@livekit/agents 1.0.43 → 1.0.45

package/dist/vad.js

@@ -32,13 +32,14 @@ class VADStream {
   outputReader;
   closed = false;
   inputClosed = false;
-  #vad;
-  #lastActivityTime = BigInt(0);
-  logger = log();
+  vad;
+  lastActivityTime = BigInt(0);
+  logger;
   deferredInputStream;
   metricsStream;
   constructor(vad) {
-    this.#vad = vad;
+    this.logger = log();
+    this.vad = vad;
     this.deferredInputStream = new DeferredReadableStream();
     this.inputWriter = this.input.writable.getWriter();
     this.inputReader = this.input.readable.getReader();
@@ -83,16 +84,16 @@ class VADStream {
       switch (value.type) {
         case 0 /* START_OF_SPEECH */:
           inferenceCount++;
-          if (inferenceCount >= 1e3 / this.#vad.capabilities.updateInterval) {
-            this.#vad.emit("metrics_collected", {
+          if (inferenceCount >= 1e3 / this.vad.capabilities.updateInterval) {
+            this.vad.emit("metrics_collected", {
               type: "vad_metrics",
               timestamp: Date.now(),
               idleTimeMs: Math.trunc(
-                Number((process.hrtime.bigint() - this.#lastActivityTime) / BigInt(1e6))
+                Number((process.hrtime.bigint() - this.lastActivityTime) / BigInt(1e6))
               ),
               inferenceDurationTotalMs,
               inferenceCount,
-              label: this.#vad.label
+              label: this.vad.label
             });
             inferenceCount = 0;
             inferenceDurationTotalMs = 0;
@@ -100,10 +101,10 @@ class VADStream {
           break;
         case 1 /* INFERENCE_DONE */:
           inferenceDurationTotalMs += Math.round(value.inferenceDuration);
-          this.#lastActivityTime = process.hrtime.bigint();
+          this.lastActivityTime = process.hrtime.bigint();
           break;
         case 2 /* END_OF_SPEECH */:
-          this.#lastActivityTime = process.hrtime.bigint();
+          this.lastActivityTime = process.hrtime.bigint();
           break;
       }
     }

package/dist/vad.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../src/vad.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type {\n  ReadableStream,\n  ReadableStreamDefaultReader,\n  WritableStreamDefaultWriter,\n} from 'node:stream/web';\nimport { log } from './log.js';\nimport type { VADMetrics } from './metrics/base.js';\nimport { DeferredReadableStream } from './stream/deferred_stream.js';\nimport { IdentityTransform } from './stream/identity_transform.js';\n\nexport enum VADEventType {\n  START_OF_SPEECH,\n  INFERENCE_DONE,\n  END_OF_SPEECH,\n  METRICS_COLLECTED,\n}\n\nexport interface VADEvent {\n  /** Type of the VAD event (e.g., start of speech, end of speech, inference done). */\n  type: VADEventType;\n  /**\n   * Index of the audio sample where the event occurred, relative to the inference sample rate.\n   */\n  samplesIndex: number;\n  /** Timestamp when the event was fired. */\n  timestamp: number;\n  /** Duration of the speech segment in seconds. */\n  speechDuration: number;\n  /** Duration of the silence segment in seconds. */\n  silenceDuration: number;\n  /**\n   * List of audio frames associated with the speech.\n   *\n   * @remarks\n   * - For `start_of_speech` events, this contains the audio chunks that triggered the detection.\n   * - For `inference_done` events, this contains the audio chunks that were processed.\n   * - For `end_of_speech` events, this contains the complete user speech.\n   */\n  frames: AudioFrame[];\n  /** Probability that speech is present (only for `INFERENCE_DONE` events). */\n  probability: number;\n  /** Time taken to perform the inference, in seconds (only for `INFERENCE_DONE` events). */\n  inferenceDuration: number;\n  /** Indicates whether speech was detected in the frames. */\n  speaking: boolean;\n  /** Threshold used to detect silence. */\n  rawAccumulatedSilence: number;\n  /** Threshold used to detect speech. */\n  rawAccumulatedSpeech: number;\n}\n\nexport interface VADCapabilities {\n  /** Duration of each VAD inference window in milliseconds. Used to batch metrics emissions to roughly once per second. */\n  updateInterval: number;\n}\n\nexport type VADCallbacks = {\n  ['metrics_collected']: (metrics: VADMetrics) => void;\n};\n\nexport abstract class VAD extends (EventEmitter as new () => TypedEmitter<VADCallbacks>) {\n  #capabilities: VADCapabilities;\n  abstract label: string;\n\n  constructor(capabilities: VADCapabilities) {\n    super();\n    this.#capabilities = capabilities;\n  }\n\n  get capabilities(): VADCapabilities {\n    return this.#capabilities;\n  }\n\n  /**\n   * Returns a {@link VADStream} that can be used to push audio frames and receive VAD events.\n   */\n  abstract stream(): VADStream;\n\n  async close(): Promise<void> {\n    return;\n  }\n}\n\nexport abstract class VADStream implements AsyncIterableIterator<VADEvent> {\n  protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n  protected input = new IdentityTransform<AudioFrame | typeof VADStream.FLUSH_SENTINEL>();\n  protected output = new IdentityTransform<VADEvent>();\n  protected inputWriter: WritableStreamDefaultWriter<AudioFrame | typeof VADStream.FLUSH_SENTINEL>;\n  protected inputReader: ReadableStreamDefaultReader<AudioFrame | typeof VADStream.FLUSH_SENTINEL>;\n  protected outputWriter: WritableStreamDefaultWriter<VADEvent>;\n  protected outputReader: ReadableStreamDefaultReader<VADEvent>;\n  protected closed = false;\n  protected inputClosed = false;\n\n  #vad: VAD;\n  #lastActivityTime = BigInt(0);\n  private logger = log();\n  private deferredInputStream: DeferredReadableStream<AudioFrame>;\n\n  private metricsStream: ReadableStream<VADEvent>;\n  constructor(vad: VAD) {\n    this.#vad = vad;\n    this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n\n    this.inputWriter = this.input.writable.getWriter();\n    this.inputReader = this.input.readable.getReader();\n    this.outputWriter = this.output.writable.getWriter();\n\n    const [outputStream, metricsStream] = this.output.readable.tee();\n    this.metricsStream = metricsStream;\n    this.outputReader = outputStream.getReader();\n\n    this.pumpDeferredStream();\n    this.monitorMetrics();\n  }\n\n  /**\n   * Reads from the deferred input stream and forwards chunks to the input writer.\n   *\n   * Note: we can't just do this.deferredInputStream.stream.pipeTo(this.input.writable)\n   * because the inputWriter locks the this.input.writable stream. All writes must go through\n   * the inputWriter.\n   */\n  private async pumpDeferredStream() {\n    const reader = this.deferredInputStream.stream.getReader();\n    try {\n      while (true) {\n        const { done, value } = await reader.read();\n        if (done) break;\n        await this.inputWriter.write(value);\n      }\n    } catch (e) {\n      this.logger.error(`Error pumping deferred stream: ${e}`);\n      throw e;\n    } finally {\n      reader.releaseLock();\n    }\n  }\n\n  protected async monitorMetrics() {\n    let inferenceDurationTotalMs = 0;\n    let inferenceCount = 0;\n    const metricsReader = this.metricsStream.getReader();\n    while (true) {\n      const { done, value } = await metricsReader.read();\n      if (done) {\n        break;\n      }\n      switch (value.type) {\n        case VADEventType.START_OF_SPEECH:\n          inferenceCount++;\n          if (inferenceCount >= 1000 / this.#vad.capabilities.updateInterval) {\n            this.#vad.emit('metrics_collected', {\n              type: 'vad_metrics',\n              timestamp: Date.now(),\n              idleTimeMs: Math.trunc(\n                Number((process.hrtime.bigint() - this.#lastActivityTime) / BigInt(1000000)),\n              ),\n              inferenceDurationTotalMs,\n              inferenceCount,\n              label: this.#vad.label,\n            });\n\n            inferenceCount = 0;\n            inferenceDurationTotalMs = 0;\n          }\n          break;\n        case VADEventType.INFERENCE_DONE:\n          inferenceDurationTotalMs += Math.round(value.inferenceDuration);\n          this.#lastActivityTime = process.hrtime.bigint();\n          break;\n        case VADEventType.END_OF_SPEECH:\n          this.#lastActivityTime = process.hrtime.bigint();\n          break;\n      }\n    }\n  }\n\n  /**\n   * Safely send a VAD event to the output stream, handling writer release errors during shutdown.\n   * @returns true if the event was sent, false if the stream is closing\n   * @throws Error if an unexpected error occurs\n   */\n  protected sendVADEvent(event: VADEvent): boolean {\n    if (this.closed) {\n      return false;\n    }\n\n    try {\n      this.outputWriter.write(event);\n      return true;\n    } catch (e) {\n      throw e;\n    }\n  }\n\n  updateInputStream(audioStream: ReadableStream<AudioFrame>) {\n    this.deferredInputStream.setSource(audioStream);\n  }\n\n  detachInputStream() {\n    this.deferredInputStream.detachSource();\n  }\n\n  /** @deprecated Use `updateInputStream` instead */\n  pushFrame(frame: AudioFrame) {\n    // TODO(AJS-395): remove this method\n    if (this.inputClosed) {\n      throw new Error('Input is closed');\n    }\n    if (this.closed) {\n      throw new Error('Stream is closed');\n    }\n    this.inputWriter.write(frame);\n  }\n\n  flush() {\n    if (this.inputClosed) {\n      throw new Error('Input is closed');\n    }\n    if (this.closed) {\n      throw new Error('Stream is closed');\n    }\n    this.inputWriter.write(VADStream.FLUSH_SENTINEL);\n  }\n\n  endInput() {\n    if (this.inputClosed) {\n      throw new Error('Input is closed');\n    }\n    if (this.closed) {\n      throw new Error('Stream is closed');\n    }\n    this.inputClosed = true;\n    this.input.writable.close();\n  }\n\n  async next(): Promise<IteratorResult<VADEvent>> {\n    return this.outputReader.read().then(({ done, value }) => {\n      if (done) {\n        return { done: true, value: undefined };\n      }\n      return { done: false, value };\n    });\n  }\n\n  close() {\n    this.outputWriter.releaseLock();\n    this.outputReader.cancel();\n    this.output.writable.close();\n    this.closed = true;\n  }\n\n  [Symbol.asyncIterator](): VADStream {\n    return this;\n  }\n}\n"],"mappings":"AAKA,SAAS,oBAAoB;AAM7B,SAAS,WAAW;AAEpB,SAAS,8BAA8B;AACvC,SAAS,yBAAyB;AAE3B,IAAK,eAAL,kBAAKA,kBAAL;AACL,EAAAA,4BAAA;AACA,EAAAA,4BAAA;AACA,EAAAA,4BAAA;AACA,EAAAA,4BAAA;AAJU,SAAAA;AAAA,GAAA;AAkDL,MAAe,YAAa,aAAsD;AAAA,EACvF;AAAA,EAGA,YAAY,cAA+B;AACzC,UAAM;AACN,SAAK,gBAAgB;AAAA,EACvB;AAAA,EAEA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA,EAOA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAEO,MAAe,UAAqD;AAAA,EACzE,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EACxD,QAAQ,IAAI,kBAAgE;AAAA,EAC5E,SAAS,IAAI,kBAA4B;AAAA,EACzC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,SAAS;AAAA,EACT,cAAc;AAAA,EAExB;AAAA,EACA,oBAAoB,OAAO,CAAC;AAAA,EACpB,SAAS,IAAI;AAAA,EACb;AAAA,EAEA;AAAA,EACR,YAAY,KAAU;AACpB,SAAK,OAAO;AACZ,SAAK,sBAAsB,IAAI,uBAAmC;AAElE,SAAK,cAAc,KAAK,MAAM,SAAS,UAAU;AACjD,SAAK,cAAc,KAAK,MAAM,SAAS,UAAU;AACjD,SAAK,eAAe,KAAK,OAAO,SAAS,UAAU;AAEnD,UAAM,CAAC,cAAc,aAAa,IAAI,KAAK,OAAO,SAAS,IAAI;AAC/D,SAAK,gBAAgB;AACrB,SAAK,eAAe,aAAa,UAAU;AAE3C,SAAK,mBAAmB;AACxB,SAAK,eAAe;AAAA,EACtB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,MAAc,qBAAqB;AACjC,UAAM,SAAS,KAAK,oBAAoB,OAAO,UAAU;AACzD,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,KAAM;AACV,cAAM,KAAK,YAAY,MAAM,KAAK;AAAA,MACpC;AAAA,IACF,SAAS,GAAG;AACV,WAAK,OAAO,MAAM,kCAAkC,CAAC,EAAE;AACvD,YAAM;AAAA,IACR,UAAE;AACA,aAAO,YAAY;AAAA,IACrB;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,QAAI,2BAA2B;AAC/B,QAAI,iBAAiB;AACrB,UAAM,gBAAgB,KAAK,cAAc,UAAU;AACnD,WAAO,MAAM;AACX,YAAM,EAAE,MAAM,MAAM,IAAI,MAAM,cAAc,KAAK;AACjD,UAAI,MAAM;AACR;AAAA,MACF;AACA,cAAQ,MAAM,MAAM;AAAA,QAClB,KAAK;AACH;AACA,cAAI,kBAAkB,MAAO,KAAK,KAAK,aAAa,gBAAgB;AAClE,iBAAK,KAAK,KAAK,qBAAqB;AAAA,cAClC,MAAM;AAAA,cACN,WAAW,KAAK,IAAI;AAAA,cACpB,YAAY,KAAK;AAAA,gBACf,QAAQ,QAAQ,OAAO,OAAO,IAAI,KAAK,qBAAqB,OAAO,GAAO,CAAC;AAAA,cAC7E;AAAA,cACA;AAAA,cACA;AAAA,cACA,OAAO,KAAK,KAAK;AAAA,YACnB,CAAC;AAED,6BAAiB;AACjB,uCAA2B;AAAA,UAC7B;AACA;AAAA,QACF,KAAK;AACH,sCAA4B,KAAK,MAAM,MAAM,iBAAiB;AAC9D,eAAK,oBAAoB,QAAQ,OAAO,OAAO;AAC/C;AAAA,QACF,KAAK;AACH,eAAK,oBAAoB,QAAQ,OAAO,OAAO;AAC/C;AAAA,MACJ;AAAA,IACF;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOU,aAAa,OAA0B;AAC/C,QAAI,KAAK,QAAQ;AACf,aAAO;AAAA,IACT;AAEA,QAAI;AACF,WAAK,aAAa,MAAM,KAAK;AAC7B,aAAO;AAAA,IACT,SAAS,GAAG;AACV,YAAM;AAAA,IACR;AAAA,EACF;AAAA,EAEA,kBAAkB,aAAyC;AACzD,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,oBAAoB;AAClB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA;AAAA,EAGA,UAAU,OAAmB;AAE3B,QAAI,KAAK,aAAa;AACpB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,YAAY,MAAM,KAAK;AAAA,EAC9B;AAAA,EAEA,QAAQ;AACN,QAAI,KAAK,aAAa;AACpB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,YAAY,MAAM,UAAU,cAAc;AAAA,EACjD;AAAA,EAEA,WAAW;AACT,QAAI,KAAK,aAAa;AACpB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,cAAc;AACnB,SAAK,MAAM,SAAS,MAAM;AAAA,EAC5B;AAAA,EAEA,MAAM,OAA0C;AAC9C,WAAO,KAAK,aAAa,KAAK,EAAE,KAAK,CAAC,EAAE,MAAM,MAAM,MAAM;AACxD,UAAI,MAAM;AACR,eAAO,EAAE,MAAM,MAAM,OAAO,OAAU;AAAA,MACxC;AACA,aAAO,EAAE,MAAM,OAAO,MAAM;AAAA,IAC9B,CAAC;AAAA,EACH;AAAA,EAEA,QAAQ;AACN,SAAK,aAAa,YAAY;AAC9B,SAAK,aAAa,OAAO;AACzB,SAAK,OAAO,SAAS,MAAM;AAC3B,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAe;AAClC,WAAO;AAAA,EACT;AACF;","names":["VADEventType"]}
+{"version":3,"sources":["../src/vad.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport { EventEmitter } from 'node:events';\nimport type {\n  ReadableStream,\n  ReadableStreamDefaultReader,\n  WritableStreamDefaultWriter,\n} from 'node:stream/web';\nimport { log } from './log.js';\nimport type { VADMetrics } from './metrics/base.js';\nimport { DeferredReadableStream } from './stream/deferred_stream.js';\nimport { IdentityTransform } from './stream/identity_transform.js';\n\nexport enum VADEventType {\n  START_OF_SPEECH,\n  INFERENCE_DONE,\n  END_OF_SPEECH,\n  METRICS_COLLECTED,\n}\n\nexport interface VADEvent {\n  /** Type of the VAD event (e.g., start of speech, end of speech, inference done). */\n  type: VADEventType;\n  /**\n   * Index of the audio sample where the event occurred, relative to the inference sample rate.\n   */\n  samplesIndex: number;\n  /** Timestamp when the event was fired. */\n  timestamp: number;\n  /** Duration of the speech segment in seconds. */\n  speechDuration: number;\n  /** Duration of the silence segment in seconds. */\n  silenceDuration: number;\n  /**\n   * List of audio frames associated with the speech.\n   *\n   * @remarks\n   * - For `start_of_speech` events, this contains the audio chunks that triggered the detection.\n   * - For `inference_done` events, this contains the audio chunks that were processed.\n   * - For `end_of_speech` events, this contains the complete user speech.\n   */\n  frames: AudioFrame[];\n  /** Probability that speech is present (only for `INFERENCE_DONE` events). */\n  probability: number;\n  /** Time taken to perform the inference, in seconds (only for `INFERENCE_DONE` events). */\n  inferenceDuration: number;\n  /** Indicates whether speech was detected in the frames. */\n  speaking: boolean;\n  /** Threshold used to detect silence. */\n  rawAccumulatedSilence: number;\n  /** Threshold used to detect speech. */\n  rawAccumulatedSpeech: number;\n}\n\nexport interface VADCapabilities {\n  /** Duration of each VAD inference window in milliseconds. Used to batch metrics emissions to roughly once per second. */\n  updateInterval: number;\n}\n\nexport type VADCallbacks = {\n  ['metrics_collected']: (metrics: VADMetrics) => void;\n};\n\nexport abstract class VAD extends (EventEmitter as new () => TypedEmitter<VADCallbacks>) {\n  #capabilities: VADCapabilities;\n  abstract label: string;\n\n  constructor(capabilities: VADCapabilities) {\n    super();\n    this.#capabilities = capabilities;\n  }\n\n  get capabilities(): VADCapabilities {\n    return this.#capabilities;\n  }\n\n  /**\n   * Returns a {@link VADStream} that can be used to push audio frames and receive VAD events.\n   */\n  abstract stream(): VADStream;\n\n  async close(): Promise<void> {\n    return;\n  }\n}\n\nexport abstract class VADStream implements AsyncIterableIterator<VADEvent> {\n  protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n  protected input = new IdentityTransform<AudioFrame | typeof VADStream.FLUSH_SENTINEL>();\n  protected output = new IdentityTransform<VADEvent>();\n  protected inputWriter: WritableStreamDefaultWriter<AudioFrame | typeof VADStream.FLUSH_SENTINEL>;\n  protected inputReader: ReadableStreamDefaultReader<AudioFrame | typeof VADStream.FLUSH_SENTINEL>;\n  protected outputWriter: WritableStreamDefaultWriter<VADEvent>;\n  protected outputReader: ReadableStreamDefaultReader<VADEvent>;\n  protected closed = false;\n  protected inputClosed = false;\n\n  protected vad: VAD;\n  protected lastActivityTime = BigInt(0);\n  protected logger;\n  protected deferredInputStream: DeferredReadableStream<AudioFrame>;\n\n  private metricsStream: ReadableStream<VADEvent>;\n  constructor(vad: VAD) {\n    this.logger = log();\n    this.vad = vad;\n    this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n\n    this.inputWriter = this.input.writable.getWriter();\n    this.inputReader = this.input.readable.getReader();\n    this.outputWriter = this.output.writable.getWriter();\n\n    const [outputStream, metricsStream] = this.output.readable.tee();\n    this.metricsStream = metricsStream;\n    this.outputReader = outputStream.getReader();\n\n    this.pumpDeferredStream();\n    this.monitorMetrics();\n  }\n\n  /**\n   * Reads from the deferred input stream and forwards chunks to the input writer.\n   *\n   * Note: we can't just do this.deferredInputStream.stream.pipeTo(this.input.writable)\n   * because the inputWriter locks the this.input.writable stream. All writes must go through\n   * the inputWriter.\n   */\n  private async pumpDeferredStream() {\n    const reader = this.deferredInputStream.stream.getReader();\n    try {\n      while (true) {\n        const { done, value } = await reader.read();\n        if (done) break;\n        await this.inputWriter.write(value);\n      }\n    } catch (e) {\n      this.logger.error(`Error pumping deferred stream: ${e}`);\n      throw e;\n    } finally {\n      reader.releaseLock();\n    }\n  }\n\n  protected async monitorMetrics() {\n    let inferenceDurationTotalMs = 0;\n    let inferenceCount = 0;\n    const metricsReader = this.metricsStream.getReader();\n    while (true) {\n      const { done, value } = await metricsReader.read();\n      if (done) {\n        break;\n      }\n      switch (value.type) {\n        case VADEventType.START_OF_SPEECH:\n          inferenceCount++;\n          if (inferenceCount >= 1000 / this.vad.capabilities.updateInterval) {\n            this.vad.emit('metrics_collected', {\n              type: 'vad_metrics',\n              timestamp: Date.now(),\n              idleTimeMs: Math.trunc(\n                Number((process.hrtime.bigint() - this.lastActivityTime) / BigInt(1000000)),\n              ),\n              inferenceDurationTotalMs,\n              inferenceCount,\n              label: this.vad.label,\n            });\n\n            inferenceCount = 0;\n            inferenceDurationTotalMs = 0;\n          }\n          break;\n        case VADEventType.INFERENCE_DONE:\n          inferenceDurationTotalMs += Math.round(value.inferenceDuration);\n          this.lastActivityTime = process.hrtime.bigint();\n          break;\n        case VADEventType.END_OF_SPEECH:\n          this.lastActivityTime = process.hrtime.bigint();\n          break;\n      }\n    }\n  }\n\n  /**\n   * Safely send a VAD event to the output stream, handling writer release errors during shutdown.\n   * @returns true if the event was sent, false if the stream is closing\n   * @throws Error if an unexpected error occurs\n   */\n  protected sendVADEvent(event: VADEvent): boolean {\n    if (this.closed) {\n      return false;\n    }\n\n    try {\n      this.outputWriter.write(event);\n      return true;\n    } catch (e) {\n      throw e;\n    }\n  }\n\n  updateInputStream(audioStream: ReadableStream<AudioFrame>) {\n    this.deferredInputStream.setSource(audioStream);\n  }\n\n  detachInputStream() {\n    this.deferredInputStream.detachSource();\n  }\n\n  /** @deprecated Use `updateInputStream` instead */\n  pushFrame(frame: AudioFrame) {\n    // TODO(AJS-395): remove this method\n    if (this.inputClosed) {\n      throw new Error('Input is closed');\n    }\n    if (this.closed) {\n      throw new Error('Stream is closed');\n    }\n    this.inputWriter.write(frame);\n  }\n\n  flush() {\n    if (this.inputClosed) {\n      throw new Error('Input is closed');\n    }\n    if (this.closed) {\n      throw new Error('Stream is closed');\n    }\n    this.inputWriter.write(VADStream.FLUSH_SENTINEL);\n  }\n\n  endInput() {\n    if (this.inputClosed) {\n      throw new Error('Input is closed');\n    }\n    if (this.closed) {\n      throw new Error('Stream is closed');\n    }\n    this.inputClosed = true;\n    this.input.writable.close();\n  }\n\n  async next(): Promise<IteratorResult<VADEvent>> {\n    return this.outputReader.read().then(({ done, value }) => {\n      if (done) {\n        return { done: true, value: undefined };\n      }\n      return { done: false, value };\n    });\n  }\n\n  close() {\n    this.outputWriter.releaseLock();\n    this.outputReader.cancel();\n    this.output.writable.close();\n    this.closed = true;\n  }\n\n  [Symbol.asyncIterator](): VADStream {\n    return this;\n  }\n}\n"],"mappings":"AAKA,SAAS,oBAAoB;AAM7B,SAAS,WAAW;AAEpB,SAAS,8BAA8B;AACvC,SAAS,yBAAyB;AAE3B,IAAK,eAAL,kBAAKA,kBAAL;AACL,EAAAA,4BAAA;AACA,EAAAA,4BAAA;AACA,EAAAA,4BAAA;AACA,EAAAA,4BAAA;AAJU,SAAAA;AAAA,GAAA;AAkDL,MAAe,YAAa,aAAsD;AAAA,EACvF;AAAA,EAGA,YAAY,cAA+B;AACzC,UAAM;AACN,SAAK,gBAAgB;AAAA,EACvB;AAAA,EAEA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA,EAOA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAEO,MAAe,UAAqD;AAAA,EACzE,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EACxD,QAAQ,IAAI,kBAAgE;AAAA,EAC5E,SAAS,IAAI,kBAA4B;AAAA,EACzC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA,SAAS;AAAA,EACT,cAAc;AAAA,EAEd;AAAA,EACA,mBAAmB,OAAO,CAAC;AAAA,EAC3B;AAAA,EACA;AAAA,EAEF;AAAA,EACR,YAAY,KAAU;AACpB,SAAK,SAAS,IAAI;AAClB,SAAK,MAAM;AACX,SAAK,sBAAsB,IAAI,uBAAmC;AAElE,SAAK,cAAc,KAAK,MAAM,SAAS,UAAU;AACjD,SAAK,cAAc,KAAK,MAAM,SAAS,UAAU;AACjD,SAAK,eAAe,KAAK,OAAO,SAAS,UAAU;AAEnD,UAAM,CAAC,cAAc,aAAa,IAAI,KAAK,OAAO,SAAS,IAAI;AAC/D,SAAK,gBAAgB;AACrB,SAAK,eAAe,aAAa,UAAU;AAE3C,SAAK,mBAAmB;AACxB,SAAK,eAAe;AAAA,EACtB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EASA,MAAc,qBAAqB;AACjC,UAAM,SAAS,KAAK,oBAAoB,OAAO,UAAU;AACzD,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,KAAM;AACV,cAAM,KAAK,YAAY,MAAM,KAAK;AAAA,MACpC;AAAA,IACF,SAAS,GAAG;AACV,WAAK,OAAO,MAAM,kCAAkC,CAAC,EAAE;AACvD,YAAM;AAAA,IACR,UAAE;AACA,aAAO,YAAY;AAAA,IACrB;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,QAAI,2BAA2B;AAC/B,QAAI,iBAAiB;AACrB,UAAM,gBAAgB,KAAK,cAAc,UAAU;AACnD,WAAO,MAAM;AACX,YAAM,EAAE,MAAM,MAAM,IAAI,MAAM,cAAc,KAAK;AACjD,UAAI,MAAM;AACR;AAAA,MACF;AACA,cAAQ,MAAM,MAAM;AAAA,QAClB,KAAK;AACH;AACA,cAAI,kBAAkB,MAAO,KAAK,IAAI,aAAa,gBAAgB;AACjE,iBAAK,IAAI,KAAK,qBAAqB;AAAA,cACjC,MAAM;AAAA,cACN,WAAW,KAAK,IAAI;AAAA,cACpB,YAAY,KAAK;AAAA,gBACf,QAAQ,QAAQ,OAAO,OAAO,IAAI,KAAK,oBAAoB,OAAO,GAAO,CAAC;AAAA,cAC5E;AAAA,cACA;AAAA,cACA;AAAA,cACA,OAAO,KAAK,IAAI;AAAA,YAClB,CAAC;AAED,6BAAiB;AACjB,uCAA2B;AAAA,UAC7B;AACA;AAAA,QACF,KAAK;AACH,sCAA4B,KAAK,MAAM,MAAM,iBAAiB;AAC9D,eAAK,mBAAmB,QAAQ,OAAO,OAAO;AAC9C;AAAA,QACF,KAAK;AACH,eAAK,mBAAmB,QAAQ,OAAO,OAAO;AAC9C;AAAA,MACJ;AAAA,IACF;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOU,aAAa,OAA0B;AAC/C,QAAI,KAAK,QAAQ;AACf,aAAO;AAAA,IACT;AAEA,QAAI;AACF,WAAK,aAAa,MAAM,KAAK;AAC7B,aAAO;AAAA,IACT,SAAS,GAAG;AACV,YAAM;AAAA,IACR;AAAA,EACF;AAAA,EAEA,kBAAkB,aAAyC;AACzD,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,oBAAoB;AAClB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA;AAAA,EAGA,UAAU,OAAmB;AAE3B,QAAI,KAAK,aAAa;AACpB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,YAAY,MAAM,KAAK;AAAA,EAC9B;AAAA,EAEA,QAAQ;AACN,QAAI,KAAK,aAAa;AACpB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,YAAY,MAAM,UAAU,cAAc;AAAA,EACjD;AAAA,EAEA,WAAW;AACT,QAAI,KAAK,aAAa;AACpB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,cAAc;AACnB,SAAK,MAAM,SAAS,MAAM;AAAA,EAC5B;AAAA,EAEA,MAAM,OAA0C;AAC9C,WAAO,KAAK,aAAa,KAAK,EAAE,KAAK,CAAC,EAAE,MAAM,MAAM,MAAM;AACxD,UAAI,MAAM;AACR,eAAO,EAAE,MAAM,MAAM,OAAO,OAAU;AAAA,MACxC;AACA,aAAO,EAAE,MAAM,OAAO,MAAM;AAAA,IAC9B,CAAC;AAAA,EACH;AAAA,EAEA,QAAQ;AACN,SAAK,aAAa,YAAY;AAC9B,SAAK,aAAa,OAAO;AACzB,SAAK,OAAO,SAAS,MAAM;AAC3B,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAe;AAClC,WAAO;AAAA,EACT;AACF;","names":["VADEventType"]}

package/dist/voice/room_io/_input.cjs

@@ -60,8 +60,10 @@ class ParticipantAudioInputStream extends import_io.AudioInput {
     if (this.participantIdentity === participantIdentity) {
       return;
     }
+    if (this.participantIdentity) {
+      this.closeStream();
+    }
     this.participantIdentity = participantIdentity;
-    this.closeStream();
     if (!participantIdentity) {
       return;
     }
@@ -97,11 +99,9 @@ class ParticipantAudioInputStream extends import_io.AudioInput {
     }
   };
   closeStream() {
-    var _a;
     if (this.deferredStream.isSourceSet) {
       this.deferredStream.detachSource();
     }
-    (_a = this.frameProcessor) == null ? void 0 : _a.close();
     this.publication = null;
   }
   onTrackSubscribed = (track, publication, participant) => {
@@ -147,10 +147,13 @@ class ParticipantAudioInputStream extends import_io.AudioInput {
     });
   }
   async close() {
+    var _a;
     this.room.off(import_rtc_node2.RoomEvent.TrackSubscribed, this.onTrackSubscribed);
     this.room.off(import_rtc_node2.RoomEvent.TrackUnpublished, this.onTrackUnpublished);
     this.room.off(import_rtc_node2.RoomEvent.TokenRefreshed, this.onTokenRefreshed);
     this.closeStream();
+    (_a = this.frameProcessor) == null ? void 0 : _a.close();
+    this.frameProcessor = void 0;
     await this.deferredStream.stream.cancel().catch(() => {
     });
   }

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 { type AudioFrame, FrameProcessor } 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 frameProcessor?: FrameProcessor<AudioFrame>;\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 | FrameProcessor<AudioFrame>;\n  }) {\n    super();\n    this.room = room;\n    this.sampleRate = sampleRate;\n    this.numChannels = numChannels;\n    if (noiseCancellation instanceof FrameProcessor) {\n      this.frameProcessor = noiseCancellation;\n    } else {\n      this.noiseCancellation = noiseCancellation;\n    }\n\n    this.room.on(RoomEvent.TrackSubscribed, this.onTrackSubscribed);\n    this.room.on(RoomEvent.TrackUnpublished, this.onTrackUnpublished);\n    this.room.on(RoomEvent.TokenRefreshed, this.onTokenRefreshed);\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    // Convert Map iterator to array for Pino serialization\n    const trackPublicationsArray = Array.from(participantValue?.trackPublications.values() ?? []);\n\n    this.logger.info(\n      {\n        participantValue: participantValue?.identity,\n        trackPublications: trackPublicationsArray,\n        lengthOfTrackPublications: trackPublicationsArray.length,\n      },\n      'participantValue.trackPublications',\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\n    this.frameProcessor?.close();\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    this.frameProcessor?.onStreamInfoUpdated({\n      participantIdentity: participant.identity,\n      roomName: this.room.name!,\n      publicationSid: publication.sid!,\n    });\n    this.frameProcessor?.onCredentialsUpdated({\n      token: this.room.token!,\n      url: this.room.serverUrl!,\n    });\n    return true;\n  };\n\n  private onTokenRefreshed = () => {\n    if (this.room.token && this.room.serverUrl) {\n      this.frameProcessor?.onCredentialsUpdated({\n        token: this.room.token,\n        url: this.room.serverUrl,\n      });\n    }\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.frameProcessor || 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.room.off(RoomEvent.TokenRefreshed, this.onTokenRefreshed);\n    this.closeStream();\n    // Ignore errors - stream may be locked by RecorderIO or already cancelled\n    await this.deferredStream.stream.cancel().catch(() => {});\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAAgD;AAChD,IAAAA,mBASO;AAEP,iBAAoB;AACpB,mBAA+B;AAC/B,gBAA2B;AAEpB,MAAM,oCAAoC,qBAAW;AAAA,EAClD;AAAA,EACA;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,QAAI,6BAA6B,gCAAgB;AAC/C,WAAK,iBAAiB;AAAA,IACxB,OAAO;AACL,WAAK,oBAAoB;AAAA,IAC3B;AAEA,SAAK,KAAK,GAAG,2BAAU,iBAAiB,KAAK,iBAAiB;AAC9D,SAAK,KAAK,GAAG,2BAAU,kBAAkB,KAAK,kBAAkB;AAChE,SAAK,KAAK,GAAG,2BAAU,gBAAgB,KAAK,gBAAgB;AAAA,EAC9D;AAAA,EAEA,eAAe,aAAgD;AAC7D,SAAK,OAAO,MAAM,EAAE,YAAY,GAAG,iCAAiC;AACpE,UAAM,sBACJ,uBAAuB,qCAAoB,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,qCACnB,cACA,KAAK,KAAK,mBAAmB,IAAI,mBAAmB;AAG1D,UAAM,yBAAyB,MAAM,MAAK,qDAAkB,kBAAkB,aAAY,CAAC,CAAC;AAE5F,SAAK,OAAO;AAAA,MACV;AAAA,QACE,kBAAkB,qDAAkB;AAAA,QACpC,mBAAmB;AAAA,QACnB,2BAA2B,uBAAuB;AAAA,MACpD;AAAA,MACA;AAAA,IACF;AAGA,QAAI,kBAAkB;AACpB,iBAAW,eAAe,iBAAiB,kBAAkB,OAAO,GAAG;AACrE,YAAI,YAAY,SAAS,YAAY,WAAW,6BAAY,mBAAmB;AAC7E,eAAK,kBAAkB,YAAY,OAAO,aAAa,gBAAgB;AACvE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,qBAAqB,CAC3B,aACA,gBACG;AApGP;AAqGI,UACE,UAAK,gBAAL,mBAAkB,SAAQ,YAAY,OACtC,YAAY,aAAa,KAAK,qBAC9B;AACA;AAAA,IACF;AACA,SAAK,YAAY;AAGjB,eAAWC,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;AAxHxB;AAyHI,QAAI,KAAK,eAAe,aAAa;AACnC,WAAK,eAAe,aAAa;AAAA,IACnC;AAEA,eAAK,mBAAL,mBAAqB;AAErB,SAAK,cAAc;AAAA,EACrB;AAAA,EAEQ,oBAAoB,CAC1B,OACA,aACA,gBACY;AAtIhB;AAuII,SAAK,OAAO,MAAM,EAAE,aAAa,YAAY,SAAS,GAAG,6BAA6B;AACtF,QACE,KAAK,wBAAwB,YAAY,YACzC,YAAY,WAAW,6BAAY,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,eAAK,mBAAL,mBAAqB,oBAAoB;AAAA,MACvC,qBAAqB,YAAY;AAAA,MACjC,UAAU,KAAK,KAAK;AAAA,MACpB,gBAAgB,YAAY;AAAA,IAC9B;AACA,eAAK,mBAAL,mBAAqB,qBAAqB;AAAA,MACxC,OAAO,KAAK,KAAK;AAAA,MACjB,KAAK,KAAK,KAAK;AAAA,IACjB;AACA,WAAO;AAAA,EACT;AAAA,EAEQ,mBAAmB,MAAM;AAnKnC;AAoKI,QAAI,KAAK,KAAK,SAAS,KAAK,KAAK,WAAW;AAC1C,iBAAK,mBAAL,mBAAqB,qBAAqB;AAAA,QACxC,OAAO,KAAK,KAAK;AAAA,QACjB,KAAK,KAAK,KAAK;AAAA,MACjB;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,aAAa,OAAgD;AACnE,WAAO,IAAI,6BAAY,OAAO;AAAA,MAC5B,YAAY,KAAK;AAAA,MACjB,aAAa,KAAK;AAAA,MAClB,mBAAmB,KAAK,kBAAkB,KAAK;AAAA;AAAA,IAEjD,CAAC;AAAA,EACH;AAAA,EAEA,MAAM,QAAQ;AACZ,SAAK,KAAK,IAAI,2BAAU,iBAAiB,KAAK,iBAAiB;AAC/D,SAAK,KAAK,IAAI,2BAAU,kBAAkB,KAAK,kBAAkB;AACjE,SAAK,KAAK,IAAI,2BAAU,gBAAgB,KAAK,gBAAgB;AAC7D,SAAK,YAAY;AAEjB,UAAM,KAAK,eAAe,OAAO,OAAO,EAAE,MAAM,MAAM;AAAA,IAAC,CAAC;AAAA,EAC1D;AACF;","names":["import_rtc_node","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, FrameProcessor } 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 frameProcessor?: FrameProcessor<AudioFrame>;\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 | FrameProcessor<AudioFrame>;\n  }) {\n    super();\n    this.room = room;\n    this.sampleRate = sampleRate;\n    this.numChannels = numChannels;\n    if (noiseCancellation instanceof FrameProcessor) {\n      this.frameProcessor = noiseCancellation;\n    } else {\n      this.noiseCancellation = noiseCancellation;\n    }\n\n    this.room.on(RoomEvent.TrackSubscribed, this.onTrackSubscribed);\n    this.room.on(RoomEvent.TrackUnpublished, this.onTrackUnpublished);\n    this.room.on(RoomEvent.TokenRefreshed, this.onTokenRefreshed);\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    if (this.participantIdentity) {\n      this.closeStream();\n    }\n    this.participantIdentity = participantIdentity;\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    // Convert Map iterator to array for Pino serialization\n    const trackPublicationsArray = Array.from(participantValue?.trackPublications.values() ?? []);\n\n    this.logger.info(\n      {\n        participantValue: participantValue?.identity,\n        trackPublications: trackPublicationsArray,\n        lengthOfTrackPublications: trackPublicationsArray.length,\n      },\n      'participantValue.trackPublications',\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\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    this.frameProcessor?.onStreamInfoUpdated({\n      participantIdentity: participant.identity,\n      roomName: this.room.name!,\n      publicationSid: publication.sid!,\n    });\n    this.frameProcessor?.onCredentialsUpdated({\n      token: this.room.token!,\n      url: this.room.serverUrl!,\n    });\n    return true;\n  };\n\n  private onTokenRefreshed = () => {\n    if (this.room.token && this.room.serverUrl) {\n      this.frameProcessor?.onCredentialsUpdated({\n        token: this.room.token,\n        url: this.room.serverUrl,\n      });\n    }\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.frameProcessor || 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.room.off(RoomEvent.TokenRefreshed, this.onTokenRefreshed);\n    this.closeStream();\n    this.frameProcessor?.close();\n    this.frameProcessor = undefined;\n    // Ignore errors - stream may be locked by RecorderIO or already cancelled\n    await this.deferredStream.stream.cancel().catch(() => {});\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAAgD;AAChD,IAAAA,mBASO;AAEP,iBAAoB;AACpB,mBAA+B;AAC/B,gBAA2B;AAEpB,MAAM,oCAAoC,qBAAW;AAAA,EAClD;AAAA,EACA;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,QAAI,6BAA6B,gCAAgB;AAC/C,WAAK,iBAAiB;AAAA,IACxB,OAAO;AACL,WAAK,oBAAoB;AAAA,IAC3B;AAEA,SAAK,KAAK,GAAG,2BAAU,iBAAiB,KAAK,iBAAiB;AAC9D,SAAK,KAAK,GAAG,2BAAU,kBAAkB,KAAK,kBAAkB;AAChE,SAAK,KAAK,GAAG,2BAAU,gBAAgB,KAAK,gBAAgB;AAAA,EAC9D;AAAA,EAEA,eAAe,aAAgD;AAC7D,SAAK,OAAO,MAAM,EAAE,YAAY,GAAG,iCAAiC;AACpE,UAAM,sBACJ,uBAAuB,qCAAoB,YAAY,WAAW;AAEpE,QAAI,KAAK,wBAAwB,qBAAqB;AACpD;AAAA,IACF;AACA,QAAI,KAAK,qBAAqB;AAC5B,WAAK,YAAY;AAAA,IACnB;AACA,SAAK,sBAAsB;AAE3B,QAAI,CAAC,qBAAqB;AACxB;AAAA,IACF;AAEA,UAAM,mBACJ,uBAAuB,qCACnB,cACA,KAAK,KAAK,mBAAmB,IAAI,mBAAmB;AAG1D,UAAM,yBAAyB,MAAM,MAAK,qDAAkB,kBAAkB,aAAY,CAAC,CAAC;AAE5F,SAAK,OAAO;AAAA,MACV;AAAA,QACE,kBAAkB,qDAAkB;AAAA,QACpC,mBAAmB;AAAA,QACnB,2BAA2B,uBAAuB;AAAA,MACpD;AAAA,MACA;AAAA,IACF;AAGA,QAAI,kBAAkB;AACpB,iBAAW,eAAe,iBAAiB,kBAAkB,OAAO,GAAG;AACrE,YAAI,YAAY,SAAS,YAAY,WAAW,6BAAY,mBAAmB;AAC7E,eAAK,kBAAkB,YAAY,OAAO,aAAa,gBAAgB;AACvE;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,qBAAqB,CAC3B,aACA,gBACG;AAtGP;AAuGI,UACE,UAAK,gBAAL,mBAAkB,SAAQ,YAAY,OACtC,YAAY,aAAa,KAAK,qBAC9B;AACA;AAAA,IACF;AACA,SAAK,YAAY;AAGjB,eAAWC,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;AAEA,SAAK,cAAc;AAAA,EACrB;AAAA,EAEQ,oBAAoB,CAC1B,OACA,aACA,gBACY;AAtIhB;AAuII,SAAK,OAAO,MAAM,EAAE,aAAa,YAAY,SAAS,GAAG,6BAA6B;AACtF,QACE,KAAK,wBAAwB,YAAY,YACzC,YAAY,WAAW,6BAAY,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,eAAK,mBAAL,mBAAqB,oBAAoB;AAAA,MACvC,qBAAqB,YAAY;AAAA,MACjC,UAAU,KAAK,KAAK;AAAA,MACpB,gBAAgB,YAAY;AAAA,IAC9B;AACA,eAAK,mBAAL,mBAAqB,qBAAqB;AAAA,MACxC,OAAO,KAAK,KAAK;AAAA,MACjB,KAAK,KAAK,KAAK;AAAA,IACjB;AACA,WAAO;AAAA,EACT;AAAA,EAEQ,mBAAmB,MAAM;AAnKnC;AAoKI,QAAI,KAAK,KAAK,SAAS,KAAK,KAAK,WAAW;AAC1C,iBAAK,mBAAL,mBAAqB,qBAAqB;AAAA,QACxC,OAAO,KAAK,KAAK;AAAA,QACjB,KAAK,KAAK,KAAK;AAAA,MACjB;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,aAAa,OAAgD;AACnE,WAAO,IAAI,6BAAY,OAAO;AAAA,MAC5B,YAAY,KAAK;AAAA,MACjB,aAAa,KAAK;AAAA,MAClB,mBAAmB,KAAK,kBAAkB,KAAK;AAAA;AAAA,IAEjD,CAAC;AAAA,EACH;AAAA,EAEA,MAAM,QAAQ;AArLhB;AAsLI,SAAK,KAAK,IAAI,2BAAU,iBAAiB,KAAK,iBAAiB;AAC/D,SAAK,KAAK,IAAI,2BAAU,kBAAkB,KAAK,kBAAkB;AACjE,SAAK,KAAK,IAAI,2BAAU,gBAAgB,KAAK,gBAAgB;AAC7D,SAAK,YAAY;AACjB,eAAK,mBAAL,mBAAqB;AACrB,SAAK,iBAAiB;AAEtB,UAAM,KAAK,eAAe,OAAO,OAAO,EAAE,MAAM,MAAM;AAAA,IAAC,CAAC;AAAA,EAC1D;AACF;","names":["import_rtc_node","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,EAAE,KAAK,UAAU,EAAE,cAAc,EAAE,MAAM,mBAAmB,CAAC;AACpE,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,cAAc,CAAC,CAA6B;IACpD,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,GAAG,cAAc,CAAC,UAAU,CAAC,CAAC;KAC3E;IAgBD,cAAc,CAAC,WAAW,EAAE,iBAAiB,GAAG,MAAM,GAAG,IAAI;IA2C7D,OAAO,CAAC,kBAAkB,CAqBxB;IAEF,OAAO,CAAC,WAAW;IAUnB,OAAO,CAAC,iBAAiB,CA+BvB;IAEF,OAAO,CAAC,gBAAgB,CAOtB;IAEF,OAAO,CAAC,YAAY;IASd,KAAK;CAQZ"}
+{"version":3,"file":"_input.d.ts","sourceRoot":"","sources":["../../../src/voice/room_io/_input.ts"],"names":[],"mappings":"AAGA,OAAO,EAAE,KAAK,UAAU,EAAE,cAAc,EAAE,MAAM,mBAAmB,CAAC;AACpE,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,cAAc,CAAC,CAA6B;IACpD,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,GAAG,cAAc,CAAC,UAAU,CAAC,CAAC;KAC3E;IAgBD,cAAc,CAAC,WAAW,EAAE,iBAAiB,GAAG,MAAM,GAAG,IAAI;IA6C7D,OAAO,CAAC,kBAAkB,CAqBxB;IAEF,OAAO,CAAC,WAAW;IAQnB,OAAO,CAAC,iBAAiB,CA+BvB;IAEF,OAAO,CAAC,gBAAgB,CAOtB;IAEF,OAAO,CAAC,YAAY;IASd,KAAK;CAUZ"}

package/dist/voice/room_io/_input.js

@@ -42,8 +42,10 @@ class ParticipantAudioInputStream extends AudioInput {
     if (this.participantIdentity === participantIdentity) {
       return;
     }
+    if (this.participantIdentity) {
+      this.closeStream();
+    }
     this.participantIdentity = participantIdentity;
-    this.closeStream();
     if (!participantIdentity) {
       return;
     }
@@ -79,11 +81,9 @@ class ParticipantAudioInputStream extends AudioInput {
     }
   };
   closeStream() {
-    var _a;
     if (this.deferredStream.isSourceSet) {
       this.deferredStream.detachSource();
     }
-    (_a = this.frameProcessor) == null ? void 0 : _a.close();
     this.publication = null;
   }
   onTrackSubscribed = (track, publication, participant) => {
@@ -129,10 +129,13 @@ class ParticipantAudioInputStream extends AudioInput {
     });
   }
   async close() {
+    var _a;
     this.room.off(RoomEvent.TrackSubscribed, this.onTrackSubscribed);
     this.room.off(RoomEvent.TrackUnpublished, this.onTrackUnpublished);
     this.room.off(RoomEvent.TokenRefreshed, this.onTokenRefreshed);
     this.closeStream();
+    (_a = this.frameProcessor) == null ? void 0 : _a.close();
+    this.frameProcessor = void 0;
     await this.deferredStream.stream.cancel().catch(() => {
     });
   }

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 { type AudioFrame, FrameProcessor } 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 frameProcessor?: FrameProcessor<AudioFrame>;\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 | FrameProcessor<AudioFrame>;\n  }) {\n    super();\n    this.room = room;\n    this.sampleRate = sampleRate;\n    this.numChannels = numChannels;\n    if (noiseCancellation instanceof FrameProcessor) {\n      this.frameProcessor = noiseCancellation;\n    } else {\n      this.noiseCancellation = noiseCancellation;\n    }\n\n    this.room.on(RoomEvent.TrackSubscribed, this.onTrackSubscribed);\n    this.room.on(RoomEvent.TrackUnpublished, this.onTrackUnpublished);\n    this.room.on(RoomEvent.TokenRefreshed, this.onTokenRefreshed);\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    // Convert Map iterator to array for Pino serialization\n    const trackPublicationsArray = Array.from(participantValue?.trackPublications.values() ?? []);\n\n    this.logger.info(\n      {\n        participantValue: participantValue?.identity,\n        trackPublications: trackPublicationsArray,\n        lengthOfTrackPublications: trackPublicationsArray.length,\n      },\n      'participantValue.trackPublications',\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\n    this.frameProcessor?.close();\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    this.frameProcessor?.onStreamInfoUpdated({\n      participantIdentity: participant.identity,\n      roomName: this.room.name!,\n      publicationSid: publication.sid!,\n    });\n    this.frameProcessor?.onCredentialsUpdated({\n      token: this.room.token!,\n      url: this.room.serverUrl!,\n    });\n    return true;\n  };\n\n  private onTokenRefreshed = () => {\n    if (this.room.token && this.room.serverUrl) {\n      this.frameProcessor?.onCredentialsUpdated({\n        token: this.room.token,\n        url: this.room.serverUrl,\n      });\n    }\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.frameProcessor || 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.room.off(RoomEvent.TokenRefreshed, this.onTokenRefreshed);\n    this.closeStream();\n    // Ignore errors - stream may be locked by RecorderIO or already cancelled\n    await this.deferredStream.stream.cancel().catch(() => {});\n  }\n}\n"],"mappings":"AAGA,SAA0B,sBAAsB;AAChD;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;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,QAAI,6BAA6B,gBAAgB;AAC/C,WAAK,iBAAiB;AAAA,IACxB,OAAO;AACL,WAAK,oBAAoB;AAAA,IAC3B;AAEA,SAAK,KAAK,GAAG,UAAU,iBAAiB,KAAK,iBAAiB;AAC9D,SAAK,KAAK,GAAG,UAAU,kBAAkB,KAAK,kBAAkB;AAChE,SAAK,KAAK,GAAG,UAAU,gBAAgB,KAAK,gBAAgB;AAAA,EAC9D;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;AAG1D,UAAM,yBAAyB,MAAM,MAAK,qDAAkB,kBAAkB,aAAY,CAAC,CAAC;AAE5F,SAAK,OAAO;AAAA,MACV;AAAA,QACE,kBAAkB,qDAAkB;AAAA,QACpC,mBAAmB;AAAA,QACnB,2BAA2B,uBAAuB;AAAA,MACpD;AAAA,MACA;AAAA,IACF;AAGA,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;AApGP;AAqGI,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;AAxHxB;AAyHI,QAAI,KAAK,eAAe,aAAa;AACnC,WAAK,eAAe,aAAa;AAAA,IACnC;AAEA,eAAK,mBAAL,mBAAqB;AAErB,SAAK,cAAc;AAAA,EACrB;AAAA,EAEQ,oBAAoB,CAC1B,OACA,aACA,gBACY;AAtIhB;AAuII,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,eAAK,mBAAL,mBAAqB,oBAAoB;AAAA,MACvC,qBAAqB,YAAY;AAAA,MACjC,UAAU,KAAK,KAAK;AAAA,MACpB,gBAAgB,YAAY;AAAA,IAC9B;AACA,eAAK,mBAAL,mBAAqB,qBAAqB;AAAA,MACxC,OAAO,KAAK,KAAK;AAAA,MACjB,KAAK,KAAK,KAAK;AAAA,IACjB;AACA,WAAO;AAAA,EACT;AAAA,EAEQ,mBAAmB,MAAM;AAnKnC;AAoKI,QAAI,KAAK,KAAK,SAAS,KAAK,KAAK,WAAW;AAC1C,iBAAK,mBAAL,mBAAqB,qBAAqB;AAAA,QACxC,OAAO,KAAK,KAAK;AAAA,QACjB,KAAK,KAAK,KAAK;AAAA,MACjB;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,aAAa,OAAgD;AACnE,WAAO,IAAI,YAAY,OAAO;AAAA,MAC5B,YAAY,KAAK;AAAA,MACjB,aAAa,KAAK;AAAA,MAClB,mBAAmB,KAAK,kBAAkB,KAAK;AAAA;AAAA,IAEjD,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,KAAK,IAAI,UAAU,gBAAgB,KAAK,gBAAgB;AAC7D,SAAK,YAAY;AAEjB,UAAM,KAAK,eAAe,OAAO,OAAO,EAAE,MAAM,MAAM;AAAA,IAAC,CAAC;AAAA,EAC1D;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, FrameProcessor } 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 frameProcessor?: FrameProcessor<AudioFrame>;\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 | FrameProcessor<AudioFrame>;\n  }) {\n    super();\n    this.room = room;\n    this.sampleRate = sampleRate;\n    this.numChannels = numChannels;\n    if (noiseCancellation instanceof FrameProcessor) {\n      this.frameProcessor = noiseCancellation;\n    } else {\n      this.noiseCancellation = noiseCancellation;\n    }\n\n    this.room.on(RoomEvent.TrackSubscribed, this.onTrackSubscribed);\n    this.room.on(RoomEvent.TrackUnpublished, this.onTrackUnpublished);\n    this.room.on(RoomEvent.TokenRefreshed, this.onTokenRefreshed);\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    if (this.participantIdentity) {\n      this.closeStream();\n    }\n    this.participantIdentity = participantIdentity;\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    // Convert Map iterator to array for Pino serialization\n    const trackPublicationsArray = Array.from(participantValue?.trackPublications.values() ?? []);\n\n    this.logger.info(\n      {\n        participantValue: participantValue?.identity,\n        trackPublications: trackPublicationsArray,\n        lengthOfTrackPublications: trackPublicationsArray.length,\n      },\n      'participantValue.trackPublications',\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\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    this.frameProcessor?.onStreamInfoUpdated({\n      participantIdentity: participant.identity,\n      roomName: this.room.name!,\n      publicationSid: publication.sid!,\n    });\n    this.frameProcessor?.onCredentialsUpdated({\n      token: this.room.token!,\n      url: this.room.serverUrl!,\n    });\n    return true;\n  };\n\n  private onTokenRefreshed = () => {\n    if (this.room.token && this.room.serverUrl) {\n      this.frameProcessor?.onCredentialsUpdated({\n        token: this.room.token,\n        url: this.room.serverUrl,\n      });\n    }\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.frameProcessor || 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.room.off(RoomEvent.TokenRefreshed, this.onTokenRefreshed);\n    this.closeStream();\n    this.frameProcessor?.close();\n    this.frameProcessor = undefined;\n    // Ignore errors - stream may be locked by RecorderIO or already cancelled\n    await this.deferredStream.stream.cancel().catch(() => {});\n  }\n}\n"],"mappings":"AAGA,SAA0B,sBAAsB;AAChD;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;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,QAAI,6BAA6B,gBAAgB;AAC/C,WAAK,iBAAiB;AAAA,IACxB,OAAO;AACL,WAAK,oBAAoB;AAAA,IAC3B;AAEA,SAAK,KAAK,GAAG,UAAU,iBAAiB,KAAK,iBAAiB;AAC9D,SAAK,KAAK,GAAG,UAAU,kBAAkB,KAAK,kBAAkB;AAChE,SAAK,KAAK,GAAG,UAAU,gBAAgB,KAAK,gBAAgB;AAAA,EAC9D;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,QAAI,KAAK,qBAAqB;AAC5B,WAAK,YAAY;AAAA,IACnB;AACA,SAAK,sBAAsB;AAE3B,QAAI,CAAC,qBAAqB;AACxB;AAAA,IACF;AAEA,UAAM,mBACJ,uBAAuB,oBACnB,cACA,KAAK,KAAK,mBAAmB,IAAI,mBAAmB;AAG1D,UAAM,yBAAyB,MAAM,MAAK,qDAAkB,kBAAkB,aAAY,CAAC,CAAC;AAE5F,SAAK,OAAO;AAAA,MACV;AAAA,QACE,kBAAkB,qDAAkB;AAAA,QACpC,mBAAmB;AAAA,QACnB,2BAA2B,uBAAuB;AAAA,MACpD;AAAA,MACA;AAAA,IACF;AAGA,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;AAtGP;AAuGI,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;AAEA,SAAK,cAAc;AAAA,EACrB;AAAA,EAEQ,oBAAoB,CAC1B,OACA,aACA,gBACY;AAtIhB;AAuII,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,eAAK,mBAAL,mBAAqB,oBAAoB;AAAA,MACvC,qBAAqB,YAAY;AAAA,MACjC,UAAU,KAAK,KAAK;AAAA,MACpB,gBAAgB,YAAY;AAAA,IAC9B;AACA,eAAK,mBAAL,mBAAqB,qBAAqB;AAAA,MACxC,OAAO,KAAK,KAAK;AAAA,MACjB,KAAK,KAAK,KAAK;AAAA,IACjB;AACA,WAAO;AAAA,EACT;AAAA,EAEQ,mBAAmB,MAAM;AAnKnC;AAoKI,QAAI,KAAK,KAAK,SAAS,KAAK,KAAK,WAAW;AAC1C,iBAAK,mBAAL,mBAAqB,qBAAqB;AAAA,QACxC,OAAO,KAAK,KAAK;AAAA,QACjB,KAAK,KAAK,KAAK;AAAA,MACjB;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,aAAa,OAAgD;AACnE,WAAO,IAAI,YAAY,OAAO;AAAA,MAC5B,YAAY,KAAK;AAAA,MACjB,aAAa,KAAK;AAAA,MAClB,mBAAmB,KAAK,kBAAkB,KAAK;AAAA;AAAA,IAEjD,CAAC;AAAA,EACH;AAAA,EAEA,MAAM,QAAQ;AArLhB;AAsLI,SAAK,KAAK,IAAI,UAAU,iBAAiB,KAAK,iBAAiB;AAC/D,SAAK,KAAK,IAAI,UAAU,kBAAkB,KAAK,kBAAkB;AACjE,SAAK,KAAK,IAAI,UAAU,gBAAgB,KAAK,gBAAgB;AAC7D,SAAK,YAAY;AACjB,eAAK,mBAAL,mBAAqB;AACrB,SAAK,iBAAiB;AAEtB,UAAM,KAAK,eAAe,OAAO,OAAO,EAAE,MAAM,MAAM;AAAA,IAAC,CAAC;AAAA,EAC1D;AACF;","names":["publication"]}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@livekit/agents",
-  "version": "1.0.43",
+  "version": "1.0.45",
   "description": "LiveKit Agents - Node.js",
   "main": "dist/index.js",
   "require": "dist/index.cjs",

package/src/inference/stt.ts

@@ -177,7 +177,7 @@ export class STT<TModel extends STTModels> extends BaseSTT {
   #logger = log();
 
   constructor(opts?: {
-    model?: TModel;
+    model?: ModelWithLanguage;
     language?: STTLanguages;
     baseURL?: string;
     encoding?: STTEncoding;
@@ -215,11 +215,29 @@ export class STT<TModel extends STTModels> extends BaseSTT {
       throw new Error('apiSecret is required: pass apiSecret or set LIVEKIT_API_SECRET');
     }
 
+    // Parse language from model string if provided: "provider/model:language"
+    let nextModel = model;
+    let nextLanguage = language;
+    if (typeof nextModel === 'string') {
+      const idx = nextModel.lastIndexOf(':');
+      if (idx !== -1) {
+        const languageFromModel = nextModel.slice(idx + 1) as STTLanguages;
+        if (nextLanguage && nextLanguage !== languageFromModel) {
+          this.#logger.warn(
+            '`language` is provided via both argument and model, using the one from the argument',
+            { language: nextLanguage, model: nextModel },
+          );
+        } else {
+          nextLanguage = languageFromModel;
+        }
+        nextModel = nextModel.slice(0, idx) as TModel;
+      }
+    }
     const normalizedFallback = fallback ? normalizeSTTFallback(fallback) : undefined;
 
     this.opts = {
-      model,
-      language,
+      model: nextModel as TModel,
+      language: nextLanguage,
       encoding,
       sampleRate,
       baseURL: lkBaseURL,

package/src/ipc/job_proc_lazy_main.ts

@@ -1,7 +1,7 @@
 // SPDX-FileCopyrightText: 2024 LiveKit, Inc.
 //
 // SPDX-License-Identifier: Apache-2.0
-import { Room, RoomEvent } from '@livekit/rtc-node';
+import { Room, RoomEvent, dispose } from '@livekit/rtc-node';
 import { EventEmitter, once } from 'node:events';
 import { pathToFileURL } from 'node:url';
 import type { Logger } from 'pino';
@@ -245,6 +245,18 @@ const startJob = (
 
     await join.await;
 
+    // Dispose native FFI resources (Rust FfiServer, tokio runtimes, libwebrtc)
+    // before process.exit() to prevent libc++abi mutex crash during teardown.
+    // Without this, process.exit() can kill the process while native threads are
+    // still running, causing: "mutex lock failed: Invalid argument"
+    // See: https://github.com/livekit/node-sdks/issues/564
+    try {
+      await dispose();
+      logger.debug('native resources disposed');
+    } catch (error) {
+      logger.warn({ error }, 'failed to dispose native resources');
+    }
+
     logger.debug('Job process shutdown');
     process.exit(0);
   }

package/src/ipc/supervised_proc.ts

@@ -182,7 +182,7 @@ export abstract class SupervisedProc {
     this.proc.send({
       case: 'initializeRequest',
       value: {
-        loggerOptions,
+        loggerOptions: loggerOptions(),
         pingInterval: this.#opts.pingInterval,
         pingTimeout: this.#opts.pingTimeout,
         highPingThreshold: this.#opts.highPingThreshold,

package/src/llm/llm.ts

@@ -174,7 +174,7 @@ export abstract class LLMStream implements AsyncIterableIterator<ChatChunk> {
             this.emitError({ error, recoverable: true });
             this.logger.warn(
               { llm: this.#llm.label(), attempt: i + 1, error },
-              `failed to generate LLM completion, retrying in ${retryInterval}s`,
+              `failed to generate LLM completion, retrying in ${retryInterval}ms`,
             );
           }
 

package/src/log.ts

@@ -13,17 +13,27 @@ export type LoggerOptions = {
   level?: string;
 };
 
-/** @internal */
-export let loggerOptions: LoggerOptions;
+// Use Symbol.for() + globalThis to create process-wide singletons.
+// This avoids the "dual package hazard". Symbol.for() returns the same Symbol
+// across all module instances, and globalThis is shared process-wide.
+const LOGGER_KEY = Symbol.for('@livekit/agents:logger');
+const LOGGER_OPTIONS_KEY = Symbol.for('@livekit/agents:loggerOptions');
+const OTEL_ENABLED_KEY = Symbol.for('@livekit/agents:otelEnabled');
 
-/** @internal */
-let logger: Logger | undefined = undefined;
+type GlobalState = {
+  [LOGGER_KEY]?: Logger;
+  [LOGGER_OPTIONS_KEY]?: LoggerOptions;
+  [OTEL_ENABLED_KEY]?: boolean;
+};
+
+const globals = globalThis as typeof globalThis & GlobalState;
 
 /** @internal */
-let otelEnabled = false;
+export const loggerOptions = (): LoggerOptions | undefined => globals[LOGGER_OPTIONS_KEY];
 
 /** @internal */
 export const log = () => {
+  const logger = globals[LOGGER_KEY];
   if (!logger) {
     throw new TypeError('logger not initialized. did you forget to run initializeLogger()?');
   }
@@ -32,8 +42,8 @@ export const log = () => {
 
 /** @internal */
 export const initializeLogger = ({ pretty, level }: LoggerOptions) => {
-  loggerOptions = { pretty, level };
-  logger = pino(
+  globals[LOGGER_OPTIONS_KEY] = { pretty, level };
+  globals[LOGGER_KEY] = pino(
     { level: level || 'info' },
     pretty ? pinoPretty({ colorize: true }) : process.stdout,
   );
@@ -65,13 +75,14 @@ class OtelDestination extends Writable {
  * @internal
  */
 export const enableOtelLogging = () => {
-  if (otelEnabled || !logger) {
+  if (globals[OTEL_ENABLED_KEY] || !globals[LOGGER_KEY]) {
     console.warn('OTEL logging already enabled or logger not initialized');
     return;
   }
-  otelEnabled = true;
+  globals[OTEL_ENABLED_KEY] = true;
 
-  const { pretty, level } = loggerOptions;
+  const opts = globals[LOGGER_OPTIONS_KEY]!;
+  const { pretty, level } = opts;
 
   const logLevel = level || 'info';
   const streams: { stream: DestinationStream; level: string }[] = [
@@ -79,5 +90,5 @@ export const enableOtelLogging = () => {
     { stream: new OtelDestination(), level: 'debug' },
   ];
 
-  logger = pino({ level: logLevel }, multistream(streams));
+  globals[LOGGER_KEY] = pino({ level: logLevel }, multistream(streams));
 };

package/src/stt/stt.ts

@@ -236,8 +236,8 @@ export abstract class SpeechStream implements AsyncIterableIterator<SpeechEvent>
             // Don't emit error event for recoverable errors during retry loop
             // to avoid ERR_UNHANDLED_ERROR or premature session termination
             this.logger.warn(
-              { tts: this.#stt.label, attempt: i + 1, error },
-              `failed to recognize speech, retrying in ${retryInterval}s`,
+              { stt: this.#stt.label, attempt: i + 1, error },
+              `failed to recognize speech, retrying in ${retryInterval}ms`,
             );
           }
 
@@ -351,6 +351,11 @@ export abstract class SpeechStream implements AsyncIterableIterator<SpeechEvent>
       }
     }
 
+    if (frame.samplesPerChannel === 0) {
+      this.input.put(frame);
+      return;
+    }
+
     if (this.resampler) {
       const frames = this.resampler.push(frame);
       for (const frame of frames) {