAAQ,CAAC,EAAE,0BAA0B,CAAC;QACtC,MAAM,CAAC,EAAE,gBAAgB,CAAC;QAC1B,uEAAuE;QACvE,SAAS,CAAC,EAAE,MAAM,CAAC;KACnB;IAeD,OAAO,CAAC,YAAY;IASpB,SAAS,CAAC,OAAO,CAAC,KAAK,EAAE,QAAQ,GAAG,IAAI;cAWxB,OAAO,IAAI,OAAO,CAAC,gBAAgB,CAAC;IAcpD,SAAS,CAAC,SAAS,IAAI,IAAI;CAM3B;AAMD,MAAM,WAAW,0BAA0B;IAC1C,GAAG,EAAE,iBAAiB,CAAC;IACvB,UAAU,EAAE,MAAM,2BAA2B,CAAC;IAC9C,GAAG,CAAC,EAAE,cAAc,CAAC;IACrB,QAAQ,CAAC,EAAE,0BAA0B,CAAC;IACtC,MAAM,CAAC,EAAE,gBAAgB,CAAC;IAC1B,8FAA8F;IAC9F,aAAa,CAAC,EAAE,MAAM,CAAC;IACvB,kFAAkF;IAClF,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,wFAAwF;IACxF,WAAW,CAAC,EAAE,MAAM,CAAC;CACrB;AAED;;;;;;;;;;;;;;;;;;;GAmBG;AACH,qBAAa,mBAAoB,SAAQ,wBAAwB;IAChE,OAAO,CAAC,QAAQ,CAAC,GAAG,CAAoB;IACxC,OAAO,CAAC,QAAQ,CAAC,UAAU,CAAoC;IAC/D,OAAO,CAAC,QAAQ,CAAC,aAAa,CAAS;IACvC,OAAO,CAAC,QAAQ,CAAC,cAAc,CAAS;IACxC,OAAO,CAAC,QAAQ,CAAC,WAAW,CAAS;IACrC,qEAAqE;IACrE,OAAO,CAAC,GAAG,CAAqC;IAChD,wDAAwD;IACxD,OAAO,CAAC,gBAAgB,CAAK;IAC7B,sDAAsD;IACtD,OAAO,CAAC,SAAS,CAAM;IACvB,yEAAyE;IACzE,OAAO,CAAC,YAAY,CAAK;IACzB,sFAAsF;IACtF,OAAO,CAAC,WAAW,CAAoC;gBAE3C,IAAI,EAAE,0BAA0B;IAe5C,OAAO,CAAC,YAAY;IAWpB,SAAS,CAAC,OAAO,CAAC,KAAK,EAAE,QAAQ,GAAG,IAAI;cAQxB,OAAO,IAAI,OAAO,CAAC,gBAAgB,CAAC;IAQpD,SAAS,CAAC,SAAS,IAAI,IAAI;IAI3B,OAAO,CAAC,YAAY;IAOpB,OAAO,CAAC,cAAc;YAIR,SAAS;CA+BvB;AAqCD,MAAM,WAAW,iCAAiC;IACjD,gFAAgF;IAChF,GAAG,CAAC,EAAE,iBAAiB,GAAG,IAAI,CAAC;IAC/B,gFAAgF;IAChF,UAAU,CAAC,EAAE,MAAM,2BAA2B,CAAC;IAC/C;;;;OAIG;IACH,gBAAgB,CAAC,EAAE,OAAO,CAAC;IAC3B,8CAA8C;IAC9C,GAAG,CAAC,EAAE,cAAc,CAAC;IACrB,6EAA6E;IAC7E,QAAQ,CAAC,EAAE,0BAA0B,CAAC;IACtC,mDAAmD;IACnD,MAAM,CAAC,EAAE,gBAAgB,CAAC;IAC1B,6EAA6E;IAC7E,QAAQ,CAAC,EAAE,IAAI,CAAC,0BAA0B,EAAE,KAAK,GAAG,YAAY,CAAC,CAAC;IAClE;;;;;OAKG;IACH,MAAM,CAAC,EAAE,oBAAoB,CAAC;CAC9B;AAED;;;;;;;;;;;GAWG;AACH,wBAAgB,0BAA0B,CACzC,IAAI,GAAE,iCAAsC,GAC1C,oBAAoB,CAsDtB"}

package/dist/services/voice/transcript-knowledge.d.ts

@@ -0,0 +1,37 @@
+/**
+ * Transcript → knowledge (documents store) mapping (#8789 transcripts).
+ *
+ * A finished transcript is mirrored into the documents/knowledge store as a
+ * private, searchable text item so it is parsable/readable from the Knowledge
+ * view, surfaced by the documents provider, and found by the document `search`
+ * action — exactly as a user would expect "search my transcripts" to work. The
+ * structured record (audio + word timings + diarization) stays the source of
+ * truth in the transcript store; this is the denormalized search copy, linked
+ * back by `metadata.transcriptId`.
+ *
+ * Pure: it builds the documents-store *fields* from a shared `Transcript`; the
+ * service combines them with the runtime UUIDs (world/room/entity/client ids)
+ * and `addedFrom` to call `DocumentService.addDocument`.
+ */
+import type { Transcript, TranscriptScope } from "@elizaos/shared/transcripts";
+/** The documents-store fields derived from a transcript (sans runtime UUIDs). */
+export interface TranscriptKnowledgePayload {
+    /** Plain, speaker-labeled transcript text — the searchable + chunked body. */
+    content: string;
+    /** Suggested filename (slugified title + `.txt`). */
+    filename: string;
+    contentType: string;
+    scope: TranscriptScope;
+    /** Metadata merged onto the document — tags + the link back to the record. */
+    metadata: Record<string, unknown>;
+}
+/** Tag every mirrored transcript carries so it's filterable as a transcript. */
+export declare const TRANSCRIPT_DOCUMENT_TAG = "transcript";
+/**
+ * Build the documents-store fields for mirroring a transcript into knowledge.
+ * The body is the speaker-labeled plain text (what gets embedded + searched);
+ * `metadata.transcriptId` links the knowledge item back to the rich record so a
+ * Knowledge-view consumer can open it in the player.
+ */
+export declare function transcriptKnowledgePayload(transcript: Transcript): TranscriptKnowledgePayload;
+//# sourceMappingURL=transcript-knowledge.d.ts.map

package/dist/services/voice/transcript-knowledge.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"transcript-knowledge.d.ts","sourceRoot":"","sources":["../../../src/services/voice/transcript-knowledge.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;GAcG;AAEH,OAAO,KAAK,EAAE,UAAU,EAAE,eAAe,EAAE,MAAM,6BAA6B,CAAC;AAG/E,iFAAiF;AACjF,MAAM,WAAW,0BAA0B;IAC1C,8EAA8E;IAC9E,OAAO,EAAE,MAAM,CAAC;IAChB,qDAAqD;IACrD,QAAQ,EAAE,MAAM,CAAC;IACjB,WAAW,EAAE,MAAM,CAAC;IACpB,KAAK,EAAE,eAAe,CAAC;IACvB,8EAA8E;IAC9E,QAAQ,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CAClC;AAED,gFAAgF;AAChF,eAAO,MAAM,uBAAuB,eAAe,CAAC;AAYpD;;;;;GAKG;AACH,wBAAgB,0BAA0B,CACzC,UAAU,EAAE,UAAU,GACpB,0BAA0B,CAsB5B"}

package/dist/services/voice/transcript-service.d.ts

@@ -0,0 +1,60 @@
+/**
+ * TranscriptService (#8789) — orchestrates a finished recording into both homes:
+ * the rich {@link TranscriptStore} record (source of truth: audio + word-timed
+ * diarized segments) AND a searchable knowledge/documents mirror, linked by
+ * `knowledgeDocumentId`. Create mirrors first (best-effort — a search-index
+ * failure must not lose the recording), then persists the record carrying the
+ * link. Delete removes the mirror too.
+ */
+import { type UUID } from "@elizaos/core";
+import type { Transcript, TranscriptSegment, TranscriptSummary } from "@elizaos/shared/transcripts";
+import { type TranscriptStoreRuntime } from "./transcript-store";
+/** Store runtime + service resolution (the real IAgentRuntime satisfies it). */
+export interface TranscriptServiceRuntime extends TranscriptStoreRuntime {
+    getService<T>(name: string): T | null;
+}
+export interface CreateTranscriptInput {
+    worldId: UUID;
+    roomId: UUID;
+    /** The owner/speaker entity the recording is attributed to. */
+    entityId: UUID;
+    transcript: Transcript;
+}
+export interface UpdateTranscriptInput {
+    /** Ids the knowledge re-mirror is attributed to (default to the agent ctx). */
+    worldId: UUID;
+    roomId: UUID;
+    entityId: UUID;
+    /** The user edit: a new title and/or replacement segments. */
+    patch: {
+        title?: string;
+        segments?: TranscriptSegment[];
+    };
+}
+export declare class TranscriptService {
+    private readonly runtime;
+    private readonly store;
+    constructor(runtime: TranscriptServiceRuntime);
+    /** Mirror the transcript text into knowledge, then persist the record. */
+    create(input: CreateTranscriptInput): Promise<Transcript>;
+    list(roomId?: UUID, limit?: number): Promise<TranscriptSummary[]>;
+    get(id: UUID): Promise<Transcript | null>;
+    /**
+     * Apply a user edit to a transcript record (title and/or segment text) and
+     * persist it, re-mirroring the new text into the knowledge index so search
+     * reflects the correction. Returns the updated record, or null if no record
+     * with that id exists. Timing/speaker metadata is re-derived from the edited
+     * segments; the knowledge mirror is replaced (old doc removed best-effort).
+     */
+    update(id: UUID, input: UpdateTranscriptInput): Promise<Transcript | null>;
+    /** Delete the record + its knowledge mirror (mirror removal best-effort). */
+    delete(id: UUID): Promise<void>;
+    /**
+     * Best-effort mirror into the documents store. Returns the stored document id
+     * to link, or undefined when no documents service is loaded or the mirror
+     * fails — the recording still persists either way (the link is a secondary
+     * search index, never the source of truth).
+     */
+    private mirrorToKnowledge;
+}
+//# sourceMappingURL=transcript-service.d.ts.map

package/dist/services/voice/transcript-service.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"transcript-service.d.ts","sourceRoot":"","sources":["../../../src/services/voice/transcript-service.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AAEH,OAAO,EAAU,KAAK,IAAI,EAAE,MAAM,eAAe,CAAC;AAClD,OAAO,KAAK,EACX,UAAU,EAEV,iBAAiB,EACjB,iBAAiB,EACjB,MAAM,6BAA6B,CAAC;AAMrC,OAAO,EAEN,KAAK,sBAAsB,EAC3B,MAAM,oBAAoB,CAAC;AAkB5B,gFAAgF;AAChF,MAAM,WAAW,wBAAyB,SAAQ,sBAAsB;IACvE,UAAU,CAAC,CAAC,EAAE,IAAI,EAAE,MAAM,GAAG,CAAC,GAAG,IAAI,CAAC;CACtC;AAED,MAAM,WAAW,qBAAqB;IACrC,OAAO,EAAE,IAAI,CAAC;IACd,MAAM,EAAE,IAAI,CAAC;IACb,+DAA+D;IAC/D,QAAQ,EAAE,IAAI,CAAC;IACf,UAAU,EAAE,UAAU,CAAC;CACvB;AAED,MAAM,WAAW,qBAAqB;IACrC,+EAA+E;IAC/E,OAAO,EAAE,IAAI,CAAC;IACd,MAAM,EAAE,IAAI,CAAC;IACb,QAAQ,EAAE,IAAI,CAAC;IACf,8DAA8D;IAC9D,KAAK,EAAE;QAAE,KAAK,CAAC,EAAE,MAAM,CAAC;QAAC,QAAQ,CAAC,EAAE,iBAAiB,EAAE,CAAA;KAAE,CAAC;CAC1D;AAED,qBAAa,iBAAiB;IAGjB,OAAO,CAAC,QAAQ,CAAC,OAAO;IAFpC,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAkB;gBAEX,OAAO,EAAE,wBAAwB;IAI9D,0EAA0E;IACpE,MAAM,CAAC,KAAK,EAAE,qBAAqB,GAAG,OAAO,CAAC,UAAU,CAAC;IAc/D,IAAI,CAAC,MAAM,CAAC,EAAE,IAAI,EAAE,KAAK,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC,iBAAiB,EAAE,CAAC;IAIjE,GAAG,CAAC,EAAE,EAAE,IAAI,GAAG,OAAO,CAAC,UAAU,GAAG,IAAI,CAAC;IAIzC;;;;;;OAMG;IACG,MAAM,CACX,EAAE,EAAE,IAAI,EACR,KAAK,EAAE,qBAAqB,GAC1B,OAAO,CAAC,UAAU,GAAG,IAAI,CAAC;IAwC7B,6EAA6E;IACvE,MAAM,CAAC,EAAE,EAAE,IAAI,GAAG,OAAO,CAAC,IAAI,CAAC;IAmBrC;;;;;OAKG;YACW,iBAAiB;CAkC/B"}

package/dist/services/voice/transcript-store.d.ts

@@ -0,0 +1,64 @@
+/**
+ * Transcript store (#8789 transcripts) — persistence for the rich transcript
+ * record (audio URL + word-timed diarized segments).
+ *
+ * Reuses the runtime's proven `memories` partition mechanism (exactly how the
+ * documents store works) rather than a new table/migration: each transcript is
+ * one memory row in the `"transcripts"` partition, with the full {@link Transcript}
+ * in `content.transcript`. The player loads a whole record by id and the list
+ * reads recent rows — no querying INSIDE segments is needed, because search is
+ * served by the knowledge mirror (see `transcript-knowledge.ts`). A custom
+ * `metadata.type` keeps it clear of the document/fragment CHECK constraints.
+ */
+import type { Memory, MemoryMetadata, UUID } from "@elizaos/core";
+import type { Transcript, TranscriptSummary } from "@elizaos/shared/transcripts";
+/** The `type` column partition transcripts live in (sibling to "messages"). */
+export declare const TRANSCRIPTS_TABLE = "transcripts";
+/** `metadata.type` marker — NOT "document"/"fragment", so no CHECK fires. */
+export declare const TRANSCRIPT_METADATA_TYPE = "transcript";
+/** The subset of `IAgentRuntime` the store needs (real runtime satisfies it). */
+export interface TranscriptStoreRuntime {
+    agentId: UUID;
+    createMemory(memory: Memory, tableName: string, unique?: boolean): Promise<UUID>;
+    getMemories(params: {
+        tableName: string;
+        roomId?: UUID;
+        count?: number;
+        orderBy?: "createdAt";
+        orderDirection?: "asc" | "desc";
+    }): Promise<Memory[]>;
+    getMemoryById(id: UUID): Promise<Memory | null>;
+    updateMemory(memory: Partial<Memory> & {
+        id: UUID;
+        metadata?: MemoryMetadata;
+    }): Promise<boolean>;
+    deleteMemory(id: UUID): Promise<void>;
+}
+export interface CreateTranscriptInput {
+    roomId: UUID;
+    /** The owner/speaker entity the recording is attributed to. */
+    entityId: UUID;
+    /** The fully-built transcript record (audio + segments + words). */
+    transcript: Transcript;
+}
+/** CRUD for transcript records over the runtime memory partition. */
+export declare class TranscriptStore {
+    private readonly runtime;
+    constructor(runtime: TranscriptStoreRuntime);
+    /** Persist a transcript record; returns it unchanged. */
+    create(input: CreateTranscriptInput): Promise<Transcript>;
+    /** List recent transcripts (newest first) as compact summaries. */
+    list(roomId?: UUID, limit?: number): Promise<TranscriptSummary[]>;
+    /** Load one full transcript record by id. */
+    get(id: UUID): Promise<Transcript | null>;
+    /**
+     * Overwrite an existing transcript record in place (same id/row) — used when
+     * the user edits the transcript text. Re-derives the preview text body and
+     * the timing/speaker metadata from the updated record so generic memory
+     * consumers and the list stay consistent. Returns the record as stored.
+     */
+    update(transcript: Transcript): Promise<Transcript>;
+    /** Delete a transcript record (the knowledge mirror is removed separately). */
+    delete(id: UUID): Promise<void>;
+}
+//# sourceMappingURL=transcript-store.d.ts.map

package/dist/services/voice/transcript-store.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"transcript-store.d.ts","sourceRoot":"","sources":["../../../src/services/voice/transcript-store.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;GAWG;AAEH,OAAO,KAAK,EAAE,MAAM,EAAE,cAAc,EAAE,IAAI,EAAE,MAAM,eAAe,CAAC;AAClE,OAAO,KAAK,EACX,UAAU,EACV,iBAAiB,EACjB,MAAM,6BAA6B,CAAC;AAMrC,+EAA+E;AAC/E,eAAO,MAAM,iBAAiB,gBAAgB,CAAC;AAC/C,6EAA6E;AAC7E,eAAO,MAAM,wBAAwB,eAAe,CAAC;AAErD,iFAAiF;AACjF,MAAM,WAAW,sBAAsB;IACtC,OAAO,EAAE,IAAI,CAAC;IACd,YAAY,CACX,MAAM,EAAE,MAAM,EACd,SAAS,EAAE,MAAM,EACjB,MAAM,CAAC,EAAE,OAAO,GACd,OAAO,CAAC,IAAI,CAAC,CAAC;IACjB,WAAW,CAAC,MAAM,EAAE;QACnB,SAAS,EAAE,MAAM,CAAC;QAClB,MAAM,CAAC,EAAE,IAAI,CAAC;QACd,KAAK,CAAC,EAAE,MAAM,CAAC;QACf,OAAO,CAAC,EAAE,WAAW,CAAC;QACtB,cAAc,CAAC,EAAE,KAAK,GAAG,MAAM,CAAC;KAChC,GAAG,OAAO,CAAC,MAAM,EAAE,CAAC,CAAC;IACtB,aAAa,CAAC,EAAE,EAAE,IAAI,GAAG,OAAO,CAAC,MAAM,GAAG,IAAI,CAAC,CAAC;IAChD,YAAY,CACX,MAAM,EAAE,OAAO,CAAC,MAAM,CAAC,GAAG;QAAE,EAAE,EAAE,IAAI,CAAC;QAAC,QAAQ,CAAC,EAAE,cAAc,CAAA;KAAE,GAC/D,OAAO,CAAC,OAAO,CAAC,CAAC;IACpB,YAAY,CAAC,EAAE,EAAE,IAAI,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC;CACtC;AAED,MAAM,WAAW,qBAAqB;IACrC,MAAM,EAAE,IAAI,CAAC;IACb,+DAA+D;IAC/D,QAAQ,EAAE,IAAI,CAAC;IACf,oEAAoE;IACpE,UAAU,EAAE,UAAU,CAAC;CACvB;AAeD,qEAAqE;AACrE,qBAAa,eAAe;IACf,OAAO,CAAC,QAAQ,CAAC,OAAO;gBAAP,OAAO,EAAE,sBAAsB;IAE5D,yDAAyD;IACnD,MAAM,CAAC,KAAK,EAAE,qBAAqB,GAAG,OAAO,CAAC,UAAU,CAAC;IA+B/D,mEAAmE;IAC7D,IAAI,CAAC,MAAM,CAAC,EAAE,IAAI,EAAE,KAAK,SAAM,GAAG,OAAO,CAAC,iBAAiB,EAAE,CAAC;IAgBpE,6CAA6C;IACvC,GAAG,CAAC,EAAE,EAAE,IAAI,GAAG,OAAO,CAAC,UAAU,GAAG,IAAI,CAAC;IAK/C;;;;;OAKG;IACG,MAAM,CAAC,UAAU,EAAE,UAAU,GAAG,OAAO,CAAC,UAAU,CAAC;IAuBzD,+EAA+E;IACzE,MAAM,CAAC,EAAE,EAAE,IAAI,GAAG,OAAO,CAAC,IAAI,CAAC;CAGrC"}

package/dist/services/voice/turn-controller.d.ts

@@ -0,0 +1,183 @@
+/**
+ * Voice turn controller — the turn-taking layer above `VoiceScheduler`.
+ *
+ * Sits between W1's `VadEvent` stream + W2's `StreamingTranscriber` events
+ * and the generation path (the runtime message handler / the local engine's
+ * `generate`, which routes through `voiceStreamingArgs` → `VoiceScheduler` →
+ * phrase chunker → TTS). Implements the brief's items A4 / A5 / A6:
+ *
+ *   - `speech-start`           → fire `prewarm(roomId)` immediately (the
+ *                                response-handler stable prefix / MTP
+ *                                slot KV-prefill) — before STT finishes.
+ *   - `speech-pause(ms > thr)` → kick a SPECULATIVE response off W2's
+ *                                current partial transcript. The generate
+ *                                call gets an `AbortSignal`; the in-flight
+ *                                generation is stashed.
+ *   - `speech-active` / a new `speech-start` / VAD re-trigger
+ *                              → ABORT the speculative generation (the abort
+ *                                propagates into `dispatcher.generate`).
+ *   - `speech-end` (no new speech)
+ *                              → finalize: flush the transcriber for the
+ *                                final transcript; if the speculative result
+ *                                is still valid against it, PROMOTE it; else
+ *                                discard and run the real turn on the
+ *                                finalized transcript.
+ *
+ * Barge-in: while the agent is speaking the controller flips
+ * `scheduler.bargeIn.setAgentSpeaking(true)` (and binds the VAD into the
+ * barge-in controller). A provisional `pause-tts` pauses TTS in the
+ * scheduler; a `blip` → `resume-tts`; ASR-confirmed words → `hard-stop` →
+ * the scheduler drains the ring buffer + flushes the chunker, and the
+ * controller aborts the in-flight turn (the same `AbortSignal` the engine
+ * threads into `dispatcher.generate`). The transcriber's `words` event is
+ * wired into `bargeIn.onWordsDetected({wordCount})` so a blip alone only
+ * pauses, but real recognized words hard-stop.
+ *
+ * No fallback sludge: `prewarm` failures surface via `onError`; a speculative
+ * abort is a real `AbortSignal.abort()`, never a swallowed flag.
+ */
+import { type EotClassifier, type VoiceTurnSignal } from "./eot-classifier";
+import type { VoiceScheduler } from "./scheduler";
+import type { StreamingTranscriber, VadEventSource, VoiceInputSource, VoiceSegment, VoiceSpeaker, VoiceTurnMetadata } from "./types";
+/** Outcome of one generation pass (speculative or final). */
+export interface VoiceTurnOutcome {
+    /** The transcript the generation ran against (so the controller can
+     *  decide whether a speculative result is still valid). */
+    transcript: string;
+    /** Voice attribution metadata for the transcript that produced this outcome. */
+    source?: VoiceInputSource;
+    speaker?: VoiceSpeaker;
+    segments?: VoiceSegment[];
+    turn?: VoiceTurnMetadata;
+    /** Final reply text the model produced (already streamed into TTS by the
+     *  generate callee). May be empty for an IGNORE turn. */
+    replyText: string;
+}
+export interface VoiceGenerateRequest {
+    /** Best transcript available at the time the request is issued. */
+    transcript: string;
+    /** Optional source/speaker metadata for attribution-only storage. */
+    source?: VoiceInputSource;
+    speaker?: VoiceSpeaker;
+    segments?: VoiceSegment[];
+    turn?: VoiceTurnMetadata;
+    /** True for the finalized turn (post `speech-end` + `flush()`), false for
+     *  a speculative pass off a partial. */
+    final: boolean;
+    /** Aborted when speech resumes (speculative) or on a hard-stop barge-in. */
+    signal: AbortSignal;
+    /**
+     * Semantic turn-taking signal available at request issue time. Response
+     * handlers can deterministically suppress/accept without waiting for another
+     * model token when this says the next speaker is not the agent.
+     */
+    turnSignal?: VoiceTurnSignal;
+}
+export interface VoiceTurnControllerDeps {
+    /** W1: the authoritative VAD event stream (a `VadDetector` is structurally one of these). */
+    vad: VadEventSource;
+    /** W2: the live streaming transcriber. The controller subscribes to its
+     *  events and calls `flush()` on `speech-end`. */
+    transcriber: StreamingTranscriber;
+    /** W9: the voice scheduler — used for the barge-in controller + agent-speaking flag. */
+    scheduler: VoiceScheduler;
+    /**
+     * KV-prefill / response-handler-prefix prewarm. Called on `speech-start`.
+     * Fire-and-forget; a rejection is surfaced via `onError`, not swallowed.
+     * (In the engine this wraps `engine.prewarmConversation(roomId, ...)` /
+     * `runtime.prewarmResponseHandler(roomId)`.)
+     */
+    prewarm?: (roomId: string) => void | Promise<void>;
+    /** Optional cached first-audio filler played immediately on speech-start. */
+    playFirstAudioFiller?: () => string | null;
+    /**
+     * Semantic turn detector layered with VAD/STT. It runs continuously on
+     * partial transcripts so `speech-pause` can decide whether to speculate or
+     * wait for the user to continue.
+     */
+    turnDetector?: EotClassifier;
+    /**
+     * Run a generation pass. The callee builds the message, calls the runtime
+     * message handler / `useModel`, and streams `replyText` into TTS via the
+     * scheduler. Must honour `request.signal` (abort = stop the LLM/drafter at
+     * the next kernel boundary). Resolves with the produced reply + the
+     * transcript it ran against. Rejecting with the request's `AbortError` is
+     * fine — the controller treats that as "aborted".
+     */
+    generate: (request: VoiceGenerateRequest) => Promise<VoiceTurnOutcome>;
+}
+export interface VoiceTurnControllerConfig {
+    /** Conversation / room id passed to `prewarm` and (implicitly) `generate`. */
+    roomId: string;
+    /**
+     * Minimum `speech-pause` duration before a speculative response is kicked.
+     * Default 300 ms — long enough that mid-sentence breath pauses don't
+     * trigger one, short enough to win latency on a real end-of-utterance.
+     */
+    speculatePauseMs?: number;
+}
+export interface VoiceTurnControllerEvents {
+    /** A speculative generation was started off a partial transcript. */
+    onSpeculativeStart?(transcript: string): void;
+    /** The in-flight speculative generation was aborted (speech resumed). */
+    onSpeculativeAbort?(): void;
+    /** The speculative result was promoted as the turn's answer (it matched the final transcript). */
+    onSpeculativePromoted?(outcome: VoiceTurnOutcome): void;
+    /** A turn finished (promoted speculative OR a fresh final run). */
+    onTurnComplete?(outcome: VoiceTurnOutcome): void;
+    /** `prewarm` rejected, or a `generate` pass rejected with a non-abort error. */
+    onError?(error: Error): void;
+    /** A VAD pause/end was suppressed because semantic turn-taking says user continues. */
+    onTurnSuppressed?(transcript: string, signal: VoiceTurnSignal): void;
+}
+export declare class VoiceTurnController {
+    private readonly deps;
+    private readonly events;
+    private readonly roomId;
+    private readonly speculatePauseMs;
+    private readonly bargeIn;
+    private speculative;
+    /** A finalize() in progress (awaiting `transcriber.flush()` + generate). */
+    private finalizing;
+    private latestPartial;
+    private latestTurnSignal;
+    private turnSignalSequence;
+    private started;
+    private vadUnsub;
+    private transcriberUnsub;
+    private bargeSignalUnsub;
+    private activeFinalController;
+    /** True once `speech-end` ran and finalize is pending/done for this segment. */
+    private segmentEnded;
+    private latestUpdate;
+    constructor(deps: VoiceTurnControllerDeps, config: VoiceTurnControllerConfig, events?: VoiceTurnControllerEvents);
+    /** Subscribe to the VAD + transcriber streams and start turn-taking. Idempotent. */
+    start(): void;
+    /** Detach from the streams and abort any in-flight speculative generation. */
+    stop(): void;
+    private onVadEvent;
+    private onTranscriberEvent;
+    /**
+     * C2 — public idle prewarm entry point. Callers (e.g. the UI when a
+     * conversation opens) invoke this to materialize the KV cache for the
+     * response-handler stable prefix BEFORE the user starts speaking, so the
+     * first speech-start has nothing left to do. Fire-and-forget: the
+     * returned promise is `void` because we don't want callers blocking on
+     * prewarm; failures surface via `onError` exactly like the speech-start
+     * path. Idempotent — repeated calls just re-prewarm.
+     */
+    prewarmOnIdle(): void;
+    private firePrewarm;
+    private playFirstAudioFiller;
+    private maybeStartSpeculative;
+    private startSpeculativeAfterTurnSignal;
+    private startSpeculative;
+    private abortSpeculative;
+    private beginFinalize;
+    private finalize;
+    private runGenerate;
+    private queueTurnSignalRefresh;
+    private ensureTurnSignal;
+    private computeTurnSignal;
+}
+//# sourceMappingURL=turn-controller.d.ts.map

package/dist/services/voice/turn-controller.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"turn-controller.d.ts","sourceRoot":"","sources":["../../../src/services/voice/turn-controller.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAqCG;AAGH,OAAO,EAEN,KAAK,aAAa,EAElB,KAAK,eAAe,EACpB,MAAM,kBAAkB,CAAC;AAC1B,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAClD,OAAO,KAAK,EACX,oBAAoB,EAIpB,cAAc,EACd,gBAAgB,EAChB,YAAY,EACZ,YAAY,EACZ,iBAAiB,EACjB,MAAM,SAAS,CAAC;AAEjB,6DAA6D;AAC7D,MAAM,WAAW,gBAAgB;IAChC;+DAC2D;IAC3D,UAAU,EAAE,MAAM,CAAC;IACnB,gFAAgF;IAChF,MAAM,CAAC,EAAE,gBAAgB,CAAC;IAC1B,OAAO,CAAC,EAAE,YAAY,CAAC;IACvB,QAAQ,CAAC,EAAE,YAAY,EAAE,CAAC;IAC1B,IAAI,CAAC,EAAE,iBAAiB,CAAC;IACzB;6DACyD;IACzD,SAAS,EAAE,MAAM,CAAC;CAClB;AAED,MAAM,WAAW,oBAAoB;IACpC,mEAAmE;IACnE,UAAU,EAAE,MAAM,CAAC;IACnB,qEAAqE;IACrE,MAAM,CAAC,EAAE,gBAAgB,CAAC;IAC1B,OAAO,CAAC,EAAE,YAAY,CAAC;IACvB,QAAQ,CAAC,EAAE,YAAY,EAAE,CAAC;IAC1B,IAAI,CAAC,EAAE,iBAAiB,CAAC;IACzB;4CACwC;IACxC,KAAK,EAAE,OAAO,CAAC;IACf,4EAA4E;IAC5E,MAAM,EAAE,WAAW,CAAC;IACpB;;;;OAIG;IACH,UAAU,CAAC,EAAE,eAAe,CAAC;CAC7B;AAED,MAAM,WAAW,uBAAuB;IACvC,6FAA6F;IAC7F,GAAG,EAAE,cAAc,CAAC;IACpB;sDACkD;IAClD,WAAW,EAAE,oBAAoB,CAAC;IAClC,wFAAwF;IACxF,SAAS,EAAE,cAAc,CAAC;IAC1B;;;;;OAKG;IACH,OAAO,CAAC,EAAE,CAAC,MAAM,EAAE,MAAM,KAAK,IAAI,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC;IACnD,6EAA6E;IAC7E,oBAAoB,CAAC,EAAE,MAAM,MAAM,GAAG,IAAI,CAAC;IAC3C;;;;OAIG;IACH,YAAY,CAAC,EAAE,aAAa,CAAC;IAC7B;;;;;;;OAOG;IACH,QAAQ,EAAE,CAAC,OAAO,EAAE,oBAAoB,KAAK,OAAO,CAAC,gBAAgB,CAAC,CAAC;CACvE;AAED,MAAM,WAAW,yBAAyB;IACzC,8EAA8E;IAC9E,MAAM,EAAE,MAAM,CAAC;IACf;;;;OAIG;IACH,gBAAgB,CAAC,EAAE,MAAM,CAAC;CAC1B;AAED,MAAM,WAAW,yBAAyB;IACzC,qEAAqE;IACrE,kBAAkB,CAAC,CAAC,UAAU,EAAE,MAAM,GAAG,IAAI,CAAC;IAC9C,yEAAyE;IACzE,kBAAkB,CAAC,IAAI,IAAI,CAAC;IAC5B,kGAAkG;IAClG,qBAAqB,CAAC,CAAC,OAAO,EAAE,gBAAgB,GAAG,IAAI,CAAC;IACxD,mEAAmE;IACnE,cAAc,CAAC,CAAC,OAAO,EAAE,gBAAgB,GAAG,IAAI,CAAC;IACjD,gFAAgF;IAChF,OAAO,CAAC,CAAC,KAAK,EAAE,KAAK,GAAG,IAAI,CAAC;IAC7B,uFAAuF;IACvF,gBAAgB,CAAC,CAAC,UAAU,EAAE,MAAM,EAAE,MAAM,EAAE,eAAe,GAAG,IAAI,CAAC;CACrE;AAWD,qBAAa,mBAAmB;IAC/B,OAAO,CAAC,QAAQ,CAAC,IAAI,CAA0B;IAC/C,OAAO,CAAC,QAAQ,CAAC,MAAM,CAA4B;IACnD,OAAO,CAAC,QAAQ,CAAC,MAAM,CAAS;IAChC,OAAO,CAAC,QAAQ,CAAC,gBAAgB,CAAS;IAC1C,OAAO,CAAC,QAAQ,CAAC,OAAO,CAAoB;IAE5C,OAAO,CAAC,WAAW,CAAmC;IACtD,4EAA4E;IAC5E,OAAO,CAAC,UAAU,CAA8B;IAChD,OAAO,CAAC,aAAa,CAAM;IAC3B,OAAO,CAAC,gBAAgB,CAIR;IAChB,OAAO,CAAC,kBAAkB,CAAK;IAC/B,OAAO,CAAC,OAAO,CAAS;IACxB,OAAO,CAAC,QAAQ,CAA6B;IAC7C,OAAO,CAAC,gBAAgB,CAA6B;IACrD,OAAO,CAAC,gBAAgB,CAA6B;IACrD,OAAO,CAAC,qBAAqB,CAAgC;IAC7D,gFAAgF;IAChF,OAAO,CAAC,YAAY,CAAS;IAC7B,OAAO,CAAC,YAAY,CAAiC;gBAGpD,IAAI,EAAE,uBAAuB,EAC7B,MAAM,EAAE,yBAAyB,EACjC,MAAM,GAAE,yBAA8B;IAYvC,oFAAoF;IACpF,KAAK,IAAI,IAAI;IAuBb,8EAA8E;IAC9E,IAAI,IAAI,IAAI;IAsBZ,OAAO,CAAC,UAAU;IA6ClB,OAAO,CAAC,kBAAkB;IA2B1B;;;;;;;;OAQG;IACH,aAAa,IAAI,IAAI;YAIP,WAAW;IASzB,OAAO,CAAC,oBAAoB;IAW5B,OAAO,CAAC,qBAAqB;YAaf,+BAA+B;IAoB7C,OAAO,CAAC,gBAAgB;IAiBxB,OAAO,CAAC,gBAAgB;IAcxB,OAAO,CAAC,aAAa;YASP,QAAQ;YA8ER,WAAW;IAczB,OAAO,CAAC,sBAAsB;YAKhB,gBAAgB;YAUhB,iBAAiB;CA2B/B"}