@livekit/agents 0.7.9 → 1.0.0-next.1

package/dist/pipeline/agent_playout.d.cts

@@ -1,40 +0,0 @@
-import type { AudioFrame, AudioSource } from '@livekit/rtc-node';
-import type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';
-import type { TextAudioSynchronizer } from '../transcription.js';
-import { Future } from '../utils.js';
-import { SynthesisHandle } from './agent_output.js';
-export declare enum AgentPlayoutEvent {
-    PLAYOUT_STARTED = 0,
-    PLAYOUT_STOPPED = 1
-}
-export type AgentPlayoutCallbacks = {
-    [AgentPlayoutEvent.PLAYOUT_STARTED]: () => void;
-    [AgentPlayoutEvent.PLAYOUT_STOPPED]: (interrupt: boolean) => void;
-};
-export declare class PlayoutHandle {
-    #private;
-    playoutSource: AsyncIterable<AudioFrame | typeof SynthesisHandle.FLUSH_SENTINEL>;
-    totalPlayedTime?: number;
-    synchronizer: TextAudioSynchronizer;
-    pushedDuration: number;
-    intFut: Future;
-    doneFut: Future;
-    constructor(speechId: string, audioSource: AudioSource, playoutSource: AsyncIterable<AudioFrame | typeof SynthesisHandle.FLUSH_SENTINEL>, synchronizer: TextAudioSynchronizer);
-    get speechId(): string;
-    get interrupted(): boolean;
-    get timePlayed(): number;
-    get done(): boolean;
-    interrupt(): void;
-    join(): Future;
-}
-declare const AgentPlayout_base: new () => TypedEmitter<AgentPlayoutCallbacks>;
-export declare class AgentPlayout extends AgentPlayout_base {
-    #private;
-    constructor(audioSource: AudioSource);
-    get targetVolume(): number;
-    set targetVolume(vol: number);
-    play(speechId: string, playoutSource: AsyncIterable<AudioFrame | typeof SynthesisHandle.FLUSH_SENTINEL>, synchronizer: TextAudioSynchronizer): PlayoutHandle;
-    close(): Promise<void>;
-}
-export {};
-//# sourceMappingURL=agent_playout.d.ts.map

package/dist/pipeline/agent_playout.d.ts

@@ -1,40 +0,0 @@
-import type { AudioFrame, AudioSource } from '@livekit/rtc-node';
-import type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';
-import type { TextAudioSynchronizer } from '../transcription.js';
-import { Future } from '../utils.js';
-import { SynthesisHandle } from './agent_output.js';
-export declare enum AgentPlayoutEvent {
-    PLAYOUT_STARTED = 0,
-    PLAYOUT_STOPPED = 1
-}
-export type AgentPlayoutCallbacks = {
-    [AgentPlayoutEvent.PLAYOUT_STARTED]: () => void;
-    [AgentPlayoutEvent.PLAYOUT_STOPPED]: (interrupt: boolean) => void;
-};
-export declare class PlayoutHandle {
-    #private;
-    playoutSource: AsyncIterable<AudioFrame | typeof SynthesisHandle.FLUSH_SENTINEL>;
-    totalPlayedTime?: number;
-    synchronizer: TextAudioSynchronizer;
-    pushedDuration: number;
-    intFut: Future;
-    doneFut: Future;
-    constructor(speechId: string, audioSource: AudioSource, playoutSource: AsyncIterable<AudioFrame | typeof SynthesisHandle.FLUSH_SENTINEL>, synchronizer: TextAudioSynchronizer);
-    get speechId(): string;
-    get interrupted(): boolean;
-    get timePlayed(): number;
-    get done(): boolean;
-    interrupt(): void;
-    join(): Future;
-}
-declare const AgentPlayout_base: new () => TypedEmitter<AgentPlayoutCallbacks>;
-export declare class AgentPlayout extends AgentPlayout_base {
-    #private;
-    constructor(audioSource: AudioSource);
-    get targetVolume(): number;
-    set targetVolume(vol: number);
-    play(speechId: string, playoutSource: AsyncIterable<AudioFrame | typeof SynthesisHandle.FLUSH_SENTINEL>, synchronizer: TextAudioSynchronizer): PlayoutHandle;
-    close(): Promise<void>;
-}
-export {};
-//# sourceMappingURL=agent_playout.d.ts.map

package/dist/pipeline/agent_playout.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"agent_playout.d.ts","sourceRoot":"","sources":["../../src/pipeline/agent_playout.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC;AACjE,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAGhF,OAAO,KAAK,EAAE,qBAAqB,EAAE,MAAM,qBAAqB,CAAC;AACjE,OAAO,EAAsB,MAAM,EAAoB,MAAM,aAAa,CAAC;AAC3E,OAAO,EAAE,eAAe,EAAE,MAAM,mBAAmB,CAAC;AAEpD,oBAAY,iBAAiB;IAC3B,eAAe,IAAA;IACf,eAAe,IAAA;CAChB;AAED,MAAM,MAAM,qBAAqB,GAAG;IAClC,CAAC,iBAAiB,CAAC,eAAe,CAAC,EAAE,MAAM,IAAI,CAAC;IAChD,CAAC,iBAAiB,CAAC,eAAe,CAAC,EAAE,CAAC,SAAS,EAAE,OAAO,KAAK,IAAI,CAAC;CACnE,CAAC;AAEF,qBAAa,aAAa;;IAGxB,aAAa,EAAE,aAAa,CAAC,UAAU,GAAG,OAAO,eAAe,CAAC,cAAc,CAAC,CAAC;IACjF,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,YAAY,EAAE,qBAAqB,CAAC;IAEpC,cAAc,SAAK;IACnB,MAAM,SAAgB;IACtB,OAAO,SAAgB;gBAGrB,QAAQ,EAAE,MAAM,EAChB,WAAW,EAAE,WAAW,EACxB,aAAa,EAAE,aAAa,CAAC,UAAU,GAAG,OAAO,eAAe,CAAC,cAAc,CAAC,EAChF,YAAY,EAAE,qBAAqB;IAQrC,IAAI,QAAQ,IAAI,MAAM,CAErB;IAED,IAAI,WAAW,IAAI,OAAO,CAEzB;IAED,IAAI,UAAU,IAAI,MAAM,CAEvB;IAED,IAAI,IAAI,IAAI,OAAO,CAElB;IAED,SAAS;IAST,IAAI,IAAI,MAAM;CAGf;2CAE4D,aAAa,qBAAqB,CAAC;AAAhG,qBAAa,YAAa,SAAQ,iBAA+D;;gBAOnF,WAAW,EAAE,WAAW;IAKpC,IAAI,YAAY,IAAI,MAAM,CAEzB;IAED,IAAI,YAAY,CAAC,GAAG,EAAE,MAAM,EAE3B;IAED,IAAI,CACF,QAAQ,EAAE,MAAM,EAChB,aAAa,EAAE,aAAa,CAAC,UAAU,GAAG,OAAO,eAAe,CAAC,cAAc,CAAC,EAChF,YAAY,EAAE,qBAAqB,GAClC,aAAa;IAyFV,KAAK;CAIZ"}

package/dist/pipeline/agent_playout.js

@@ -1,139 +0,0 @@
-import EventEmitter from "node:events";
-import { log } from "../log.js";
-import { CancellablePromise, Future, gracefullyCancel } from "../utils.js";
-import { SynthesisHandle } from "./agent_output.js";
-var AgentPlayoutEvent = /* @__PURE__ */ ((AgentPlayoutEvent2) => {
-  AgentPlayoutEvent2[AgentPlayoutEvent2["PLAYOUT_STARTED"] = 0] = "PLAYOUT_STARTED";
-  AgentPlayoutEvent2[AgentPlayoutEvent2["PLAYOUT_STOPPED"] = 1] = "PLAYOUT_STOPPED";
-  return AgentPlayoutEvent2;
-})(AgentPlayoutEvent || {});
-class PlayoutHandle {
-  #speechId;
-  #audioSource;
-  playoutSource;
-  totalPlayedTime;
-  synchronizer;
-  #interrupted = false;
-  pushedDuration = 0;
-  intFut = new Future();
-  doneFut = new Future();
-  constructor(speechId, audioSource, playoutSource, synchronizer) {
-    this.#speechId = speechId;
-    this.#audioSource = audioSource;
-    this.playoutSource = playoutSource;
-    this.synchronizer = synchronizer;
-  }
-  get speechId() {
-    return this.#speechId;
-  }
-  get interrupted() {
-    return this.#interrupted;
-  }
-  get timePlayed() {
-    return this.totalPlayedTime || this.pushedDuration - this.#audioSource.queuedDuration;
-  }
-  get done() {
-    return this.doneFut.done || this.#interrupted;
-  }
-  interrupt() {
-    if (this.done) {
-      return;
-    }
-    this.intFut.resolve();
-    this.#interrupted = true;
-  }
-  join() {
-    return this.doneFut;
-  }
-}
-class AgentPlayout extends EventEmitter {
-  #closed = false;
-  #audioSource;
-  #targetVolume = 1;
-  #playoutTask;
-  #logger = log();
-  constructor(audioSource) {
-    super();
-    this.#audioSource = audioSource;
-  }
-  get targetVolume() {
-    return this.#targetVolume;
-  }
-  set targetVolume(vol) {
-    this.#targetVolume = vol;
-  }
-  play(speechId, playoutSource, synchronizer) {
-    if (this.#closed) {
-      throw new Error("source closed");
-    }
-    const handle = new PlayoutHandle(speechId, this.#audioSource, playoutSource, synchronizer);
-    this.#playoutTask = this.#playout(handle, this.#playoutTask);
-    return handle;
-  }
-  #playout(handle, oldTask) {
-    return new CancellablePromise(async (resolve, _, onCancel) => {
-      const cancel = () => {
-        captureTask.cancel();
-        handle.totalPlayedTime = handle.pushedDuration - this.#audioSource.queuedDuration;
-        if (handle.interrupted || captureTask.error) {
-          handle.synchronizer.close(true);
-          this.#audioSource.clearQueue();
-        }
-        if (!firstFrame) {
-          this.emit(1 /* PLAYOUT_STOPPED */, handle.interrupted);
-        }
-        handle.doneFut.resolve();
-        this.#logger.child({ speechId: handle.speechId, interrupted: handle.interrupted }).debug("playout finished");
-      };
-      onCancel(() => {
-        cancel();
-      });
-      if (oldTask) {
-        await gracefullyCancel(oldTask);
-      }
-      if (this.#audioSource.queuedDuration > 0) {
-        this.#logger.child({ speechId: handle.speechId, queuedDuration: this.#audioSource.queuedDuration }).warn("new playout while the source is still playing");
-      }
-      let firstFrame = true;
-      const captureTask = new CancellablePromise(async (resolve2, _2, onCancel2) => {
-        let cancelled = false;
-        onCancel2(() => {
-          cancelled = true;
-        });
-        for await (const frame of handle.playoutSource) {
-          if (cancelled || frame === SynthesisHandle.FLUSH_SENTINEL) {
-            break;
-          }
-          if (firstFrame) {
-            this.#logger.child({ speechId: handle.speechId }).debug("started playing the first time");
-            this.emit(0 /* PLAYOUT_STARTED */);
-            handle.synchronizer.segmentPlayoutStarted();
-            firstFrame = false;
-          }
-          handle.pushedDuration += frame.samplesPerChannel / frame.sampleRate * 1e3;
-          handle.synchronizer.pushAudio(frame);
-          await this.#audioSource.captureFrame(frame);
-        }
-        await this.#audioSource.waitForPlayout();
-        handle.synchronizer.close(false);
-        resolve2();
-      });
-      try {
-        await Promise.any([captureTask, handle.intFut.await]);
-      } finally {
-        cancel();
-        resolve();
-      }
-    });
-  }
-  async close() {
-    this.#closed = true;
-    await this.#playoutTask;
-  }
-}
-export {
-  AgentPlayout,
-  AgentPlayoutEvent,
-  PlayoutHandle
-};
-//# sourceMappingURL=agent_playout.js.map

package/dist/pipeline/agent_playout.js.map

@@ -1 +0,0 @@
-{"version":3,"sources":["../../src/pipeline/agent_playout.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame, AudioSource } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport EventEmitter from 'node:events';\nimport { log } from '../log.js';\nimport type { TextAudioSynchronizer } from '../transcription.js';\nimport { CancellablePromise, Future, gracefullyCancel } from '../utils.js';\nimport { SynthesisHandle } from './agent_output.js';\n\nexport enum AgentPlayoutEvent {\n  PLAYOUT_STARTED,\n  PLAYOUT_STOPPED,\n}\n\nexport type AgentPlayoutCallbacks = {\n  [AgentPlayoutEvent.PLAYOUT_STARTED]: () => void;\n  [AgentPlayoutEvent.PLAYOUT_STOPPED]: (interrupt: boolean) => void;\n};\n\nexport class PlayoutHandle {\n  #speechId: string;\n  #audioSource: AudioSource;\n  playoutSource: AsyncIterable<AudioFrame | typeof SynthesisHandle.FLUSH_SENTINEL>;\n  totalPlayedTime?: number;\n  synchronizer: TextAudioSynchronizer;\n  #interrupted = false;\n  pushedDuration = 0;\n  intFut = new Future();\n  doneFut = new Future();\n\n  constructor(\n    speechId: string,\n    audioSource: AudioSource,\n    playoutSource: AsyncIterable<AudioFrame | typeof SynthesisHandle.FLUSH_SENTINEL>,\n    synchronizer: TextAudioSynchronizer,\n  ) {\n    this.#speechId = speechId;\n    this.#audioSource = audioSource;\n    this.playoutSource = playoutSource;\n    this.synchronizer = synchronizer;\n  }\n\n  get speechId(): string {\n    return this.#speechId;\n  }\n\n  get interrupted(): boolean {\n    return this.#interrupted;\n  }\n\n  get timePlayed(): number {\n    return this.totalPlayedTime || this.pushedDuration - this.#audioSource.queuedDuration;\n  }\n\n  get done(): boolean {\n    return this.doneFut.done || this.#interrupted;\n  }\n\n  interrupt() {\n    if (this.done) {\n      return;\n    }\n\n    this.intFut.resolve();\n    this.#interrupted = true;\n  }\n\n  join(): Future {\n    return this.doneFut;\n  }\n}\n\nexport class AgentPlayout extends (EventEmitter as new () => TypedEmitter<AgentPlayoutCallbacks>) {\n  #closed = false;\n  #audioSource: AudioSource;\n  #targetVolume = 1;\n  #playoutTask?: CancellablePromise<void>;\n  #logger = log();\n\n  constructor(audioSource: AudioSource) {\n    super();\n    this.#audioSource = audioSource;\n  }\n\n  get targetVolume(): number {\n    return this.#targetVolume;\n  }\n\n  set targetVolume(vol: number) {\n    this.#targetVolume = vol;\n  }\n\n  play(\n    speechId: string,\n    playoutSource: AsyncIterable<AudioFrame | typeof SynthesisHandle.FLUSH_SENTINEL>,\n    synchronizer: TextAudioSynchronizer,\n  ): PlayoutHandle {\n    if (this.#closed) {\n      throw new Error('source closed');\n    }\n\n    const handle = new PlayoutHandle(speechId, this.#audioSource, playoutSource, synchronizer);\n\n    this.#playoutTask = this.#playout(handle, this.#playoutTask);\n    return handle;\n  }\n\n  #playout(handle: PlayoutHandle, oldTask?: CancellablePromise<void>): CancellablePromise<void> {\n    return new CancellablePromise(async (resolve, _, onCancel) => {\n      const cancel = () => {\n        captureTask.cancel();\n        handle.totalPlayedTime = handle.pushedDuration - this.#audioSource.queuedDuration;\n\n        if (handle.interrupted || captureTask.error) {\n          handle.synchronizer.close(true);\n          this.#audioSource.clearQueue(); // make sure to remove any queued frames\n        }\n\n        if (!firstFrame) {\n          this.emit(AgentPlayoutEvent.PLAYOUT_STOPPED, handle.interrupted);\n        }\n\n        handle.doneFut.resolve();\n\n        this.#logger\n          .child({ speechId: handle.speechId, interrupted: handle.interrupted })\n          .debug('playout finished');\n      };\n\n      onCancel(() => {\n        cancel();\n      });\n\n      if (oldTask) {\n        await gracefullyCancel(oldTask);\n      }\n\n      if (this.#audioSource.queuedDuration > 0) {\n        // this should not happen, but log it just in case\n        this.#logger\n          .child({ speechId: handle.speechId, queuedDuration: this.#audioSource.queuedDuration })\n          .warn('new playout while the source is still playing');\n      }\n\n      let firstFrame = true;\n\n      // eslint-disable-next-line @typescript-eslint/no-unused-vars\n      const captureTask = new CancellablePromise<void>(async (resolve, _, onCancel) => {\n        let cancelled = false;\n        onCancel(() => {\n          cancelled = true;\n        });\n\n        for await (const frame of handle.playoutSource) {\n          if (cancelled || frame === SynthesisHandle.FLUSH_SENTINEL) {\n            break;\n          }\n          if (firstFrame) {\n            this.#logger\n              .child({ speechId: handle.speechId })\n              .debug('started playing the first time');\n            this.emit(AgentPlayoutEvent.PLAYOUT_STARTED);\n            handle.synchronizer.segmentPlayoutStarted();\n            firstFrame = false;\n          }\n          handle.pushedDuration += (frame.samplesPerChannel / frame.sampleRate) * 1000;\n          handle.synchronizer.pushAudio(frame);\n          await this.#audioSource.captureFrame(frame);\n        }\n\n        await this.#audioSource.waitForPlayout();\n\n        handle.synchronizer.close(false);\n        resolve();\n      });\n\n      try {\n        await Promise.any([captureTask, handle.intFut.await]);\n      } finally {\n        cancel();\n        resolve();\n      }\n    });\n  }\n\n  async close() {\n    this.#closed = true;\n    await this.#playoutTask;\n  }\n}\n"],"mappings":"AAKA,OAAO,kBAAkB;AACzB,SAAS,WAAW;AAEpB,SAAS,oBAAoB,QAAQ,wBAAwB;AAC7D,SAAS,uBAAuB;AAEzB,IAAK,oBAAL,kBAAKA,uBAAL;AACL,EAAAA,sCAAA;AACA,EAAAA,sCAAA;AAFU,SAAAA;AAAA,GAAA;AAUL,MAAM,cAAc;AAAA,EACzB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,eAAe;AAAA,EACf,iBAAiB;AAAA,EACjB,SAAS,IAAI,OAAO;AAAA,EACpB,UAAU,IAAI,OAAO;AAAA,EAErB,YACE,UACA,aACA,eACA,cACA;AACA,SAAK,YAAY;AACjB,SAAK,eAAe;AACpB,SAAK,gBAAgB;AACrB,SAAK,eAAe;AAAA,EACtB;AAAA,EAEA,IAAI,WAAmB;AACrB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,cAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,aAAqB;AACvB,WAAO,KAAK,mBAAmB,KAAK,iBAAiB,KAAK,aAAa;AAAA,EACzE;AAAA,EAEA,IAAI,OAAgB;AAClB,WAAO,KAAK,QAAQ,QAAQ,KAAK;AAAA,EACnC;AAAA,EAEA,YAAY;AACV,QAAI,KAAK,MAAM;AACb;AAAA,IACF;AAEA,SAAK,OAAO,QAAQ;AACpB,SAAK,eAAe;AAAA,EACtB;AAAA,EAEA,OAAe;AACb,WAAO,KAAK;AAAA,EACd;AACF;AAEO,MAAM,qBAAsB,aAA+D;AAAA,EAChG,UAAU;AAAA,EACV;AAAA,EACA,gBAAgB;AAAA,EAChB;AAAA,EACA,UAAU,IAAI;AAAA,EAEd,YAAY,aAA0B;AACpC,UAAM;AACN,SAAK,eAAe;AAAA,EACtB;AAAA,EAEA,IAAI,eAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,aAAa,KAAa;AAC5B,SAAK,gBAAgB;AAAA,EACvB;AAAA,EAEA,KACE,UACA,eACA,cACe;AACf,QAAI,KAAK,SAAS;AAChB,YAAM,IAAI,MAAM,eAAe;AAAA,IACjC;AAEA,UAAM,SAAS,IAAI,cAAc,UAAU,KAAK,cAAc,eAAe,YAAY;AAEzF,SAAK,eAAe,KAAK,SAAS,QAAQ,KAAK,YAAY;AAC3D,WAAO;AAAA,EACT;AAAA,EAEA,SAAS,QAAuB,SAA8D;AAC5F,WAAO,IAAI,mBAAmB,OAAO,SAAS,GAAG,aAAa;AAC5D,YAAM,SAAS,MAAM;AACnB,oBAAY,OAAO;AACnB,eAAO,kBAAkB,OAAO,iBAAiB,KAAK,aAAa;AAEnE,YAAI,OAAO,eAAe,YAAY,OAAO;AAC3C,iBAAO,aAAa,MAAM,IAAI;AAC9B,eAAK,aAAa,WAAW;AAAA,QAC/B;AAEA,YAAI,CAAC,YAAY;AACf,eAAK,KAAK,yBAAmC,OAAO,WAAW;AAAA,QACjE;AAEA,eAAO,QAAQ,QAAQ;AAEvB,aAAK,QACF,MAAM,EAAE,UAAU,OAAO,UAAU,aAAa,OAAO,YAAY,CAAC,EACpE,MAAM,kBAAkB;AAAA,MAC7B;AAEA,eAAS,MAAM;AACb,eAAO;AAAA,MACT,CAAC;AAED,UAAI,SAAS;AACX,cAAM,iBAAiB,OAAO;AAAA,MAChC;AAEA,UAAI,KAAK,aAAa,iBAAiB,GAAG;AAExC,aAAK,QACF,MAAM,EAAE,UAAU,OAAO,UAAU,gBAAgB,KAAK,aAAa,eAAe,CAAC,EACrF,KAAK,+CAA+C;AAAA,MACzD;AAEA,UAAI,aAAa;AAGjB,YAAM,cAAc,IAAI,mBAAyB,OAAOC,UAASC,IAAGC,cAAa;AAC/E,YAAI,YAAY;AAChB,QAAAA,UAAS,MAAM;AACb,sBAAY;AAAA,QACd,CAAC;AAED,yBAAiB,SAAS,OAAO,eAAe;AAC9C,cAAI,aAAa,UAAU,gBAAgB,gBAAgB;AACzD;AAAA,UACF;AACA,cAAI,YAAY;AACd,iBAAK,QACF,MAAM,EAAE,UAAU,OAAO,SAAS,CAAC,EACnC,MAAM,gCAAgC;AACzC,iBAAK,KAAK,uBAAiC;AAC3C,mBAAO,aAAa,sBAAsB;AAC1C,yBAAa;AAAA,UACf;AACA,iBAAO,kBAAmB,MAAM,oBAAoB,MAAM,aAAc;AACxE,iBAAO,aAAa,UAAU,KAAK;AACnC,gBAAM,KAAK,aAAa,aAAa,KAAK;AAAA,QAC5C;AAEA,cAAM,KAAK,aAAa,eAAe;AAEvC,eAAO,aAAa,MAAM,KAAK;AAC/B,QAAAF,SAAQ;AAAA,MACV,CAAC;AAED,UAAI;AACF,cAAM,QAAQ,IAAI,CAAC,aAAa,OAAO,OAAO,KAAK,CAAC;AAAA,MACtD,UAAE;AACA,eAAO;AACP,gBAAQ;AAAA,MACV;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAEA,MAAM,QAAQ;AACZ,SAAK,UAAU;AACf,UAAM,KAAK;AAAA,EACb;AACF;","names":["AgentPlayoutEvent","resolve","_","onCancel"]}

package/dist/pipeline/human_input.cjs

@@ -1,171 +0,0 @@
-"use strict";
-var __defProp = Object.defineProperty;
-var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
-var __getOwnPropNames = Object.getOwnPropertyNames;
-var __hasOwnProp = Object.prototype.hasOwnProperty;
-var __export = (target, all) => {
-  for (var name in all)
-    __defProp(target, name, { get: all[name], enumerable: true });
-};
-var __copyProps = (to, from, except, desc) => {
-  if (from && typeof from === "object" || typeof from === "function") {
-    for (let key of __getOwnPropNames(from))
-      if (!__hasOwnProp.call(to, key) && key !== except)
-        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
-  }
-  return to;
-};
-var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
-var human_input_exports = {};
-__export(human_input_exports, {
-  HumanInput: () => HumanInput,
-  HumanInputEvent: () => HumanInputEvent
-});
-module.exports = __toCommonJS(human_input_exports);
-var import_rtc_node = require("@livekit/rtc-node");
-var import_node_events = require("node:events");
-var import_log = require("../log.cjs");
-var import_stt = require("../stt/stt.cjs");
-var import_utils = require("../utils.cjs");
-var import_vad = require("../vad.cjs");
-var HumanInputEvent = /* @__PURE__ */ ((HumanInputEvent2) => {
-  HumanInputEvent2[HumanInputEvent2["START_OF_SPEECH"] = 0] = "START_OF_SPEECH";
-  HumanInputEvent2[HumanInputEvent2["VAD_INFERENCE_DONE"] = 1] = "VAD_INFERENCE_DONE";
-  HumanInputEvent2[HumanInputEvent2["END_OF_SPEECH"] = 2] = "END_OF_SPEECH";
-  HumanInputEvent2[HumanInputEvent2["FINAL_TRANSCRIPT"] = 3] = "FINAL_TRANSCRIPT";
-  HumanInputEvent2[HumanInputEvent2["INTERIM_TRANSCRIPT"] = 4] = "INTERIM_TRANSCRIPT";
-  return HumanInputEvent2;
-})(HumanInputEvent || {});
-class HumanInput extends import_node_events.EventEmitter {
-  #closed = false;
-  #room;
-  #vad;
-  #stt;
-  #participant;
-  #subscribedTrack;
-  #recognizeTask;
-  #speaking = false;
-  #speechProbability = 0;
-  #logger = (0, import_log.log)();
-  #noiseCancellation;
-  constructor(room, vad, stt, participant, noiseCancellation) {
-    super();
-    this.#room = room;
-    this.#vad = vad;
-    this.#stt = stt;
-    this.#participant = participant;
-    this.#noiseCancellation = noiseCancellation;
-    this.#room.on(import_rtc_node.RoomEvent.TrackPublished, this.#subscribeToMicrophone.bind(this));
-    this.#room.on(import_rtc_node.RoomEvent.TrackSubscribed, this.#subscribeToMicrophone.bind(this));
-    this.#subscribeToMicrophone();
-  }
-  get participant() {
-    return this.#participant;
-  }
-  get subscribedTrack() {
-    return this.#subscribedTrack;
-  }
-  #subscribeToMicrophone() {
-    if (!this.#participant) {
-      this.#logger.error("Participant is not set");
-      return;
-    }
-    let microphonePublication = void 0;
-    for (const publication of this.#participant.trackPublications.values()) {
-      if (publication.source === import_rtc_node.TrackSource.SOURCE_MICROPHONE) {
-        microphonePublication = publication;
-        break;
-      }
-    }
-    if (!microphonePublication) {
-      return;
-    }
-    if (!microphonePublication.subscribed) {
-      microphonePublication.setSubscribed(true);
-    }
-    const track = microphonePublication.track;
-    if (track && track !== this.#subscribedTrack) {
-      this.#subscribedTrack = track;
-      if (this.#recognizeTask) {
-        this.#recognizeTask.cancel();
-      }
-      const audioStreamOptions = {
-        sampleRate: 16e3,
-        numChannels: 1,
-        ...this.#noiseCancellation ? { noiseCancellation: this.#noiseCancellation } : {}
-      };
-      const audioStream = new import_rtc_node.AudioStream(track, audioStreamOptions);
-      this.#recognizeTask = new import_utils.CancellablePromise(async (resolve, _, onCancel) => {
-        let cancelled = false;
-        onCancel(() => {
-          cancelled = true;
-        });
-        const sttStream = this.#stt.stream();
-        const vadStream = this.#vad.stream();
-        const audioStreamCo = async () => {
-          for await (const ev of audioStream) {
-            if (cancelled) return;
-            sttStream.pushFrame(ev);
-            vadStream.pushFrame(ev);
-          }
-        };
-        const vadStreamCo = async () => {
-          for await (const ev of vadStream) {
-            if (cancelled) return;
-            switch (ev.type) {
-              case import_vad.VADEventType.START_OF_SPEECH:
-                this.#speaking = true;
-                this.emit(0 /* START_OF_SPEECH */, ev);
-                break;
-              case import_vad.VADEventType.INFERENCE_DONE:
-                this.#speechProbability = ev.probability;
-                this.emit(1 /* VAD_INFERENCE_DONE */, ev);
-                break;
-              case import_vad.VADEventType.END_OF_SPEECH:
-                this.#speaking = false;
-                this.emit(2 /* END_OF_SPEECH */, ev);
-                break;
-            }
-          }
-        };
-        const sttStreamCo = async () => {
-          for await (const ev of sttStream) {
-            if (cancelled) return;
-            if (ev.type === import_stt.SpeechEventType.FINAL_TRANSCRIPT) {
-              this.emit(3 /* FINAL_TRANSCRIPT */, ev);
-            } else if (ev.type == import_stt.SpeechEventType.INTERIM_TRANSCRIPT) {
-              this.emit(4 /* INTERIM_TRANSCRIPT */, ev);
-            }
-          }
-        };
-        await Promise.all([audioStreamCo(), vadStreamCo(), sttStreamCo()]);
-        sttStream.close();
-        vadStream.close();
-        resolve();
-      });
-    }
-  }
-  get speaking() {
-    return this.#speaking;
-  }
-  get speakingProbability() {
-    return this.#speechProbability;
-  }
-  async close() {
-    if (this.#closed) {
-      throw new Error("HumanInput already closed");
-    }
-    this.#closed = true;
-    this.#room.removeAllListeners();
-    this.#speaking = false;
-    if (this.#recognizeTask) {
-      await (0, import_utils.gracefullyCancel)(this.#recognizeTask);
-    }
-  }
-}
-// Annotate the CommonJS export names for ESM import in node:
-0 && (module.exports = {
-  HumanInput,
-  HumanInputEvent
-});
-//# sourceMappingURL=human_input.cjs.map

package/dist/pipeline/human_input.cjs.map

@@ -1 +0,0 @@
-{"version":3,"sources":["../../src/pipeline/human_input.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type {\n  NoiseCancellationOptions,\n  RemoteAudioTrack,\n  RemoteParticipant,\n  RemoteTrackPublication,\n  Room,\n} from '@livekit/rtc-node';\nimport { AudioStream, RoomEvent, TrackSource } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport { log } from '../log.js';\nimport type { STT, SpeechEvent } from '../stt/stt.js';\nimport { SpeechEventType } from '../stt/stt.js';\nimport { CancellablePromise, gracefullyCancel } from '../utils.js';\nimport type { VAD, VADEvent } from '../vad.js';\nimport { VADEventType } from '../vad.js';\n\nexport enum HumanInputEvent {\n  START_OF_SPEECH,\n  VAD_INFERENCE_DONE,\n  END_OF_SPEECH,\n  FINAL_TRANSCRIPT,\n  INTERIM_TRANSCRIPT,\n}\n\nexport type HumanInputCallbacks = {\n  [HumanInputEvent.START_OF_SPEECH]: (event: VADEvent) => void;\n  [HumanInputEvent.VAD_INFERENCE_DONE]: (event: VADEvent) => void;\n  [HumanInputEvent.END_OF_SPEECH]: (event: VADEvent) => void;\n  [HumanInputEvent.FINAL_TRANSCRIPT]: (event: SpeechEvent) => void;\n  [HumanInputEvent.INTERIM_TRANSCRIPT]: (event: SpeechEvent) => void;\n};\n\nexport class HumanInput extends (EventEmitter as new () => TypedEmitter<HumanInputCallbacks>) {\n  #closed = false;\n  #room: Room;\n  #vad: VAD;\n  #stt: STT;\n  #participant: RemoteParticipant;\n  #subscribedTrack?: RemoteAudioTrack;\n  #recognizeTask?: CancellablePromise<void>;\n  #speaking = false;\n  #speechProbability = 0;\n  #logger = log();\n  #noiseCancellation?: NoiseCancellationOptions;\n\n  constructor(\n    room: Room,\n    vad: VAD,\n    stt: STT,\n    participant: RemoteParticipant,\n    noiseCancellation?: NoiseCancellationOptions,\n  ) {\n    super();\n    this.#room = room;\n    this.#vad = vad;\n    this.#stt = stt;\n    this.#participant = participant;\n    this.#noiseCancellation = noiseCancellation;\n\n    this.#room.on(RoomEvent.TrackPublished, this.#subscribeToMicrophone.bind(this));\n    this.#room.on(RoomEvent.TrackSubscribed, this.#subscribeToMicrophone.bind(this));\n    this.#subscribeToMicrophone();\n  }\n\n  get participant(): RemoteParticipant {\n    return this.#participant;\n  }\n\n  get subscribedTrack(): RemoteAudioTrack | undefined {\n    return this.#subscribedTrack;\n  }\n\n  #subscribeToMicrophone(): void {\n    if (!this.#participant) {\n      this.#logger.error('Participant is not set');\n      return;\n    }\n\n    let microphonePublication: RemoteTrackPublication | undefined = undefined;\n    for (const publication of this.#participant.trackPublications.values()) {\n      if (publication.source === TrackSource.SOURCE_MICROPHONE) {\n        microphonePublication = publication;\n        break;\n      }\n    }\n    if (!microphonePublication) {\n      return;\n    }\n\n    if (!microphonePublication.subscribed) {\n      microphonePublication.setSubscribed(true);\n    }\n\n    const track = microphonePublication.track;\n    if (track && track !== this.#subscribedTrack) {\n      this.#subscribedTrack = track;\n      if (this.#recognizeTask) {\n        this.#recognizeTask.cancel();\n      }\n\n      const audioStreamOptions = {\n        sampleRate: 16000,\n        numChannels: 1,\n        ...(this.#noiseCancellation ? { noiseCancellation: this.#noiseCancellation } : {}),\n      };\n      const audioStream = new AudioStream(track, audioStreamOptions);\n\n      // eslint-disable-next-line @typescript-eslint/no-unused-vars\n      this.#recognizeTask = new CancellablePromise(async (resolve, _, onCancel) => {\n        let cancelled = false;\n        onCancel(() => {\n          cancelled = true;\n        });\n\n        const sttStream = this.#stt.stream();\n        const vadStream = this.#vad.stream();\n\n        const audioStreamCo = async () => {\n          for await (const ev of audioStream) {\n            if (cancelled) return;\n            sttStream.pushFrame(ev);\n            vadStream.pushFrame(ev);\n          }\n        };\n\n        const vadStreamCo = async () => {\n          for await (const ev of vadStream) {\n            if (cancelled) return;\n            switch (ev.type) {\n              case VADEventType.START_OF_SPEECH:\n                this.#speaking = true;\n                this.emit(HumanInputEvent.START_OF_SPEECH, ev);\n                break;\n              case VADEventType.INFERENCE_DONE:\n                this.#speechProbability = ev.probability;\n                this.emit(HumanInputEvent.VAD_INFERENCE_DONE, ev);\n                break;\n              case VADEventType.END_OF_SPEECH:\n                this.#speaking = false;\n                this.emit(HumanInputEvent.END_OF_SPEECH, ev);\n                break;\n            }\n          }\n        };\n\n        const sttStreamCo = async () => {\n          for await (const ev of sttStream) {\n            if (cancelled) return;\n            if (ev.type === SpeechEventType.FINAL_TRANSCRIPT) {\n              this.emit(HumanInputEvent.FINAL_TRANSCRIPT, ev);\n            } else if (ev.type == SpeechEventType.INTERIM_TRANSCRIPT) {\n              this.emit(HumanInputEvent.INTERIM_TRANSCRIPT, ev);\n            }\n          }\n        };\n\n        await Promise.all([audioStreamCo(), vadStreamCo(), sttStreamCo()]);\n        sttStream.close();\n        vadStream.close();\n        resolve();\n      });\n    }\n  }\n\n  get speaking(): boolean {\n    return this.#speaking;\n  }\n\n  get speakingProbability(): number {\n    return this.#speechProbability;\n  }\n\n  async close() {\n    if (this.#closed) {\n      throw new Error('HumanInput already closed');\n    }\n    this.#closed = true;\n    this.#room.removeAllListeners();\n    this.#speaking = false;\n    if (this.#recognizeTask) {\n      await gracefullyCancel(this.#recognizeTask);\n    }\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAUA,sBAAoD;AAEpD,yBAA6B;AAC7B,iBAAoB;AAEpB,iBAAgC;AAChC,mBAAqD;AAErD,iBAA6B;AAEtB,IAAK,kBAAL,kBAAKA,qBAAL;AACL,EAAAA,kCAAA;AACA,EAAAA,kCAAA;AACA,EAAAA,kCAAA;AACA,EAAAA,kCAAA;AACA,EAAAA,kCAAA;AALU,SAAAA;AAAA,GAAA;AAgBL,MAAM,mBAAoB,gCAA6D;AAAA,EAC5F,UAAU;AAAA,EACV;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,YAAY;AAAA,EACZ,qBAAqB;AAAA,EACrB,cAAU,gBAAI;AAAA,EACd;AAAA,EAEA,YACE,MACA,KACA,KACA,aACA,mBACA;AACA,UAAM;AACN,SAAK,QAAQ;AACb,SAAK,OAAO;AACZ,SAAK,OAAO;AACZ,SAAK,eAAe;AACpB,SAAK,qBAAqB;AAE1B,SAAK,MAAM,GAAG,0BAAU,gBAAgB,KAAK,uBAAuB,KAAK,IAAI,CAAC;AAC9E,SAAK,MAAM,GAAG,0BAAU,iBAAiB,KAAK,uBAAuB,KAAK,IAAI,CAAC;AAC/E,SAAK,uBAAuB;AAAA,EAC9B;AAAA,EAEA,IAAI,cAAiC;AACnC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,kBAAgD;AAClD,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,yBAA+B;AAC7B,QAAI,CAAC,KAAK,cAAc;AACtB,WAAK,QAAQ,MAAM,wBAAwB;AAC3C;AAAA,IACF;AAEA,QAAI,wBAA4D;AAChE,eAAW,eAAe,KAAK,aAAa,kBAAkB,OAAO,GAAG;AACtE,UAAI,YAAY,WAAW,4BAAY,mBAAmB;AACxD,gCAAwB;AACxB;AAAA,MACF;AAAA,IACF;AACA,QAAI,CAAC,uBAAuB;AAC1B;AAAA,IACF;AAEA,QAAI,CAAC,sBAAsB,YAAY;AACrC,4BAAsB,cAAc,IAAI;AAAA,IAC1C;AAEA,UAAM,QAAQ,sBAAsB;AACpC,QAAI,SAAS,UAAU,KAAK,kBAAkB;AAC5C,WAAK,mBAAmB;AACxB,UAAI,KAAK,gBAAgB;AACvB,aAAK,eAAe,OAAO;AAAA,MAC7B;AAEA,YAAM,qBAAqB;AAAA,QACzB,YAAY;AAAA,QACZ,aAAa;AAAA,QACb,GAAI,KAAK,qBAAqB,EAAE,mBAAmB,KAAK,mBAAmB,IAAI,CAAC;AAAA,MAClF;AACA,YAAM,cAAc,IAAI,4BAAY,OAAO,kBAAkB;AAG7D,WAAK,iBAAiB,IAAI,gCAAmB,OAAO,SAAS,GAAG,aAAa;AAC3E,YAAI,YAAY;AAChB,iBAAS,MAAM;AACb,sBAAY;AAAA,QACd,CAAC;AAED,cAAM,YAAY,KAAK,KAAK,OAAO;AACnC,cAAM,YAAY,KAAK,KAAK,OAAO;AAEnC,cAAM,gBAAgB,YAAY;AAChC,2BAAiB,MAAM,aAAa;AAClC,gBAAI,UAAW;AACf,sBAAU,UAAU,EAAE;AACtB,sBAAU,UAAU,EAAE;AAAA,UACxB;AAAA,QACF;AAEA,cAAM,cAAc,YAAY;AAC9B,2BAAiB,MAAM,WAAW;AAChC,gBAAI,UAAW;AACf,oBAAQ,GAAG,MAAM;AAAA,cACf,KAAK,wBAAa;AAChB,qBAAK,YAAY;AACjB,qBAAK,KAAK,yBAAiC,EAAE;AAC7C;AAAA,cACF,KAAK,wBAAa;AAChB,qBAAK,qBAAqB,GAAG;AAC7B,qBAAK,KAAK,4BAAoC,EAAE;AAChD;AAAA,cACF,KAAK,wBAAa;AAChB,qBAAK,YAAY;AACjB,qBAAK,KAAK,uBAA+B,EAAE;AAC3C;AAAA,YACJ;AAAA,UACF;AAAA,QACF;AAEA,cAAM,cAAc,YAAY;AAC9B,2BAAiB,MAAM,WAAW;AAChC,gBAAI,UAAW;AACf,gBAAI,GAAG,SAAS,2BAAgB,kBAAkB;AAChD,mBAAK,KAAK,0BAAkC,EAAE;AAAA,YAChD,WAAW,GAAG,QAAQ,2BAAgB,oBAAoB;AACxD,mBAAK,KAAK,4BAAoC,EAAE;AAAA,YAClD;AAAA,UACF;AAAA,QACF;AAEA,cAAM,QAAQ,IAAI,CAAC,cAAc,GAAG,YAAY,GAAG,YAAY,CAAC,CAAC;AACjE,kBAAU,MAAM;AAChB,kBAAU,MAAM;AAChB,gBAAQ;AAAA,MACV,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,IAAI,WAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,sBAA8B;AAChC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,QAAQ;AACZ,QAAI,KAAK,SAAS;AAChB,YAAM,IAAI,MAAM,2BAA2B;AAAA,IAC7C;AACA,SAAK,UAAU;AACf,SAAK,MAAM,mBAAmB;AAC9B,SAAK,YAAY;AACjB,QAAI,KAAK,gBAAgB;AACvB,gBAAM,+BAAiB,KAAK,cAAc;AAAA,IAC5C;AAAA,EACF;AACF;","names":["HumanInputEvent"]}

package/dist/pipeline/human_input.d.cts

@@ -1,30 +0,0 @@
-import type { NoiseCancellationOptions, RemoteAudioTrack, RemoteParticipant, Room } from '@livekit/rtc-node';
-import type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';
-import type { STT, SpeechEvent } from '../stt/stt.js';
-import type { VAD, VADEvent } from '../vad.js';
-export declare enum HumanInputEvent {
-    START_OF_SPEECH = 0,
-    VAD_INFERENCE_DONE = 1,
-    END_OF_SPEECH = 2,
-    FINAL_TRANSCRIPT = 3,
-    INTERIM_TRANSCRIPT = 4
-}
-export type HumanInputCallbacks = {
-    [HumanInputEvent.START_OF_SPEECH]: (event: VADEvent) => void;
-    [HumanInputEvent.VAD_INFERENCE_DONE]: (event: VADEvent) => void;
-    [HumanInputEvent.END_OF_SPEECH]: (event: VADEvent) => void;
-    [HumanInputEvent.FINAL_TRANSCRIPT]: (event: SpeechEvent) => void;
-    [HumanInputEvent.INTERIM_TRANSCRIPT]: (event: SpeechEvent) => void;
-};
-declare const HumanInput_base: new () => TypedEmitter<HumanInputCallbacks>;
-export declare class HumanInput extends HumanInput_base {
-    #private;
-    constructor(room: Room, vad: VAD, stt: STT, participant: RemoteParticipant, noiseCancellation?: NoiseCancellationOptions);
-    get participant(): RemoteParticipant;
-    get subscribedTrack(): RemoteAudioTrack | undefined;
-    get speaking(): boolean;
-    get speakingProbability(): number;
-    close(): Promise<void>;
-}
-export {};
-//# sourceMappingURL=human_input.d.ts.map

package/dist/pipeline/human_input.d.ts

@@ -1,30 +0,0 @@
-import type { NoiseCancellationOptions, RemoteAudioTrack, RemoteParticipant, Room } from '@livekit/rtc-node';
-import type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';
-import type { STT, SpeechEvent } from '../stt/stt.js';
-import type { VAD, VADEvent } from '../vad.js';
-export declare enum HumanInputEvent {
-    START_OF_SPEECH = 0,
-    VAD_INFERENCE_DONE = 1,
-    END_OF_SPEECH = 2,
-    FINAL_TRANSCRIPT = 3,
-    INTERIM_TRANSCRIPT = 4
-}
-export type HumanInputCallbacks = {
-    [HumanInputEvent.START_OF_SPEECH]: (event: VADEvent) => void;
-    [HumanInputEvent.VAD_INFERENCE_DONE]: (event: VADEvent) => void;
-    [HumanInputEvent.END_OF_SPEECH]: (event: VADEvent) => void;
-    [HumanInputEvent.FINAL_TRANSCRIPT]: (event: SpeechEvent) => void;
-    [HumanInputEvent.INTERIM_TRANSCRIPT]: (event: SpeechEvent) => void;
-};
-declare const HumanInput_base: new () => TypedEmitter<HumanInputCallbacks>;
-export declare class HumanInput extends HumanInput_base {
-    #private;
-    constructor(room: Room, vad: VAD, stt: STT, participant: RemoteParticipant, noiseCancellation?: NoiseCancellationOptions);
-    get participant(): RemoteParticipant;
-    get subscribedTrack(): RemoteAudioTrack | undefined;
-    get speaking(): boolean;
-    get speakingProbability(): number;
-    close(): Promise<void>;
-}
-export {};
-//# sourceMappingURL=human_input.d.ts.map

package/dist/pipeline/human_input.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"human_input.d.ts","sourceRoot":"","sources":["../../src/pipeline/human_input.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EACV,wBAAwB,EACxB,gBAAgB,EAChB,iBAAiB,EAEjB,IAAI,EACL,MAAM,mBAAmB,CAAC;AAE3B,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAGhF,OAAO,KAAK,EAAE,GAAG,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC;AAGtD,OAAO,KAAK,EAAE,GAAG,EAAE,QAAQ,EAAE,MAAM,WAAW,CAAC;AAG/C,oBAAY,eAAe;IACzB,eAAe,IAAA;IACf,kBAAkB,IAAA;IAClB,aAAa,IAAA;IACb,gBAAgB,IAAA;IAChB,kBAAkB,IAAA;CACnB;AAED,MAAM,MAAM,mBAAmB,GAAG;IAChC,CAAC,eAAe,CAAC,eAAe,CAAC,EAAE,CAAC,KAAK,EAAE,QAAQ,KAAK,IAAI,CAAC;IAC7D,CAAC,eAAe,CAAC,kBAAkB,CAAC,EAAE,CAAC,KAAK,EAAE,QAAQ,KAAK,IAAI,CAAC;IAChE,CAAC,eAAe,CAAC,aAAa,CAAC,EAAE,CAAC,KAAK,EAAE,QAAQ,KAAK,IAAI,CAAC;IAC3D,CAAC,eAAe,CAAC,gBAAgB,CAAC,EAAE,CAAC,KAAK,EAAE,WAAW,KAAK,IAAI,CAAC;IACjE,CAAC,eAAe,CAAC,kBAAkB,CAAC,EAAE,CAAC,KAAK,EAAE,WAAW,KAAK,IAAI,CAAC;CACpE,CAAC;yCAEyD,aAAa,mBAAmB,CAAC;AAA5F,qBAAa,UAAW,SAAQ,eAA6D;;gBAczF,IAAI,EAAE,IAAI,EACV,GAAG,EAAE,GAAG,EACR,GAAG,EAAE,GAAG,EACR,WAAW,EAAE,iBAAiB,EAC9B,iBAAiB,CAAC,EAAE,wBAAwB;IAc9C,IAAI,WAAW,IAAI,iBAAiB,CAEnC;IAED,IAAI,eAAe,IAAI,gBAAgB,GAAG,SAAS,CAElD;IA8FD,IAAI,QAAQ,IAAI,OAAO,CAEtB;IAED,IAAI,mBAAmB,IAAI,MAAM,CAEhC;IAEK,KAAK;CAWZ"}

package/dist/pipeline/human_input.js

@@ -1,146 +0,0 @@
-import { AudioStream, RoomEvent, TrackSource } from "@livekit/rtc-node";
-import { EventEmitter } from "node:events";
-import { log } from "../log.js";
-import { SpeechEventType } from "../stt/stt.js";
-import { CancellablePromise, gracefullyCancel } from "../utils.js";
-import { VADEventType } from "../vad.js";
-var HumanInputEvent = /* @__PURE__ */ ((HumanInputEvent2) => {
-  HumanInputEvent2[HumanInputEvent2["START_OF_SPEECH"] = 0] = "START_OF_SPEECH";
-  HumanInputEvent2[HumanInputEvent2["VAD_INFERENCE_DONE"] = 1] = "VAD_INFERENCE_DONE";
-  HumanInputEvent2[HumanInputEvent2["END_OF_SPEECH"] = 2] = "END_OF_SPEECH";
-  HumanInputEvent2[HumanInputEvent2["FINAL_TRANSCRIPT"] = 3] = "FINAL_TRANSCRIPT";
-  HumanInputEvent2[HumanInputEvent2["INTERIM_TRANSCRIPT"] = 4] = "INTERIM_TRANSCRIPT";
-  return HumanInputEvent2;
-})(HumanInputEvent || {});
-class HumanInput extends EventEmitter {
-  #closed = false;
-  #room;
-  #vad;
-  #stt;
-  #participant;
-  #subscribedTrack;
-  #recognizeTask;
-  #speaking = false;
-  #speechProbability = 0;
-  #logger = log();
-  #noiseCancellation;
-  constructor(room, vad, stt, participant, noiseCancellation) {
-    super();
-    this.#room = room;
-    this.#vad = vad;
-    this.#stt = stt;
-    this.#participant = participant;
-    this.#noiseCancellation = noiseCancellation;
-    this.#room.on(RoomEvent.TrackPublished, this.#subscribeToMicrophone.bind(this));
-    this.#room.on(RoomEvent.TrackSubscribed, this.#subscribeToMicrophone.bind(this));
-    this.#subscribeToMicrophone();
-  }
-  get participant() {
-    return this.#participant;
-  }
-  get subscribedTrack() {
-    return this.#subscribedTrack;
-  }
-  #subscribeToMicrophone() {
-    if (!this.#participant) {
-      this.#logger.error("Participant is not set");
-      return;
-    }
-    let microphonePublication = void 0;
-    for (const publication of this.#participant.trackPublications.values()) {
-      if (publication.source === TrackSource.SOURCE_MICROPHONE) {
-        microphonePublication = publication;
-        break;
-      }
-    }
-    if (!microphonePublication) {
-      return;
-    }
-    if (!microphonePublication.subscribed) {
-      microphonePublication.setSubscribed(true);
-    }
-    const track = microphonePublication.track;
-    if (track && track !== this.#subscribedTrack) {
-      this.#subscribedTrack = track;
-      if (this.#recognizeTask) {
-        this.#recognizeTask.cancel();
-      }
-      const audioStreamOptions = {
-        sampleRate: 16e3,
-        numChannels: 1,
-        ...this.#noiseCancellation ? { noiseCancellation: this.#noiseCancellation } : {}
-      };
-      const audioStream = new AudioStream(track, audioStreamOptions);
-      this.#recognizeTask = new CancellablePromise(async (resolve, _, onCancel) => {
-        let cancelled = false;
-        onCancel(() => {
-          cancelled = true;
-        });
-        const sttStream = this.#stt.stream();
-        const vadStream = this.#vad.stream();
-        const audioStreamCo = async () => {
-          for await (const ev of audioStream) {
-            if (cancelled) return;
-            sttStream.pushFrame(ev);
-            vadStream.pushFrame(ev);
-          }
-        };
-        const vadStreamCo = async () => {
-          for await (const ev of vadStream) {
-            if (cancelled) return;
-            switch (ev.type) {
-              case VADEventType.START_OF_SPEECH:
-                this.#speaking = true;
-                this.emit(0 /* START_OF_SPEECH */, ev);
-                break;
-              case VADEventType.INFERENCE_DONE:
-                this.#speechProbability = ev.probability;
-                this.emit(1 /* VAD_INFERENCE_DONE */, ev);
-                break;
-              case VADEventType.END_OF_SPEECH:
-                this.#speaking = false;
-                this.emit(2 /* END_OF_SPEECH */, ev);
-                break;
-            }
-          }
-        };
-        const sttStreamCo = async () => {
-          for await (const ev of sttStream) {
-            if (cancelled) return;
-            if (ev.type === SpeechEventType.FINAL_TRANSCRIPT) {
-              this.emit(3 /* FINAL_TRANSCRIPT */, ev);
-            } else if (ev.type == SpeechEventType.INTERIM_TRANSCRIPT) {
-              this.emit(4 /* INTERIM_TRANSCRIPT */, ev);
-            }
-          }
-        };
-        await Promise.all([audioStreamCo(), vadStreamCo(), sttStreamCo()]);
-        sttStream.close();
-        vadStream.close();
-        resolve();
-      });
-    }
-  }
-  get speaking() {
-    return this.#speaking;
-  }
-  get speakingProbability() {
-    return this.#speechProbability;
-  }
-  async close() {
-    if (this.#closed) {
-      throw new Error("HumanInput already closed");
-    }
-    this.#closed = true;
-    this.#room.removeAllListeners();
-    this.#speaking = false;
-    if (this.#recognizeTask) {
-      await gracefullyCancel(this.#recognizeTask);
-    }
-  }
-}
-export {
-  HumanInput,
-  HumanInputEvent
-};
-//# sourceMappingURL=human_input.js.map