@livekit/agents 0.0.0-20260120144724

package/dist/voice/io.cjs.map

@@ -0,0 +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>;\n\n/**\n * A string with timing information for word-level alignment.\n */\nexport interface TimedString {\n  text: string;\n  startTime?: number; // seconds\n  endTime?: number; // seconds\n}\n\nexport interface AudioOutputCapabilities {\n  /** Whether this output supports pause/resume functionality */\n  pause: boolean;\n}\n\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  protected readonly capabilities: AudioOutputCapabilities;\n\n  constructor(\n    public sampleRate?: number,\n    protected readonly nextInChain?: AudioOutput,\n    capabilities: AudioOutputCapabilities = { pause: false },\n  ) {\n    super();\n    this.capabilities = capabilities;\n    if (this.nextInChain) {\n      this.nextInChain.on(AudioOutput.EVENT_PLAYBACK_FINISHED, (ev: PlaybackFinishedEvent) =>\n        this.onPlaybackFinished(ev),\n      );\n    }\n  }\n\n  /**\n   * Whether this output and all outputs in the chain support pause/resume.\n   */\n  get canPause(): boolean {\n    return this.capabilities.pause && (this.nextInChain?.canPause ?? true);\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  /**\n   * Pause the audio playback\n   */\n  pause(): void {\n    if (this.nextInChain) {\n      this.nextInChain.pause();\n    }\n  }\n\n  /**\n   * Resume the audio playback\n   */\n  resume(): void {\n    if (this.nextInChain) {\n      this.nextInChain.resume();\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;AAiChB,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,EAcrD,YACS,YACY,aACnB,eAAwC,EAAE,OAAO,MAAM,GACvD;AACA,UAAM;AAJC;AACY;AAInB,SAAK,eAAe;AACpB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY;AAAA,QAAG,YAAY;AAAA,QAAyB,CAAC,OACxD,KAAK,mBAAmB,EAAE;AAAA,MAC5B;AAAA,IACF;AAAA,EACF;AAAA,EAzBA,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,EACJ;AAAA;AAAA;AAAA;AAAA,EAmBnB,IAAI,WAAoB;AAzF1B;AA0FI,WAAO,KAAK,aAAa,YAAU,UAAK,gBAAL,mBAAkB,aAAY;AAAA,EACnE;AAAA;AAAA;AAAA;AAAA,EAKA,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;AAAA;AAAA;AAAA;AAAA,EAKA,QAAc;AACZ,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,MAAM;AAAA,IACzB;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKA,SAAe;AACb,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,OAAO;AAAA,IAC1B;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

@@ -0,0 +1,126 @@
+/// <reference types="node" resolution-mode="require"/>
+/// <reference types="node" resolution-mode="require"/>
+import type { AudioFrame } from '@livekit/rtc-node';
+import { EventEmitter } from 'node:events';
+import type { ReadableStream } from 'node:stream/web';
+import type { ChatContext } from '../llm/chat_context.js';
+import type { ChatChunk } from '../llm/llm.js';
+import type { ToolContext } from '../llm/tool_context.js';
+import { DeferredReadableStream } from '../stream/deferred_stream.js';
+import type { SpeechEvent } from '../stt/stt.js';
+import type { ModelSettings } from './agent.js';
+export type STTNode = (audio: ReadableStream<AudioFrame>, modelSettings: ModelSettings) => Promise<ReadableStream<SpeechEvent | string> | null>;
+export type LLMNode = (chatCtx: ChatContext, toolCtx: ToolContext, modelSettings: ModelSettings) => Promise<ReadableStream<ChatChunk | string> | null>;
+export type TTSNode = (text: ReadableStream<string>, modelSettings: ModelSettings) => Promise<ReadableStream<AudioFrame> | null>;
+/**
+ * A string with timing information for word-level alignment.
+ */
+export interface TimedString {
+    text: string;
+    startTime?: number;
+    endTime?: number;
+}
+export interface AudioOutputCapabilities {
+    /** Whether this output supports pause/resume functionality */
+    pause: boolean;
+}
+export declare abstract class AudioInput {
+    protected deferredStream: DeferredReadableStream<AudioFrame>;
+    get stream(): ReadableStream<AudioFrame>;
+    onAttached(): void;
+    onDetached(): void;
+}
+export declare abstract class AudioOutput extends EventEmitter {
+    sampleRate?: number | undefined;
+    protected readonly nextInChain?: AudioOutput | undefined;
+    static readonly EVENT_PLAYBACK_FINISHED = "playbackFinished";
+    private playbackFinishedFuture;
+    private _capturing;
+    private playbackFinishedCount;
+    private playbackSegmentsCount;
+    private lastPlaybackEvent;
+    protected logger: import("pino").Logger;
+    protected readonly capabilities: AudioOutputCapabilities;
+    constructor(sampleRate?: number | undefined, nextInChain?: AudioOutput | undefined, capabilities?: AudioOutputCapabilities);
+    /**
+     * Whether this output and all outputs in the chain support pause/resume.
+     */
+    get canPause(): boolean;
+    /**
+     * Capture an audio frame for playback, frames can be pushed faster than real-time
+     */
+    captureFrame(_frame: AudioFrame): Promise<void>;
+    /**
+     * Wait for the past audio segments to finish playing out.
+     *
+     * @returns The event that was emitted when the audio finished playing out (only the last segment information)
+     */
+    waitForPlayout(): Promise<PlaybackFinishedEvent>;
+    /**
+     * Developers building audio sinks must call this method when a playback/segment is finished.
+     * Segments are segmented by calls to flush() or clearBuffer()
+     */
+    onPlaybackFinished(options: PlaybackFinishedEvent): void;
+    flush(): void;
+    /**
+     * Clear the buffer, stopping playback immediately
+     */
+    abstract clearBuffer(): void;
+    onAttached(): void;
+    onDetached(): void;
+    /**
+     * Pause the audio playback
+     */
+    pause(): void;
+    /**
+     * Resume the audio playback
+     */
+    resume(): void;
+}
+export interface PlaybackFinishedEvent {
+    playbackPosition: number;
+    interrupted: boolean;
+    synchronizedTranscript?: string;
+}
+export declare abstract class TextOutput {
+    protected readonly nextInChain?: TextOutput | undefined;
+    constructor(nextInChain?: TextOutput | undefined);
+    /**
+     * Capture a text segment (Used by the output of LLM nodes)
+     */
+    abstract captureText(text: string): Promise<void>;
+    /**
+     * Mark the current text segment as complete (e.g LLM generation is complete)
+     */
+    abstract flush(): void;
+    onAttached(): void;
+    onDetached(): void;
+}
+export declare class AgentInput {
+    private readonly audioChanged;
+    private _audioStream;
+    private _audioEnabled;
+    constructor(audioChanged: () => void);
+    setAudioEnabled(enable: boolean): void;
+    get audioEnabled(): boolean;
+    get audio(): AudioInput | null;
+    set audio(stream: AudioInput | null);
+}
+export declare class AgentOutput {
+    private readonly audioChanged;
+    private readonly transcriptionChanged;
+    private _audioSink;
+    private _transcriptionSink;
+    private _audioEnabled;
+    private _transcriptionEnabled;
+    constructor(audioChanged: () => void, transcriptionChanged: () => void);
+    setAudioEnabled(enabled: boolean): void;
+    setTranscriptionEnabled(enabled: boolean): void;
+    get audioEnabled(): boolean;
+    get transcriptionEnabled(): boolean;
+    get audio(): AudioOutput | null;
+    set audio(sink: AudioOutput | null);
+    get transcription(): TextOutput | null;
+    set transcription(sink: TextOutput | null);
+}
+//# sourceMappingURL=io.d.ts.map

package/dist/voice/io.d.ts

@@ -0,0 +1,126 @@
+/// <reference types="node" resolution-mode="require"/>
+/// <reference types="node" resolution-mode="require"/>
+import type { AudioFrame } from '@livekit/rtc-node';
+import { EventEmitter } from 'node:events';
+import type { ReadableStream } from 'node:stream/web';
+import type { ChatContext } from '../llm/chat_context.js';
+import type { ChatChunk } from '../llm/llm.js';
+import type { ToolContext } from '../llm/tool_context.js';
+import { DeferredReadableStream } from '../stream/deferred_stream.js';
+import type { SpeechEvent } from '../stt/stt.js';
+import type { ModelSettings } from './agent.js';
+export type STTNode = (audio: ReadableStream<AudioFrame>, modelSettings: ModelSettings) => Promise<ReadableStream<SpeechEvent | string> | null>;
+export type LLMNode = (chatCtx: ChatContext, toolCtx: ToolContext, modelSettings: ModelSettings) => Promise<ReadableStream<ChatChunk | string> | null>;
+export type TTSNode = (text: ReadableStream<string>, modelSettings: ModelSettings) => Promise<ReadableStream<AudioFrame> | null>;
+/**
+ * A string with timing information for word-level alignment.
+ */
+export interface TimedString {
+    text: string;
+    startTime?: number;
+    endTime?: number;
+}
+export interface AudioOutputCapabilities {
+    /** Whether this output supports pause/resume functionality */
+    pause: boolean;
+}
+export declare abstract class AudioInput {
+    protected deferredStream: DeferredReadableStream<AudioFrame>;
+    get stream(): ReadableStream<AudioFrame>;
+    onAttached(): void;
+    onDetached(): void;
+}
+export declare abstract class AudioOutput extends EventEmitter {
+    sampleRate?: number | undefined;
+    protected readonly nextInChain?: AudioOutput | undefined;
+    static readonly EVENT_PLAYBACK_FINISHED = "playbackFinished";
+    private playbackFinishedFuture;
+    private _capturing;
+    private playbackFinishedCount;
+    private playbackSegmentsCount;
+    private lastPlaybackEvent;
+    protected logger: import("pino").Logger;
+    protected readonly capabilities: AudioOutputCapabilities;
+    constructor(sampleRate?: number | undefined, nextInChain?: AudioOutput | undefined, capabilities?: AudioOutputCapabilities);
+    /**
+     * Whether this output and all outputs in the chain support pause/resume.
+     */
+    get canPause(): boolean;
+    /**
+     * Capture an audio frame for playback, frames can be pushed faster than real-time
+     */
+    captureFrame(_frame: AudioFrame): Promise<void>;
+    /**
+     * Wait for the past audio segments to finish playing out.
+     *
+     * @returns The event that was emitted when the audio finished playing out (only the last segment information)
+     */
+    waitForPlayout(): Promise<PlaybackFinishedEvent>;
+    /**
+     * Developers building audio sinks must call this method when a playback/segment is finished.
+     * Segments are segmented by calls to flush() or clearBuffer()
+     */
+    onPlaybackFinished(options: PlaybackFinishedEvent): void;
+    flush(): void;
+    /**
+     * Clear the buffer, stopping playback immediately
+     */
+    abstract clearBuffer(): void;
+    onAttached(): void;
+    onDetached(): void;
+    /**
+     * Pause the audio playback
+     */
+    pause(): void;
+    /**
+     * Resume the audio playback
+     */
+    resume(): void;
+}
+export interface PlaybackFinishedEvent {
+    playbackPosition: number;
+    interrupted: boolean;
+    synchronizedTranscript?: string;
+}
+export declare abstract class TextOutput {
+    protected readonly nextInChain?: TextOutput | undefined;
+    constructor(nextInChain?: TextOutput | undefined);
+    /**
+     * Capture a text segment (Used by the output of LLM nodes)
+     */
+    abstract captureText(text: string): Promise<void>;
+    /**
+     * Mark the current text segment as complete (e.g LLM generation is complete)
+     */
+    abstract flush(): void;
+    onAttached(): void;
+    onDetached(): void;
+}
+export declare class AgentInput {
+    private readonly audioChanged;
+    private _audioStream;
+    private _audioEnabled;
+    constructor(audioChanged: () => void);
+    setAudioEnabled(enable: boolean): void;
+    get audioEnabled(): boolean;
+    get audio(): AudioInput | null;
+    set audio(stream: AudioInput | null);
+}
+export declare class AgentOutput {
+    private readonly audioChanged;
+    private readonly transcriptionChanged;
+    private _audioSink;
+    private _transcriptionSink;
+    private _audioEnabled;
+    private _transcriptionEnabled;
+    constructor(audioChanged: () => void, transcriptionChanged: () => void);
+    setAudioEnabled(enabled: boolean): void;
+    setTranscriptionEnabled(enabled: boolean): void;
+    get audioEnabled(): boolean;
+    get transcriptionEnabled(): boolean;
+    get audio(): AudioOutput | null;
+    set audio(sink: AudioOutput | null);
+    get transcription(): TextOutput | null;
+    set transcription(sink: TextOutput | null);
+}
+//# sourceMappingURL=io.d.ts.map

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

@@ -0,0 +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;AAEhD;;GAEG;AACH,MAAM,WAAW,WAAW;IAC1B,IAAI,EAAE,MAAM,CAAC;IACb,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,OAAO,CAAC,EAAE,MAAM,CAAC;CAClB;AAED,MAAM,WAAW,uBAAuB;IACtC,8DAA8D;IAC9D,KAAK,EAAE,OAAO,CAAC;CAChB;AAED,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;IAe3C,UAAU,CAAC;IAClB,SAAS,CAAC,QAAQ,CAAC,WAAW,CAAC;IAfjC,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;IACzB,SAAS,CAAC,QAAQ,CAAC,YAAY,EAAE,uBAAuB,CAAC;gBAGhD,UAAU,CAAC,oBAAQ,EACP,WAAW,CAAC,yBAAa,EAC5C,YAAY,GAAE,uBAA0C;IAW1D;;OAEG;IACH,IAAI,QAAQ,IAAI,OAAO,CAEtB;IAED;;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;IAMlB;;OAEG;IACH,KAAK,IAAI,IAAI;IAMb;;OAEG;IACH,MAAM,IAAI,IAAI;CAKf;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

@@ -0,0 +1,242 @@
+import { EventEmitter } from "node:events";
+import { log } from "../log.js";
+import { DeferredReadableStream } from "../stream/deferred_stream.js";
+import { Future } from "../utils.js";
+class AudioInput {
+  deferredStream = new DeferredReadableStream();
+  get stream() {
+    return this.deferredStream.stream;
+  }
+  onAttached() {
+  }
+  onDetached() {
+  }
+}
+class AudioOutput extends EventEmitter {
+  constructor(sampleRate, nextInChain, capabilities = { pause: false }) {
+    super();
+    this.sampleRate = sampleRate;
+    this.nextInChain = nextInChain;
+    this.capabilities = capabilities;
+    if (this.nextInChain) {
+      this.nextInChain.on(
+        AudioOutput.EVENT_PLAYBACK_FINISHED,
+        (ev) => this.onPlaybackFinished(ev)
+      );
+    }
+  }
+  static EVENT_PLAYBACK_FINISHED = "playbackFinished";
+  playbackFinishedFuture = new Future();
+  _capturing = false;
+  playbackFinishedCount = 0;
+  playbackSegmentsCount = 0;
+  lastPlaybackEvent = {
+    playbackPosition: 0,
+    interrupted: false
+  };
+  logger = log();
+  capabilities;
+  /**
+   * Whether this output and all outputs in the chain support pause/resume.
+   */
+  get canPause() {
+    var _a;
+    return this.capabilities.pause && (((_a = this.nextInChain) == null ? void 0 : _a.canPause) ?? true);
+  }
+  /**
+   * Capture an audio frame for playback, frames can be pushed faster than real-time
+   */
+  async captureFrame(_frame) {
+    if (!this._capturing) {
+      this._capturing = true;
+      this.playbackSegmentsCount++;
+    }
+  }
+  /**
+   * Wait for the past audio segments to finish playing out.
+   *
+   * @returns The event that was emitted when the audio finished playing out (only the last segment information)
+   */
+  async waitForPlayout() {
+    const target = this.playbackSegmentsCount;
+    while (this.playbackFinishedCount < target) {
+      await this.playbackFinishedFuture.await;
+      this.playbackFinishedFuture = new Future();
+    }
+    return this.lastPlaybackEvent;
+  }
+  /**
+   * Developers building audio sinks must call this method when a playback/segment is finished.
+   * Segments are segmented by calls to flush() or clearBuffer()
+   */
+  onPlaybackFinished(options) {
+    if (this.playbackFinishedCount >= this.playbackSegmentsCount) {
+      this.logger.warn("playback_finished called more times than playback segments were captured");
+      return;
+    }
+    this.lastPlaybackEvent = options;
+    this.playbackFinishedCount++;
+    this.playbackFinishedFuture.resolve();
+    this.emit(AudioOutput.EVENT_PLAYBACK_FINISHED, options);
+  }
+  flush() {
+    this._capturing = false;
+  }
+  onAttached() {
+    if (this.nextInChain) {
+      this.nextInChain.onAttached();
+    }
+  }
+  onDetached() {
+    if (this.nextInChain) {
+      this.nextInChain.onDetached();
+    }
+  }
+  /**
+   * Pause the audio playback
+   */
+  pause() {
+    if (this.nextInChain) {
+      this.nextInChain.pause();
+    }
+  }
+  /**
+   * Resume the audio playback
+   */
+  resume() {
+    if (this.nextInChain) {
+      this.nextInChain.resume();
+    }
+  }
+}
+class TextOutput {
+  constructor(nextInChain) {
+    this.nextInChain = nextInChain;
+  }
+  onAttached() {
+    if (this.nextInChain) {
+      this.nextInChain.onAttached();
+    }
+  }
+  onDetached() {
+    if (this.nextInChain) {
+      this.nextInChain.onDetached();
+    }
+  }
+}
+class AgentInput {
+  constructor(audioChanged) {
+    this.audioChanged = audioChanged;
+  }
+  _audioStream = null;
+  // enabled by default
+  _audioEnabled = true;
+  setAudioEnabled(enable) {
+    if (enable === this._audioEnabled) {
+      return;
+    }
+    this._audioEnabled = enable;
+    if (!this._audioStream) {
+      return;
+    }
+    if (enable) {
+      this._audioStream.onAttached();
+    } else {
+      this._audioStream.onDetached();
+    }
+  }
+  get audioEnabled() {
+    return this._audioEnabled;
+  }
+  get audio() {
+    return this._audioStream;
+  }
+  set audio(stream) {
+    this._audioStream = stream;
+    this.audioChanged();
+  }
+}
+class AgentOutput {
+  constructor(audioChanged, transcriptionChanged) {
+    this.audioChanged = audioChanged;
+    this.transcriptionChanged = transcriptionChanged;
+  }
+  _audioSink = null;
+  _transcriptionSink = null;
+  _audioEnabled = true;
+  _transcriptionEnabled = true;
+  setAudioEnabled(enabled) {
+    if (enabled === this._audioEnabled) {
+      return;
+    }
+    this._audioEnabled = enabled;
+    if (!this._audioSink) {
+      return;
+    }
+    if (enabled) {
+      this._audioSink.onAttached();
+    } else {
+      this._audioSink.onDetached();
+    }
+  }
+  setTranscriptionEnabled(enabled) {
+    if (enabled === this._transcriptionEnabled) {
+      return;
+    }
+    this._transcriptionEnabled = enabled;
+    if (!this._transcriptionSink) {
+      return;
+    }
+    if (enabled) {
+      this._transcriptionSink.onAttached();
+    } else {
+      this._transcriptionSink.onDetached();
+    }
+  }
+  get audioEnabled() {
+    return this._audioEnabled;
+  }
+  get transcriptionEnabled() {
+    return this._transcriptionEnabled;
+  }
+  get audio() {
+    return this._audioSink;
+  }
+  set audio(sink) {
+    if (sink === this._audioSink) {
+      return;
+    }
+    if (this._audioSink) {
+      this._audioSink.onDetached();
+    }
+    this._audioSink = sink;
+    this.audioChanged();
+    if (this._audioSink) {
+      this._audioSink.onAttached();
+    }
+  }
+  get transcription() {
+    return this._transcriptionSink;
+  }
+  set transcription(sink) {
+    if (sink === this._transcriptionSink) {
+      return;
+    }
+    if (this._transcriptionSink) {
+      this._transcriptionSink.onDetached();
+    }
+    this._transcriptionSink = sink;
+    this.transcriptionChanged();
+    if (this._transcriptionSink) {
+      this._transcriptionSink.onAttached();
+    }
+  }
+}
+export {
+  AgentInput,
+  AgentOutput,
+  AudioInput,
+  AudioOutput,
+  TextOutput
+};
+//# sourceMappingURL=io.js.map

package/dist/voice/io.js.map

@@ -0,0 +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>;\n\n/**\n * A string with timing information for word-level alignment.\n */\nexport interface TimedString {\n  text: string;\n  startTime?: number; // seconds\n  endTime?: number; // seconds\n}\n\nexport interface AudioOutputCapabilities {\n  /** Whether this output supports pause/resume functionality */\n  pause: boolean;\n}\n\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  protected readonly capabilities: AudioOutputCapabilities;\n\n  constructor(\n    public sampleRate?: number,\n    protected readonly nextInChain?: AudioOutput,\n    capabilities: AudioOutputCapabilities = { pause: false },\n  ) {\n    super();\n    this.capabilities = capabilities;\n    if (this.nextInChain) {\n      this.nextInChain.on(AudioOutput.EVENT_PLAYBACK_FINISHED, (ev: PlaybackFinishedEvent) =>\n        this.onPlaybackFinished(ev),\n      );\n    }\n  }\n\n  /**\n   * Whether this output and all outputs in the chain support pause/resume.\n   */\n  get canPause(): boolean {\n    return this.capabilities.pause && (this.nextInChain?.canPause ?? true);\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  /**\n   * Pause the audio playback\n   */\n  pause(): void {\n    if (this.nextInChain) {\n      this.nextInChain.pause();\n    }\n  }\n\n  /**\n   * Resume the audio playback\n   */\n  resume(): void {\n    if (this.nextInChain) {\n      this.nextInChain.resume();\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;AAiChB,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,EAcrD,YACS,YACY,aACnB,eAAwC,EAAE,OAAO,MAAM,GACvD;AACA,UAAM;AAJC;AACY;AAInB,SAAK,eAAe;AACpB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY;AAAA,QAAG,YAAY;AAAA,QAAyB,CAAC,OACxD,KAAK,mBAAmB,EAAE;AAAA,MAC5B;AAAA,IACF;AAAA,EACF;AAAA,EAzBA,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,EACJ;AAAA;AAAA;AAAA;AAAA,EAmBnB,IAAI,WAAoB;AAzF1B;AA0FI,WAAO,KAAK,aAAa,YAAU,UAAK,gBAAL,mBAAkB,aAAY;AAAA,EACnE;AAAA;AAAA;AAAA;AAAA,EAKA,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;AAAA;AAAA;AAAA;AAAA,EAKA,QAAc;AACZ,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,MAAM;AAAA,IACzB;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKA,SAAe;AACb,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,OAAO;AAAA,IAC1B;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/recorder_io/index.cjs

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

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

@@ -0,0 +1 @@
+{"version":3,"sources":["../../../src/voice/recorder_io/index.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nexport * from './recorder_io.js';\n"],"mappings":";;;;;;;;;;;;;;;AAAA;AAAA;AAGA,gCAAc,6BAHd;","names":[]}

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

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

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

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

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

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

package/dist/voice/recorder_io/index.js

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

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

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