@livekit/agents 0.7.8 → 1.0.0-next.0

package/dist/voice/speech_handle.d.ts

@@ -0,0 +1,46 @@
+import type { ChatMessage } from '../llm/index.js';
+export declare class SpeechHandle {
+    readonly id: string;
+    readonly allowInterruptions: boolean;
+    readonly stepIndex: number;
+    readonly parent?: SpeechHandle | undefined;
+    /** Priority for messages that should be played after all other messages in the queue */
+    static SPEECH_PRIORITY_LOW: number;
+    /** Every speech generates by the VoiceAgent defaults to this priority. */
+    static SPEECH_PRIORITY_NORMAL: number;
+    /** Priority for important messages that should be played before others. */
+    static SPEECH_PRIORITY_HIGH: number;
+    private interruptFut;
+    private authorizeFut;
+    private playoutDoneFut;
+    private _chatMessage?;
+    constructor(id: string, allowInterruptions: boolean, stepIndex: number, parent?: SpeechHandle | undefined);
+    static create(options: {
+        allowInterruptions?: boolean;
+        stepIndex?: number;
+        parent?: SpeechHandle;
+    }): SpeechHandle;
+    get interrupted(): boolean;
+    get done(): boolean;
+    get chatMessage(): ChatMessage | undefined;
+    /**
+     * Interrupt the current speech generation.
+     *
+     * @throws Error If this speech handle does not allow interruptions.
+     *
+     * @returns The same speech handle that was interrupted.
+     */
+    interrupt(): SpeechHandle;
+    then(callback: (sh: SpeechHandle) => void): Promise<void>;
+    waitForPlayout(): Promise<void>;
+    waitIfNotInterrupted(aw: Promise<unknown>[]): Promise<void>;
+    /** @internal */
+    _setChatMessage(chatMessage: ChatMessage): void;
+    /** @internal */
+    _authorizePlayout(): void;
+    /** @internal */
+    _waitForAuthorization(): Promise<void>;
+    /** @internal */
+    _markPlayoutDone(): void;
+}
+//# sourceMappingURL=speech_handle.d.ts.map

package/dist/voice/speech_handle.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"speech_handle.d.ts","sourceRoot":"","sources":["../../src/voice/speech_handle.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,iBAAiB,CAAC;AAGnD,qBAAa,YAAY;IAerB,QAAQ,CAAC,EAAE,EAAE,MAAM;IACnB,QAAQ,CAAC,kBAAkB,EAAE,OAAO;IACpC,QAAQ,CAAC,SAAS,EAAE,MAAM;IAC1B,QAAQ,CAAC,MAAM,CAAC;IAjBlB,wFAAwF;IACxF,MAAM,CAAC,mBAAmB,SAAK;IAC/B,0EAA0E;IAC1E,MAAM,CAAC,sBAAsB,SAAK;IAClC,2EAA2E;IAC3E,MAAM,CAAC,oBAAoB,SAAM;IAEjC,OAAO,CAAC,YAAY,CAAgB;IACpC,OAAO,CAAC,YAAY,CAAgB;IACpC,OAAO,CAAC,cAAc,CAAgB;IAEtC,OAAO,CAAC,YAAY,CAAC,CAAc;gBAGxB,EAAE,EAAE,MAAM,EACV,kBAAkB,EAAE,OAAO,EAC3B,SAAS,EAAE,MAAM,EACjB,MAAM,CAAC,0BAAc;IAGhC,MAAM,CAAC,MAAM,CAAC,OAAO,EAAE;QACrB,kBAAkB,CAAC,EAAE,OAAO,CAAC;QAC7B,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,MAAM,CAAC,EAAE,YAAY,CAAC;KACvB;IAMD,IAAI,WAAW,IAAI,OAAO,CAEzB;IAED,IAAI,IAAI,IAAI,OAAO,CAElB;IAED,IAAI,WAAW,IAAI,WAAW,GAAG,SAAS,CAEzC;IAED;;;;;;OAMG;IACH,SAAS,IAAI,YAAY;IAWzB,IAAI,CAAC,QAAQ,EAAE,CAAC,EAAE,EAAE,YAAY,KAAK,IAAI;IAInC,cAAc;IAId,oBAAoB,CAAC,EAAE,EAAE,OAAO,CAAC,OAAO,CAAC,EAAE,GAAG,OAAO,CAAC,IAAI,CAAC;IAMjE,gBAAgB;IAChB,eAAe,CAAC,WAAW,EAAE,WAAW;IAOxC,gBAAgB;IAChB,iBAAiB;IAIjB,gBAAgB;IACV,qBAAqB;IAI3B,gBAAgB;IAChB,gBAAgB;CAGjB"}

package/dist/voice/speech_handle.js

@@ -0,0 +1,81 @@
+import { Future, shortuuid } from "../utils.js";
+class SpeechHandle {
+  constructor(id, allowInterruptions, stepIndex, parent) {
+    this.id = id;
+    this.allowInterruptions = allowInterruptions;
+    this.stepIndex = stepIndex;
+    this.parent = parent;
+  }
+  /** Priority for messages that should be played after all other messages in the queue */
+  static SPEECH_PRIORITY_LOW = 0;
+  /** Every speech generates by the VoiceAgent defaults to this priority. */
+  static SPEECH_PRIORITY_NORMAL = 5;
+  /** Priority for important messages that should be played before others. */
+  static SPEECH_PRIORITY_HIGH = 10;
+  interruptFut = new Future();
+  authorizeFut = new Future();
+  playoutDoneFut = new Future();
+  _chatMessage;
+  static create(options) {
+    const { allowInterruptions = false, stepIndex = 0, parent } = options ?? {};
+    return new SpeechHandle(shortuuid("speech_"), allowInterruptions, stepIndex, parent);
+  }
+  get interrupted() {
+    return this.interruptFut.done;
+  }
+  get done() {
+    return this.playoutDoneFut.done;
+  }
+  get chatMessage() {
+    return this._chatMessage;
+  }
+  /**
+   * Interrupt the current speech generation.
+   *
+   * @throws Error If this speech handle does not allow interruptions.
+   *
+   * @returns The same speech handle that was interrupted.
+   */
+  interrupt() {
+    if (!this.allowInterruptions) {
+      throw new Error("interruptions are not allowed");
+    }
+    if (this.done) return this;
+    this.interruptFut.resolve();
+    return this;
+  }
+  then(callback) {
+    return this.playoutDoneFut.await.finally(() => callback(this));
+  }
+  async waitForPlayout() {
+    return this.playoutDoneFut.await;
+  }
+  async waitIfNotInterrupted(aw) {
+    const allTasksPromise = Promise.all(aw);
+    const fs = [allTasksPromise, this.interruptFut.await];
+    await Promise.race(fs);
+  }
+  /** @internal */
+  _setChatMessage(chatMessage) {
+    if (this.done) {
+      throw new Error("cannot set chat message after speech has been played");
+    }
+    this._chatMessage = chatMessage;
+  }
+  /** @internal */
+  _authorizePlayout() {
+    this.authorizeFut.resolve();
+  }
+  /** @internal */
+  async _waitForAuthorization() {
+    return this.authorizeFut.await;
+  }
+  /** @internal */
+  _markPlayoutDone() {
+    this.playoutDoneFut.resolve();
+  }
+}
+export {
+  SpeechHandle
+};
+//# sourceMappingURL=speech_handle.js.map

package/dist/voice/speech_handle.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../src/voice/speech_handle.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { ChatMessage } from '../llm/index.js';\nimport { Future, shortuuid } from '../utils.js';\n\nexport class SpeechHandle {\n  /** Priority for messages that should be played after all other messages in the queue */\n  static SPEECH_PRIORITY_LOW = 0;\n  /** Every speech generates by the VoiceAgent defaults to this priority. */\n  static SPEECH_PRIORITY_NORMAL = 5;\n  /** Priority for important messages that should be played before others. */\n  static SPEECH_PRIORITY_HIGH = 10;\n\n  private interruptFut = new Future();\n  private authorizeFut = new Future();\n  private playoutDoneFut = new Future();\n\n  private _chatMessage?: ChatMessage;\n\n  constructor(\n    readonly id: string,\n    readonly allowInterruptions: boolean,\n    readonly stepIndex: number,\n    readonly parent?: SpeechHandle,\n  ) {}\n\n  static create(options: {\n    allowInterruptions?: boolean;\n    stepIndex?: number;\n    parent?: SpeechHandle;\n  }) {\n    const { allowInterruptions = false, stepIndex = 0, parent } = options ?? {};\n\n    return new SpeechHandle(shortuuid('speech_'), allowInterruptions, stepIndex, parent);\n  }\n\n  get interrupted(): boolean {\n    return this.interruptFut.done;\n  }\n\n  get done(): boolean {\n    return this.playoutDoneFut.done;\n  }\n\n  get chatMessage(): ChatMessage | undefined {\n    return this._chatMessage;\n  }\n\n  /**\n   * Interrupt the current speech generation.\n   *\n   * @throws Error If this speech handle does not allow interruptions.\n   *\n   * @returns The same speech handle that was interrupted.\n   */\n  interrupt(): SpeechHandle {\n    if (!this.allowInterruptions) {\n      throw new Error('interruptions are not allowed');\n    }\n\n    if (this.done) return this;\n\n    this.interruptFut.resolve();\n    return this;\n  }\n\n  then(callback: (sh: SpeechHandle) => void) {\n    return this.playoutDoneFut.await.finally(() => callback(this));\n  }\n\n  async waitForPlayout() {\n    return this.playoutDoneFut.await;\n  }\n\n  async waitIfNotInterrupted(aw: Promise<unknown>[]): Promise<void> {\n    const allTasksPromise = Promise.all(aw);\n    const fs: Promise<unknown>[] = [allTasksPromise, this.interruptFut.await];\n    await Promise.race(fs);\n  }\n\n  /** @internal */\n  _setChatMessage(chatMessage: ChatMessage) {\n    if (this.done) {\n      throw new Error('cannot set chat message after speech has been played');\n    }\n    this._chatMessage = chatMessage;\n  }\n\n  /** @internal */\n  _authorizePlayout() {\n    this.authorizeFut.resolve();\n  }\n\n  /** @internal */\n  async _waitForAuthorization() {\n    return this.authorizeFut.await;\n  }\n\n  /** @internal */\n  _markPlayoutDone() {\n    this.playoutDoneFut.resolve();\n  }\n}\n"],"mappings":"AAIA,SAAS,QAAQ,iBAAiB;AAE3B,MAAM,aAAa;AAAA,EAcxB,YACW,IACA,oBACA,WACA,QACT;AAJS;AACA;AACA;AACA;AAAA,EACR;AAAA;AAAA,EAjBH,OAAO,sBAAsB;AAAA;AAAA,EAE7B,OAAO,yBAAyB;AAAA;AAAA,EAEhC,OAAO,uBAAuB;AAAA,EAEtB,eAAe,IAAI,OAAO;AAAA,EAC1B,eAAe,IAAI,OAAO;AAAA,EAC1B,iBAAiB,IAAI,OAAO;AAAA,EAE5B;AAAA,EASR,OAAO,OAAO,SAIX;AACD,UAAM,EAAE,qBAAqB,OAAO,YAAY,GAAG,OAAO,IAAI,WAAW,CAAC;AAE1E,WAAO,IAAI,aAAa,UAAU,SAAS,GAAG,oBAAoB,WAAW,MAAM;AAAA,EACrF;AAAA,EAEA,IAAI,cAAuB;AACzB,WAAO,KAAK,aAAa;AAAA,EAC3B;AAAA,EAEA,IAAI,OAAgB;AAClB,WAAO,KAAK,eAAe;AAAA,EAC7B;AAAA,EAEA,IAAI,cAAuC;AACzC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,YAA0B;AACxB,QAAI,CAAC,KAAK,oBAAoB;AAC5B,YAAM,IAAI,MAAM,+BAA+B;AAAA,IACjD;AAEA,QAAI,KAAK,KAAM,QAAO;AAEtB,SAAK,aAAa,QAAQ;AAC1B,WAAO;AAAA,EACT;AAAA,EAEA,KAAK,UAAsC;AACzC,WAAO,KAAK,eAAe,MAAM,QAAQ,MAAM,SAAS,IAAI,CAAC;AAAA,EAC/D;AAAA,EAEA,MAAM,iBAAiB;AACrB,WAAO,KAAK,eAAe;AAAA,EAC7B;AAAA,EAEA,MAAM,qBAAqB,IAAuC;AAChE,UAAM,kBAAkB,QAAQ,IAAI,EAAE;AACtC,UAAM,KAAyB,CAAC,iBAAiB,KAAK,aAAa,KAAK;AACxE,UAAM,QAAQ,KAAK,EAAE;AAAA,EACvB;AAAA;AAAA,EAGA,gBAAgB,aAA0B;AACxC,QAAI,KAAK,MAAM;AACb,YAAM,IAAI,MAAM,sDAAsD;AAAA,IACxE;AACA,SAAK,eAAe;AAAA,EACtB;AAAA;AAAA,EAGA,oBAAoB;AAClB,SAAK,aAAa,QAAQ;AAAA,EAC5B;AAAA;AAAA,EAGA,MAAM,wBAAwB;AAC5B,WAAO,KAAK,aAAa;AAAA,EAC3B;AAAA;AAAA,EAGA,mBAAmB;AACjB,SAAK,eAAe,QAAQ;AAAA,EAC9B;AACF;","names":[]}

package/dist/voice/transcription/_utils.cjs

@@ -0,0 +1,45 @@
+"use strict";
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+var utils_exports = {};
+__export(utils_exports, {
+  findMicrophoneTrackId: () => findMicrophoneTrackId
+});
+module.exports = __toCommonJS(utils_exports);
+var import_rtc_node = require("@livekit/rtc-node");
+function findMicrophoneTrackId(room, identity) {
+  var _a;
+  let p = room.remoteParticipants.get(identity) ?? null;
+  if (identity === ((_a = room.localParticipant) == null ? void 0 : _a.identity)) {
+    p = room.localParticipant;
+  }
+  if (p === null) {
+    throw new Error(`Participant ${identity} not found`);
+  }
+  for (const track of p.trackPublications.values()) {
+    if (track.source === import_rtc_node.TrackSource.SOURCE_MICROPHONE && track.sid) {
+      return track.sid;
+    }
+  }
+  throw new Error(`Participant ${identity} does not have a microphone track`);
+}
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  findMicrophoneTrackId
+});
+//# sourceMappingURL=_utils.cjs.map

package/dist/voice/transcription/_utils.cjs.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/voice/transcription/_utils.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { LocalParticipant, Participant, Room } from '@livekit/rtc-node';\nimport { TrackSource } from '@livekit/rtc-node';\n\nexport function findMicrophoneTrackId(room: Room, identity: string): string {\n  let p: Participant | LocalParticipant | null = room.remoteParticipants.get(identity) ?? null;\n  if (identity === room.localParticipant?.identity) {\n    p = room.localParticipant;\n  }\n\n  if (p === null) {\n    throw new Error(`Participant ${identity} not found`);\n  }\n\n  for (const track of p.trackPublications.values()) {\n    if (track.source === TrackSource.SOURCE_MICROPHONE && track.sid) {\n      // find the first microphone track\n      return track.sid;\n    }\n  }\n\n  throw new Error(`Participant ${identity} does not have a microphone track`);\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,sBAA4B;AAErB,SAAS,sBAAsB,MAAY,UAA0B;AAN5E;AAOE,MAAI,IAA2C,KAAK,mBAAmB,IAAI,QAAQ,KAAK;AACxF,MAAI,eAAa,UAAK,qBAAL,mBAAuB,WAAU;AAChD,QAAI,KAAK;AAAA,EACX;AAEA,MAAI,MAAM,MAAM;AACd,UAAM,IAAI,MAAM,eAAe,QAAQ,YAAY;AAAA,EACrD;AAEA,aAAW,SAAS,EAAE,kBAAkB,OAAO,GAAG;AAChD,QAAI,MAAM,WAAW,4BAAY,qBAAqB,MAAM,KAAK;AAE/D,aAAO,MAAM;AAAA,IACf;AAAA,EACF;AAEA,QAAM,IAAI,MAAM,eAAe,QAAQ,mCAAmC;AAC5E;","names":[]}

package/dist/voice/transcription/_utils.d.cts

@@ -0,0 +1,3 @@
+import type { Room } from '@livekit/rtc-node';
+export declare function findMicrophoneTrackId(room: Room, identity: string): string;
+//# sourceMappingURL=_utils.d.ts.map

package/dist/voice/transcription/_utils.d.ts

@@ -0,0 +1,3 @@
+import type { Room } from '@livekit/rtc-node';
+export declare function findMicrophoneTrackId(room: Room, identity: string): string;
+//# sourceMappingURL=_utils.d.ts.map

package/dist/voice/transcription/_utils.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"_utils.d.ts","sourceRoot":"","sources":["../../../src/voice/transcription/_utils.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAiC,IAAI,EAAE,MAAM,mBAAmB,CAAC;AAG7E,wBAAgB,qBAAqB,CAAC,IAAI,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,GAAG,MAAM,CAkB1E"}

package/dist/voice/transcription/_utils.js

@@ -0,0 +1,21 @@
+import { TrackSource } from "@livekit/rtc-node";
+function findMicrophoneTrackId(room, identity) {
+  var _a;
+  let p = room.remoteParticipants.get(identity) ?? null;
+  if (identity === ((_a = room.localParticipant) == null ? void 0 : _a.identity)) {
+    p = room.localParticipant;
+  }
+  if (p === null) {
+    throw new Error(`Participant ${identity} not found`);
+  }
+  for (const track of p.trackPublications.values()) {
+    if (track.source === TrackSource.SOURCE_MICROPHONE && track.sid) {
+      return track.sid;
+    }
+  }
+  throw new Error(`Participant ${identity} does not have a microphone track`);
+}
+export {
+  findMicrophoneTrackId
+};
+//# sourceMappingURL=_utils.js.map

package/dist/voice/transcription/_utils.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/voice/transcription/_utils.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { LocalParticipant, Participant, Room } from '@livekit/rtc-node';\nimport { TrackSource } from '@livekit/rtc-node';\n\nexport function findMicrophoneTrackId(room: Room, identity: string): string {\n  let p: Participant | LocalParticipant | null = room.remoteParticipants.get(identity) ?? null;\n  if (identity === room.localParticipant?.identity) {\n    p = room.localParticipant;\n  }\n\n  if (p === null) {\n    throw new Error(`Participant ${identity} not found`);\n  }\n\n  for (const track of p.trackPublications.values()) {\n    if (track.source === TrackSource.SOURCE_MICROPHONE && track.sid) {\n      // find the first microphone track\n      return track.sid;\n    }\n  }\n\n  throw new Error(`Participant ${identity} does not have a microphone track`);\n}\n"],"mappings":"AAIA,SAAS,mBAAmB;AAErB,SAAS,sBAAsB,MAAY,UAA0B;AAN5E;AAOE,MAAI,IAA2C,KAAK,mBAAmB,IAAI,QAAQ,KAAK;AACxF,MAAI,eAAa,UAAK,qBAAL,mBAAuB,WAAU;AAChD,QAAI,KAAK;AAAA,EACX;AAEA,MAAI,MAAM,MAAM;AACd,UAAM,IAAI,MAAM,eAAe,QAAQ,YAAY;AAAA,EACrD;AAEA,aAAW,SAAS,EAAE,kBAAkB,OAAO,GAAG;AAChD,QAAI,MAAM,WAAW,YAAY,qBAAqB,MAAM,KAAK;AAE/D,aAAO,MAAM;AAAA,IACf;AAAA,EACF;AAEA,QAAM,IAAI,MAAM,eAAe,QAAQ,mCAAmC;AAC5E;","names":[]}

package/dist/voice/transcription/index.cjs

@@ -0,0 +1,23 @@
+"use strict";
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __reExport = (target, mod, secondTarget) => (__copyProps(target, mod, "default"), secondTarget && __copyProps(secondTarget, mod, "default"));
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+var transcription_exports = {};
+module.exports = __toCommonJS(transcription_exports);
+__reExport(transcription_exports, require("./_utils.cjs"), module.exports);
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  ...require("./_utils.cjs")
+});
+//# sourceMappingURL=index.cjs.map

package/dist/voice/transcription/index.cjs.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/voice/transcription/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nexport * from './_utils.js';\n"],"mappings":";;;;;;;;;;;;;;;AAAA;AAAA;AAGA,kCAAc,wBAHd;","names":[]}

package/dist/voice/transcription/index.d.cts

@@ -0,0 +1,2 @@
+export * from './_utils.js';
+//# sourceMappingURL=index.d.ts.map

package/dist/voice/transcription/index.d.ts

@@ -0,0 +1,2 @@
+export * from './_utils.js';
+//# sourceMappingURL=index.d.ts.map

package/dist/voice/transcription/index.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/voice/transcription/index.ts"],"names":[],"mappings":"AAGA,cAAc,aAAa,CAAC"}

package/dist/voice/transcription/index.js

@@ -0,0 +1,2 @@
+export * from "./_utils.js";
+//# sourceMappingURL=index.js.map

package/dist/voice/transcription/index.js.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/voice/transcription/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nexport * from './_utils.js';\n"],"mappings":"AAGA,cAAc;","names":[]}

package/dist/voice/transcription/synchronizer.cjs

@@ -0,0 +1,380 @@
+"use strict";
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+var synchronizer_exports = {};
+__export(synchronizer_exports, {
+  TranscriptionSynchronizer: () => TranscriptionSynchronizer,
+  defaultTextSyncOptions: () => defaultTextSyncOptions
+});
+module.exports = __toCommonJS(synchronizer_exports);
+var import_async = require("@std/async");
+var import_log = require("../../log.cjs");
+var import_identity_transform = require("../../stream/identity_transform.cjs");
+var import_tokenize = require("../../tokenize/index.cjs");
+var import_utils = require("../../utils.cjs");
+var import_io = require("../io.cjs");
+const STANDARD_SPEECH_RATE = 3.83;
+class SegmentSynchronizerImpl {
+  constructor(options, nextInChain) {
+    this.options = options;
+    this.nextInChain = nextInChain;
+    this.speed = options.speed * STANDARD_SPEECH_RATE;
+    this.textData = {
+      sentenceStream: options.sentenceTokenizer.stream(),
+      pushedText: "",
+      done: false,
+      forwardedHyphens: 0,
+      forwardedText: ""
+    };
+    this.audioData = {
+      pushedDuration: 0,
+      done: false
+    };
+    this.outputStream = new import_identity_transform.IdentityTransform();
+    this.outputStreamWriter = this.outputStream.writable.getWriter();
+    this.mainTask().then(() => {
+      this.outputStreamWriter.close();
+    }).catch((error) => {
+      this.logger.error({ error }, "mainTask SegmentSynchronizerImpl");
+    });
+    this.captureTask = this.captureTaskImpl();
+  }
+  textData;
+  audioData;
+  speed;
+  outputStream;
+  outputStreamWriter;
+  captureTask;
+  startWallTime;
+  startFuture = new import_utils.Future();
+  closedFuture = new import_utils.Future();
+  playbackCompleted = false;
+  logger = (0, import_log.log)();
+  get closed() {
+    return this.closedFuture.done;
+  }
+  get audioInputEnded() {
+    return this.audioData.done;
+  }
+  get textInputEnded() {
+    return this.textData.done;
+  }
+  get readable() {
+    return this.outputStream.readable;
+  }
+  pushAudio(frame) {
+    if (this.closed) {
+      this.logger.warn("SegmentSynchronizerImpl.pushAudio called after close");
+      return;
+    }
+    const frameDuration = frame.samplesPerChannel / frame.sampleRate;
+    if (!this.startWallTime && frameDuration > 0) {
+      this.startWallTime = Date.now();
+      this.startFuture.resolve();
+    }
+    this.audioData.pushedDuration += frameDuration;
+  }
+  endAudioInput() {
+    if (this.closed) {
+      this.logger.warn("SegmentSynchronizerImpl.endAudioInput called after close");
+      return;
+    }
+    this.audioData.done = true;
+  }
+  pushText(text) {
+    if (this.closed) {
+      this.logger.warn("SegmentSynchronizerImpl.pushText called after close");
+      return;
+    }
+    this.textData.sentenceStream.pushText(text);
+    this.textData.pushedText += text;
+  }
+  endTextInput() {
+    if (this.closed) {
+      this.logger.warn("SegmentSynchronizerImpl.endTextInput called after close");
+      return;
+    }
+    this.textData.done = true;
+    this.textData.sentenceStream.endInput();
+  }
+  markPlaybackFinished(_playbackPosition, interrupted) {
+    if (this.closed) {
+      this.logger.warn("SegmentSynchronizerImpl.markPlaybackFinished called after close");
+      return;
+    }
+    if (!this.textData.done || !this.audioData.done) {
+      this.logger.warn(
+        { textDone: this.textData.done, audioDone: this.audioData.done },
+        "SegmentSynchronizerImpl.markPlaybackFinished called before text/audio input is done"
+      );
+      return;
+    }
+    if (!interrupted) {
+      this.playbackCompleted = true;
+    }
+  }
+  get synchronizedTranscript() {
+    if (this.playbackCompleted) {
+      return this.textData.pushedText;
+    }
+    return this.textData.forwardedText;
+  }
+  async captureTaskImpl() {
+    const reader = this.outputStream.readable.getReader();
+    while (true) {
+      const { done, value: text } = await reader.read();
+      if (done) {
+        break;
+      }
+      this.textData.forwardedText += text;
+      await this.nextInChain.captureText(text);
+    }
+    reader.releaseLock();
+    this.nextInChain.flush();
+  }
+  async mainTask() {
+    await this.startFuture.await;
+    if (this.closed && !this.playbackCompleted) {
+      return;
+    }
+    if (!this.startWallTime) {
+      throw new Error("startWallTime is not set when starting SegmentSynchronizerImpl.mainTask");
+    }
+    for await (const textSegment of this.textData.sentenceStream) {
+      const sentence = textSegment.token;
+      let textCursor = 0;
+      if (this.closed && !this.playbackCompleted) {
+        return;
+      }
+      for (const [word, _, endPos] of this.options.splitWords(sentence)) {
+        if (this.closed && !this.playbackCompleted) {
+          return;
+        }
+        if (this.playbackCompleted) {
+          this.outputStreamWriter.write(sentence.slice(textCursor, endPos));
+          textCursor = endPos;
+          continue;
+        }
+        const wordHphens = this.options.hyphenateWord(word).length;
+        const elapsedSeconds = (Date.now() - this.startWallTime) / 1e3;
+        const targetHyphens = elapsedSeconds * this.options.speed;
+        const hyphensBehind = Math.max(0, targetHyphens - this.textData.forwardedHyphens);
+        let delay2 = Math.max(0, wordHphens - hyphensBehind) / this.speed;
+        if (this.playbackCompleted) {
+          delay2 = 0;
+        }
+        await this.sleepIfNotClosed(delay2 / 2);
+        this.outputStreamWriter.write(sentence.slice(textCursor, endPos));
+        await this.sleepIfNotClosed(delay2 / 2);
+        this.textData.forwardedHyphens += wordHphens;
+        textCursor = endPos;
+      }
+      if (textCursor < sentence.length) {
+        const remaining = sentence.slice(textCursor);
+        this.outputStreamWriter.write(remaining);
+      }
+    }
+  }
+  async sleepIfNotClosed(sleepTimeSeconds) {
+    if (this.closed) {
+      return;
+    }
+    await (0, import_async.delay)(sleepTimeSeconds * 1e3);
+  }
+  async close() {
+    if (this.closed) {
+      return;
+    }
+    this.closedFuture.resolve();
+    this.startFuture.resolve();
+    this.textData.sentenceStream.close();
+    await this.captureTask;
+  }
+}
+const defaultTextSyncOptions = {
+  speed: 1,
+  hyphenateWord: import_tokenize.basic.hyphenateWord,
+  splitWords: import_tokenize.basic.splitWords,
+  sentenceTokenizer: new import_tokenize.basic.SentenceTokenizer({
+    retainFormat: true
+  })
+};
+class TranscriptionSynchronizer {
+  audioOutput;
+  textOutput;
+  options;
+  rotateSegmentTask;
+  _enabled = true;
+  closed = false;
+  /* @internal */
+  _impl;
+  logger = (0, import_log.log)();
+  constructor(nextInChainAudio, nextInChainText, options = defaultTextSyncOptions) {
+    this.audioOutput = new SyncedAudioOutput(this, nextInChainAudio);
+    this.textOutput = new SyncedTextOutput(this, nextInChainText);
+    this.options = {
+      speed: options.speed,
+      hyphenateWord: options.hyphenateWord,
+      splitWords: options.splitWords,
+      sentenceTokenizer: options.sentenceTokenizer
+    };
+    this._impl = new SegmentSynchronizerImpl(this.options, nextInChainText);
+    this.rotateSegmentTask = import_utils.Task.from(
+      (controller) => this.rotateSegmentTaskImpl(controller.signal)
+    );
+  }
+  get enabled() {
+    return this._enabled;
+  }
+  set enabled(enabled) {
+    if (this._enabled === enabled) {
+      return;
+    }
+    this._enabled = enabled;
+    this.rotateSegment();
+  }
+  rotateSegment() {
+    if (this.closed) {
+      return;
+    }
+    if (!this.rotateSegmentTask.done) {
+      this.logger.warn("rotateSegment called while previous segment is still being rotated");
+    }
+    this.rotateSegmentTask = import_utils.Task.from(
+      (controller) => this.rotateSegmentTaskImpl(controller.signal, this.rotateSegmentTask)
+    );
+  }
+  async close() {
+    this.closed = true;
+    await this.rotateSegmentTask.cancelAndWait();
+    await this._impl.close();
+  }
+  async barrier() {
+    if (this.rotateSegmentTask.done) {
+      return;
+    }
+    await this.rotateSegmentTask.result;
+  }
+  async rotateSegmentTaskImpl(abort, oldTask) {
+    if (oldTask) {
+      await oldTask.result;
+    }
+    if (abort.aborted) {
+      return;
+    }
+    await this._impl.close();
+    this._impl = new SegmentSynchronizerImpl(this.options, this.textOutput.nextInChain);
+  }
+}
+class SyncedAudioOutput extends import_io.AudioOutput {
+  constructor(synchronizer, nextInChainAudio) {
+    super(nextInChainAudio.sampleRate, nextInChainAudio);
+    this.synchronizer = synchronizer;
+    this.nextInChainAudio = nextInChainAudio;
+  }
+  pushedDuration = 0;
+  async captureFrame(frame) {
+    await this.synchronizer.barrier();
+    await super.captureFrame(frame);
+    await this.nextInChainAudio.captureFrame(frame);
+    this.pushedDuration += frame.samplesPerChannel / frame.sampleRate;
+    if (!this.synchronizer.enabled) {
+      return;
+    }
+    if (this.synchronizer._impl.audioInputEnded) {
+      this.logger.warn(
+        "SegmentSynchronizerImpl audio marked as ended in capture audio, rotating segment"
+      );
+      this.synchronizer.rotateSegment();
+      await this.synchronizer.barrier();
+    }
+    this.synchronizer._impl.pushAudio(frame);
+  }
+  flush() {
+    super.flush();
+    this.nextInChainAudio.flush();
+    if (!this.synchronizer.enabled) {
+      return;
+    }
+    if (!this.pushedDuration) {
+      this.synchronizer.rotateSegment();
+      return;
+    }
+    this.synchronizer._impl.endAudioInput();
+  }
+  clearBuffer() {
+    this.nextInChainAudio.clearBuffer();
+  }
+  // this is going to be automatically called by the next_in_chain
+  onPlaybackFinished(ev) {
+    if (!this.synchronizer.enabled) {
+      super.onPlaybackFinished(ev);
+      return;
+    }
+    this.synchronizer._impl.markPlaybackFinished(ev.playbackPosition, ev.interrupted);
+    super.onPlaybackFinished({
+      playbackPosition: ev.playbackPosition,
+      interrupted: ev.interrupted,
+      synchronizedTranscript: this.synchronizer._impl.synchronizedTranscript
+    });
+    this.synchronizer.rotateSegment();
+    this.pushedDuration = 0;
+  }
+}
+class SyncedTextOutput extends import_io.TextOutput {
+  constructor(synchronizer, nextInChain) {
+    super(nextInChain);
+    this.synchronizer = synchronizer;
+    this.nextInChain = nextInChain;
+  }
+  capturing = false;
+  logger = (0, import_log.log)();
+  async captureText(text) {
+    await this.synchronizer.barrier();
+    if (!this.synchronizer.enabled) {
+      await this.nextInChain.captureText(text);
+      return;
+    }
+    this.capturing = true;
+    if (this.synchronizer._impl.textInputEnded) {
+      this.logger.warn(
+        "SegmentSynchronizerImpl text marked as ended in capture text, rotating segment"
+      );
+      this.synchronizer.rotateSegment();
+      await this.synchronizer.barrier();
+    }
+    this.synchronizer._impl.pushText(text);
+  }
+  flush() {
+    if (!this.synchronizer.enabled) {
+      this.nextInChain.flush();
+      return;
+    }
+    if (!this.capturing) {
+      return;
+    }
+    this.capturing = false;
+    this.synchronizer._impl.endTextInput();
+  }
+}
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  TranscriptionSynchronizer,
+  defaultTextSyncOptions
+});
+//# sourceMappingURL=synchronizer.cjs.map