@livekit/agents 1.0.3 → 1.0.5

package/dist/voice/io.cjs.map

@@ -1 +1 @@
-{"version":3,"sources":["../../src/voice/io.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport type { ChatContext } from '../llm/chat_context.js';\nimport type { ChatChunk } from '../llm/llm.js';\nimport type { ToolContext } from '../llm/tool_context.js';\nimport { log } from '../log.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport type { SpeechEvent } from '../stt/stt.js';\nimport { Future } from '../utils.js';\nimport type { ModelSettings } from './agent.js';\n\nexport type STTNode = (\n  audio: ReadableStream<AudioFrame>,\n  modelSettings: ModelSettings,\n) => Promise<ReadableStream<SpeechEvent | string> | null>;\n\nexport type LLMNode = (\n  chatCtx: ChatContext,\n  toolCtx: ToolContext,\n  modelSettings: ModelSettings,\n) => Promise<ReadableStream<ChatChunk | string> | null>;\n\nexport type TTSNode = (\n  text: ReadableStream<string>,\n  modelSettings: ModelSettings,\n) => Promise<ReadableStream<AudioFrame> | null>;\nexport abstract class AudioInput {\n  protected deferredStream: DeferredReadableStream<AudioFrame> =\n    new DeferredReadableStream<AudioFrame>();\n\n  get stream(): ReadableStream<AudioFrame> {\n    return this.deferredStream.stream;\n  }\n\n  onAttached(): void {}\n\n  onDetached(): void {}\n}\n\nexport abstract class AudioOutput extends EventEmitter {\n  static readonly EVENT_PLAYBACK_FINISHED = 'playbackFinished';\n\n  private playbackFinishedFuture: Future<void> = new Future();\n  private _capturing: boolean = false;\n  private playbackFinishedCount: number = 0;\n  private playbackSegmentsCount: number = 0;\n  private lastPlaybackEvent: PlaybackFinishedEvent = {\n    playbackPosition: 0,\n    interrupted: false,\n  };\n  protected logger = log();\n\n  constructor(\n    readonly sampleRate?: number,\n    protected readonly nextInChain?: AudioOutput,\n  ) {\n    super();\n    if (this.nextInChain) {\n      this.nextInChain.on(AudioOutput.EVENT_PLAYBACK_FINISHED, (ev: PlaybackFinishedEvent) =>\n        this.onPlaybackFinished(ev),\n      );\n    }\n  }\n\n  /**\n   * Capture an audio frame for playback, frames can be pushed faster than real-time\n   */\n  async captureFrame(_frame: AudioFrame): Promise<void> {\n    if (!this._capturing) {\n      this._capturing = true;\n      this.playbackSegmentsCount++;\n    }\n  }\n\n  /**\n   * Wait for the past audio segments to finish playing out.\n   *\n   * @returns The event that was emitted when the audio finished playing out (only the last segment information)\n   */\n  async waitForPlayout(): Promise<PlaybackFinishedEvent> {\n    const target = this.playbackSegmentsCount;\n\n    while (this.playbackFinishedCount < target) {\n      await this.playbackFinishedFuture.await;\n      this.playbackFinishedFuture = new Future();\n    }\n\n    return this.lastPlaybackEvent;\n  }\n\n  /**\n   * Developers building audio sinks must call this method when a playback/segment is finished.\n   * Segments are segmented by calls to flush() or clearBuffer()\n   */\n  onPlaybackFinished(options: PlaybackFinishedEvent) {\n    if (this.playbackFinishedCount >= this.playbackSegmentsCount) {\n      this.logger.warn('playback_finished called more times than playback segments were captured');\n      return;\n    }\n\n    this.lastPlaybackEvent = options;\n    this.playbackFinishedCount++;\n    this.playbackFinishedFuture.resolve();\n    this.emit(AudioOutput.EVENT_PLAYBACK_FINISHED, options);\n  }\n\n  flush(): void {\n    this._capturing = false;\n  }\n\n  /**\n   * Clear the buffer, stopping playback immediately\n   */\n  abstract clearBuffer(): void;\n\n  onAttached(): void {\n    if (this.nextInChain) {\n      this.nextInChain.onAttached();\n    }\n  }\n\n  onDetached(): void {\n    if (this.nextInChain) {\n      this.nextInChain.onDetached();\n    }\n  }\n}\n\nexport interface PlaybackFinishedEvent {\n  // How much of the audio was played back\n  playbackPosition: number;\n  // Interrupted is True if playback was interrupted (clearBuffer() was called)\n  interrupted: boolean;\n  // Transcript synced with playback; may be partial if the audio was interrupted\n  // When null, the transcript is not synchronized with the playback\n  synchronizedTranscript?: string;\n}\n\nexport abstract class TextOutput {\n  constructor(protected readonly nextInChain?: TextOutput) {}\n\n  /**\n   * Capture a text segment (Used by the output of LLM nodes)\n   */\n  abstract captureText(text: string): Promise<void>;\n\n  /**\n   * Mark the current text segment as complete (e.g LLM generation is complete)\n   */\n  abstract flush(): void;\n\n  onAttached(): void {\n    if (this.nextInChain) {\n      this.nextInChain.onAttached();\n    }\n  }\n\n  onDetached(): void {\n    if (this.nextInChain) {\n      this.nextInChain.onDetached();\n    }\n  }\n}\n\nexport class AgentInput {\n  private _audioStream: AudioInput | null = null;\n  // enabled by default\n  private _audioEnabled: boolean = true;\n\n  constructor(private readonly audioChanged: () => void) {}\n\n  setAudioEnabled(enable: boolean): void {\n    if (enable === this._audioEnabled) {\n      return;\n    }\n\n    this._audioEnabled = enable;\n\n    if (!this._audioStream) {\n      return;\n    }\n\n    if (enable) {\n      this._audioStream.onAttached();\n    } else {\n      this._audioStream.onDetached();\n    }\n  }\n\n  get audioEnabled(): boolean {\n    return this._audioEnabled;\n  }\n\n  get audio(): AudioInput | null {\n    return this._audioStream;\n  }\n\n  set audio(stream: AudioInput | null) {\n    this._audioStream = stream;\n    this.audioChanged();\n  }\n}\n\nexport class AgentOutput {\n  private _audioSink: AudioOutput | null = null;\n  private _transcriptionSink: TextOutput | null = null;\n  private _audioEnabled: boolean = true;\n  private _transcriptionEnabled: boolean = true;\n\n  constructor(\n    private readonly audioChanged: () => void,\n    private readonly transcriptionChanged: () => void,\n  ) {}\n\n  setAudioEnabled(enabled: boolean): void {\n    if (enabled === this._audioEnabled) {\n      return;\n    }\n\n    this._audioEnabled = enabled;\n\n    if (!this._audioSink) {\n      return;\n    }\n\n    if (enabled) {\n      this._audioSink.onAttached();\n    } else {\n      this._audioSink.onDetached();\n    }\n  }\n\n  setTranscriptionEnabled(enabled: boolean): void {\n    if (enabled === this._transcriptionEnabled) {\n      return;\n    }\n\n    this._transcriptionEnabled = enabled;\n\n    if (!this._transcriptionSink) {\n      return;\n    }\n\n    if (enabled) {\n      this._transcriptionSink.onAttached();\n    } else {\n      this._transcriptionSink.onDetached();\n    }\n  }\n\n  get audioEnabled(): boolean {\n    return this._audioEnabled;\n  }\n\n  get transcriptionEnabled(): boolean {\n    return this._transcriptionEnabled;\n  }\n\n  get audio(): AudioOutput | null {\n    return this._audioSink;\n  }\n\n  set audio(sink: AudioOutput | null) {\n    if (sink === this._audioSink) {\n      return;\n    }\n\n    if (this._audioSink) {\n      this._audioSink.onDetached();\n    }\n\n    this._audioSink = sink;\n    this.audioChanged();\n\n    if (this._audioSink) {\n      this._audioSink.onAttached();\n    }\n  }\n\n  get transcription(): TextOutput | null {\n    return this._transcriptionSink;\n  }\n\n  set transcription(sink: TextOutput | null) {\n    if (sink === this._transcriptionSink) {\n      return;\n    }\n\n    if (this._transcriptionSink) {\n      this._transcriptionSink.onDetached();\n    }\n\n    this._transcriptionSink = sink;\n    this.transcriptionChanged();\n\n    if (this._transcriptionSink) {\n      this._transcriptionSink.onAttached();\n    }\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,yBAA6B;AAK7B,iBAAoB;AACpB,6BAAuC;AAEvC,mBAAuB;AAkBhB,MAAe,WAAW;AAAA,EACrB,iBACR,IAAI,8CAAmC;AAAA,EAEzC,IAAI,SAAqC;AACvC,WAAO,KAAK,eAAe;AAAA,EAC7B;AAAA,EAEA,aAAmB;AAAA,EAAC;AAAA,EAEpB,aAAmB;AAAA,EAAC;AACtB;AAEO,MAAe,oBAAoB,gCAAa;AAAA,EAarD,YACW,YACU,aACnB;AACA,UAAM;AAHG;AACU;AAGnB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY;AAAA,QAAG,YAAY;AAAA,QAAyB,CAAC,OACxD,KAAK,mBAAmB,EAAE;AAAA,MAC5B;AAAA,IACF;AAAA,EACF;AAAA,EAtBA,OAAgB,0BAA0B;AAAA,EAElC,yBAAuC,IAAI,oBAAO;AAAA,EAClD,aAAsB;AAAA,EACtB,wBAAgC;AAAA,EAChC,wBAAgC;AAAA,EAChC,oBAA2C;AAAA,IACjD,kBAAkB;AAAA,IAClB,aAAa;AAAA,EACf;AAAA,EACU,aAAS,gBAAI;AAAA;AAAA;AAAA;AAAA,EAiBvB,MAAM,aAAa,QAAmC;AACpD,QAAI,CAAC,KAAK,YAAY;AACpB,WAAK,aAAa;AAClB,WAAK;AAAA,IACP;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,MAAM,iBAAiD;AACrD,UAAM,SAAS,KAAK;AAEpB,WAAO,KAAK,wBAAwB,QAAQ;AAC1C,YAAM,KAAK,uBAAuB;AAClC,WAAK,yBAAyB,IAAI,oBAAO;AAAA,IAC3C;AAEA,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,mBAAmB,SAAgC;AACjD,QAAI,KAAK,yBAAyB,KAAK,uBAAuB;AAC5D,WAAK,OAAO,KAAK,0EAA0E;AAC3F;AAAA,IACF;AAEA,SAAK,oBAAoB;AACzB,SAAK;AACL,SAAK,uBAAuB,QAAQ;AACpC,SAAK,KAAK,YAAY,yBAAyB,OAAO;AAAA,EACxD;AAAA,EAEA,QAAc;AACZ,SAAK,aAAa;AAAA,EACpB;AAAA,EAOA,aAAmB;AACjB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,WAAW;AAAA,IAC9B;AAAA,EACF;AAAA,EAEA,aAAmB;AACjB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,WAAW;AAAA,IAC9B;AAAA,EACF;AACF;AAYO,MAAe,WAAW;AAAA,EAC/B,YAA+B,aAA0B;AAA1B;AAAA,EAA2B;AAAA,EAY1D,aAAmB;AACjB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,WAAW;AAAA,IAC9B;AAAA,EACF;AAAA,EAEA,aAAmB;AACjB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,WAAW;AAAA,IAC9B;AAAA,EACF;AACF;AAEO,MAAM,WAAW;AAAA,EAKtB,YAA6B,cAA0B;AAA1B;AAAA,EAA2B;AAAA,EAJhD,eAAkC;AAAA;AAAA,EAElC,gBAAyB;AAAA,EAIjC,gBAAgB,QAAuB;AACrC,QAAI,WAAW,KAAK,eAAe;AACjC;AAAA,IACF;AAEA,SAAK,gBAAgB;AAErB,QAAI,CAAC,KAAK,cAAc;AACtB;AAAA,IACF;AAEA,QAAI,QAAQ;AACV,WAAK,aAAa,WAAW;AAAA,IAC/B,OAAO;AACL,WAAK,aAAa,WAAW;AAAA,IAC/B;AAAA,EACF;AAAA,EAEA,IAAI,eAAwB;AAC1B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,QAA2B;AAC7B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAM,QAA2B;AACnC,SAAK,eAAe;AACpB,SAAK,aAAa;AAAA,EACpB;AACF;AAEO,MAAM,YAAY;AAAA,EAMvB,YACmB,cACA,sBACjB;AAFiB;AACA;AAAA,EAChB;AAAA,EARK,aAAiC;AAAA,EACjC,qBAAwC;AAAA,EACxC,gBAAyB;AAAA,EACzB,wBAAiC;AAAA,EAOzC,gBAAgB,SAAwB;AACtC,QAAI,YAAY,KAAK,eAAe;AAClC;AAAA,IACF;AAEA,SAAK,gBAAgB;AAErB,QAAI,CAAC,KAAK,YAAY;AACpB;AAAA,IACF;AAEA,QAAI,SAAS;AACX,WAAK,WAAW,WAAW;AAAA,IAC7B,OAAO;AACL,WAAK,WAAW,WAAW;AAAA,IAC7B;AAAA,EACF;AAAA,EAEA,wBAAwB,SAAwB;AAC9C,QAAI,YAAY,KAAK,uBAAuB;AAC1C;AAAA,IACF;AAEA,SAAK,wBAAwB;AAE7B,QAAI,CAAC,KAAK,oBAAoB;AAC5B;AAAA,IACF;AAEA,QAAI,SAAS;AACX,WAAK,mBAAmB,WAAW;AAAA,IACrC,OAAO;AACL,WAAK,mBAAmB,WAAW;AAAA,IACrC;AAAA,EACF;AAAA,EAEA,IAAI,eAAwB;AAC1B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,uBAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,QAA4B;AAC9B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAM,MAA0B;AAClC,QAAI,SAAS,KAAK,YAAY;AAC5B;AAAA,IACF;AAEA,QAAI,KAAK,YAAY;AACnB,WAAK,WAAW,WAAW;AAAA,IAC7B;AAEA,SAAK,aAAa;AAClB,SAAK,aAAa;AAElB,QAAI,KAAK,YAAY;AACnB,WAAK,WAAW,WAAW;AAAA,IAC7B;AAAA,EACF;AAAA,EAEA,IAAI,gBAAmC;AACrC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,cAAc,MAAyB;AACzC,QAAI,SAAS,KAAK,oBAAoB;AACpC;AAAA,IACF;AAEA,QAAI,KAAK,oBAAoB;AAC3B,WAAK,mBAAmB,WAAW;AAAA,IACrC;AAEA,SAAK,qBAAqB;AAC1B,SAAK,qBAAqB;AAE1B,QAAI,KAAK,oBAAoB;AAC3B,WAAK,mBAAmB,WAAW;AAAA,IACrC;AAAA,EACF;AACF;","names":[]}
+{"version":3,"sources":["../../src/voice/io.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport type { ChatContext } from '../llm/chat_context.js';\nimport type { ChatChunk } from '../llm/llm.js';\nimport type { ToolContext } from '../llm/tool_context.js';\nimport { log } from '../log.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport type { SpeechEvent } from '../stt/stt.js';\nimport { Future } from '../utils.js';\nimport type { ModelSettings } from './agent.js';\n\nexport type STTNode = (\n  audio: ReadableStream<AudioFrame>,\n  modelSettings: ModelSettings,\n) => Promise<ReadableStream<SpeechEvent | string> | null>;\n\nexport type LLMNode = (\n  chatCtx: ChatContext,\n  toolCtx: ToolContext,\n  modelSettings: ModelSettings,\n) => Promise<ReadableStream<ChatChunk | string> | null>;\n\nexport type TTSNode = (\n  text: ReadableStream<string>,\n  modelSettings: ModelSettings,\n) => Promise<ReadableStream<AudioFrame> | null>;\nexport abstract class AudioInput {\n  protected deferredStream: DeferredReadableStream<AudioFrame> =\n    new DeferredReadableStream<AudioFrame>();\n\n  get stream(): ReadableStream<AudioFrame> {\n    return this.deferredStream.stream;\n  }\n\n  onAttached(): void {}\n\n  onDetached(): void {}\n}\n\nexport abstract class AudioOutput extends EventEmitter {\n  static readonly EVENT_PLAYBACK_FINISHED = 'playbackFinished';\n\n  private playbackFinishedFuture: Future<void> = new Future();\n  private _capturing: boolean = false;\n  private playbackFinishedCount: number = 0;\n  private playbackSegmentsCount: number = 0;\n  private lastPlaybackEvent: PlaybackFinishedEvent = {\n    playbackPosition: 0,\n    interrupted: false,\n  };\n  protected logger = log();\n\n  constructor(\n    public sampleRate?: number,\n    protected readonly nextInChain?: AudioOutput,\n  ) {\n    super();\n    if (this.nextInChain) {\n      this.nextInChain.on(AudioOutput.EVENT_PLAYBACK_FINISHED, (ev: PlaybackFinishedEvent) =>\n        this.onPlaybackFinished(ev),\n      );\n    }\n  }\n\n  /**\n   * Capture an audio frame for playback, frames can be pushed faster than real-time\n   */\n  async captureFrame(_frame: AudioFrame): Promise<void> {\n    if (!this._capturing) {\n      this._capturing = true;\n      this.playbackSegmentsCount++;\n    }\n  }\n\n  /**\n   * Wait for the past audio segments to finish playing out.\n   *\n   * @returns The event that was emitted when the audio finished playing out (only the last segment information)\n   */\n  async waitForPlayout(): Promise<PlaybackFinishedEvent> {\n    const target = this.playbackSegmentsCount;\n\n    while (this.playbackFinishedCount < target) {\n      await this.playbackFinishedFuture.await;\n      this.playbackFinishedFuture = new Future();\n    }\n\n    return this.lastPlaybackEvent;\n  }\n\n  /**\n   * Developers building audio sinks must call this method when a playback/segment is finished.\n   * Segments are segmented by calls to flush() or clearBuffer()\n   */\n  onPlaybackFinished(options: PlaybackFinishedEvent) {\n    if (this.playbackFinishedCount >= this.playbackSegmentsCount) {\n      this.logger.warn('playback_finished called more times than playback segments were captured');\n      return;\n    }\n\n    this.lastPlaybackEvent = options;\n    this.playbackFinishedCount++;\n    this.playbackFinishedFuture.resolve();\n    this.emit(AudioOutput.EVENT_PLAYBACK_FINISHED, options);\n  }\n\n  flush(): void {\n    this._capturing = false;\n  }\n\n  /**\n   * Clear the buffer, stopping playback immediately\n   */\n  abstract clearBuffer(): void;\n\n  onAttached(): void {\n    if (this.nextInChain) {\n      this.nextInChain.onAttached();\n    }\n  }\n\n  onDetached(): void {\n    if (this.nextInChain) {\n      this.nextInChain.onDetached();\n    }\n  }\n}\n\nexport interface PlaybackFinishedEvent {\n  // How much of the audio was played back\n  playbackPosition: number;\n  // Interrupted is True if playback was interrupted (clearBuffer() was called)\n  interrupted: boolean;\n  // Transcript synced with playback; may be partial if the audio was interrupted\n  // When null, the transcript is not synchronized with the playback\n  synchronizedTranscript?: string;\n}\n\nexport abstract class TextOutput {\n  constructor(protected readonly nextInChain?: TextOutput) {}\n\n  /**\n   * Capture a text segment (Used by the output of LLM nodes)\n   */\n  abstract captureText(text: string): Promise<void>;\n\n  /**\n   * Mark the current text segment as complete (e.g LLM generation is complete)\n   */\n  abstract flush(): void;\n\n  onAttached(): void {\n    if (this.nextInChain) {\n      this.nextInChain.onAttached();\n    }\n  }\n\n  onDetached(): void {\n    if (this.nextInChain) {\n      this.nextInChain.onDetached();\n    }\n  }\n}\n\nexport class AgentInput {\n  private _audioStream: AudioInput | null = null;\n  // enabled by default\n  private _audioEnabled: boolean = true;\n\n  constructor(private readonly audioChanged: () => void) {}\n\n  setAudioEnabled(enable: boolean): void {\n    if (enable === this._audioEnabled) {\n      return;\n    }\n\n    this._audioEnabled = enable;\n\n    if (!this._audioStream) {\n      return;\n    }\n\n    if (enable) {\n      this._audioStream.onAttached();\n    } else {\n      this._audioStream.onDetached();\n    }\n  }\n\n  get audioEnabled(): boolean {\n    return this._audioEnabled;\n  }\n\n  get audio(): AudioInput | null {\n    return this._audioStream;\n  }\n\n  set audio(stream: AudioInput | null) {\n    this._audioStream = stream;\n    this.audioChanged();\n  }\n}\n\nexport class AgentOutput {\n  private _audioSink: AudioOutput | null = null;\n  private _transcriptionSink: TextOutput | null = null;\n  private _audioEnabled: boolean = true;\n  private _transcriptionEnabled: boolean = true;\n\n  constructor(\n    private readonly audioChanged: () => void,\n    private readonly transcriptionChanged: () => void,\n  ) {}\n\n  setAudioEnabled(enabled: boolean): void {\n    if (enabled === this._audioEnabled) {\n      return;\n    }\n\n    this._audioEnabled = enabled;\n\n    if (!this._audioSink) {\n      return;\n    }\n\n    if (enabled) {\n      this._audioSink.onAttached();\n    } else {\n      this._audioSink.onDetached();\n    }\n  }\n\n  setTranscriptionEnabled(enabled: boolean): void {\n    if (enabled === this._transcriptionEnabled) {\n      return;\n    }\n\n    this._transcriptionEnabled = enabled;\n\n    if (!this._transcriptionSink) {\n      return;\n    }\n\n    if (enabled) {\n      this._transcriptionSink.onAttached();\n    } else {\n      this._transcriptionSink.onDetached();\n    }\n  }\n\n  get audioEnabled(): boolean {\n    return this._audioEnabled;\n  }\n\n  get transcriptionEnabled(): boolean {\n    return this._transcriptionEnabled;\n  }\n\n  get audio(): AudioOutput | null {\n    return this._audioSink;\n  }\n\n  set audio(sink: AudioOutput | null) {\n    if (sink === this._audioSink) {\n      return;\n    }\n\n    if (this._audioSink) {\n      this._audioSink.onDetached();\n    }\n\n    this._audioSink = sink;\n    this.audioChanged();\n\n    if (this._audioSink) {\n      this._audioSink.onAttached();\n    }\n  }\n\n  get transcription(): TextOutput | null {\n    return this._transcriptionSink;\n  }\n\n  set transcription(sink: TextOutput | null) {\n    if (sink === this._transcriptionSink) {\n      return;\n    }\n\n    if (this._transcriptionSink) {\n      this._transcriptionSink.onDetached();\n    }\n\n    this._transcriptionSink = sink;\n    this.transcriptionChanged();\n\n    if (this._transcriptionSink) {\n      this._transcriptionSink.onAttached();\n    }\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,yBAA6B;AAK7B,iBAAoB;AACpB,6BAAuC;AAEvC,mBAAuB;AAkBhB,MAAe,WAAW;AAAA,EACrB,iBACR,IAAI,8CAAmC;AAAA,EAEzC,IAAI,SAAqC;AACvC,WAAO,KAAK,eAAe;AAAA,EAC7B;AAAA,EAEA,aAAmB;AAAA,EAAC;AAAA,EAEpB,aAAmB;AAAA,EAAC;AACtB;AAEO,MAAe,oBAAoB,gCAAa;AAAA,EAarD,YACS,YACY,aACnB;AACA,UAAM;AAHC;AACY;AAGnB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY;AAAA,QAAG,YAAY;AAAA,QAAyB,CAAC,OACxD,KAAK,mBAAmB,EAAE;AAAA,MAC5B;AAAA,IACF;AAAA,EACF;AAAA,EAtBA,OAAgB,0BAA0B;AAAA,EAElC,yBAAuC,IAAI,oBAAO;AAAA,EAClD,aAAsB;AAAA,EACtB,wBAAgC;AAAA,EAChC,wBAAgC;AAAA,EAChC,oBAA2C;AAAA,IACjD,kBAAkB;AAAA,IAClB,aAAa;AAAA,EACf;AAAA,EACU,aAAS,gBAAI;AAAA;AAAA;AAAA;AAAA,EAiBvB,MAAM,aAAa,QAAmC;AACpD,QAAI,CAAC,KAAK,YAAY;AACpB,WAAK,aAAa;AAClB,WAAK;AAAA,IACP;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,MAAM,iBAAiD;AACrD,UAAM,SAAS,KAAK;AAEpB,WAAO,KAAK,wBAAwB,QAAQ;AAC1C,YAAM,KAAK,uBAAuB;AAClC,WAAK,yBAAyB,IAAI,oBAAO;AAAA,IAC3C;AAEA,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,mBAAmB,SAAgC;AACjD,QAAI,KAAK,yBAAyB,KAAK,uBAAuB;AAC5D,WAAK,OAAO,KAAK,0EAA0E;AAC3F;AAAA,IACF;AAEA,SAAK,oBAAoB;AACzB,SAAK;AACL,SAAK,uBAAuB,QAAQ;AACpC,SAAK,KAAK,YAAY,yBAAyB,OAAO;AAAA,EACxD;AAAA,EAEA,QAAc;AACZ,SAAK,aAAa;AAAA,EACpB;AAAA,EAOA,aAAmB;AACjB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,WAAW;AAAA,IAC9B;AAAA,EACF;AAAA,EAEA,aAAmB;AACjB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,WAAW;AAAA,IAC9B;AAAA,EACF;AACF;AAYO,MAAe,WAAW;AAAA,EAC/B,YAA+B,aAA0B;AAA1B;AAAA,EAA2B;AAAA,EAY1D,aAAmB;AACjB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,WAAW;AAAA,IAC9B;AAAA,EACF;AAAA,EAEA,aAAmB;AACjB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,WAAW;AAAA,IAC9B;AAAA,EACF;AACF;AAEO,MAAM,WAAW;AAAA,EAKtB,YAA6B,cAA0B;AAA1B;AAAA,EAA2B;AAAA,EAJhD,eAAkC;AAAA;AAAA,EAElC,gBAAyB;AAAA,EAIjC,gBAAgB,QAAuB;AACrC,QAAI,WAAW,KAAK,eAAe;AACjC;AAAA,IACF;AAEA,SAAK,gBAAgB;AAErB,QAAI,CAAC,KAAK,cAAc;AACtB;AAAA,IACF;AAEA,QAAI,QAAQ;AACV,WAAK,aAAa,WAAW;AAAA,IAC/B,OAAO;AACL,WAAK,aAAa,WAAW;AAAA,IAC/B;AAAA,EACF;AAAA,EAEA,IAAI,eAAwB;AAC1B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,QAA2B;AAC7B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAM,QAA2B;AACnC,SAAK,eAAe;AACpB,SAAK,aAAa;AAAA,EACpB;AACF;AAEO,MAAM,YAAY;AAAA,EAMvB,YACmB,cACA,sBACjB;AAFiB;AACA;AAAA,EAChB;AAAA,EARK,aAAiC;AAAA,EACjC,qBAAwC;AAAA,EACxC,gBAAyB;AAAA,EACzB,wBAAiC;AAAA,EAOzC,gBAAgB,SAAwB;AACtC,QAAI,YAAY,KAAK,eAAe;AAClC;AAAA,IACF;AAEA,SAAK,gBAAgB;AAErB,QAAI,CAAC,KAAK,YAAY;AACpB;AAAA,IACF;AAEA,QAAI,SAAS;AACX,WAAK,WAAW,WAAW;AAAA,IAC7B,OAAO;AACL,WAAK,WAAW,WAAW;AAAA,IAC7B;AAAA,EACF;AAAA,EAEA,wBAAwB,SAAwB;AAC9C,QAAI,YAAY,KAAK,uBAAuB;AAC1C;AAAA,IACF;AAEA,SAAK,wBAAwB;AAE7B,QAAI,CAAC,KAAK,oBAAoB;AAC5B;AAAA,IACF;AAEA,QAAI,SAAS;AACX,WAAK,mBAAmB,WAAW;AAAA,IACrC,OAAO;AACL,WAAK,mBAAmB,WAAW;AAAA,IACrC;AAAA,EACF;AAAA,EAEA,IAAI,eAAwB;AAC1B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,uBAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,QAA4B;AAC9B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAM,MAA0B;AAClC,QAAI,SAAS,KAAK,YAAY;AAC5B;AAAA,IACF;AAEA,QAAI,KAAK,YAAY;AACnB,WAAK,WAAW,WAAW;AAAA,IAC7B;AAEA,SAAK,aAAa;AAClB,SAAK,aAAa;AAElB,QAAI,KAAK,YAAY;AACnB,WAAK,WAAW,WAAW;AAAA,IAC7B;AAAA,EACF;AAAA,EAEA,IAAI,gBAAmC;AACrC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,cAAc,MAAyB;AACzC,QAAI,SAAS,KAAK,oBAAoB;AACpC;AAAA,IACF;AAEA,QAAI,KAAK,oBAAoB;AAC3B,WAAK,mBAAmB,WAAW;AAAA,IACrC;AAEA,SAAK,qBAAqB;AAC1B,SAAK,qBAAqB;AAE1B,QAAI,KAAK,oBAAoB;AAC3B,WAAK,mBAAmB,WAAW;AAAA,IACrC;AAAA,EACF;AACF;","names":[]}

package/dist/voice/io.d.cts

@@ -19,7 +19,7 @@ export declare abstract class AudioInput {
     onDetached(): void;
 }
 export declare abstract class AudioOutput extends EventEmitter {
-    readonly sampleRate?: number | undefined;
+    sampleRate?: number | undefined;
     protected readonly nextInChain?: AudioOutput | undefined;
     static readonly EVENT_PLAYBACK_FINISHED = "playbackFinished";
     private playbackFinishedFuture;

package/dist/voice/io.d.ts

@@ -19,7 +19,7 @@ export declare abstract class AudioInput {
     onDetached(): void;
 }
 export declare abstract class AudioOutput extends EventEmitter {
-    readonly sampleRate?: number | undefined;
+    sampleRate?: number | undefined;
     protected readonly nextInChain?: AudioOutput | undefined;
     static readonly EVENT_PLAYBACK_FINISHED = "playbackFinished";
     private playbackFinishedFuture;

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

@@ -1 +1 @@
-{"version":3,"file":"io.d.ts","sourceRoot":"","sources":["../../src/voice/io.ts"],"names":[],"mappings":";;AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AACpD,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AACtD,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,wBAAwB,CAAC;AAC1D,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,eAAe,CAAC;AAC/C,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,wBAAwB,CAAC;AAE1D,OAAO,EAAE,sBAAsB,EAAE,MAAM,8BAA8B,CAAC;AACtE,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC;AAEjD,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,YAAY,CAAC;AAEhD,MAAM,MAAM,OAAO,GAAG,CACpB,KAAK,EAAE,cAAc,CAAC,UAAU,CAAC,EACjC,aAAa,EAAE,aAAa,KACzB,OAAO,CAAC,cAAc,CAAC,WAAW,GAAG,MAAM,CAAC,GAAG,IAAI,CAAC,CAAC;AAE1D,MAAM,MAAM,OAAO,GAAG,CACpB,OAAO,EAAE,WAAW,EACpB,OAAO,EAAE,WAAW,EACpB,aAAa,EAAE,aAAa,KACzB,OAAO,CAAC,cAAc,CAAC,SAAS,GAAG,MAAM,CAAC,GAAG,IAAI,CAAC,CAAC;AAExD,MAAM,MAAM,OAAO,GAAG,CACpB,IAAI,EAAE,cAAc,CAAC,MAAM,CAAC,EAC5B,aAAa,EAAE,aAAa,KACzB,OAAO,CAAC,cAAc,CAAC,UAAU,CAAC,GAAG,IAAI,CAAC,CAAC;AAChD,8BAAsB,UAAU;IAC9B,SAAS,CAAC,cAAc,EAAE,sBAAsB,CAAC,UAAU,CAAC,CACjB;IAE3C,IAAI,MAAM,IAAI,cAAc,CAAC,UAAU,CAAC,CAEvC;IAED,UAAU,IAAI,IAAI;IAElB,UAAU,IAAI,IAAI;CACnB;AAED,8BAAsB,WAAY,SAAQ,YAAY;IAclD,QAAQ,CAAC,UAAU,CAAC;IACpB,SAAS,CAAC,QAAQ,CAAC,WAAW,CAAC;IAdjC,MAAM,CAAC,QAAQ,CAAC,uBAAuB,sBAAsB;IAE7D,OAAO,CAAC,sBAAsB,CAA8B;IAC5D,OAAO,CAAC,UAAU,CAAkB;IACpC,OAAO,CAAC,qBAAqB,CAAa;IAC1C,OAAO,CAAC,qBAAqB,CAAa;IAC1C,OAAO,CAAC,iBAAiB,CAGvB;IACF,SAAS,CAAC,MAAM,wBAAS;gBAGd,UAAU,CAAC,oBAAQ,EACT,WAAW,CAAC,yBAAa;IAU9C;;OAEG;IACG,YAAY,CAAC,MAAM,EAAE,UAAU,GAAG,OAAO,CAAC,IAAI,CAAC;IAOrD;;;;OAIG;IACG,cAAc,IAAI,OAAO,CAAC,qBAAqB,CAAC;IAWtD;;;OAGG;IACH,kBAAkB,CAAC,OAAO,EAAE,qBAAqB;IAYjD,KAAK,IAAI,IAAI;IAIb;;OAEG;IACH,QAAQ,CAAC,WAAW,IAAI,IAAI;IAE5B,UAAU,IAAI,IAAI;IAMlB,UAAU,IAAI,IAAI;CAKnB;AAED,MAAM,WAAW,qBAAqB;IAEpC,gBAAgB,EAAE,MAAM,CAAC;IAEzB,WAAW,EAAE,OAAO,CAAC;IAGrB,sBAAsB,CAAC,EAAE,MAAM,CAAC;CACjC;AAED,8BAAsB,UAAU;IAClB,SAAS,CAAC,QAAQ,CAAC,WAAW,CAAC;gBAAZ,WAAW,CAAC,wBAAY;IAEvD;;OAEG;IACH,QAAQ,CAAC,WAAW,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC;IAEjD;;OAEG;IACH,QAAQ,CAAC,KAAK,IAAI,IAAI;IAEtB,UAAU,IAAI,IAAI;IAMlB,UAAU,IAAI,IAAI;CAKnB;AAED,qBAAa,UAAU;IAKT,OAAO,CAAC,QAAQ,CAAC,YAAY;IAJzC,OAAO,CAAC,YAAY,CAA2B;IAE/C,OAAO,CAAC,aAAa,CAAiB;gBAET,YAAY,EAAE,MAAM,IAAI;IAErD,eAAe,CAAC,MAAM,EAAE,OAAO,GAAG,IAAI;IAkBtC,IAAI,YAAY,IAAI,OAAO,CAE1B;IAED,IAAI,KAAK,IAAI,UAAU,GAAG,IAAI,CAE7B;IAED,IAAI,KAAK,CAAC,MAAM,EAAE,UAAU,GAAG,IAAI,EAGlC;CACF;AAED,qBAAa,WAAW;IAOpB,OAAO,CAAC,QAAQ,CAAC,YAAY;IAC7B,OAAO,CAAC,QAAQ,CAAC,oBAAoB;IAPvC,OAAO,CAAC,UAAU,CAA4B;IAC9C,OAAO,CAAC,kBAAkB,CAA2B;IACrD,OAAO,CAAC,aAAa,CAAiB;IACtC,OAAO,CAAC,qBAAqB,CAAiB;gBAG3B,YAAY,EAAE,MAAM,IAAI,EACxB,oBAAoB,EAAE,MAAM,IAAI;IAGnD,eAAe,CAAC,OAAO,EAAE,OAAO,GAAG,IAAI;IAkBvC,uBAAuB,CAAC,OAAO,EAAE,OAAO,GAAG,IAAI;IAkB/C,IAAI,YAAY,IAAI,OAAO,CAE1B;IAED,IAAI,oBAAoB,IAAI,OAAO,CAElC;IAED,IAAI,KAAK,IAAI,WAAW,GAAG,IAAI,CAE9B;IAED,IAAI,KAAK,CAAC,IAAI,EAAE,WAAW,GAAG,IAAI,EAejC;IAED,IAAI,aAAa,IAAI,UAAU,GAAG,IAAI,CAErC;IAED,IAAI,aAAa,CAAC,IAAI,EAAE,UAAU,GAAG,IAAI,EAexC;CACF"}
+{"version":3,"file":"io.d.ts","sourceRoot":"","sources":["../../src/voice/io.ts"],"names":[],"mappings":";;AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AACpD,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAC3C,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AACtD,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,wBAAwB,CAAC;AAC1D,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,eAAe,CAAC;AAC/C,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,wBAAwB,CAAC;AAE1D,OAAO,EAAE,sBAAsB,EAAE,MAAM,8BAA8B,CAAC;AACtE,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC;AAEjD,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,YAAY,CAAC;AAEhD,MAAM,MAAM,OAAO,GAAG,CACpB,KAAK,EAAE,cAAc,CAAC,UAAU,CAAC,EACjC,aAAa,EAAE,aAAa,KACzB,OAAO,CAAC,cAAc,CAAC,WAAW,GAAG,MAAM,CAAC,GAAG,IAAI,CAAC,CAAC;AAE1D,MAAM,MAAM,OAAO,GAAG,CACpB,OAAO,EAAE,WAAW,EACpB,OAAO,EAAE,WAAW,EACpB,aAAa,EAAE,aAAa,KACzB,OAAO,CAAC,cAAc,CAAC,SAAS,GAAG,MAAM,CAAC,GAAG,IAAI,CAAC,CAAC;AAExD,MAAM,MAAM,OAAO,GAAG,CACpB,IAAI,EAAE,cAAc,CAAC,MAAM,CAAC,EAC5B,aAAa,EAAE,aAAa,KACzB,OAAO,CAAC,cAAc,CAAC,UAAU,CAAC,GAAG,IAAI,CAAC,CAAC;AAChD,8BAAsB,UAAU;IAC9B,SAAS,CAAC,cAAc,EAAE,sBAAsB,CAAC,UAAU,CAAC,CACjB;IAE3C,IAAI,MAAM,IAAI,cAAc,CAAC,UAAU,CAAC,CAEvC;IAED,UAAU,IAAI,IAAI;IAElB,UAAU,IAAI,IAAI;CACnB;AAED,8BAAsB,WAAY,SAAQ,YAAY;IAc3C,UAAU,CAAC;IAClB,SAAS,CAAC,QAAQ,CAAC,WAAW,CAAC;IAdjC,MAAM,CAAC,QAAQ,CAAC,uBAAuB,sBAAsB;IAE7D,OAAO,CAAC,sBAAsB,CAA8B;IAC5D,OAAO,CAAC,UAAU,CAAkB;IACpC,OAAO,CAAC,qBAAqB,CAAa;IAC1C,OAAO,CAAC,qBAAqB,CAAa;IAC1C,OAAO,CAAC,iBAAiB,CAGvB;IACF,SAAS,CAAC,MAAM,wBAAS;gBAGhB,UAAU,CAAC,oBAAQ,EACP,WAAW,CAAC,yBAAa;IAU9C;;OAEG;IACG,YAAY,CAAC,MAAM,EAAE,UAAU,GAAG,OAAO,CAAC,IAAI,CAAC;IAOrD;;;;OAIG;IACG,cAAc,IAAI,OAAO,CAAC,qBAAqB,CAAC;IAWtD;;;OAGG;IACH,kBAAkB,CAAC,OAAO,EAAE,qBAAqB;IAYjD,KAAK,IAAI,IAAI;IAIb;;OAEG;IACH,QAAQ,CAAC,WAAW,IAAI,IAAI;IAE5B,UAAU,IAAI,IAAI;IAMlB,UAAU,IAAI,IAAI;CAKnB;AAED,MAAM,WAAW,qBAAqB;IAEpC,gBAAgB,EAAE,MAAM,CAAC;IAEzB,WAAW,EAAE,OAAO,CAAC;IAGrB,sBAAsB,CAAC,EAAE,MAAM,CAAC;CACjC;AAED,8BAAsB,UAAU;IAClB,SAAS,CAAC,QAAQ,CAAC,WAAW,CAAC;gBAAZ,WAAW,CAAC,wBAAY;IAEvD;;OAEG;IACH,QAAQ,CAAC,WAAW,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC;IAEjD;;OAEG;IACH,QAAQ,CAAC,KAAK,IAAI,IAAI;IAEtB,UAAU,IAAI,IAAI;IAMlB,UAAU,IAAI,IAAI;CAKnB;AAED,qBAAa,UAAU;IAKT,OAAO,CAAC,QAAQ,CAAC,YAAY;IAJzC,OAAO,CAAC,YAAY,CAA2B;IAE/C,OAAO,CAAC,aAAa,CAAiB;gBAET,YAAY,EAAE,MAAM,IAAI;IAErD,eAAe,CAAC,MAAM,EAAE,OAAO,GAAG,IAAI;IAkBtC,IAAI,YAAY,IAAI,OAAO,CAE1B;IAED,IAAI,KAAK,IAAI,UAAU,GAAG,IAAI,CAE7B;IAED,IAAI,KAAK,CAAC,MAAM,EAAE,UAAU,GAAG,IAAI,EAGlC;CACF;AAED,qBAAa,WAAW;IAOpB,OAAO,CAAC,QAAQ,CAAC,YAAY;IAC7B,OAAO,CAAC,QAAQ,CAAC,oBAAoB;IAPvC,OAAO,CAAC,UAAU,CAA4B;IAC9C,OAAO,CAAC,kBAAkB,CAA2B;IACrD,OAAO,CAAC,aAAa,CAAiB;IACtC,OAAO,CAAC,qBAAqB,CAAiB;gBAG3B,YAAY,EAAE,MAAM,IAAI,EACxB,oBAAoB,EAAE,MAAM,IAAI;IAGnD,eAAe,CAAC,OAAO,EAAE,OAAO,GAAG,IAAI;IAkBvC,uBAAuB,CAAC,OAAO,EAAE,OAAO,GAAG,IAAI;IAkB/C,IAAI,YAAY,IAAI,OAAO,CAE1B;IAED,IAAI,oBAAoB,IAAI,OAAO,CAElC;IAED,IAAI,KAAK,IAAI,WAAW,GAAG,IAAI,CAE9B;IAED,IAAI,KAAK,CAAC,IAAI,EAAE,WAAW,GAAG,IAAI,EAejC;IAED,IAAI,aAAa,IAAI,UAAU,GAAG,IAAI,CAErC;IAED,IAAI,aAAa,CAAC,IAAI,EAAE,UAAU,GAAG,IAAI,EAexC;CACF"}

package/dist/voice/io.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../../src/voice/io.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport type { ChatContext } from '../llm/chat_context.js';\nimport type { ChatChunk } from '../llm/llm.js';\nimport type { ToolContext } from '../llm/tool_context.js';\nimport { log } from '../log.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport type { SpeechEvent } from '../stt/stt.js';\nimport { Future } from '../utils.js';\nimport type { ModelSettings } from './agent.js';\n\nexport type STTNode = (\n  audio: ReadableStream<AudioFrame>,\n  modelSettings: ModelSettings,\n) => Promise<ReadableStream<SpeechEvent | string> | null>;\n\nexport type LLMNode = (\n  chatCtx: ChatContext,\n  toolCtx: ToolContext,\n  modelSettings: ModelSettings,\n) => Promise<ReadableStream<ChatChunk | string> | null>;\n\nexport type TTSNode = (\n  text: ReadableStream<string>,\n  modelSettings: ModelSettings,\n) => Promise<ReadableStream<AudioFrame> | null>;\nexport abstract class AudioInput {\n  protected deferredStream: DeferredReadableStream<AudioFrame> =\n    new DeferredReadableStream<AudioFrame>();\n\n  get stream(): ReadableStream<AudioFrame> {\n    return this.deferredStream.stream;\n  }\n\n  onAttached(): void {}\n\n  onDetached(): void {}\n}\n\nexport abstract class AudioOutput extends EventEmitter {\n  static readonly EVENT_PLAYBACK_FINISHED = 'playbackFinished';\n\n  private playbackFinishedFuture: Future<void> = new Future();\n  private _capturing: boolean = false;\n  private playbackFinishedCount: number = 0;\n  private playbackSegmentsCount: number = 0;\n  private lastPlaybackEvent: PlaybackFinishedEvent = {\n    playbackPosition: 0,\n    interrupted: false,\n  };\n  protected logger = log();\n\n  constructor(\n    readonly sampleRate?: number,\n    protected readonly nextInChain?: AudioOutput,\n  ) {\n    super();\n    if (this.nextInChain) {\n      this.nextInChain.on(AudioOutput.EVENT_PLAYBACK_FINISHED, (ev: PlaybackFinishedEvent) =>\n        this.onPlaybackFinished(ev),\n      );\n    }\n  }\n\n  /**\n   * Capture an audio frame for playback, frames can be pushed faster than real-time\n   */\n  async captureFrame(_frame: AudioFrame): Promise<void> {\n    if (!this._capturing) {\n      this._capturing = true;\n      this.playbackSegmentsCount++;\n    }\n  }\n\n  /**\n   * Wait for the past audio segments to finish playing out.\n   *\n   * @returns The event that was emitted when the audio finished playing out (only the last segment information)\n   */\n  async waitForPlayout(): Promise<PlaybackFinishedEvent> {\n    const target = this.playbackSegmentsCount;\n\n    while (this.playbackFinishedCount < target) {\n      await this.playbackFinishedFuture.await;\n      this.playbackFinishedFuture = new Future();\n    }\n\n    return this.lastPlaybackEvent;\n  }\n\n  /**\n   * Developers building audio sinks must call this method when a playback/segment is finished.\n   * Segments are segmented by calls to flush() or clearBuffer()\n   */\n  onPlaybackFinished(options: PlaybackFinishedEvent) {\n    if (this.playbackFinishedCount >= this.playbackSegmentsCount) {\n      this.logger.warn('playback_finished called more times than playback segments were captured');\n      return;\n    }\n\n    this.lastPlaybackEvent = options;\n    this.playbackFinishedCount++;\n    this.playbackFinishedFuture.resolve();\n    this.emit(AudioOutput.EVENT_PLAYBACK_FINISHED, options);\n  }\n\n  flush(): void {\n    this._capturing = false;\n  }\n\n  /**\n   * Clear the buffer, stopping playback immediately\n   */\n  abstract clearBuffer(): void;\n\n  onAttached(): void {\n    if (this.nextInChain) {\n      this.nextInChain.onAttached();\n    }\n  }\n\n  onDetached(): void {\n    if (this.nextInChain) {\n      this.nextInChain.onDetached();\n    }\n  }\n}\n\nexport interface PlaybackFinishedEvent {\n  // How much of the audio was played back\n  playbackPosition: number;\n  // Interrupted is True if playback was interrupted (clearBuffer() was called)\n  interrupted: boolean;\n  // Transcript synced with playback; may be partial if the audio was interrupted\n  // When null, the transcript is not synchronized with the playback\n  synchronizedTranscript?: string;\n}\n\nexport abstract class TextOutput {\n  constructor(protected readonly nextInChain?: TextOutput) {}\n\n  /**\n   * Capture a text segment (Used by the output of LLM nodes)\n   */\n  abstract captureText(text: string): Promise<void>;\n\n  /**\n   * Mark the current text segment as complete (e.g LLM generation is complete)\n   */\n  abstract flush(): void;\n\n  onAttached(): void {\n    if (this.nextInChain) {\n      this.nextInChain.onAttached();\n    }\n  }\n\n  onDetached(): void {\n    if (this.nextInChain) {\n      this.nextInChain.onDetached();\n    }\n  }\n}\n\nexport class AgentInput {\n  private _audioStream: AudioInput | null = null;\n  // enabled by default\n  private _audioEnabled: boolean = true;\n\n  constructor(private readonly audioChanged: () => void) {}\n\n  setAudioEnabled(enable: boolean): void {\n    if (enable === this._audioEnabled) {\n      return;\n    }\n\n    this._audioEnabled = enable;\n\n    if (!this._audioStream) {\n      return;\n    }\n\n    if (enable) {\n      this._audioStream.onAttached();\n    } else {\n      this._audioStream.onDetached();\n    }\n  }\n\n  get audioEnabled(): boolean {\n    return this._audioEnabled;\n  }\n\n  get audio(): AudioInput | null {\n    return this._audioStream;\n  }\n\n  set audio(stream: AudioInput | null) {\n    this._audioStream = stream;\n    this.audioChanged();\n  }\n}\n\nexport class AgentOutput {\n  private _audioSink: AudioOutput | null = null;\n  private _transcriptionSink: TextOutput | null = null;\n  private _audioEnabled: boolean = true;\n  private _transcriptionEnabled: boolean = true;\n\n  constructor(\n    private readonly audioChanged: () => void,\n    private readonly transcriptionChanged: () => void,\n  ) {}\n\n  setAudioEnabled(enabled: boolean): void {\n    if (enabled === this._audioEnabled) {\n      return;\n    }\n\n    this._audioEnabled = enabled;\n\n    if (!this._audioSink) {\n      return;\n    }\n\n    if (enabled) {\n      this._audioSink.onAttached();\n    } else {\n      this._audioSink.onDetached();\n    }\n  }\n\n  setTranscriptionEnabled(enabled: boolean): void {\n    if (enabled === this._transcriptionEnabled) {\n      return;\n    }\n\n    this._transcriptionEnabled = enabled;\n\n    if (!this._transcriptionSink) {\n      return;\n    }\n\n    if (enabled) {\n      this._transcriptionSink.onAttached();\n    } else {\n      this._transcriptionSink.onDetached();\n    }\n  }\n\n  get audioEnabled(): boolean {\n    return this._audioEnabled;\n  }\n\n  get transcriptionEnabled(): boolean {\n    return this._transcriptionEnabled;\n  }\n\n  get audio(): AudioOutput | null {\n    return this._audioSink;\n  }\n\n  set audio(sink: AudioOutput | null) {\n    if (sink === this._audioSink) {\n      return;\n    }\n\n    if (this._audioSink) {\n      this._audioSink.onDetached();\n    }\n\n    this._audioSink = sink;\n    this.audioChanged();\n\n    if (this._audioSink) {\n      this._audioSink.onAttached();\n    }\n  }\n\n  get transcription(): TextOutput | null {\n    return this._transcriptionSink;\n  }\n\n  set transcription(sink: TextOutput | null) {\n    if (sink === this._transcriptionSink) {\n      return;\n    }\n\n    if (this._transcriptionSink) {\n      this._transcriptionSink.onDetached();\n    }\n\n    this._transcriptionSink = sink;\n    this.transcriptionChanged();\n\n    if (this._transcriptionSink) {\n      this._transcriptionSink.onAttached();\n    }\n  }\n}\n"],"mappings":"AAIA,SAAS,oBAAoB;AAK7B,SAAS,WAAW;AACpB,SAAS,8BAA8B;AAEvC,SAAS,cAAc;AAkBhB,MAAe,WAAW;AAAA,EACrB,iBACR,IAAI,uBAAmC;AAAA,EAEzC,IAAI,SAAqC;AACvC,WAAO,KAAK,eAAe;AAAA,EAC7B;AAAA,EAEA,aAAmB;AAAA,EAAC;AAAA,EAEpB,aAAmB;AAAA,EAAC;AACtB;AAEO,MAAe,oBAAoB,aAAa;AAAA,EAarD,YACW,YACU,aACnB;AACA,UAAM;AAHG;AACU;AAGnB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY;AAAA,QAAG,YAAY;AAAA,QAAyB,CAAC,OACxD,KAAK,mBAAmB,EAAE;AAAA,MAC5B;AAAA,IACF;AAAA,EACF;AAAA,EAtBA,OAAgB,0BAA0B;AAAA,EAElC,yBAAuC,IAAI,OAAO;AAAA,EAClD,aAAsB;AAAA,EACtB,wBAAgC;AAAA,EAChC,wBAAgC;AAAA,EAChC,oBAA2C;AAAA,IACjD,kBAAkB;AAAA,IAClB,aAAa;AAAA,EACf;AAAA,EACU,SAAS,IAAI;AAAA;AAAA;AAAA;AAAA,EAiBvB,MAAM,aAAa,QAAmC;AACpD,QAAI,CAAC,KAAK,YAAY;AACpB,WAAK,aAAa;AAClB,WAAK;AAAA,IACP;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,MAAM,iBAAiD;AACrD,UAAM,SAAS,KAAK;AAEpB,WAAO,KAAK,wBAAwB,QAAQ;AAC1C,YAAM,KAAK,uBAAuB;AAClC,WAAK,yBAAyB,IAAI,OAAO;AAAA,IAC3C;AAEA,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,mBAAmB,SAAgC;AACjD,QAAI,KAAK,yBAAyB,KAAK,uBAAuB;AAC5D,WAAK,OAAO,KAAK,0EAA0E;AAC3F;AAAA,IACF;AAEA,SAAK,oBAAoB;AACzB,SAAK;AACL,SAAK,uBAAuB,QAAQ;AACpC,SAAK,KAAK,YAAY,yBAAyB,OAAO;AAAA,EACxD;AAAA,EAEA,QAAc;AACZ,SAAK,aAAa;AAAA,EACpB;AAAA,EAOA,aAAmB;AACjB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,WAAW;AAAA,IAC9B;AAAA,EACF;AAAA,EAEA,aAAmB;AACjB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,WAAW;AAAA,IAC9B;AAAA,EACF;AACF;AAYO,MAAe,WAAW;AAAA,EAC/B,YAA+B,aAA0B;AAA1B;AAAA,EAA2B;AAAA,EAY1D,aAAmB;AACjB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,WAAW;AAAA,IAC9B;AAAA,EACF;AAAA,EAEA,aAAmB;AACjB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,WAAW;AAAA,IAC9B;AAAA,EACF;AACF;AAEO,MAAM,WAAW;AAAA,EAKtB,YAA6B,cAA0B;AAA1B;AAAA,EAA2B;AAAA,EAJhD,eAAkC;AAAA;AAAA,EAElC,gBAAyB;AAAA,EAIjC,gBAAgB,QAAuB;AACrC,QAAI,WAAW,KAAK,eAAe;AACjC;AAAA,IACF;AAEA,SAAK,gBAAgB;AAErB,QAAI,CAAC,KAAK,cAAc;AACtB;AAAA,IACF;AAEA,QAAI,QAAQ;AACV,WAAK,aAAa,WAAW;AAAA,IAC/B,OAAO;AACL,WAAK,aAAa,WAAW;AAAA,IAC/B;AAAA,EACF;AAAA,EAEA,IAAI,eAAwB;AAC1B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,QAA2B;AAC7B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAM,QAA2B;AACnC,SAAK,eAAe;AACpB,SAAK,aAAa;AAAA,EACpB;AACF;AAEO,MAAM,YAAY;AAAA,EAMvB,YACmB,cACA,sBACjB;AAFiB;AACA;AAAA,EAChB;AAAA,EARK,aAAiC;AAAA,EACjC,qBAAwC;AAAA,EACxC,gBAAyB;AAAA,EACzB,wBAAiC;AAAA,EAOzC,gBAAgB,SAAwB;AACtC,QAAI,YAAY,KAAK,eAAe;AAClC;AAAA,IACF;AAEA,SAAK,gBAAgB;AAErB,QAAI,CAAC,KAAK,YAAY;AACpB;AAAA,IACF;AAEA,QAAI,SAAS;AACX,WAAK,WAAW,WAAW;AAAA,IAC7B,OAAO;AACL,WAAK,WAAW,WAAW;AAAA,IAC7B;AAAA,EACF;AAAA,EAEA,wBAAwB,SAAwB;AAC9C,QAAI,YAAY,KAAK,uBAAuB;AAC1C;AAAA,IACF;AAEA,SAAK,wBAAwB;AAE7B,QAAI,CAAC,KAAK,oBAAoB;AAC5B;AAAA,IACF;AAEA,QAAI,SAAS;AACX,WAAK,mBAAmB,WAAW;AAAA,IACrC,OAAO;AACL,WAAK,mBAAmB,WAAW;AAAA,IACrC;AAAA,EACF;AAAA,EAEA,IAAI,eAAwB;AAC1B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,uBAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,QAA4B;AAC9B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAM,MAA0B;AAClC,QAAI,SAAS,KAAK,YAAY;AAC5B;AAAA,IACF;AAEA,QAAI,KAAK,YAAY;AACnB,WAAK,WAAW,WAAW;AAAA,IAC7B;AAEA,SAAK,aAAa;AAClB,SAAK,aAAa;AAElB,QAAI,KAAK,YAAY;AACnB,WAAK,WAAW,WAAW;AAAA,IAC7B;AAAA,EACF;AAAA,EAEA,IAAI,gBAAmC;AACrC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,cAAc,MAAyB;AACzC,QAAI,SAAS,KAAK,oBAAoB;AACpC;AAAA,IACF;AAEA,QAAI,KAAK,oBAAoB;AAC3B,WAAK,mBAAmB,WAAW;AAAA,IACrC;AAEA,SAAK,qBAAqB;AAC1B,SAAK,qBAAqB;AAE1B,QAAI,KAAK,oBAAoB;AAC3B,WAAK,mBAAmB,WAAW;AAAA,IACrC;AAAA,EACF;AACF;","names":[]}
+{"version":3,"sources":["../../src/voice/io.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport type { ChatContext } from '../llm/chat_context.js';\nimport type { ChatChunk } from '../llm/llm.js';\nimport type { ToolContext } from '../llm/tool_context.js';\nimport { log } from '../log.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport type { SpeechEvent } from '../stt/stt.js';\nimport { Future } from '../utils.js';\nimport type { ModelSettings } from './agent.js';\n\nexport type STTNode = (\n  audio: ReadableStream<AudioFrame>,\n  modelSettings: ModelSettings,\n) => Promise<ReadableStream<SpeechEvent | string> | null>;\n\nexport type LLMNode = (\n  chatCtx: ChatContext,\n  toolCtx: ToolContext,\n  modelSettings: ModelSettings,\n) => Promise<ReadableStream<ChatChunk | string> | null>;\n\nexport type TTSNode = (\n  text: ReadableStream<string>,\n  modelSettings: ModelSettings,\n) => Promise<ReadableStream<AudioFrame> | null>;\nexport abstract class AudioInput {\n  protected deferredStream: DeferredReadableStream<AudioFrame> =\n    new DeferredReadableStream<AudioFrame>();\n\n  get stream(): ReadableStream<AudioFrame> {\n    return this.deferredStream.stream;\n  }\n\n  onAttached(): void {}\n\n  onDetached(): void {}\n}\n\nexport abstract class AudioOutput extends EventEmitter {\n  static readonly EVENT_PLAYBACK_FINISHED = 'playbackFinished';\n\n  private playbackFinishedFuture: Future<void> = new Future();\n  private _capturing: boolean = false;\n  private playbackFinishedCount: number = 0;\n  private playbackSegmentsCount: number = 0;\n  private lastPlaybackEvent: PlaybackFinishedEvent = {\n    playbackPosition: 0,\n    interrupted: false,\n  };\n  protected logger = log();\n\n  constructor(\n    public sampleRate?: number,\n    protected readonly nextInChain?: AudioOutput,\n  ) {\n    super();\n    if (this.nextInChain) {\n      this.nextInChain.on(AudioOutput.EVENT_PLAYBACK_FINISHED, (ev: PlaybackFinishedEvent) =>\n        this.onPlaybackFinished(ev),\n      );\n    }\n  }\n\n  /**\n   * Capture an audio frame for playback, frames can be pushed faster than real-time\n   */\n  async captureFrame(_frame: AudioFrame): Promise<void> {\n    if (!this._capturing) {\n      this._capturing = true;\n      this.playbackSegmentsCount++;\n    }\n  }\n\n  /**\n   * Wait for the past audio segments to finish playing out.\n   *\n   * @returns The event that was emitted when the audio finished playing out (only the last segment information)\n   */\n  async waitForPlayout(): Promise<PlaybackFinishedEvent> {\n    const target = this.playbackSegmentsCount;\n\n    while (this.playbackFinishedCount < target) {\n      await this.playbackFinishedFuture.await;\n      this.playbackFinishedFuture = new Future();\n    }\n\n    return this.lastPlaybackEvent;\n  }\n\n  /**\n   * Developers building audio sinks must call this method when a playback/segment is finished.\n   * Segments are segmented by calls to flush() or clearBuffer()\n   */\n  onPlaybackFinished(options: PlaybackFinishedEvent) {\n    if (this.playbackFinishedCount >= this.playbackSegmentsCount) {\n      this.logger.warn('playback_finished called more times than playback segments were captured');\n      return;\n    }\n\n    this.lastPlaybackEvent = options;\n    this.playbackFinishedCount++;\n    this.playbackFinishedFuture.resolve();\n    this.emit(AudioOutput.EVENT_PLAYBACK_FINISHED, options);\n  }\n\n  flush(): void {\n    this._capturing = false;\n  }\n\n  /**\n   * Clear the buffer, stopping playback immediately\n   */\n  abstract clearBuffer(): void;\n\n  onAttached(): void {\n    if (this.nextInChain) {\n      this.nextInChain.onAttached();\n    }\n  }\n\n  onDetached(): void {\n    if (this.nextInChain) {\n      this.nextInChain.onDetached();\n    }\n  }\n}\n\nexport interface PlaybackFinishedEvent {\n  // How much of the audio was played back\n  playbackPosition: number;\n  // Interrupted is True if playback was interrupted (clearBuffer() was called)\n  interrupted: boolean;\n  // Transcript synced with playback; may be partial if the audio was interrupted\n  // When null, the transcript is not synchronized with the playback\n  synchronizedTranscript?: string;\n}\n\nexport abstract class TextOutput {\n  constructor(protected readonly nextInChain?: TextOutput) {}\n\n  /**\n   * Capture a text segment (Used by the output of LLM nodes)\n   */\n  abstract captureText(text: string): Promise<void>;\n\n  /**\n   * Mark the current text segment as complete (e.g LLM generation is complete)\n   */\n  abstract flush(): void;\n\n  onAttached(): void {\n    if (this.nextInChain) {\n      this.nextInChain.onAttached();\n    }\n  }\n\n  onDetached(): void {\n    if (this.nextInChain) {\n      this.nextInChain.onDetached();\n    }\n  }\n}\n\nexport class AgentInput {\n  private _audioStream: AudioInput | null = null;\n  // enabled by default\n  private _audioEnabled: boolean = true;\n\n  constructor(private readonly audioChanged: () => void) {}\n\n  setAudioEnabled(enable: boolean): void {\n    if (enable === this._audioEnabled) {\n      return;\n    }\n\n    this._audioEnabled = enable;\n\n    if (!this._audioStream) {\n      return;\n    }\n\n    if (enable) {\n      this._audioStream.onAttached();\n    } else {\n      this._audioStream.onDetached();\n    }\n  }\n\n  get audioEnabled(): boolean {\n    return this._audioEnabled;\n  }\n\n  get audio(): AudioInput | null {\n    return this._audioStream;\n  }\n\n  set audio(stream: AudioInput | null) {\n    this._audioStream = stream;\n    this.audioChanged();\n  }\n}\n\nexport class AgentOutput {\n  private _audioSink: AudioOutput | null = null;\n  private _transcriptionSink: TextOutput | null = null;\n  private _audioEnabled: boolean = true;\n  private _transcriptionEnabled: boolean = true;\n\n  constructor(\n    private readonly audioChanged: () => void,\n    private readonly transcriptionChanged: () => void,\n  ) {}\n\n  setAudioEnabled(enabled: boolean): void {\n    if (enabled === this._audioEnabled) {\n      return;\n    }\n\n    this._audioEnabled = enabled;\n\n    if (!this._audioSink) {\n      return;\n    }\n\n    if (enabled) {\n      this._audioSink.onAttached();\n    } else {\n      this._audioSink.onDetached();\n    }\n  }\n\n  setTranscriptionEnabled(enabled: boolean): void {\n    if (enabled === this._transcriptionEnabled) {\n      return;\n    }\n\n    this._transcriptionEnabled = enabled;\n\n    if (!this._transcriptionSink) {\n      return;\n    }\n\n    if (enabled) {\n      this._transcriptionSink.onAttached();\n    } else {\n      this._transcriptionSink.onDetached();\n    }\n  }\n\n  get audioEnabled(): boolean {\n    return this._audioEnabled;\n  }\n\n  get transcriptionEnabled(): boolean {\n    return this._transcriptionEnabled;\n  }\n\n  get audio(): AudioOutput | null {\n    return this._audioSink;\n  }\n\n  set audio(sink: AudioOutput | null) {\n    if (sink === this._audioSink) {\n      return;\n    }\n\n    if (this._audioSink) {\n      this._audioSink.onDetached();\n    }\n\n    this._audioSink = sink;\n    this.audioChanged();\n\n    if (this._audioSink) {\n      this._audioSink.onAttached();\n    }\n  }\n\n  get transcription(): TextOutput | null {\n    return this._transcriptionSink;\n  }\n\n  set transcription(sink: TextOutput | null) {\n    if (sink === this._transcriptionSink) {\n      return;\n    }\n\n    if (this._transcriptionSink) {\n      this._transcriptionSink.onDetached();\n    }\n\n    this._transcriptionSink = sink;\n    this.transcriptionChanged();\n\n    if (this._transcriptionSink) {\n      this._transcriptionSink.onAttached();\n    }\n  }\n}\n"],"mappings":"AAIA,SAAS,oBAAoB;AAK7B,SAAS,WAAW;AACpB,SAAS,8BAA8B;AAEvC,SAAS,cAAc;AAkBhB,MAAe,WAAW;AAAA,EACrB,iBACR,IAAI,uBAAmC;AAAA,EAEzC,IAAI,SAAqC;AACvC,WAAO,KAAK,eAAe;AAAA,EAC7B;AAAA,EAEA,aAAmB;AAAA,EAAC;AAAA,EAEpB,aAAmB;AAAA,EAAC;AACtB;AAEO,MAAe,oBAAoB,aAAa;AAAA,EAarD,YACS,YACY,aACnB;AACA,UAAM;AAHC;AACY;AAGnB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY;AAAA,QAAG,YAAY;AAAA,QAAyB,CAAC,OACxD,KAAK,mBAAmB,EAAE;AAAA,MAC5B;AAAA,IACF;AAAA,EACF;AAAA,EAtBA,OAAgB,0BAA0B;AAAA,EAElC,yBAAuC,IAAI,OAAO;AAAA,EAClD,aAAsB;AAAA,EACtB,wBAAgC;AAAA,EAChC,wBAAgC;AAAA,EAChC,oBAA2C;AAAA,IACjD,kBAAkB;AAAA,IAClB,aAAa;AAAA,EACf;AAAA,EACU,SAAS,IAAI;AAAA;AAAA;AAAA;AAAA,EAiBvB,MAAM,aAAa,QAAmC;AACpD,QAAI,CAAC,KAAK,YAAY;AACpB,WAAK,aAAa;AAClB,WAAK;AAAA,IACP;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,MAAM,iBAAiD;AACrD,UAAM,SAAS,KAAK;AAEpB,WAAO,KAAK,wBAAwB,QAAQ;AAC1C,YAAM,KAAK,uBAAuB;AAClC,WAAK,yBAAyB,IAAI,OAAO;AAAA,IAC3C;AAEA,WAAO,KAAK;AAAA,EACd;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,mBAAmB,SAAgC;AACjD,QAAI,KAAK,yBAAyB,KAAK,uBAAuB;AAC5D,WAAK,OAAO,KAAK,0EAA0E;AAC3F;AAAA,IACF;AAEA,SAAK,oBAAoB;AACzB,SAAK;AACL,SAAK,uBAAuB,QAAQ;AACpC,SAAK,KAAK,YAAY,yBAAyB,OAAO;AAAA,EACxD;AAAA,EAEA,QAAc;AACZ,SAAK,aAAa;AAAA,EACpB;AAAA,EAOA,aAAmB;AACjB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,WAAW;AAAA,IAC9B;AAAA,EACF;AAAA,EAEA,aAAmB;AACjB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,WAAW;AAAA,IAC9B;AAAA,EACF;AACF;AAYO,MAAe,WAAW;AAAA,EAC/B,YAA+B,aAA0B;AAA1B;AAAA,EAA2B;AAAA,EAY1D,aAAmB;AACjB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,WAAW;AAAA,IAC9B;AAAA,EACF;AAAA,EAEA,aAAmB;AACjB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,WAAW;AAAA,IAC9B;AAAA,EACF;AACF;AAEO,MAAM,WAAW;AAAA,EAKtB,YAA6B,cAA0B;AAA1B;AAAA,EAA2B;AAAA,EAJhD,eAAkC;AAAA;AAAA,EAElC,gBAAyB;AAAA,EAIjC,gBAAgB,QAAuB;AACrC,QAAI,WAAW,KAAK,eAAe;AACjC;AAAA,IACF;AAEA,SAAK,gBAAgB;AAErB,QAAI,CAAC,KAAK,cAAc;AACtB;AAAA,IACF;AAEA,QAAI,QAAQ;AACV,WAAK,aAAa,WAAW;AAAA,IAC/B,OAAO;AACL,WAAK,aAAa,WAAW;AAAA,IAC/B;AAAA,EACF;AAAA,EAEA,IAAI,eAAwB;AAC1B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,QAA2B;AAC7B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAM,QAA2B;AACnC,SAAK,eAAe;AACpB,SAAK,aAAa;AAAA,EACpB;AACF;AAEO,MAAM,YAAY;AAAA,EAMvB,YACmB,cACA,sBACjB;AAFiB;AACA;AAAA,EAChB;AAAA,EARK,aAAiC;AAAA,EACjC,qBAAwC;AAAA,EACxC,gBAAyB;AAAA,EACzB,wBAAiC;AAAA,EAOzC,gBAAgB,SAAwB;AACtC,QAAI,YAAY,KAAK,eAAe;AAClC;AAAA,IACF;AAEA,SAAK,gBAAgB;AAErB,QAAI,CAAC,KAAK,YAAY;AACpB;AAAA,IACF;AAEA,QAAI,SAAS;AACX,WAAK,WAAW,WAAW;AAAA,IAC7B,OAAO;AACL,WAAK,WAAW,WAAW;AAAA,IAC7B;AAAA,EACF;AAAA,EAEA,wBAAwB,SAAwB;AAC9C,QAAI,YAAY,KAAK,uBAAuB;AAC1C;AAAA,IACF;AAEA,SAAK,wBAAwB;AAE7B,QAAI,CAAC,KAAK,oBAAoB;AAC5B;AAAA,IACF;AAEA,QAAI,SAAS;AACX,WAAK,mBAAmB,WAAW;AAAA,IACrC,OAAO;AACL,WAAK,mBAAmB,WAAW;AAAA,IACrC;AAAA,EACF;AAAA,EAEA,IAAI,eAAwB;AAC1B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,uBAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,QAA4B;AAC9B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAM,MAA0B;AAClC,QAAI,SAAS,KAAK,YAAY;AAC5B;AAAA,IACF;AAEA,QAAI,KAAK,YAAY;AACnB,WAAK,WAAW,WAAW;AAAA,IAC7B;AAEA,SAAK,aAAa;AAClB,SAAK,aAAa;AAElB,QAAI,KAAK,YAAY;AACnB,WAAK,WAAW,WAAW;AAAA,IAC7B;AAAA,EACF;AAAA,EAEA,IAAI,gBAAmC;AACrC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,cAAc,MAAyB;AACzC,QAAI,SAAS,KAAK,oBAAoB;AACpC;AAAA,IACF;AAEA,QAAI,KAAK,oBAAoB;AAC3B,WAAK,mBAAmB,WAAW;AAAA,IACrC;AAEA,SAAK,qBAAqB;AAC1B,SAAK,qBAAqB;AAE1B,QAAI,KAAK,oBAAoB;AAC3B,WAAK,mBAAmB,WAAW;AAAA,IACrC;AAAA,EACF;AACF;","names":[]}

package/dist/voice/room_io/_input.cjs

@@ -60,7 +60,7 @@ class ParticipantAudioInputStream extends import_io.AudioInput {
     }
     const participantValue = participant instanceof import_rtc_node.RemoteParticipant ? participant : this.room.remoteParticipants.get(participantIdentity);
     if (participantValue) {
-      for (const publication of Object.values(participantValue.trackPublications)) {
+      for (const publication of participantValue.trackPublications.values()) {
         if (publication.track && publication.source === import_rtc_node.TrackSource.SOURCE_MICROPHONE) {
           this.onTrackSubscribed(publication.track, publication, participantValue);
           break;
@@ -87,6 +87,7 @@ class ParticipantAudioInputStream extends import_io.AudioInput {
     this.publication = null;
   }
   onTrackSubscribed = (track, publication, participant) => {
+    this.logger.debug({ participant: participant.identity }, "onTrackSubscribed in _input");
     if (this.participantIdentity !== participant.identity || publication.source !== import_rtc_node.TrackSource.SOURCE_MICROPHONE || this.publication && this.publication.sid === publication.sid) {
       return false;
     }
@@ -105,6 +106,7 @@ class ParticipantAudioInputStream extends import_io.AudioInput {
       sampleRate: this.sampleRate,
       numChannels: this.numChannels,
       noiseCancellation: this.noiseCancellation
+      // TODO(AJS-269): resolve compatibility issue with node-sdk to remove the forced type casting
     });
   }
   async close() {

package/dist/voice/room_io/_input.cjs.map

@@ -1 +1 @@
-{"version":3,"sources":["../../../src/voice/room_io/_input.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n  AudioStream,\n  type NoiseCancellationOptions,\n  RemoteParticipant,\n  type RemoteTrack,\n  type RemoteTrackPublication,\n  type Room,\n  RoomEvent,\n  TrackSource,\n} from '@livekit/rtc-node';\nimport { log } from '../../log.js';\nimport { resampleStream } from '../../utils.js';\nimport { AudioInput } from '../io.js';\n\nexport class ParticipantAudioInputStream extends AudioInput {\n  private room: Room;\n  private sampleRate: number;\n  private numChannels: number;\n  private noiseCancellation?: NoiseCancellationOptions;\n  private publication: RemoteTrackPublication | null = null;\n  private participantIdentity: string | null = null;\n  private logger = log();\n  constructor({\n    room,\n    sampleRate,\n    numChannels,\n    noiseCancellation,\n  }: {\n    room: Room;\n    sampleRate: number;\n    numChannels: number;\n    noiseCancellation?: NoiseCancellationOptions;\n  }) {\n    super();\n    this.room = room;\n    this.sampleRate = sampleRate;\n    this.numChannels = numChannels;\n    this.noiseCancellation = noiseCancellation;\n\n    this.room.on(RoomEvent.TrackSubscribed, this.onTrackSubscribed);\n    this.room.on(RoomEvent.TrackUnpublished, this.onTrackUnpublished);\n  }\n\n  setParticipant(participant: RemoteParticipant | string | null) {\n    this.logger.debug({ participant }, 'setting participant audio input');\n    const participantIdentity =\n      participant instanceof RemoteParticipant ? participant.identity : participant;\n\n    if (this.participantIdentity === participantIdentity) {\n      return;\n    }\n    this.participantIdentity = participantIdentity;\n    this.closeStream();\n\n    if (!participantIdentity) {\n      return;\n    }\n\n    const participantValue =\n      participant instanceof RemoteParticipant\n        ? participant\n        : this.room.remoteParticipants.get(participantIdentity);\n\n    if (participantValue) {\n      for (const publication of Object.values(participantValue.trackPublications)) {\n        if (publication.track && publication.source === TrackSource.SOURCE_MICROPHONE) {\n          this.onTrackSubscribed(publication.track, publication, participantValue);\n          break;\n        }\n      }\n    }\n  }\n\n  private onTrackUnpublished = (\n    publication: RemoteTrackPublication,\n    participant: RemoteParticipant,\n  ) => {\n    if (\n      this.publication?.sid !== publication.sid ||\n      participant.identity !== this.participantIdentity\n    ) {\n      return;\n    }\n    this.closeStream();\n\n    // subscribe to the first available track\n    for (const publication of participant.trackPublications.values()) {\n      if (\n        publication.track &&\n        this.onTrackSubscribed(publication.track, publication, participant)\n      ) {\n        return;\n      }\n    }\n  };\n\n  private closeStream() {\n    if (this.deferredStream.isSourceSet) {\n      this.deferredStream.detachSource();\n    }\n    this.publication = null;\n  }\n\n  private onTrackSubscribed = (\n    track: RemoteTrack,\n    publication: RemoteTrackPublication,\n    participant: RemoteParticipant,\n  ): boolean => {\n    if (\n      this.participantIdentity !== participant.identity ||\n      publication.source !== TrackSource.SOURCE_MICROPHONE ||\n      (this.publication && this.publication.sid === publication.sid)\n    ) {\n      return false;\n    }\n    this.closeStream();\n    this.publication = publication;\n    this.deferredStream.setSource(\n      resampleStream({\n        stream: this.createStream(track),\n        outputRate: this.sampleRate,\n      }),\n    );\n    return true;\n  };\n\n  private createStream(track: RemoteTrack) {\n    return new AudioStream(track, {\n      sampleRate: this.sampleRate,\n      numChannels: this.numChannels,\n      noiseCancellation: this.noiseCancellation,\n    });\n  }\n\n  async close() {\n    this.room.off(RoomEvent.TrackSubscribed, this.onTrackSubscribed);\n    this.room.off(RoomEvent.TrackUnpublished, this.onTrackUnpublished);\n    this.closeStream();\n    this.deferredStream.stream.cancel();\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBASO;AACP,iBAAoB;AACpB,mBAA+B;AAC/B,gBAA2B;AAEpB,MAAM,oCAAoC,qBAAW;AAAA,EAClD;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,cAA6C;AAAA,EAC7C,sBAAqC;AAAA,EACrC,aAAS,gBAAI;AAAA,EACrB,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAKG;AACD,UAAM;AACN,SAAK,OAAO;AACZ,SAAK,aAAa;AAClB,SAAK,cAAc;AACnB,SAAK,oBAAoB;AAEzB,SAAK,KAAK,GAAG,0BAAU,iBAAiB,KAAK,iBAAiB;AAC9D,SAAK,KAAK,GAAG,0BAAU,kBAAkB,KAAK,kBAAkB;AAAA,EAClE;AAAA,EAEA,eAAe,aAAgD;AAC7D,SAAK,OAAO,MAAM,EAAE,YAAY,GAAG,iCAAiC;AACpE,UAAM,sBACJ,uBAAuB,oCAAoB,YAAY,WAAW;AAEpE,QAAI,KAAK,wBAAwB,qBAAqB;AACpD;AAAA,IACF;AACA,SAAK,sBAAsB;AAC3B,SAAK,YAAY;AAEjB,QAAI,CAAC,qBAAqB;AACxB;AAAA,IACF;AAEA,UAAM,mBACJ,uBAAuB,oCACnB,cACA,KAAK,KAAK,mBAAmB,IAAI,mBAAmB;AAE1D,QAAI,kBAAkB;AACpB,iBAAW,eAAe,OAAO,OAAO,iBAAiB,iBAAiB,GAAG;AAC3E,YAAI,YAAY,SAAS,YAAY,WAAW,4BAAY,mBAAmB;AAC7E,eAAK,kBAAkB,YAAY,OAAO,aAAa,gBAAgB;AACvE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,qBAAqB,CAC3B,aACA,gBACG;AA/EP;AAgFI,UACE,UAAK,gBAAL,mBAAkB,SAAQ,YAAY,OACtC,YAAY,aAAa,KAAK,qBAC9B;AACA;AAAA,IACF;AACA,SAAK,YAAY;AAGjB,eAAWA,gBAAe,YAAY,kBAAkB,OAAO,GAAG;AAChE,UACEA,aAAY,SACZ,KAAK,kBAAkBA,aAAY,OAAOA,cAAa,WAAW,GAClE;AACA;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,cAAc;AACpB,QAAI,KAAK,eAAe,aAAa;AACnC,WAAK,eAAe,aAAa;AAAA,IACnC;AACA,SAAK,cAAc;AAAA,EACrB;AAAA,EAEQ,oBAAoB,CAC1B,OACA,aACA,gBACY;AACZ,QACE,KAAK,wBAAwB,YAAY,YACzC,YAAY,WAAW,4BAAY,qBAClC,KAAK,eAAe,KAAK,YAAY,QAAQ,YAAY,KAC1D;AACA,aAAO;AAAA,IACT;AACA,SAAK,YAAY;AACjB,SAAK,cAAc;AACnB,SAAK,eAAe;AAAA,UAClB,6BAAe;AAAA,QACb,QAAQ,KAAK,aAAa,KAAK;AAAA,QAC/B,YAAY,KAAK;AAAA,MACnB,CAAC;AAAA,IACH;AACA,WAAO;AAAA,EACT;AAAA,EAEQ,aAAa,OAAoB;AACvC,WAAO,IAAI,4BAAY,OAAO;AAAA,MAC5B,YAAY,KAAK;AAAA,MACjB,aAAa,KAAK;AAAA,MAClB,mBAAmB,KAAK;AAAA,IAC1B,CAAC;AAAA,EACH;AAAA,EAEA,MAAM,QAAQ;AACZ,SAAK,KAAK,IAAI,0BAAU,iBAAiB,KAAK,iBAAiB;AAC/D,SAAK,KAAK,IAAI,0BAAU,kBAAkB,KAAK,kBAAkB;AACjE,SAAK,YAAY;AACjB,SAAK,eAAe,OAAO,OAAO;AAAA,EACpC;AACF;","names":["publication"]}
+{"version":3,"sources":["../../../src/voice/room_io/_input.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport {\n  AudioStream,\n  type NoiseCancellationOptions,\n  RemoteParticipant,\n  type RemoteTrack,\n  type RemoteTrackPublication,\n  type Room,\n  RoomEvent,\n  TrackSource,\n} from '@livekit/rtc-node';\nimport type { ReadableStream } from 'node:stream/web';\nimport { log } from '../../log.js';\nimport { resampleStream } from '../../utils.js';\nimport { AudioInput } from '../io.js';\n\nexport class ParticipantAudioInputStream extends AudioInput {\n  private room: Room;\n  private sampleRate: number;\n  private numChannels: number;\n  private noiseCancellation?: NoiseCancellationOptions;\n  private publication: RemoteTrackPublication | null = null;\n  private participantIdentity: string | null = null;\n  private logger = log();\n  constructor({\n    room,\n    sampleRate,\n    numChannels,\n    noiseCancellation,\n  }: {\n    room: Room;\n    sampleRate: number;\n    numChannels: number;\n    noiseCancellation?: NoiseCancellationOptions;\n  }) {\n    super();\n    this.room = room;\n    this.sampleRate = sampleRate;\n    this.numChannels = numChannels;\n    this.noiseCancellation = noiseCancellation;\n\n    this.room.on(RoomEvent.TrackSubscribed, this.onTrackSubscribed);\n    this.room.on(RoomEvent.TrackUnpublished, this.onTrackUnpublished);\n  }\n\n  setParticipant(participant: RemoteParticipant | string | null) {\n    this.logger.debug({ participant }, 'setting participant audio input');\n    const participantIdentity =\n      participant instanceof RemoteParticipant ? participant.identity : participant;\n\n    if (this.participantIdentity === participantIdentity) {\n      return;\n    }\n    this.participantIdentity = participantIdentity;\n    this.closeStream();\n\n    if (!participantIdentity) {\n      return;\n    }\n\n    const participantValue =\n      participant instanceof RemoteParticipant\n        ? participant\n        : this.room.remoteParticipants.get(participantIdentity);\n\n    // We need to check if the participant has a microphone track and subscribe to it\n    // in case we miss the tracksubscribed event\n    if (participantValue) {\n      for (const publication of participantValue.trackPublications.values()) {\n        if (publication.track && publication.source === TrackSource.SOURCE_MICROPHONE) {\n          this.onTrackSubscribed(publication.track, publication, participantValue);\n          break;\n        }\n      }\n    }\n  }\n\n  private onTrackUnpublished = (\n    publication: RemoteTrackPublication,\n    participant: RemoteParticipant,\n  ) => {\n    if (\n      this.publication?.sid !== publication.sid ||\n      participant.identity !== this.participantIdentity\n    ) {\n      return;\n    }\n    this.closeStream();\n\n    // subscribe to the first available track\n    for (const publication of participant.trackPublications.values()) {\n      if (\n        publication.track &&\n        this.onTrackSubscribed(publication.track, publication, participant)\n      ) {\n        return;\n      }\n    }\n  };\n\n  private closeStream() {\n    if (this.deferredStream.isSourceSet) {\n      this.deferredStream.detachSource();\n    }\n    this.publication = null;\n  }\n\n  private onTrackSubscribed = (\n    track: RemoteTrack,\n    publication: RemoteTrackPublication,\n    participant: RemoteParticipant,\n  ): boolean => {\n    this.logger.debug({ participant: participant.identity }, 'onTrackSubscribed in _input');\n    if (\n      this.participantIdentity !== participant.identity ||\n      publication.source !== TrackSource.SOURCE_MICROPHONE ||\n      (this.publication && this.publication.sid === publication.sid)\n    ) {\n      return false;\n    }\n    this.closeStream();\n    this.publication = publication;\n    this.deferredStream.setSource(\n      resampleStream({\n        stream: this.createStream(track),\n        outputRate: this.sampleRate,\n      }),\n    );\n    return true;\n  };\n\n  private createStream(track: RemoteTrack): ReadableStream<AudioFrame> {\n    return new AudioStream(track, {\n      sampleRate: this.sampleRate,\n      numChannels: this.numChannels,\n      noiseCancellation: this.noiseCancellation,\n      // TODO(AJS-269): resolve compatibility issue with node-sdk to remove the forced type casting\n    }) as unknown as ReadableStream<AudioFrame>;\n  }\n\n  async close() {\n    this.room.off(RoomEvent.TrackSubscribed, this.onTrackSubscribed);\n    this.room.off(RoomEvent.TrackUnpublished, this.onTrackUnpublished);\n    this.closeStream();\n    this.deferredStream.stream.cancel();\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAIA,sBASO;AAEP,iBAAoB;AACpB,mBAA+B;AAC/B,gBAA2B;AAEpB,MAAM,oCAAoC,qBAAW;AAAA,EAClD;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,cAA6C;AAAA,EAC7C,sBAAqC;AAAA,EACrC,aAAS,gBAAI;AAAA,EACrB,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAKG;AACD,UAAM;AACN,SAAK,OAAO;AACZ,SAAK,aAAa;AAClB,SAAK,cAAc;AACnB,SAAK,oBAAoB;AAEzB,SAAK,KAAK,GAAG,0BAAU,iBAAiB,KAAK,iBAAiB;AAC9D,SAAK,KAAK,GAAG,0BAAU,kBAAkB,KAAK,kBAAkB;AAAA,EAClE;AAAA,EAEA,eAAe,aAAgD;AAC7D,SAAK,OAAO,MAAM,EAAE,YAAY,GAAG,iCAAiC;AACpE,UAAM,sBACJ,uBAAuB,oCAAoB,YAAY,WAAW;AAEpE,QAAI,KAAK,wBAAwB,qBAAqB;AACpD;AAAA,IACF;AACA,SAAK,sBAAsB;AAC3B,SAAK,YAAY;AAEjB,QAAI,CAAC,qBAAqB;AACxB;AAAA,IACF;AAEA,UAAM,mBACJ,uBAAuB,oCACnB,cACA,KAAK,KAAK,mBAAmB,IAAI,mBAAmB;AAI1D,QAAI,kBAAkB;AACpB,iBAAW,eAAe,iBAAiB,kBAAkB,OAAO,GAAG;AACrE,YAAI,YAAY,SAAS,YAAY,WAAW,4BAAY,mBAAmB;AAC7E,eAAK,kBAAkB,YAAY,OAAO,aAAa,gBAAgB;AACvE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,qBAAqB,CAC3B,aACA,gBACG;AAnFP;AAoFI,UACE,UAAK,gBAAL,mBAAkB,SAAQ,YAAY,OACtC,YAAY,aAAa,KAAK,qBAC9B;AACA;AAAA,IACF;AACA,SAAK,YAAY;AAGjB,eAAWA,gBAAe,YAAY,kBAAkB,OAAO,GAAG;AAChE,UACEA,aAAY,SACZ,KAAK,kBAAkBA,aAAY,OAAOA,cAAa,WAAW,GAClE;AACA;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,cAAc;AACpB,QAAI,KAAK,eAAe,aAAa;AACnC,WAAK,eAAe,aAAa;AAAA,IACnC;AACA,SAAK,cAAc;AAAA,EACrB;AAAA,EAEQ,oBAAoB,CAC1B,OACA,aACA,gBACY;AACZ,SAAK,OAAO,MAAM,EAAE,aAAa,YAAY,SAAS,GAAG,6BAA6B;AACtF,QACE,KAAK,wBAAwB,YAAY,YACzC,YAAY,WAAW,4BAAY,qBAClC,KAAK,eAAe,KAAK,YAAY,QAAQ,YAAY,KAC1D;AACA,aAAO;AAAA,IACT;AACA,SAAK,YAAY;AACjB,SAAK,cAAc;AACnB,SAAK,eAAe;AAAA,UAClB,6BAAe;AAAA,QACb,QAAQ,KAAK,aAAa,KAAK;AAAA,QAC/B,YAAY,KAAK;AAAA,MACnB,CAAC;AAAA,IACH;AACA,WAAO;AAAA,EACT;AAAA,EAEQ,aAAa,OAAgD;AACnE,WAAO,IAAI,4BAAY,OAAO;AAAA,MAC5B,YAAY,KAAK;AAAA,MACjB,aAAa,KAAK;AAAA,MAClB,mBAAmB,KAAK;AAAA;AAAA,IAE1B,CAAC;AAAA,EACH;AAAA,EAEA,MAAM,QAAQ;AACZ,SAAK,KAAK,IAAI,0BAAU,iBAAiB,KAAK,iBAAiB;AAC/D,SAAK,KAAK,IAAI,0BAAU,kBAAkB,KAAK,kBAAkB;AACjE,SAAK,YAAY;AACjB,SAAK,eAAe,OAAO,OAAO;AAAA,EACpC;AACF;","names":["publication"]}

package/dist/voice/room_io/_input.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"_input.d.ts","sourceRoot":"","sources":["../../../src/voice/room_io/_input.ts"],"names":[],"mappings":"AAGA,OAAO,EAEL,KAAK,wBAAwB,EAC7B,iBAAiB,EAGjB,KAAK,IAAI,EAGV,MAAM,mBAAmB,CAAC;AAG3B,OAAO,EAAE,UAAU,EAAE,MAAM,UAAU,CAAC;AAEtC,qBAAa,2BAA4B,SAAQ,UAAU;IACzD,OAAO,CAAC,IAAI,CAAO;IACnB,OAAO,CAAC,UAAU,CAAS;IAC3B,OAAO,CAAC,WAAW,CAAS;IAC5B,OAAO,CAAC,iBAAiB,CAAC,CAA2B;IACrD,OAAO,CAAC,WAAW,CAAuC;IAC1D,OAAO,CAAC,mBAAmB,CAAuB;IAClD,OAAO,CAAC,MAAM,CAAS;gBACX,EACV,IAAI,EACJ,UAAU,EACV,WAAW,EACX,iBAAiB,GAClB,EAAE;QACD,IAAI,EAAE,IAAI,CAAC;QACX,UAAU,EAAE,MAAM,CAAC;QACnB,WAAW,EAAE,MAAM,CAAC;QACpB,iBAAiB,CAAC,EAAE,wBAAwB,CAAC;KAC9C;IAWD,cAAc,CAAC,WAAW,EAAE,iBAAiB,GAAG,MAAM,GAAG,IAAI;IA8B7D,OAAO,CAAC,kBAAkB,CAqBxB;IAEF,OAAO,CAAC,WAAW;IAOnB,OAAO,CAAC,iBAAiB,CAqBvB;IAEF,OAAO,CAAC,YAAY;IAQd,KAAK;CAMZ"}
+{"version":3,"file":"_input.d.ts","sourceRoot":"","sources":["../../../src/voice/room_io/_input.ts"],"names":[],"mappings":"AAIA,OAAO,EAEL,KAAK,wBAAwB,EAC7B,iBAAiB,EAGjB,KAAK,IAAI,EAGV,MAAM,mBAAmB,CAAC;AAI3B,OAAO,EAAE,UAAU,EAAE,MAAM,UAAU,CAAC;AAEtC,qBAAa,2BAA4B,SAAQ,UAAU;IACzD,OAAO,CAAC,IAAI,CAAO;IACnB,OAAO,CAAC,UAAU,CAAS;IAC3B,OAAO,CAAC,WAAW,CAAS;IAC5B,OAAO,CAAC,iBAAiB,CAAC,CAA2B;IACrD,OAAO,CAAC,WAAW,CAAuC;IAC1D,OAAO,CAAC,mBAAmB,CAAuB;IAClD,OAAO,CAAC,MAAM,CAAS;gBACX,EACV,IAAI,EACJ,UAAU,EACV,WAAW,EACX,iBAAiB,GAClB,EAAE;QACD,IAAI,EAAE,IAAI,CAAC;QACX,UAAU,EAAE,MAAM,CAAC;QACnB,WAAW,EAAE,MAAM,CAAC;QACpB,iBAAiB,CAAC,EAAE,wBAAwB,CAAC;KAC9C;IAWD,cAAc,CAAC,WAAW,EAAE,iBAAiB,GAAG,MAAM,GAAG,IAAI;IAgC7D,OAAO,CAAC,kBAAkB,CAqBxB;IAEF,OAAO,CAAC,WAAW;IAOnB,OAAO,CAAC,iBAAiB,CAsBvB;IAEF,OAAO,CAAC,YAAY;IASd,KAAK;CAMZ"}

package/dist/voice/room_io/_input.js

@@ -42,7 +42,7 @@ class ParticipantAudioInputStream extends AudioInput {
     }
     const participantValue = participant instanceof RemoteParticipant ? participant : this.room.remoteParticipants.get(participantIdentity);
     if (participantValue) {
-      for (const publication of Object.values(participantValue.trackPublications)) {
+      for (const publication of participantValue.trackPublications.values()) {
         if (publication.track && publication.source === TrackSource.SOURCE_MICROPHONE) {
           this.onTrackSubscribed(publication.track, publication, participantValue);
           break;
@@ -69,6 +69,7 @@ class ParticipantAudioInputStream extends AudioInput {
     this.publication = null;
   }
   onTrackSubscribed = (track, publication, participant) => {
+    this.logger.debug({ participant: participant.identity }, "onTrackSubscribed in _input");
     if (this.participantIdentity !== participant.identity || publication.source !== TrackSource.SOURCE_MICROPHONE || this.publication && this.publication.sid === publication.sid) {
       return false;
     }
@@ -87,6 +88,7 @@ class ParticipantAudioInputStream extends AudioInput {
       sampleRate: this.sampleRate,
       numChannels: this.numChannels,
       noiseCancellation: this.noiseCancellation
+      // TODO(AJS-269): resolve compatibility issue with node-sdk to remove the forced type casting
     });
   }
   async close() {

package/dist/voice/room_io/_input.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../../../src/voice/room_io/_input.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport {\n  AudioStream,\n  type NoiseCancellationOptions,\n  RemoteParticipant,\n  type RemoteTrack,\n  type RemoteTrackPublication,\n  type Room,\n  RoomEvent,\n  TrackSource,\n} from '@livekit/rtc-node';\nimport { log } from '../../log.js';\nimport { resampleStream } from '../../utils.js';\nimport { AudioInput } from '../io.js';\n\nexport class ParticipantAudioInputStream extends AudioInput {\n  private room: Room;\n  private sampleRate: number;\n  private numChannels: number;\n  private noiseCancellation?: NoiseCancellationOptions;\n  private publication: RemoteTrackPublication | null = null;\n  private participantIdentity: string | null = null;\n  private logger = log();\n  constructor({\n    room,\n    sampleRate,\n    numChannels,\n    noiseCancellation,\n  }: {\n    room: Room;\n    sampleRate: number;\n    numChannels: number;\n    noiseCancellation?: NoiseCancellationOptions;\n  }) {\n    super();\n    this.room = room;\n    this.sampleRate = sampleRate;\n    this.numChannels = numChannels;\n    this.noiseCancellation = noiseCancellation;\n\n    this.room.on(RoomEvent.TrackSubscribed, this.onTrackSubscribed);\n    this.room.on(RoomEvent.TrackUnpublished, this.onTrackUnpublished);\n  }\n\n  setParticipant(participant: RemoteParticipant | string | null) {\n    this.logger.debug({ participant }, 'setting participant audio input');\n    const participantIdentity =\n      participant instanceof RemoteParticipant ? participant.identity : participant;\n\n    if (this.participantIdentity === participantIdentity) {\n      return;\n    }\n    this.participantIdentity = participantIdentity;\n    this.closeStream();\n\n    if (!participantIdentity) {\n      return;\n    }\n\n    const participantValue =\n      participant instanceof RemoteParticipant\n        ? participant\n        : this.room.remoteParticipants.get(participantIdentity);\n\n    if (participantValue) {\n      for (const publication of Object.values(participantValue.trackPublications)) {\n        if (publication.track && publication.source === TrackSource.SOURCE_MICROPHONE) {\n          this.onTrackSubscribed(publication.track, publication, participantValue);\n          break;\n        }\n      }\n    }\n  }\n\n  private onTrackUnpublished = (\n    publication: RemoteTrackPublication,\n    participant: RemoteParticipant,\n  ) => {\n    if (\n      this.publication?.sid !== publication.sid ||\n      participant.identity !== this.participantIdentity\n    ) {\n      return;\n    }\n    this.closeStream();\n\n    // subscribe to the first available track\n    for (const publication of participant.trackPublications.values()) {\n      if (\n        publication.track &&\n        this.onTrackSubscribed(publication.track, publication, participant)\n      ) {\n        return;\n      }\n    }\n  };\n\n  private closeStream() {\n    if (this.deferredStream.isSourceSet) {\n      this.deferredStream.detachSource();\n    }\n    this.publication = null;\n  }\n\n  private onTrackSubscribed = (\n    track: RemoteTrack,\n    publication: RemoteTrackPublication,\n    participant: RemoteParticipant,\n  ): boolean => {\n    if (\n      this.participantIdentity !== participant.identity ||\n      publication.source !== TrackSource.SOURCE_MICROPHONE ||\n      (this.publication && this.publication.sid === publication.sid)\n    ) {\n      return false;\n    }\n    this.closeStream();\n    this.publication = publication;\n    this.deferredStream.setSource(\n      resampleStream({\n        stream: this.createStream(track),\n        outputRate: this.sampleRate,\n      }),\n    );\n    return true;\n  };\n\n  private createStream(track: RemoteTrack) {\n    return new AudioStream(track, {\n      sampleRate: this.sampleRate,\n      numChannels: this.numChannels,\n      noiseCancellation: this.noiseCancellation,\n    });\n  }\n\n  async close() {\n    this.room.off(RoomEvent.TrackSubscribed, this.onTrackSubscribed);\n    this.room.off(RoomEvent.TrackUnpublished, this.onTrackUnpublished);\n    this.closeStream();\n    this.deferredStream.stream.cancel();\n  }\n}\n"],"mappings":"AAGA;AAAA,EACE;AAAA,EAEA;AAAA,EAIA;AAAA,EACA;AAAA,OACK;AACP,SAAS,WAAW;AACpB,SAAS,sBAAsB;AAC/B,SAAS,kBAAkB;AAEpB,MAAM,oCAAoC,WAAW;AAAA,EAClD;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,cAA6C;AAAA,EAC7C,sBAAqC;AAAA,EACrC,SAAS,IAAI;AAAA,EACrB,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAKG;AACD,UAAM;AACN,SAAK,OAAO;AACZ,SAAK,aAAa;AAClB,SAAK,cAAc;AACnB,SAAK,oBAAoB;AAEzB,SAAK,KAAK,GAAG,UAAU,iBAAiB,KAAK,iBAAiB;AAC9D,SAAK,KAAK,GAAG,UAAU,kBAAkB,KAAK,kBAAkB;AAAA,EAClE;AAAA,EAEA,eAAe,aAAgD;AAC7D,SAAK,OAAO,MAAM,EAAE,YAAY,GAAG,iCAAiC;AACpE,UAAM,sBACJ,uBAAuB,oBAAoB,YAAY,WAAW;AAEpE,QAAI,KAAK,wBAAwB,qBAAqB;AACpD;AAAA,IACF;AACA,SAAK,sBAAsB;AAC3B,SAAK,YAAY;AAEjB,QAAI,CAAC,qBAAqB;AACxB;AAAA,IACF;AAEA,UAAM,mBACJ,uBAAuB,oBACnB,cACA,KAAK,KAAK,mBAAmB,IAAI,mBAAmB;AAE1D,QAAI,kBAAkB;AACpB,iBAAW,eAAe,OAAO,OAAO,iBAAiB,iBAAiB,GAAG;AAC3E,YAAI,YAAY,SAAS,YAAY,WAAW,YAAY,mBAAmB;AAC7E,eAAK,kBAAkB,YAAY,OAAO,aAAa,gBAAgB;AACvE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,qBAAqB,CAC3B,aACA,gBACG;AA/EP;AAgFI,UACE,UAAK,gBAAL,mBAAkB,SAAQ,YAAY,OACtC,YAAY,aAAa,KAAK,qBAC9B;AACA;AAAA,IACF;AACA,SAAK,YAAY;AAGjB,eAAWA,gBAAe,YAAY,kBAAkB,OAAO,GAAG;AAChE,UACEA,aAAY,SACZ,KAAK,kBAAkBA,aAAY,OAAOA,cAAa,WAAW,GAClE;AACA;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,cAAc;AACpB,QAAI,KAAK,eAAe,aAAa;AACnC,WAAK,eAAe,aAAa;AAAA,IACnC;AACA,SAAK,cAAc;AAAA,EACrB;AAAA,EAEQ,oBAAoB,CAC1B,OACA,aACA,gBACY;AACZ,QACE,KAAK,wBAAwB,YAAY,YACzC,YAAY,WAAW,YAAY,qBAClC,KAAK,eAAe,KAAK,YAAY,QAAQ,YAAY,KAC1D;AACA,aAAO;AAAA,IACT;AACA,SAAK,YAAY;AACjB,SAAK,cAAc;AACnB,SAAK,eAAe;AAAA,MAClB,eAAe;AAAA,QACb,QAAQ,KAAK,aAAa,KAAK;AAAA,QAC/B,YAAY,KAAK;AAAA,MACnB,CAAC;AAAA,IACH;AACA,WAAO;AAAA,EACT;AAAA,EAEQ,aAAa,OAAoB;AACvC,WAAO,IAAI,YAAY,OAAO;AAAA,MAC5B,YAAY,KAAK;AAAA,MACjB,aAAa,KAAK;AAAA,MAClB,mBAAmB,KAAK;AAAA,IAC1B,CAAC;AAAA,EACH;AAAA,EAEA,MAAM,QAAQ;AACZ,SAAK,KAAK,IAAI,UAAU,iBAAiB,KAAK,iBAAiB;AAC/D,SAAK,KAAK,IAAI,UAAU,kBAAkB,KAAK,kBAAkB;AACjE,SAAK,YAAY;AACjB,SAAK,eAAe,OAAO,OAAO;AAAA,EACpC;AACF;","names":["publication"]}
+{"version":3,"sources":["../../../src/voice/room_io/_input.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport {\n  AudioStream,\n  type NoiseCancellationOptions,\n  RemoteParticipant,\n  type RemoteTrack,\n  type RemoteTrackPublication,\n  type Room,\n  RoomEvent,\n  TrackSource,\n} from '@livekit/rtc-node';\nimport type { ReadableStream } from 'node:stream/web';\nimport { log } from '../../log.js';\nimport { resampleStream } from '../../utils.js';\nimport { AudioInput } from '../io.js';\n\nexport class ParticipantAudioInputStream extends AudioInput {\n  private room: Room;\n  private sampleRate: number;\n  private numChannels: number;\n  private noiseCancellation?: NoiseCancellationOptions;\n  private publication: RemoteTrackPublication | null = null;\n  private participantIdentity: string | null = null;\n  private logger = log();\n  constructor({\n    room,\n    sampleRate,\n    numChannels,\n    noiseCancellation,\n  }: {\n    room: Room;\n    sampleRate: number;\n    numChannels: number;\n    noiseCancellation?: NoiseCancellationOptions;\n  }) {\n    super();\n    this.room = room;\n    this.sampleRate = sampleRate;\n    this.numChannels = numChannels;\n    this.noiseCancellation = noiseCancellation;\n\n    this.room.on(RoomEvent.TrackSubscribed, this.onTrackSubscribed);\n    this.room.on(RoomEvent.TrackUnpublished, this.onTrackUnpublished);\n  }\n\n  setParticipant(participant: RemoteParticipant | string | null) {\n    this.logger.debug({ participant }, 'setting participant audio input');\n    const participantIdentity =\n      participant instanceof RemoteParticipant ? participant.identity : participant;\n\n    if (this.participantIdentity === participantIdentity) {\n      return;\n    }\n    this.participantIdentity = participantIdentity;\n    this.closeStream();\n\n    if (!participantIdentity) {\n      return;\n    }\n\n    const participantValue =\n      participant instanceof RemoteParticipant\n        ? participant\n        : this.room.remoteParticipants.get(participantIdentity);\n\n    // We need to check if the participant has a microphone track and subscribe to it\n    // in case we miss the tracksubscribed event\n    if (participantValue) {\n      for (const publication of participantValue.trackPublications.values()) {\n        if (publication.track && publication.source === TrackSource.SOURCE_MICROPHONE) {\n          this.onTrackSubscribed(publication.track, publication, participantValue);\n          break;\n        }\n      }\n    }\n  }\n\n  private onTrackUnpublished = (\n    publication: RemoteTrackPublication,\n    participant: RemoteParticipant,\n  ) => {\n    if (\n      this.publication?.sid !== publication.sid ||\n      participant.identity !== this.participantIdentity\n    ) {\n      return;\n    }\n    this.closeStream();\n\n    // subscribe to the first available track\n    for (const publication of participant.trackPublications.values()) {\n      if (\n        publication.track &&\n        this.onTrackSubscribed(publication.track, publication, participant)\n      ) {\n        return;\n      }\n    }\n  };\n\n  private closeStream() {\n    if (this.deferredStream.isSourceSet) {\n      this.deferredStream.detachSource();\n    }\n    this.publication = null;\n  }\n\n  private onTrackSubscribed = (\n    track: RemoteTrack,\n    publication: RemoteTrackPublication,\n    participant: RemoteParticipant,\n  ): boolean => {\n    this.logger.debug({ participant: participant.identity }, 'onTrackSubscribed in _input');\n    if (\n      this.participantIdentity !== participant.identity ||\n      publication.source !== TrackSource.SOURCE_MICROPHONE ||\n      (this.publication && this.publication.sid === publication.sid)\n    ) {\n      return false;\n    }\n    this.closeStream();\n    this.publication = publication;\n    this.deferredStream.setSource(\n      resampleStream({\n        stream: this.createStream(track),\n        outputRate: this.sampleRate,\n      }),\n    );\n    return true;\n  };\n\n  private createStream(track: RemoteTrack): ReadableStream<AudioFrame> {\n    return new AudioStream(track, {\n      sampleRate: this.sampleRate,\n      numChannels: this.numChannels,\n      noiseCancellation: this.noiseCancellation,\n      // TODO(AJS-269): resolve compatibility issue with node-sdk to remove the forced type casting\n    }) as unknown as ReadableStream<AudioFrame>;\n  }\n\n  async close() {\n    this.room.off(RoomEvent.TrackSubscribed, this.onTrackSubscribed);\n    this.room.off(RoomEvent.TrackUnpublished, this.onTrackUnpublished);\n    this.closeStream();\n    this.deferredStream.stream.cancel();\n  }\n}\n"],"mappings":"AAIA;AAAA,EACE;AAAA,EAEA;AAAA,EAIA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,WAAW;AACpB,SAAS,sBAAsB;AAC/B,SAAS,kBAAkB;AAEpB,MAAM,oCAAoC,WAAW;AAAA,EAClD;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,cAA6C;AAAA,EAC7C,sBAAqC;AAAA,EACrC,SAAS,IAAI;AAAA,EACrB,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAKG;AACD,UAAM;AACN,SAAK,OAAO;AACZ,SAAK,aAAa;AAClB,SAAK,cAAc;AACnB,SAAK,oBAAoB;AAEzB,SAAK,KAAK,GAAG,UAAU,iBAAiB,KAAK,iBAAiB;AAC9D,SAAK,KAAK,GAAG,UAAU,kBAAkB,KAAK,kBAAkB;AAAA,EAClE;AAAA,EAEA,eAAe,aAAgD;AAC7D,SAAK,OAAO,MAAM,EAAE,YAAY,GAAG,iCAAiC;AACpE,UAAM,sBACJ,uBAAuB,oBAAoB,YAAY,WAAW;AAEpE,QAAI,KAAK,wBAAwB,qBAAqB;AACpD;AAAA,IACF;AACA,SAAK,sBAAsB;AAC3B,SAAK,YAAY;AAEjB,QAAI,CAAC,qBAAqB;AACxB;AAAA,IACF;AAEA,UAAM,mBACJ,uBAAuB,oBACnB,cACA,KAAK,KAAK,mBAAmB,IAAI,mBAAmB;AAI1D,QAAI,kBAAkB;AACpB,iBAAW,eAAe,iBAAiB,kBAAkB,OAAO,GAAG;AACrE,YAAI,YAAY,SAAS,YAAY,WAAW,YAAY,mBAAmB;AAC7E,eAAK,kBAAkB,YAAY,OAAO,aAAa,gBAAgB;AACvE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,qBAAqB,CAC3B,aACA,gBACG;AAnFP;AAoFI,UACE,UAAK,gBAAL,mBAAkB,SAAQ,YAAY,OACtC,YAAY,aAAa,KAAK,qBAC9B;AACA;AAAA,IACF;AACA,SAAK,YAAY;AAGjB,eAAWA,gBAAe,YAAY,kBAAkB,OAAO,GAAG;AAChE,UACEA,aAAY,SACZ,KAAK,kBAAkBA,aAAY,OAAOA,cAAa,WAAW,GAClE;AACA;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,cAAc;AACpB,QAAI,KAAK,eAAe,aAAa;AACnC,WAAK,eAAe,aAAa;AAAA,IACnC;AACA,SAAK,cAAc;AAAA,EACrB;AAAA,EAEQ,oBAAoB,CAC1B,OACA,aACA,gBACY;AACZ,SAAK,OAAO,MAAM,EAAE,aAAa,YAAY,SAAS,GAAG,6BAA6B;AACtF,QACE,KAAK,wBAAwB,YAAY,YACzC,YAAY,WAAW,YAAY,qBAClC,KAAK,eAAe,KAAK,YAAY,QAAQ,YAAY,KAC1D;AACA,aAAO;AAAA,IACT;AACA,SAAK,YAAY;AACjB,SAAK,cAAc;AACnB,SAAK,eAAe;AAAA,MAClB,eAAe;AAAA,QACb,QAAQ,KAAK,aAAa,KAAK;AAAA,QAC/B,YAAY,KAAK;AAAA,MACnB,CAAC;AAAA,IACH;AACA,WAAO;AAAA,EACT;AAAA,EAEQ,aAAa,OAAgD;AACnE,WAAO,IAAI,YAAY,OAAO;AAAA,MAC5B,YAAY,KAAK;AAAA,MACjB,aAAa,KAAK;AAAA,MAClB,mBAAmB,KAAK;AAAA;AAAA,IAE1B,CAAC;AAAA,EACH;AAAA,EAEA,MAAM,QAAQ;AACZ,SAAK,KAAK,IAAI,UAAU,iBAAiB,KAAK,iBAAiB;AAC/D,SAAK,KAAK,IAAI,UAAU,kBAAkB,KAAK,kBAAkB;AACjE,SAAK,YAAY;AACjB,SAAK,eAAe,OAAO,OAAO;AAAA,EACpC;AACF;","names":["publication"]}

package/dist/voice/run_context.cjs

@@ -26,10 +26,23 @@ class RunContext {
     this.session = session;
     this.speechHandle = speechHandle;
     this.functionCall = functionCall;
+    this.initialStepIdx = speechHandle.numSteps - 1;
   }
+  initialStepIdx;
   get userData() {
     return this.session.userData;
   }
+  /**
+   * Waits for the speech playout corresponding to this function call step.
+   *
+   * Unlike {@link SpeechHandle.waitForPlayout}, which waits for the full
+   * assistant turn to complete (including all function tools),
+   * this method only waits for the assistant's spoken response prior to running
+   * this tool to finish playing.
+   */
+  async waitForPlayout() {
+    return this.speechHandle._waitForGeneration(this.initialStepIdx);
+  }
 }
 // Annotate the CommonJS export names for ESM import in node:
 0 && (module.exports = {

package/dist/voice/run_context.cjs.map

@@ -1 +1 @@
-{"version":3,"sources":["../../src/voice/run_context.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { FunctionCall } from '../llm/chat_context.js';\nimport type { AgentSession } from './agent_session.js';\nimport type { SpeechHandle } from './speech_handle.js';\n\nexport type UnknownUserData = unknown;\n\nexport class RunContext<UserData = UnknownUserData> {\n  constructor(\n    public readonly session: AgentSession<UserData>,\n    public readonly speechHandle: SpeechHandle,\n    public readonly functionCall: FunctionCall,\n  ) {}\n\n  get userData(): UserData {\n    return this.session.userData;\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AASO,MAAM,WAAuC;AAAA,EAClD,YACkB,SACA,cACA,cAChB;AAHgB;AACA;AACA;AAAA,EACf;AAAA,EAEH,IAAI,WAAqB;AACvB,WAAO,KAAK,QAAQ;AAAA,EACtB;AACF;","names":[]}
+{"version":3,"sources":["../../src/voice/run_context.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { FunctionCall } from '../llm/chat_context.js';\nimport type { AgentSession } from './agent_session.js';\nimport type { SpeechHandle } from './speech_handle.js';\n\nexport type UnknownUserData = unknown;\n\nexport class RunContext<UserData = UnknownUserData> {\n  private readonly initialStepIdx: number;\n  constructor(\n    public readonly session: AgentSession<UserData>,\n    public readonly speechHandle: SpeechHandle,\n    public readonly functionCall: FunctionCall,\n  ) {\n    this.initialStepIdx = speechHandle.numSteps - 1;\n  }\n  get userData(): UserData {\n    return this.session.userData;\n  }\n\n  /**\n   * Waits for the speech playout corresponding to this function call step.\n   *\n   * Unlike {@link SpeechHandle.waitForPlayout}, which waits for the full\n   * assistant turn to complete (including all function tools),\n   * this method only waits for the assistant's spoken response prior to running\n   * this tool to finish playing.\n   */\n  async waitForPlayout() {\n    return this.speechHandle._waitForGeneration(this.initialStepIdx);\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AASO,MAAM,WAAuC;AAAA,EAElD,YACkB,SACA,cACA,cAChB;AAHgB;AACA;AACA;AAEhB,SAAK,iBAAiB,aAAa,WAAW;AAAA,EAChD;AAAA,EAPiB;AAAA,EAQjB,IAAI,WAAqB;AACvB,WAAO,KAAK,QAAQ;AAAA,EACtB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUA,MAAM,iBAAiB;AACrB,WAAO,KAAK,aAAa,mBAAmB,KAAK,cAAc;AAAA,EACjE;AACF;","names":[]}

package/dist/voice/run_context.d.cts

@@ -6,7 +6,17 @@ export declare class RunContext<UserData = UnknownUserData> {
     readonly session: AgentSession<UserData>;
     readonly speechHandle: SpeechHandle;
     readonly functionCall: FunctionCall;
+    private readonly initialStepIdx;
     constructor(session: AgentSession<UserData>, speechHandle: SpeechHandle, functionCall: FunctionCall);
     get userData(): UserData;
+    /**
+     * Waits for the speech playout corresponding to this function call step.
+     *
+     * Unlike {@link SpeechHandle.waitForPlayout}, which waits for the full
+     * assistant turn to complete (including all function tools),
+     * this method only waits for the assistant's spoken response prior to running
+     * this tool to finish playing.
+     */
+    waitForPlayout(): Promise<void>;
 }
 //# sourceMappingURL=run_context.d.ts.map

package/dist/voice/run_context.d.ts

@@ -6,7 +6,17 @@ export declare class RunContext<UserData = UnknownUserData> {
     readonly session: AgentSession<UserData>;
     readonly speechHandle: SpeechHandle;
     readonly functionCall: FunctionCall;
+    private readonly initialStepIdx;
     constructor(session: AgentSession<UserData>, speechHandle: SpeechHandle, functionCall: FunctionCall);
     get userData(): UserData;
+    /**
+     * Waits for the speech playout corresponding to this function call step.
+     *
+     * Unlike {@link SpeechHandle.waitForPlayout}, which waits for the full
+     * assistant turn to complete (including all function tools),
+     * this method only waits for the assistant's spoken response prior to running
+     * this tool to finish playing.
+     */
+    waitForPlayout(): Promise<void>;
 }
 //# sourceMappingURL=run_context.d.ts.map

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

@@ -1 +1 @@
-{"version":3,"file":"run_context.d.ts","sourceRoot":"","sources":["../../src/voice/run_context.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAC3D,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AACvD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAEvD,MAAM,MAAM,eAAe,GAAG,OAAO,CAAC;AAEtC,qBAAa,UAAU,CAAC,QAAQ,GAAG,eAAe;aAE9B,OAAO,EAAE,YAAY,CAAC,QAAQ,CAAC;aAC/B,YAAY,EAAE,YAAY;aAC1B,YAAY,EAAE,YAAY;gBAF1B,OAAO,EAAE,YAAY,CAAC,QAAQ,CAAC,EAC/B,YAAY,EAAE,YAAY,EAC1B,YAAY,EAAE,YAAY;IAG5C,IAAI,QAAQ,IAAI,QAAQ,CAEvB;CACF"}
+{"version":3,"file":"run_context.d.ts","sourceRoot":"","sources":["../../src/voice/run_context.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAC3D,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AACvD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAEvD,MAAM,MAAM,eAAe,GAAG,OAAO,CAAC;AAEtC,qBAAa,UAAU,CAAC,QAAQ,GAAG,eAAe;aAG9B,OAAO,EAAE,YAAY,CAAC,QAAQ,CAAC;aAC/B,YAAY,EAAE,YAAY;aAC1B,YAAY,EAAE,YAAY;IAJ5C,OAAO,CAAC,QAAQ,CAAC,cAAc,CAAS;gBAEtB,OAAO,EAAE,YAAY,CAAC,QAAQ,CAAC,EAC/B,YAAY,EAAE,YAAY,EAC1B,YAAY,EAAE,YAAY;IAI5C,IAAI,QAAQ,IAAI,QAAQ,CAEvB;IAED;;;;;;;OAOG;IACG,cAAc;CAGrB"}

package/dist/voice/run_context.js

@@ -3,10 +3,23 @@ class RunContext {
     this.session = session;
     this.speechHandle = speechHandle;
     this.functionCall = functionCall;
+    this.initialStepIdx = speechHandle.numSteps - 1;
   }
+  initialStepIdx;
   get userData() {
     return this.session.userData;
   }
+  /**
+   * Waits for the speech playout corresponding to this function call step.
+   *
+   * Unlike {@link SpeechHandle.waitForPlayout}, which waits for the full
+   * assistant turn to complete (including all function tools),
+   * this method only waits for the assistant's spoken response prior to running
+   * this tool to finish playing.
+   */
+  async waitForPlayout() {
+    return this.speechHandle._waitForGeneration(this.initialStepIdx);
+  }
 }
 export {
   RunContext

package/dist/voice/run_context.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../../src/voice/run_context.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { FunctionCall } from '../llm/chat_context.js';\nimport type { AgentSession } from './agent_session.js';\nimport type { SpeechHandle } from './speech_handle.js';\n\nexport type UnknownUserData = unknown;\n\nexport class RunContext<UserData = UnknownUserData> {\n  constructor(\n    public readonly session: AgentSession<UserData>,\n    public readonly speechHandle: SpeechHandle,\n    public readonly functionCall: FunctionCall,\n  ) {}\n\n  get userData(): UserData {\n    return this.session.userData;\n  }\n}\n"],"mappings":"AASO,MAAM,WAAuC;AAAA,EAClD,YACkB,SACA,cACA,cAChB;AAHgB;AACA;AACA;AAAA,EACf;AAAA,EAEH,IAAI,WAAqB;AACvB,WAAO,KAAK,QAAQ;AAAA,EACtB;AACF;","names":[]}
+{"version":3,"sources":["../../src/voice/run_context.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { FunctionCall } from '../llm/chat_context.js';\nimport type { AgentSession } from './agent_session.js';\nimport type { SpeechHandle } from './speech_handle.js';\n\nexport type UnknownUserData = unknown;\n\nexport class RunContext<UserData = UnknownUserData> {\n  private readonly initialStepIdx: number;\n  constructor(\n    public readonly session: AgentSession<UserData>,\n    public readonly speechHandle: SpeechHandle,\n    public readonly functionCall: FunctionCall,\n  ) {\n    this.initialStepIdx = speechHandle.numSteps - 1;\n  }\n  get userData(): UserData {\n    return this.session.userData;\n  }\n\n  /**\n   * Waits for the speech playout corresponding to this function call step.\n   *\n   * Unlike {@link SpeechHandle.waitForPlayout}, which waits for the full\n   * assistant turn to complete (including all function tools),\n   * this method only waits for the assistant's spoken response prior to running\n   * this tool to finish playing.\n   */\n  async waitForPlayout() {\n    return this.speechHandle._waitForGeneration(this.initialStepIdx);\n  }\n}\n"],"mappings":"AASO,MAAM,WAAuC;AAAA,EAElD,YACkB,SACA,cACA,cAChB;AAHgB;AACA;AACA;AAEhB,SAAK,iBAAiB,aAAa,WAAW;AAAA,EAChD;AAAA,EAPiB;AAAA,EAQjB,IAAI,WAAqB;AACvB,WAAO,KAAK,QAAQ;AAAA,EACtB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAUA,MAAM,iBAAiB;AACrB,WAAO,KAAK,aAAa,mBAAmB,KAAK,cAAc;AAAA,EACjE;AACF;","names":[]}

package/dist/voice/speech_handle.cjs

@@ -22,12 +22,18 @@ __export(speech_handle_exports, {
 });
 module.exports = __toCommonJS(speech_handle_exports);
 var import_utils = require("../utils.cjs");
+var import_agent = require("./agent.cjs");
 class SpeechHandle {
-  constructor(id, allowInterruptions, stepIndex, parent) {
-    this.id = id;
-    this.allowInterruptions = allowInterruptions;
-    this.stepIndex = stepIndex;
+  constructor(_id, _allowInterruptions, _stepIndex, parent) {
+    this._id = _id;
+    this._allowInterruptions = _allowInterruptions;
+    this._stepIndex = _stepIndex;
     this.parent = parent;
+    this.doneFut.await.finally(() => {
+      for (const callback of this.doneCallbacks) {
+        callback(this);
+      }
+    });
   }
   /** Priority for messages that should be played after all other messages in the queue */
   static SPEECH_PRIORITY_LOW = 0;
@@ -36,21 +42,58 @@ class SpeechHandle {
   /** Priority for important messages that should be played before others. */
   static SPEECH_PRIORITY_HIGH = 10;
   interruptFut = new import_utils.Future();
-  authorizeFut = new import_utils.Future();
-  playoutDoneFut = new import_utils.Future();
-  _chatMessage;
+  authorizedEvent = new import_utils.Event();
+  scheduledFut = new import_utils.Future();
+  doneFut = new import_utils.Future();
+  generations = [];
+  /** @internal */
+  _tasks = [];
+  _chatItems = [];
+  _numSteps = 1;
+  itemAddedCallbacks = /* @__PURE__ */ new Set();
+  doneCallbacks = /* @__PURE__ */ new Set();
   static create(options) {
-    const { allowInterruptions = false, stepIndex = 0, parent } = options ?? {};
+    const { allowInterruptions = true, stepIndex = 0, parent } = options ?? {};
     return new SpeechHandle((0, import_utils.shortuuid)("speech_"), allowInterruptions, stepIndex, parent);
   }
   get interrupted() {
     return this.interruptFut.done;
   }
-  get done() {
-    return this.playoutDoneFut.done;
+  get numSteps() {
+    return this._numSteps;
+  }
+  get id() {
+    return this._id;
+  }
+  get scheduled() {
+    return this.scheduledFut.done;
+  }
+  get allowInterruptions() {
+    return this._allowInterruptions;
+  }
+  /**
+   * Allow or disallow interruptions on this SpeechHandle.
+   *
+   * When set to false, the SpeechHandle will no longer accept any incoming
+   * interruption requests until re-enabled. If the handle is already
+   * interrupted, clearing interruptions is not allowed.
+   *
+   * @param value - true to allow interruptions, false to disallow
+   * @throws Error If attempting to disable interruptions when already interrupted
+   */
+  set allowInterruptions(value) {
+    if (this.interrupted && !value) {
+      throw new Error(
+        "Cannot set allow_interruptions to False, the SpeechHandle is already interrupted"
+      );
+    }
+    this._allowInterruptions = value;
+  }
+  done() {
+    return this.doneFut.done;
   }
-  get chatMessage() {
-    return this._chatMessage;
+  get chatItems() {
+    return this._chatItems;
   }
   /**
    * Interrupt the current speech generation.
@@ -59,43 +102,122 @@ class SpeechHandle {
    *
    * @returns The same speech handle that was interrupted.
    */
-  interrupt() {
-    if (!this.allowInterruptions) {
-      throw new Error("interruptions are not allowed");
+  interrupt(force = false) {
+    if (!force && !this.allowInterruptions) {
+      throw new Error("This generation handle does not allow interruptions");
     }
-    if (this.done) return this;
-    this.interruptFut.resolve();
+    this._cancel();
     return this;
   }
-  then(callback) {
-    return this.playoutDoneFut.await.finally(() => callback(this));
-  }
+  /**
+   * Waits for the entire assistant turn to complete playback.
+   *
+   * This method waits until the assistant has fully finished speaking,
+   * including any finalization steps beyond initial response generation.
+   * This is appropriate to call when you want to ensure the speech output
+   * has entirely played out, including any tool calls and response follow-ups.
+   */
   async waitForPlayout() {
-    return this.playoutDoneFut.await;
+    const store = import_agent.asyncLocalStorage.getStore();
+    if (store && (store == null ? void 0 : store.functionCall)) {
+      throw new Error(
+        `Cannot call 'SpeechHandle.waitForPlayout()' from inside the function tool '${store.functionCall.name}'. This creates a circular wait: the speech handle is waiting for the function tool to complete, while the function tool is simultaneously waiting for the speech handle.
+To wait for the assistant's spoken response prior to running this tool, use RunContext.wait_for_playout() instead.`
+      );
+    }
   }
   async waitIfNotInterrupted(aw) {
     const allTasksPromise = Promise.all(aw);
     const fs = [allTasksPromise, this.interruptFut.await];
     await Promise.race(fs);
   }
+  addDoneCallback(callback) {
+    this.doneCallbacks.add(callback);
+  }
+  removeDoneCallback(callback) {
+    this.doneCallbacks.delete(callback);
+  }
   /** @internal */
-  _setChatMessage(chatMessage) {
-    if (this.done) {
-      throw new Error("cannot set chat message after speech has been played");
+  _cancel() {
+    if (this.done()) {
+      return this;
+    }
+    if (!this.interruptFut.done) {
+      this.interruptFut.resolve();
     }
-    this._chatMessage = chatMessage;
+    return this;
+  }
+  /** @internal */
+  _authorizeGeneration() {
+    const fut = new import_utils.Future();
+    this.generations.push(fut);
+    this.authorizedEvent.set();
   }
   /** @internal */
-  _authorizePlayout() {
-    this.authorizeFut.resolve();
+  _clearAuthorization() {
+    this.authorizedEvent.clear();
   }
   /** @internal */
   async _waitForAuthorization() {
-    return this.authorizeFut.await;
+    await this.authorizedEvent.wait();
   }
   /** @internal */
-  _markPlayoutDone() {
-    this.playoutDoneFut.resolve();
+  async _waitForGeneration(stepIdx = -1) {
+    if (this.generations.length === 0) {
+      throw new Error("cannot use wait_for_generation: no active generation is running.");
+    }
+    const index = stepIdx === -1 ? this.generations.length - 1 : stepIdx;
+    const generation = this.generations[index];
+    if (!generation) {
+      throw new Error(`Generation at index ${index} not found.`);
+    }
+    return generation.await;
+  }
+  /** @internal */
+  async _waitForScheduled() {
+    return this.scheduledFut.await;
+  }
+  /** @internal */
+  _markGenerationDone() {
+    if (this.generations.length === 0) {
+      throw new Error("cannot use mark_generation_done: no active generation is running.");
+    }
+    const lastGeneration = this.generations[this.generations.length - 1];
+    if (lastGeneration && !lastGeneration.done) {
+      lastGeneration.resolve();
+    }
+  }
+  /** @internal */
+  _markDone() {
+    if (!this.doneFut.done) {
+      this.doneFut.resolve();
+      if (this.generations.length > 0) {
+        this._markGenerationDone();
+      }
+    }
+  }
+  /** @internal */
+  _markScheduled() {
+    if (!this.scheduledFut.done) {
+      this.scheduledFut.resolve();
+    }
+  }
+  /** @internal */
+  _addItemAddedCallback(callback) {
+    this.itemAddedCallbacks.add(callback);
+  }
+  /** @internal */
+  _removeItemAddedCallback(callback) {
+    this.itemAddedCallbacks.delete(callback);
+  }
+  /** @internal */
+  _itemAdded(items) {
+    for (const item of items) {
+      for (const cb of this.itemAddedCallbacks) {
+        cb(item);
+      }
+      this._chatItems.push(item);
+    }
   }
 }
 // Annotate the CommonJS export names for ESM import in node: