@livekit/agents 1.0.21 → 1.0.22

package/dist/tts/tts.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../../src/tts/tts.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 { ReadableStream } from 'node:stream/web';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { log } from '../log.js';\nimport type { TTSMetrics } from '../metrics/base.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { AsyncIterableQueue, delay, mergeFrames, startSoon, toError } from '../utils.js';\n\n/** SynthesizedAudio is a packet of speech synthesis as returned by the TTS. */\nexport interface SynthesizedAudio {\n  /** Request ID (one segment could be made up of multiple requests) */\n  requestId: string;\n  /** Segment ID, each segment is separated by a flush */\n  segmentId: string;\n  /** Synthesized audio frame */\n  frame: AudioFrame;\n  /** Current segment of the synthesized audio */\n  deltaText?: string;\n  /** Whether this is the last frame of the segment (streaming only) */\n  final: boolean;\n}\n\n/**\n * Describes the capabilities of the TTS provider.\n *\n * @remarks\n * At present, only `streaming` is supplied to this interface, and the framework only supports\n * providers that do have a streaming endpoint.\n */\nexport interface TTSCapabilities {\n  streaming: boolean;\n}\n\nexport interface TTSError {\n  type: 'tts_error';\n  timestamp: number;\n  label: string;\n  error: Error;\n  recoverable: boolean;\n}\n\nexport type TTSCallbacks = {\n  ['metrics_collected']: (metrics: TTSMetrics) => void;\n  ['error']: (error: TTSError) => void;\n};\n\n/**\n * An instance of a text-to-speech adapter.\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child TTS class, which inherits this class's methods.\n */\nexport abstract class TTS extends (EventEmitter as new () => TypedEmitter<TTSCallbacks>) {\n  #capabilities: TTSCapabilities;\n  #sampleRate: number;\n  #numChannels: number;\n  abstract label: string;\n\n  constructor(sampleRate: number, numChannels: number, capabilities: TTSCapabilities) {\n    super();\n    this.#capabilities = capabilities;\n    this.#sampleRate = sampleRate;\n    this.#numChannels = numChannels;\n  }\n\n  /** Returns this TTS's capabilities */\n  get capabilities(): TTSCapabilities {\n    return this.#capabilities;\n  }\n\n  /** Returns the sample rate of audio frames returned by this TTS */\n  get sampleRate(): number {\n    return this.#sampleRate;\n  }\n\n  /** Returns the channel count of audio frames returned by this TTS */\n  get numChannels(): number {\n    return this.#numChannels;\n  }\n\n  /**\n   * Receives text and returns synthesis in the form of a {@link ChunkedStream}\n   */\n  abstract synthesize(text: string): ChunkedStream;\n\n  /**\n   * Returns a {@link SynthesizeStream} that can be used to push text and receive audio data\n   */\n  abstract stream(): SynthesizeStream;\n\n  async close(): Promise<void> {\n    return;\n  }\n}\n\n/**\n * An instance of a text-to-speech stream, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n *   await source.captureFrame(event.frame);\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child SynthesizeStream class, which inherits this class's methods.\n */\nexport abstract class SynthesizeStream\n  implements AsyncIterableIterator<SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM>\n{\n  protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n  static readonly END_OF_STREAM = Symbol('END_OF_STREAM');\n  protected input = new AsyncIterableQueue<string | typeof SynthesizeStream.FLUSH_SENTINEL>();\n  protected queue = new AsyncIterableQueue<\n    SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM\n  >();\n  protected output = new AsyncIterableQueue<\n    SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM\n  >();\n  protected closed = false;\n  abstract label: string;\n  #tts: TTS;\n  #metricsPendingTexts: string[] = [];\n  #metricsText = '';\n  #monitorMetricsTask?: Promise<void>;\n  private _connOptions: APIConnectOptions;\n  protected abortController = new AbortController();\n\n  private deferredInputStream: DeferredReadableStream<\n    string | typeof SynthesizeStream.FLUSH_SENTINEL\n  >;\n  private logger = log();\n\n  constructor(tts: TTS, connOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS) {\n    this.#tts = tts;\n    this._connOptions = connOptions;\n    this.deferredInputStream = new DeferredReadableStream();\n    this.pumpInput();\n    this.abortController.signal.addEventListener('abort', () => {\n      this.deferredInputStream.detachSource();\n      // TODO (AJS-36) clean this up when we refactor with streams\n      this.input.close();\n      this.output.close();\n      this.closed = true;\n    });\n\n    // this is a hack to immitate asyncio.create_task so that mainTask\n    // is run **after** the constructor has finished. Otherwise we get\n    // runtime error when trying to access class variables in the\n    // `run` method.\n    startSoon(() => this.mainTask().then(() => this.queue.close()));\n  }\n\n  private async mainTask() {\n    // TODO(brian): PR3 - Add span wrapping: tracer.startActiveSpan('tts_request', ..., { endOnExit: false })\n    for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n      try {\n        // TODO(brian): PR3 - Add span for retry attempts: tracer.startActiveSpan('tts_request_run', ...)\n        return await this.run();\n      } catch (error) {\n        if (error instanceof APIError) {\n          const retryInterval = this._connOptions._intervalForRetry(i);\n\n          if (this._connOptions.maxRetry === 0 || !error.retryable) {\n            this.emitError({ error, recoverable: false });\n            throw error;\n          } else if (i === this._connOptions.maxRetry) {\n            this.emitError({ error, recoverable: false });\n            throw new APIConnectionError({\n              message: `failed to generate TTS completion after ${this._connOptions.maxRetry + 1} attempts`,\n              options: { retryable: false },\n            });\n          } else {\n            // Don't emit error event for recoverable errors during retry loop\n            // to avoid ERR_UNHANDLED_ERROR or premature session termination\n            this.logger.warn(\n              { tts: this.#tts.label, attempt: i + 1, error },\n              `failed to synthesize speech, retrying in  ${retryInterval}s`,\n            );\n          }\n\n          if (retryInterval > 0) {\n            await delay(retryInterval);\n          }\n        } else {\n          this.emitError({ error: toError(error), recoverable: false });\n          throw error;\n        }\n      }\n    }\n  }\n\n  private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n    this.#tts.emit('error', {\n      type: 'tts_error',\n      timestamp: Date.now(),\n      label: this.#tts.label,\n      error,\n      recoverable,\n    });\n  }\n\n  // NOTE(AJS-37): The implementation below uses an AsyncIterableQueue (`this.input`)\n  // bridged from a DeferredReadableStream (`this.deferredInputStream`) rather than\n  // consuming the stream directly.\n  //\n  // A full refactor to native Web Streams was considered but is currently deferred.\n  // The primary reason is to maintain architectural parity with the Python SDK,\n  // which is a key design goal for the project. This ensures a consistent developer\n  // experience across both platforms.\n  //\n  // For more context, see the discussion in GitHub issue # 844.\n  protected async pumpInput() {\n    const reader = this.deferredInputStream.stream.getReader();\n    try {\n      while (true) {\n        const { done, value } = await reader.read();\n        if (done || value === SynthesizeStream.FLUSH_SENTINEL) {\n          break;\n        }\n        this.pushText(value);\n      }\n      this.endInput();\n    } catch (error) {\n      this.logger.error(error, 'Error reading deferred input stream');\n    } finally {\n      reader.releaseLock();\n      // Ensure output is closed when the stream ends\n      if (!this.#monitorMetricsTask) {\n        // No text was received, close the output directly\n        this.output.close();\n      }\n    }\n  }\n\n  protected async monitorMetrics() {\n    const startTime = process.hrtime.bigint();\n    let audioDurationMs = 0;\n    let ttfb: bigint = BigInt(-1);\n    let requestId = '';\n\n    const emit = () => {\n      if (this.#metricsPendingTexts.length) {\n        const text = this.#metricsPendingTexts.shift()!;\n        const duration = process.hrtime.bigint() - startTime;\n        const roundedAudioDurationMs = Math.round(audioDurationMs);\n        const metrics: TTSMetrics = {\n          type: 'tts_metrics',\n          timestamp: Date.now(),\n          requestId,\n          ttfbMs: ttfb === BigInt(-1) ? -1 : Math.trunc(Number(ttfb / BigInt(1000000))),\n          durationMs: Math.trunc(Number(duration / BigInt(1000000))),\n          charactersCount: text.length,\n          audioDurationMs: roundedAudioDurationMs,\n          cancelled: this.abortController.signal.aborted,\n          label: this.#tts.label,\n          streamed: false,\n        };\n        this.#tts.emit('metrics_collected', metrics);\n      }\n    };\n\n    for await (const audio of this.queue) {\n      if (this.abortController.signal.aborted) {\n        break;\n      }\n      this.output.put(audio);\n      if (audio === SynthesizeStream.END_OF_STREAM) continue;\n      requestId = audio.requestId;\n      if (ttfb === BigInt(-1)) {\n        ttfb = process.hrtime.bigint() - startTime;\n      }\n      // TODO(AJS-102): use frame.durationMs once available in rtc-node\n      audioDurationMs += (audio.frame.samplesPerChannel / audio.frame.sampleRate) * 1000;\n      if (audio.final) {\n        emit();\n      }\n    }\n\n    if (requestId) {\n      emit();\n    }\n  }\n\n  protected abstract run(): Promise<void>;\n\n  updateInputStream(text: ReadableStream<string>) {\n    this.deferredInputStream.setSource(text);\n  }\n\n  /** Push a string of text to the TTS */\n  /** @deprecated Use `updateInputStream` instead */\n  pushText(text: string) {\n    if (!this.#monitorMetricsTask) {\n      this.#monitorMetricsTask = this.monitorMetrics();\n      // Close output when metrics task completes\n      this.#monitorMetricsTask.finally(() => this.output.close());\n    }\n    this.#metricsText += text;\n\n    if (this.input.closed) {\n      throw new Error('Input is closed');\n    }\n    if (this.closed) {\n      throw new Error('Stream is closed');\n    }\n    this.input.put(text);\n  }\n\n  /** Flush the TTS, causing it to process all pending text */\n  flush() {\n    if (this.#metricsText) {\n      this.#metricsPendingTexts.push(this.#metricsText);\n      this.#metricsText = '';\n    }\n    if (this.input.closed) {\n      throw new Error('Input is closed');\n    }\n    if (this.closed) {\n      throw new Error('Stream is closed');\n    }\n    this.input.put(SynthesizeStream.FLUSH_SENTINEL);\n  }\n\n  /** Mark the input as ended and forbid additional pushes */\n  endInput() {\n    this.flush();\n    if (this.input.closed) {\n      throw new Error('Input is closed');\n    }\n    if (this.closed) {\n      throw new Error('Stream is closed');\n    }\n    this.input.close();\n  }\n\n  next(): Promise<IteratorResult<SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM>> {\n    return this.output.next();\n  }\n\n  /** Close both the input and output of the TTS stream */\n  close() {\n    this.abortController.abort();\n  }\n\n  [Symbol.asyncIterator](): SynthesizeStream {\n    return this;\n  }\n}\n\n/**\n * An instance of a text-to-speech response, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n *   await source.captureFrame(event.frame);\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child ChunkedStream class, which inherits this class's methods.\n */\nexport abstract class ChunkedStream implements AsyncIterableIterator<SynthesizedAudio> {\n  protected queue = new AsyncIterableQueue<SynthesizedAudio>();\n  protected output = new AsyncIterableQueue<SynthesizedAudio>();\n  protected closed = false;\n  abstract label: string;\n  #text: string;\n  #tts: TTS;\n  private _connOptions: APIConnectOptions;\n  private logger = log();\n\n  constructor(\n    text: string,\n    tts: TTS,\n    connOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,\n  ) {\n    this.#text = text;\n    this.#tts = tts;\n    this._connOptions = connOptions;\n\n    this.monitorMetrics();\n\n    // this is a hack to immitate asyncio.create_task so that mainTask\n    // is run **after** the constructor has finished. Otherwise we get\n    // runtime error when trying to access class variables in the\n    // `run` method.\n    Promise.resolve().then(() => this.mainTask().then(() => this.queue.close()));\n  }\n\n  private async mainTask() {\n    // TODO(brian): PR3 - Add span wrapping: tracer.startActiveSpan('tts_request', ..., { endOnExit: false })\n    for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n      try {\n        // TODO(brian): PR3 - Add span for retry attempts: tracer.startActiveSpan('tts_request_run', ...)\n        return await this.run();\n      } catch (error) {\n        if (error instanceof APIError) {\n          const retryInterval = this._connOptions._intervalForRetry(i);\n\n          if (this._connOptions.maxRetry === 0 || !error.retryable) {\n            this.emitError({ error, recoverable: false });\n            throw error;\n          } else if (i === this._connOptions.maxRetry) {\n            this.emitError({ error, recoverable: false });\n            throw new APIConnectionError({\n              message: `failed to generate TTS completion after ${this._connOptions.maxRetry + 1} attempts`,\n              options: { retryable: false },\n            });\n          } else {\n            // Don't emit error event for recoverable errors during retry loop\n            // to avoid ERR_UNHANDLED_ERROR or premature session termination\n            this.logger.warn(\n              { tts: this.#tts.label, attempt: i + 1, error },\n              `failed to generate TTS completion, retrying in ${retryInterval}s`,\n            );\n          }\n\n          if (retryInterval > 0) {\n            await delay(retryInterval);\n          }\n        } else {\n          this.emitError({ error: toError(error), recoverable: false });\n          throw error;\n        }\n      }\n    }\n  }\n\n  private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n    this.#tts.emit('error', {\n      type: 'tts_error',\n      timestamp: Date.now(),\n      label: this.#tts.label,\n      error,\n      recoverable,\n    });\n  }\n\n  protected abstract run(): Promise<void>;\n\n  get inputText(): string {\n    return this.#text;\n  }\n\n  protected async monitorMetrics() {\n    const startTime = process.hrtime.bigint();\n    let audioDurationMs = 0;\n    let ttfb: bigint = BigInt(-1);\n    let requestId = '';\n\n    for await (const audio of this.queue) {\n      this.output.put(audio);\n      requestId = audio.requestId;\n      if (ttfb === BigInt(-1)) {\n        ttfb = process.hrtime.bigint() - startTime;\n      }\n      audioDurationMs += (audio.frame.samplesPerChannel / audio.frame.sampleRate) * 1000;\n    }\n    this.output.close();\n\n    const duration = process.hrtime.bigint() - startTime;\n    const metrics: TTSMetrics = {\n      type: 'tts_metrics',\n      timestamp: Date.now(),\n      requestId,\n      ttfbMs: ttfb === BigInt(-1) ? -1 : Math.trunc(Number(ttfb / BigInt(1000000))),\n      durationMs: Math.trunc(Number(duration / BigInt(1000000))),\n      charactersCount: this.#text.length,\n      audioDurationMs: Math.round(audioDurationMs),\n      cancelled: false, // TODO(AJS-186): support ChunkedStream with 1.0 - add this.abortController.signal.aborted here\n      label: this.#tts.label,\n      streamed: false,\n    };\n    this.#tts.emit('metrics_collected', metrics);\n  }\n\n  /** Collect every frame into one in a single call */\n  async collect(): Promise<AudioFrame> {\n    const frames = [];\n    for await (const event of this) {\n      frames.push(event.frame);\n    }\n    return mergeFrames(frames);\n  }\n\n  next(): Promise<IteratorResult<SynthesizedAudio>> {\n    return this.output.next();\n  }\n\n  /** Close both the input and output of the TTS stream */\n  close() {\n    this.queue.close();\n    this.output.close();\n    this.closed = true;\n  }\n\n  [Symbol.asyncIterator](): ChunkedStream {\n    return this;\n  }\n}\n"],"mappings":"AAKA,SAAS,oBAAoB;AAE7B,SAAS,oBAAoB,gBAAgB;AAC7C,SAAS,WAAW;AAEpB,SAAS,8BAA8B;AACvC,SAAiC,mCAAmC;AACpE,SAAS,oBAAoB,OAAO,aAAa,WAAW,eAAe;AA+CpE,MAAe,YAAa,aAAsD;AAAA,EACvF;AAAA,EACA;AAAA,EACA;AAAA,EAGA,YAAY,YAAoB,aAAqB,cAA+B;AAClF,UAAM;AACN,SAAK,gBAAgB;AACrB,SAAK,cAAc;AACnB,SAAK,eAAe;AAAA,EACtB;AAAA;AAAA,EAGA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,aAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,cAAsB;AACxB,WAAO,KAAK;AAAA,EACd;AAAA,EAYA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAgBO,MAAe,iBAEtB;AAAA,EACE,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EAClE,OAAgB,gBAAgB,OAAO,eAAe;AAAA,EAC5C,QAAQ,IAAI,mBAAoE;AAAA,EAChF,QAAQ,IAAI,mBAEpB;AAAA,EACQ,SAAS,IAAI,mBAErB;AAAA,EACQ,SAAS;AAAA,EAEnB;AAAA,EACA,uBAAiC,CAAC;AAAA,EAClC,eAAe;AAAA,EACf;AAAA,EACQ;AAAA,EACE,kBAAkB,IAAI,gBAAgB;AAAA,EAExC;AAAA,EAGA,SAAS,IAAI;AAAA,EAErB,YAAY,KAAU,cAAiC,6BAA6B;AAClF,SAAK,OAAO;AACZ,SAAK,eAAe;AACpB,SAAK,sBAAsB,IAAI,uBAAuB;AACtD,SAAK,UAAU;AACf,SAAK,gBAAgB,OAAO,iBAAiB,SAAS,MAAM;AAC1D,WAAK,oBAAoB,aAAa;AAEtC,WAAK,MAAM,MAAM;AACjB,WAAK,OAAO,MAAM;AAClB,WAAK,SAAS;AAAA,IAChB,CAAC;AAMD,cAAU,MAAM,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,WAAW;AAEvB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AAEF,eAAO,MAAM,KAAK,IAAI;AAAA,MACxB,SAAS,OAAO;AACd,YAAI,iBAAiB,UAAU;AAC7B,gBAAM,gBAAgB,KAAK,aAAa,kBAAkB,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,mBAAmB;AAAA,cAC3B,SAAS,2CAA2C,KAAK,aAAa,WAAW,CAAC;AAAA,cAClF,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,6CAA6C,aAAa;AAAA,YAC5D;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,kBAAM,MAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,OAAO,QAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,MAAgB,YAAY;AAC1B,UAAM,SAAS,KAAK,oBAAoB,OAAO,UAAU;AACzD,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,QAAQ,UAAU,iBAAiB,gBAAgB;AACrD;AAAA,QACF;AACA,aAAK,SAAS,KAAK;AAAA,MACrB;AACA,WAAK,SAAS;AAAA,IAChB,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,OAAO,qCAAqC;AAAA,IAChE,UAAE;AACA,aAAO,YAAY;AAEnB,UAAI,CAAC,KAAK,qBAAqB;AAE7B,aAAK,OAAO,MAAM;AAAA,MACpB;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,kBAAkB;AACtB,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAEhB,UAAM,OAAO,MAAM;AACjB,UAAI,KAAK,qBAAqB,QAAQ;AACpC,cAAM,OAAO,KAAK,qBAAqB,MAAM;AAC7C,cAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,cAAM,yBAAyB,KAAK,MAAM,eAAe;AACzD,cAAM,UAAsB;AAAA,UAC1B,MAAM;AAAA,UACN,WAAW,KAAK,IAAI;AAAA,UACpB;AAAA,UACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,UAC5E,YAAY,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAAA,UACzD,iBAAiB,KAAK;AAAA,UACtB,iBAAiB;AAAA,UACjB,WAAW,KAAK,gBAAgB,OAAO;AAAA,UACvC,OAAO,KAAK,KAAK;AAAA,UACjB,UAAU;AAAA,QACZ;AACA,aAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,MAC7C;AAAA,IACF;AAEA,qBAAiB,SAAS,KAAK,OAAO;AACpC,UAAI,KAAK,gBAAgB,OAAO,SAAS;AACvC;AAAA,MACF;AACA,WAAK,OAAO,IAAI,KAAK;AACrB,UAAI,UAAU,iBAAiB,cAAe;AAC9C,kBAAY,MAAM;AAClB,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AAAA,MACnC;AAEA,yBAAoB,MAAM,MAAM,oBAAoB,MAAM,MAAM,aAAc;AAC9E,UAAI,MAAM,OAAO;AACf,aAAK;AAAA,MACP;AAAA,IACF;AAEA,QAAI,WAAW;AACb,WAAK;AAAA,IACP;AAAA,EACF;AAAA,EAIA,kBAAkB,MAA8B;AAC9C,SAAK,oBAAoB,UAAU,IAAI;AAAA,EACzC;AAAA;AAAA;AAAA,EAIA,SAAS,MAAc;AACrB,QAAI,CAAC,KAAK,qBAAqB;AAC7B,WAAK,sBAAsB,KAAK,eAAe;AAE/C,WAAK,oBAAoB,QAAQ,MAAM,KAAK,OAAO,MAAM,CAAC;AAAA,IAC5D;AACA,SAAK,gBAAgB;AAErB,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,IAAI,IAAI;AAAA,EACrB;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,cAAc;AACrB,WAAK,qBAAqB,KAAK,KAAK,YAAY;AAChD,WAAK,eAAe;AAAA,IACtB;AACA,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,IAAI,iBAAiB,cAAc;AAAA,EAChD;AAAA;AAAA,EAGA,WAAW;AACT,SAAK,MAAM;AACX,QAAI,KAAK,MAAM,QAAQ;AACrB,YAAM,IAAI,MAAM,iBAAiB;AAAA,IACnC;AACA,QAAI,KAAK,QAAQ;AACf,YAAM,IAAI,MAAM,kBAAkB;AAAA,IACpC;AACA,SAAK,MAAM,MAAM;AAAA,EACnB;AAAA,EAEA,OAA0F;AACxF,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AAAA,EAEA,CAAC,OAAO,aAAa,IAAsB;AACzC,WAAO;AAAA,EACT;AACF;AAgBO,MAAe,cAAiE;AAAA,EAC3E,QAAQ,IAAI,mBAAqC;AAAA,EACjD,SAAS,IAAI,mBAAqC;AAAA,EAClD,SAAS;AAAA,EAEnB;AAAA,EACA;AAAA,EACQ;AAAA,EACA,SAAS,IAAI;AAAA,EAErB,YACE,MACA,KACA,cAAiC,6BACjC;AACA,SAAK,QAAQ;AACb,SAAK,OAAO;AACZ,SAAK,eAAe;AAEpB,SAAK,eAAe;AAMpB,YAAQ,QAAQ,EAAE,KAAK,MAAM,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAC7E;AAAA,EAEA,MAAc,WAAW;AAEvB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AAEF,eAAO,MAAM,KAAK,IAAI;AAAA,MACxB,SAAS,OAAO;AACd,YAAI,iBAAiB,UAAU;AAC7B,gBAAM,gBAAgB,KAAK,aAAa,kBAAkB,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,mBAAmB;AAAA,cAC3B,SAAS,2CAA2C,KAAK,aAAa,WAAW,CAAC;AAAA,cAClF,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,kDAAkD,aAAa;AAAA,YACjE;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,kBAAM,MAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,OAAO,QAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAIA,IAAI,YAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,kBAAkB;AACtB,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAEhB,qBAAiB,SAAS,KAAK,OAAO;AACpC,WAAK,OAAO,IAAI,KAAK;AACrB,kBAAY,MAAM;AAClB,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AAAA,MACnC;AACA,yBAAoB,MAAM,MAAM,oBAAoB,MAAM,MAAM,aAAc;AAAA,IAChF;AACA,SAAK,OAAO,MAAM;AAElB,UAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,UAAM,UAAsB;AAAA,MAC1B,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,MAC5E,YAAY,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAAA,MACzD,iBAAiB,KAAK,MAAM;AAAA,MAC5B,iBAAiB,KAAK,MAAM,eAAe;AAAA,MAC3C,WAAW;AAAA;AAAA,MACX,OAAO,KAAK,KAAK;AAAA,MACjB,UAAU;AAAA,IACZ;AACA,SAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,EAC7C;AAAA;AAAA,EAGA,MAAM,UAA+B;AACnC,UAAM,SAAS,CAAC;AAChB,qBAAiB,SAAS,MAAM;AAC9B,aAAO,KAAK,MAAM,KAAK;AAAA,IACzB;AACA,WAAO,YAAY,MAAM;AAAA,EAC3B;AAAA,EAEA,OAAkD;AAChD,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,SAAK,MAAM,MAAM;AACjB,SAAK,OAAO,MAAM;AAClB,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAmB;AACtC,WAAO;AAAA,EACT;AACF;","names":[]}
+{"version":3,"sources":["../../src/tts/tts.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 { ReadableStream } from 'node:stream/web';\nimport { APIConnectionError, APIError } from '../_exceptions.js';\nimport { log } from '../log.js';\nimport type { TTSMetrics } from '../metrics/base.js';\nimport { DeferredReadableStream } from '../stream/deferred_stream.js';\nimport { type APIConnectOptions, DEFAULT_API_CONNECT_OPTIONS } from '../types.js';\nimport { AsyncIterableQueue, delay, mergeFrames, startSoon, toError } from '../utils.js';\n\n/** SynthesizedAudio is a packet of speech synthesis as returned by the TTS. */\nexport interface SynthesizedAudio {\n  /** Request ID (one segment could be made up of multiple requests) */\n  requestId: string;\n  /** Segment ID, each segment is separated by a flush */\n  segmentId: string;\n  /** Synthesized audio frame */\n  frame: AudioFrame;\n  /** Current segment of the synthesized audio */\n  deltaText?: string;\n  /** Whether this is the last frame of the segment (streaming only) */\n  final: boolean;\n}\n\n/**\n * Describes the capabilities of the TTS provider.\n *\n * @remarks\n * At present, only `streaming` is supplied to this interface, and the framework only supports\n * providers that do have a streaming endpoint.\n */\nexport interface TTSCapabilities {\n  streaming: boolean;\n}\n\nexport interface TTSError {\n  type: 'tts_error';\n  timestamp: number;\n  label: string;\n  error: Error;\n  recoverable: boolean;\n}\n\nexport type TTSCallbacks = {\n  ['metrics_collected']: (metrics: TTSMetrics) => void;\n  ['error']: (error: TTSError) => void;\n};\n\n/**\n * An instance of a text-to-speech adapter.\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child TTS class, which inherits this class's methods.\n */\nexport abstract class TTS extends (EventEmitter as new () => TypedEmitter<TTSCallbacks>) {\n  #capabilities: TTSCapabilities;\n  #sampleRate: number;\n  #numChannels: number;\n  abstract label: string;\n\n  constructor(sampleRate: number, numChannels: number, capabilities: TTSCapabilities) {\n    super();\n    this.#capabilities = capabilities;\n    this.#sampleRate = sampleRate;\n    this.#numChannels = numChannels;\n  }\n\n  /** Returns this TTS's capabilities */\n  get capabilities(): TTSCapabilities {\n    return this.#capabilities;\n  }\n\n  /** Returns the sample rate of audio frames returned by this TTS */\n  get sampleRate(): number {\n    return this.#sampleRate;\n  }\n\n  /** Returns the channel count of audio frames returned by this TTS */\n  get numChannels(): number {\n    return this.#numChannels;\n  }\n\n  /**\n   * Receives text and returns synthesis in the form of a {@link ChunkedStream}\n   */\n  abstract synthesize(text: string): ChunkedStream;\n\n  /**\n   * Returns a {@link SynthesizeStream} that can be used to push text and receive audio data\n   */\n  abstract stream(): SynthesizeStream;\n\n  async close(): Promise<void> {\n    return;\n  }\n}\n\n/**\n * An instance of a text-to-speech stream, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n *   await source.captureFrame(event.frame);\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child SynthesizeStream class, which inherits this class's methods.\n */\nexport abstract class SynthesizeStream\n  implements AsyncIterableIterator<SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM>\n{\n  protected static readonly FLUSH_SENTINEL = Symbol('FLUSH_SENTINEL');\n  static readonly END_OF_STREAM = Symbol('END_OF_STREAM');\n  protected input = new AsyncIterableQueue<string | typeof SynthesizeStream.FLUSH_SENTINEL>();\n  protected queue = new AsyncIterableQueue<\n    SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM\n  >();\n  protected output = new AsyncIterableQueue<\n    SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM\n  >();\n  protected closed = false;\n  abstract label: string;\n  #tts: TTS;\n  #metricsPendingTexts: string[] = [];\n  #metricsText = '';\n  #monitorMetricsTask?: Promise<void>;\n  private _connOptions: APIConnectOptions;\n  protected abortController = new AbortController();\n\n  private deferredInputStream: DeferredReadableStream<\n    string | typeof SynthesizeStream.FLUSH_SENTINEL\n  >;\n  private logger = log();\n\n  constructor(tts: TTS, connOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS) {\n    this.#tts = tts;\n    this._connOptions = connOptions;\n    this.deferredInputStream = new DeferredReadableStream();\n    this.pumpInput();\n    this.abortController.signal.addEventListener('abort', () => {\n      this.deferredInputStream.detachSource();\n      // TODO (AJS-36) clean this up when we refactor with streams\n      this.input.close();\n      this.output.close();\n      this.closed = true;\n    });\n\n    // this is a hack to immitate asyncio.create_task so that mainTask\n    // is run **after** the constructor has finished. Otherwise we get\n    // runtime error when trying to access class variables in the\n    // `run` method.\n    startSoon(() => this.mainTask().then(() => this.queue.close()));\n  }\n\n  private async mainTask() {\n    // TODO(brian): PR3 - Add span wrapping: tracer.startActiveSpan('tts_request', ..., { endOnExit: false })\n    for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n      try {\n        // TODO(brian): PR3 - Add span for retry attempts: tracer.startActiveSpan('tts_request_run', ...)\n        return await this.run();\n      } catch (error) {\n        if (error instanceof APIError) {\n          const retryInterval = this._connOptions._intervalForRetry(i);\n\n          if (this._connOptions.maxRetry === 0 || !error.retryable) {\n            this.emitError({ error, recoverable: false });\n            throw error;\n          } else if (i === this._connOptions.maxRetry) {\n            this.emitError({ error, recoverable: false });\n            throw new APIConnectionError({\n              message: `failed to generate TTS completion after ${this._connOptions.maxRetry + 1} attempts`,\n              options: { retryable: false },\n            });\n          } else {\n            // Don't emit error event for recoverable errors during retry loop\n            // to avoid ERR_UNHANDLED_ERROR or premature session termination\n            this.logger.warn(\n              { tts: this.#tts.label, attempt: i + 1, error },\n              `failed to synthesize speech, retrying in  ${retryInterval}s`,\n            );\n          }\n\n          if (retryInterval > 0) {\n            await delay(retryInterval);\n          }\n        } else {\n          this.emitError({ error: toError(error), recoverable: false });\n          throw error;\n        }\n      }\n    }\n  }\n\n  private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n    this.#tts.emit('error', {\n      type: 'tts_error',\n      timestamp: Date.now(),\n      label: this.#tts.label,\n      error,\n      recoverable,\n    });\n  }\n\n  // NOTE(AJS-37): The implementation below uses an AsyncIterableQueue (`this.input`)\n  // bridged from a DeferredReadableStream (`this.deferredInputStream`) rather than\n  // consuming the stream directly.\n  //\n  // A full refactor to native Web Streams was considered but is currently deferred.\n  // The primary reason is to maintain architectural parity with the Python SDK,\n  // which is a key design goal for the project. This ensures a consistent developer\n  // experience across both platforms.\n  //\n  // For more context, see the discussion in GitHub issue # 844.\n  protected async pumpInput() {\n    const reader = this.deferredInputStream.stream.getReader();\n    try {\n      while (true) {\n        const { done, value } = await reader.read();\n        if (done || value === SynthesizeStream.FLUSH_SENTINEL) {\n          break;\n        }\n        this.pushText(value);\n      }\n      this.endInput();\n    } catch (error) {\n      this.logger.error(error, 'Error reading deferred input stream');\n    } finally {\n      reader.releaseLock();\n      // Ensure output is closed when the stream ends\n      if (!this.#monitorMetricsTask) {\n        // No text was received, close the output directly\n        this.output.close();\n      }\n    }\n  }\n\n  protected async monitorMetrics() {\n    const startTime = process.hrtime.bigint();\n    let audioDurationMs = 0;\n    let ttfb: bigint = BigInt(-1);\n    let requestId = '';\n\n    const emit = () => {\n      if (this.#metricsPendingTexts.length) {\n        const text = this.#metricsPendingTexts.shift()!;\n        const duration = process.hrtime.bigint() - startTime;\n        const roundedAudioDurationMs = Math.round(audioDurationMs);\n        const metrics: TTSMetrics = {\n          type: 'tts_metrics',\n          timestamp: Date.now(),\n          requestId,\n          ttfbMs: ttfb === BigInt(-1) ? -1 : Math.trunc(Number(ttfb / BigInt(1000000))),\n          durationMs: Math.trunc(Number(duration / BigInt(1000000))),\n          charactersCount: text.length,\n          audioDurationMs: roundedAudioDurationMs,\n          cancelled: this.abortController.signal.aborted,\n          label: this.#tts.label,\n          streamed: false,\n        };\n        this.#tts.emit('metrics_collected', metrics);\n      }\n    };\n\n    for await (const audio of this.queue) {\n      if (this.abortController.signal.aborted) {\n        break;\n      }\n      this.output.put(audio);\n      if (audio === SynthesizeStream.END_OF_STREAM) continue;\n      requestId = audio.requestId;\n      if (ttfb === BigInt(-1)) {\n        ttfb = process.hrtime.bigint() - startTime;\n      }\n      // TODO(AJS-102): use frame.durationMs once available in rtc-node\n      audioDurationMs += (audio.frame.samplesPerChannel / audio.frame.sampleRate) * 1000;\n      if (audio.final) {\n        emit();\n      }\n    }\n\n    if (requestId) {\n      emit();\n    }\n  }\n\n  protected abstract run(): Promise<void>;\n\n  updateInputStream(text: ReadableStream<string>) {\n    this.deferredInputStream.setSource(text);\n  }\n\n  /** Push a string of text to the TTS */\n  /** @deprecated Use `updateInputStream` instead */\n  pushText(text: string) {\n    if (!this.#monitorMetricsTask) {\n      this.#monitorMetricsTask = this.monitorMetrics();\n      // Close output when metrics task completes\n      this.#monitorMetricsTask.finally(() => this.output.close());\n    }\n    this.#metricsText += text;\n\n    if (this.input.closed || this.closed) {\n      // Stream was aborted/closed, silently skip\n      return;\n    }\n\n    this.input.put(text);\n  }\n\n  /** Flush the TTS, causing it to process all pending text */\n  flush() {\n    if (this.#metricsText) {\n      this.#metricsPendingTexts.push(this.#metricsText);\n      this.#metricsText = '';\n    }\n\n    if (this.input.closed || this.closed) {\n      // Stream was aborted/closed, silently skip\n      return;\n    }\n\n    this.input.put(SynthesizeStream.FLUSH_SENTINEL);\n  }\n\n  /** Mark the input as ended and forbid additional pushes */\n  endInput() {\n    this.flush();\n\n    if (this.input.closed || this.closed) {\n      // Stream was aborted/closed, silently skip\n      return;\n    }\n\n    this.input.close();\n  }\n\n  next(): Promise<IteratorResult<SynthesizedAudio | typeof SynthesizeStream.END_OF_STREAM>> {\n    return this.output.next();\n  }\n\n  /** Close both the input and output of the TTS stream */\n  close() {\n    this.abortController.abort();\n  }\n\n  [Symbol.asyncIterator](): SynthesizeStream {\n    return this;\n  }\n}\n\n/**\n * An instance of a text-to-speech response, as an asynchronous iterable iterator.\n *\n * @example Looping through frames\n * ```ts\n * for await (const event of stream) {\n *   await source.captureFrame(event.frame);\n * }\n * ```\n *\n * @remarks\n * This class is abstract, and as such cannot be used directly. Instead, use a provider plugin that\n * exports its own child ChunkedStream class, which inherits this class's methods.\n */\nexport abstract class ChunkedStream implements AsyncIterableIterator<SynthesizedAudio> {\n  protected queue = new AsyncIterableQueue<SynthesizedAudio>();\n  protected output = new AsyncIterableQueue<SynthesizedAudio>();\n  protected closed = false;\n  abstract label: string;\n  #text: string;\n  #tts: TTS;\n  private _connOptions: APIConnectOptions;\n  private logger = log();\n\n  constructor(\n    text: string,\n    tts: TTS,\n    connOptions: APIConnectOptions = DEFAULT_API_CONNECT_OPTIONS,\n  ) {\n    this.#text = text;\n    this.#tts = tts;\n    this._connOptions = connOptions;\n\n    this.monitorMetrics();\n\n    // this is a hack to immitate asyncio.create_task so that mainTask\n    // is run **after** the constructor has finished. Otherwise we get\n    // runtime error when trying to access class variables in the\n    // `run` method.\n    Promise.resolve().then(() => this.mainTask().then(() => this.queue.close()));\n  }\n\n  private async mainTask() {\n    // TODO(brian): PR3 - Add span wrapping: tracer.startActiveSpan('tts_request', ..., { endOnExit: false })\n    for (let i = 0; i < this._connOptions.maxRetry + 1; i++) {\n      try {\n        // TODO(brian): PR3 - Add span for retry attempts: tracer.startActiveSpan('tts_request_run', ...)\n        return await this.run();\n      } catch (error) {\n        if (error instanceof APIError) {\n          const retryInterval = this._connOptions._intervalForRetry(i);\n\n          if (this._connOptions.maxRetry === 0 || !error.retryable) {\n            this.emitError({ error, recoverable: false });\n            throw error;\n          } else if (i === this._connOptions.maxRetry) {\n            this.emitError({ error, recoverable: false });\n            throw new APIConnectionError({\n              message: `failed to generate TTS completion after ${this._connOptions.maxRetry + 1} attempts`,\n              options: { retryable: false },\n            });\n          } else {\n            // Don't emit error event for recoverable errors during retry loop\n            // to avoid ERR_UNHANDLED_ERROR or premature session termination\n            this.logger.warn(\n              { tts: this.#tts.label, attempt: i + 1, error },\n              `failed to generate TTS completion, retrying in ${retryInterval}s`,\n            );\n          }\n\n          if (retryInterval > 0) {\n            await delay(retryInterval);\n          }\n        } else {\n          this.emitError({ error: toError(error), recoverable: false });\n          throw error;\n        }\n      }\n    }\n  }\n\n  private emitError({ error, recoverable }: { error: Error; recoverable: boolean }) {\n    this.#tts.emit('error', {\n      type: 'tts_error',\n      timestamp: Date.now(),\n      label: this.#tts.label,\n      error,\n      recoverable,\n    });\n  }\n\n  protected abstract run(): Promise<void>;\n\n  get inputText(): string {\n    return this.#text;\n  }\n\n  protected async monitorMetrics() {\n    const startTime = process.hrtime.bigint();\n    let audioDurationMs = 0;\n    let ttfb: bigint = BigInt(-1);\n    let requestId = '';\n\n    for await (const audio of this.queue) {\n      this.output.put(audio);\n      requestId = audio.requestId;\n      if (ttfb === BigInt(-1)) {\n        ttfb = process.hrtime.bigint() - startTime;\n      }\n      audioDurationMs += (audio.frame.samplesPerChannel / audio.frame.sampleRate) * 1000;\n    }\n    this.output.close();\n\n    const duration = process.hrtime.bigint() - startTime;\n    const metrics: TTSMetrics = {\n      type: 'tts_metrics',\n      timestamp: Date.now(),\n      requestId,\n      ttfbMs: ttfb === BigInt(-1) ? -1 : Math.trunc(Number(ttfb / BigInt(1000000))),\n      durationMs: Math.trunc(Number(duration / BigInt(1000000))),\n      charactersCount: this.#text.length,\n      audioDurationMs: Math.round(audioDurationMs),\n      cancelled: false, // TODO(AJS-186): support ChunkedStream with 1.0 - add this.abortController.signal.aborted here\n      label: this.#tts.label,\n      streamed: false,\n    };\n    this.#tts.emit('metrics_collected', metrics);\n  }\n\n  /** Collect every frame into one in a single call */\n  async collect(): Promise<AudioFrame> {\n    const frames = [];\n    for await (const event of this) {\n      frames.push(event.frame);\n    }\n    return mergeFrames(frames);\n  }\n\n  next(): Promise<IteratorResult<SynthesizedAudio>> {\n    return this.output.next();\n  }\n\n  /** Close both the input and output of the TTS stream */\n  close() {\n    this.queue.close();\n    this.output.close();\n    this.closed = true;\n  }\n\n  [Symbol.asyncIterator](): ChunkedStream {\n    return this;\n  }\n}\n"],"mappings":"AAKA,SAAS,oBAAoB;AAE7B,SAAS,oBAAoB,gBAAgB;AAC7C,SAAS,WAAW;AAEpB,SAAS,8BAA8B;AACvC,SAAiC,mCAAmC;AACpE,SAAS,oBAAoB,OAAO,aAAa,WAAW,eAAe;AA+CpE,MAAe,YAAa,aAAsD;AAAA,EACvF;AAAA,EACA;AAAA,EACA;AAAA,EAGA,YAAY,YAAoB,aAAqB,cAA+B;AAClF,UAAM;AACN,SAAK,gBAAgB;AACrB,SAAK,cAAc;AACnB,SAAK,eAAe;AAAA,EACtB;AAAA;AAAA,EAGA,IAAI,eAAgC;AAClC,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,aAAqB;AACvB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,cAAsB;AACxB,WAAO,KAAK;AAAA,EACd;AAAA,EAYA,MAAM,QAAuB;AAC3B;AAAA,EACF;AACF;AAgBO,MAAe,iBAEtB;AAAA,EACE,OAA0B,iBAAiB,OAAO,gBAAgB;AAAA,EAClE,OAAgB,gBAAgB,OAAO,eAAe;AAAA,EAC5C,QAAQ,IAAI,mBAAoE;AAAA,EAChF,QAAQ,IAAI,mBAEpB;AAAA,EACQ,SAAS,IAAI,mBAErB;AAAA,EACQ,SAAS;AAAA,EAEnB;AAAA,EACA,uBAAiC,CAAC;AAAA,EAClC,eAAe;AAAA,EACf;AAAA,EACQ;AAAA,EACE,kBAAkB,IAAI,gBAAgB;AAAA,EAExC;AAAA,EAGA,SAAS,IAAI;AAAA,EAErB,YAAY,KAAU,cAAiC,6BAA6B;AAClF,SAAK,OAAO;AACZ,SAAK,eAAe;AACpB,SAAK,sBAAsB,IAAI,uBAAuB;AACtD,SAAK,UAAU;AACf,SAAK,gBAAgB,OAAO,iBAAiB,SAAS,MAAM;AAC1D,WAAK,oBAAoB,aAAa;AAEtC,WAAK,MAAM,MAAM;AACjB,WAAK,OAAO,MAAM;AAClB,WAAK,SAAS;AAAA,IAChB,CAAC;AAMD,cAAU,MAAM,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,WAAW;AAEvB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AAEF,eAAO,MAAM,KAAK,IAAI;AAAA,MACxB,SAAS,OAAO;AACd,YAAI,iBAAiB,UAAU;AAC7B,gBAAM,gBAAgB,KAAK,aAAa,kBAAkB,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,mBAAmB;AAAA,cAC3B,SAAS,2CAA2C,KAAK,aAAa,WAAW,CAAC;AAAA,cAClF,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,6CAA6C,aAAa;AAAA,YAC5D;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,kBAAM,MAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,OAAO,QAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYA,MAAgB,YAAY;AAC1B,UAAM,SAAS,KAAK,oBAAoB,OAAO,UAAU;AACzD,QAAI;AACF,aAAO,MAAM;AACX,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,QAAQ,UAAU,iBAAiB,gBAAgB;AACrD;AAAA,QACF;AACA,aAAK,SAAS,KAAK;AAAA,MACrB;AACA,WAAK,SAAS;AAAA,IAChB,SAAS,OAAO;AACd,WAAK,OAAO,MAAM,OAAO,qCAAqC;AAAA,IAChE,UAAE;AACA,aAAO,YAAY;AAEnB,UAAI,CAAC,KAAK,qBAAqB;AAE7B,aAAK,OAAO,MAAM;AAAA,MACpB;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,kBAAkB;AACtB,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAEhB,UAAM,OAAO,MAAM;AACjB,UAAI,KAAK,qBAAqB,QAAQ;AACpC,cAAM,OAAO,KAAK,qBAAqB,MAAM;AAC7C,cAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,cAAM,yBAAyB,KAAK,MAAM,eAAe;AACzD,cAAM,UAAsB;AAAA,UAC1B,MAAM;AAAA,UACN,WAAW,KAAK,IAAI;AAAA,UACpB;AAAA,UACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,UAC5E,YAAY,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAAA,UACzD,iBAAiB,KAAK;AAAA,UACtB,iBAAiB;AAAA,UACjB,WAAW,KAAK,gBAAgB,OAAO;AAAA,UACvC,OAAO,KAAK,KAAK;AAAA,UACjB,UAAU;AAAA,QACZ;AACA,aAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,MAC7C;AAAA,IACF;AAEA,qBAAiB,SAAS,KAAK,OAAO;AACpC,UAAI,KAAK,gBAAgB,OAAO,SAAS;AACvC;AAAA,MACF;AACA,WAAK,OAAO,IAAI,KAAK;AACrB,UAAI,UAAU,iBAAiB,cAAe;AAC9C,kBAAY,MAAM;AAClB,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AAAA,MACnC;AAEA,yBAAoB,MAAM,MAAM,oBAAoB,MAAM,MAAM,aAAc;AAC9E,UAAI,MAAM,OAAO;AACf,aAAK;AAAA,MACP;AAAA,IACF;AAEA,QAAI,WAAW;AACb,WAAK;AAAA,IACP;AAAA,EACF;AAAA,EAIA,kBAAkB,MAA8B;AAC9C,SAAK,oBAAoB,UAAU,IAAI;AAAA,EACzC;AAAA;AAAA;AAAA,EAIA,SAAS,MAAc;AACrB,QAAI,CAAC,KAAK,qBAAqB;AAC7B,WAAK,sBAAsB,KAAK,eAAe;AAE/C,WAAK,oBAAoB,QAAQ,MAAM,KAAK,OAAO,MAAM,CAAC;AAAA,IAC5D;AACA,SAAK,gBAAgB;AAErB,QAAI,KAAK,MAAM,UAAU,KAAK,QAAQ;AAEpC;AAAA,IACF;AAEA,SAAK,MAAM,IAAI,IAAI;AAAA,EACrB;AAAA;AAAA,EAGA,QAAQ;AACN,QAAI,KAAK,cAAc;AACrB,WAAK,qBAAqB,KAAK,KAAK,YAAY;AAChD,WAAK,eAAe;AAAA,IACtB;AAEA,QAAI,KAAK,MAAM,UAAU,KAAK,QAAQ;AAEpC;AAAA,IACF;AAEA,SAAK,MAAM,IAAI,iBAAiB,cAAc;AAAA,EAChD;AAAA;AAAA,EAGA,WAAW;AACT,SAAK,MAAM;AAEX,QAAI,KAAK,MAAM,UAAU,KAAK,QAAQ;AAEpC;AAAA,IACF;AAEA,SAAK,MAAM,MAAM;AAAA,EACnB;AAAA,EAEA,OAA0F;AACxF,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,SAAK,gBAAgB,MAAM;AAAA,EAC7B;AAAA,EAEA,CAAC,OAAO,aAAa,IAAsB;AACzC,WAAO;AAAA,EACT;AACF;AAgBO,MAAe,cAAiE;AAAA,EAC3E,QAAQ,IAAI,mBAAqC;AAAA,EACjD,SAAS,IAAI,mBAAqC;AAAA,EAClD,SAAS;AAAA,EAEnB;AAAA,EACA;AAAA,EACQ;AAAA,EACA,SAAS,IAAI;AAAA,EAErB,YACE,MACA,KACA,cAAiC,6BACjC;AACA,SAAK,QAAQ;AACb,SAAK,OAAO;AACZ,SAAK,eAAe;AAEpB,SAAK,eAAe;AAMpB,YAAQ,QAAQ,EAAE,KAAK,MAAM,KAAK,SAAS,EAAE,KAAK,MAAM,KAAK,MAAM,MAAM,CAAC,CAAC;AAAA,EAC7E;AAAA,EAEA,MAAc,WAAW;AAEvB,aAAS,IAAI,GAAG,IAAI,KAAK,aAAa,WAAW,GAAG,KAAK;AACvD,UAAI;AAEF,eAAO,MAAM,KAAK,IAAI;AAAA,MACxB,SAAS,OAAO;AACd,YAAI,iBAAiB,UAAU;AAC7B,gBAAM,gBAAgB,KAAK,aAAa,kBAAkB,CAAC;AAE3D,cAAI,KAAK,aAAa,aAAa,KAAK,CAAC,MAAM,WAAW;AACxD,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM;AAAA,UACR,WAAW,MAAM,KAAK,aAAa,UAAU;AAC3C,iBAAK,UAAU,EAAE,OAAO,aAAa,MAAM,CAAC;AAC5C,kBAAM,IAAI,mBAAmB;AAAA,cAC3B,SAAS,2CAA2C,KAAK,aAAa,WAAW,CAAC;AAAA,cAClF,SAAS,EAAE,WAAW,MAAM;AAAA,YAC9B,CAAC;AAAA,UACH,OAAO;AAGL,iBAAK,OAAO;AAAA,cACV,EAAE,KAAK,KAAK,KAAK,OAAO,SAAS,IAAI,GAAG,MAAM;AAAA,cAC9C,kDAAkD,aAAa;AAAA,YACjE;AAAA,UACF;AAEA,cAAI,gBAAgB,GAAG;AACrB,kBAAM,MAAM,aAAa;AAAA,UAC3B;AAAA,QACF,OAAO;AACL,eAAK,UAAU,EAAE,OAAO,QAAQ,KAAK,GAAG,aAAa,MAAM,CAAC;AAC5D,gBAAM;AAAA,QACR;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEQ,UAAU,EAAE,OAAO,YAAY,GAA2C;AAChF,SAAK,KAAK,KAAK,SAAS;AAAA,MACtB,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB,OAAO,KAAK,KAAK;AAAA,MACjB;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA,EAIA,IAAI,YAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAgB,iBAAiB;AAC/B,UAAM,YAAY,QAAQ,OAAO,OAAO;AACxC,QAAI,kBAAkB;AACtB,QAAI,OAAe,OAAO,EAAE;AAC5B,QAAI,YAAY;AAEhB,qBAAiB,SAAS,KAAK,OAAO;AACpC,WAAK,OAAO,IAAI,KAAK;AACrB,kBAAY,MAAM;AAClB,UAAI,SAAS,OAAO,EAAE,GAAG;AACvB,eAAO,QAAQ,OAAO,OAAO,IAAI;AAAA,MACnC;AACA,yBAAoB,MAAM,MAAM,oBAAoB,MAAM,MAAM,aAAc;AAAA,IAChF;AACA,SAAK,OAAO,MAAM;AAElB,UAAM,WAAW,QAAQ,OAAO,OAAO,IAAI;AAC3C,UAAM,UAAsB;AAAA,MAC1B,MAAM;AAAA,MACN,WAAW,KAAK,IAAI;AAAA,MACpB;AAAA,MACA,QAAQ,SAAS,OAAO,EAAE,IAAI,KAAK,KAAK,MAAM,OAAO,OAAO,OAAO,GAAO,CAAC,CAAC;AAAA,MAC5E,YAAY,KAAK,MAAM,OAAO,WAAW,OAAO,GAAO,CAAC,CAAC;AAAA,MACzD,iBAAiB,KAAK,MAAM;AAAA,MAC5B,iBAAiB,KAAK,MAAM,eAAe;AAAA,MAC3C,WAAW;AAAA;AAAA,MACX,OAAO,KAAK,KAAK;AAAA,MACjB,UAAU;AAAA,IACZ;AACA,SAAK,KAAK,KAAK,qBAAqB,OAAO;AAAA,EAC7C;AAAA;AAAA,EAGA,MAAM,UAA+B;AACnC,UAAM,SAAS,CAAC;AAChB,qBAAiB,SAAS,MAAM;AAC9B,aAAO,KAAK,MAAM,KAAK;AAAA,IACzB;AACA,WAAO,YAAY,MAAM;AAAA,EAC3B;AAAA,EAEA,OAAkD;AAChD,WAAO,KAAK,OAAO,KAAK;AAAA,EAC1B;AAAA;AAAA,EAGA,QAAQ;AACN,SAAK,MAAM,MAAM;AACjB,SAAK,OAAO,MAAM;AAClB,SAAK,SAAS;AAAA,EAChB;AAAA,EAEA,CAAC,OAAO,aAAa,IAAmB;AACtC,WAAO;AAAA,EACT;AACF;","names":[]}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@livekit/agents",
-  "version": "1.0.21",
+  "version": "1.0.22",
   "description": "LiveKit Agents - Node.js",
   "main": "dist/index.js",
   "require": "dist/index.cjs",

package/src/inference/stt.ts

@@ -459,52 +459,67 @@ export class SpeechStream<TModel extends STTModels> extends BaseSpeechStream {
   }
 
   private processTranscript(data: Record<string, any>, isFinal: boolean) {
+    // Check if queue is closed to avoid race condition during disconnect
+    if (this.queue.closed) return;
+
     const requestId = data.request_id ?? this.requestId;
     const text = data.transcript ?? '';
     const language = data.language ?? this.opts.language ?? 'en';
 
     if (!text && !isFinal) return;
 
-    // We'll have a more accurate way of detecting when speech started when we have VAD
-    if (!this.speaking) {
-      this.speaking = true;
-      this.queue.put({ type: SpeechEventType.START_OF_SPEECH });
-    }
+    try {
+      // We'll have a more accurate way of detecting when speech started when we have VAD
+      if (!this.speaking) {
+        this.speaking = true;
+        this.queue.put({ type: SpeechEventType.START_OF_SPEECH });
+      }
 
-    const speechData: SpeechData = {
-      language,
-      startTime: data.start ?? 0,
-      endTime: data.duration ?? 0,
-      confidence: data.confidence ?? 1.0,
-      text,
-    };
+      const speechData: SpeechData = {
+        language,
+        startTime: data.start ?? 0,
+        endTime: data.duration ?? 0,
+        confidence: data.confidence ?? 1.0,
+        text,
+      };
+
+      if (isFinal) {
+        if (this.speechDuration > 0) {
+          this.queue.put({
+            type: SpeechEventType.RECOGNITION_USAGE,
+            requestId,
+            recognitionUsage: { audioDuration: this.speechDuration },
+          });
+          this.speechDuration = 0;
+        }
 
-    if (isFinal) {
-      if (this.speechDuration > 0) {
         this.queue.put({
-          type: SpeechEventType.RECOGNITION_USAGE,
+          type: SpeechEventType.FINAL_TRANSCRIPT,
           requestId,
-          recognitionUsage: { audioDuration: this.speechDuration },
+          alternatives: [speechData],
         });
-        this.speechDuration = 0;
-      }
-
-      this.queue.put({
-        type: SpeechEventType.FINAL_TRANSCRIPT,
-        requestId,
-        alternatives: [speechData],
-      });
 
-      if (this.speaking) {
-        this.speaking = false;
-        this.queue.put({ type: SpeechEventType.END_OF_SPEECH });
+        if (this.speaking) {
+          this.speaking = false;
+          this.queue.put({ type: SpeechEventType.END_OF_SPEECH });
+        }
+      } else {
+        this.queue.put({
+          type: SpeechEventType.INTERIM_TRANSCRIPT,
+          requestId,
+          alternatives: [speechData],
+        });
+      }
+    } catch (e) {
+      if (e instanceof Error && e.message.includes('Queue is closed')) {
+        // Expected behavior on disconnect, log as warning
+        this.#logger.warn(
+          { err: e },
+          'Queue closed during transcript processing (expected during disconnect)',
+        );
+      } else {
+        this.#logger.error({ err: e }, 'Error putting transcript to queue');
       }
-    } else {
-      this.queue.put({
-        type: SpeechEventType.INTERIM_TRANSCRIPT,
-        requestId,
-        alternatives: [speechData],
-      });
     }
   }
 }

package/src/inference/tts.ts

@@ -277,6 +277,9 @@ export class SynthesizeStream<TModel extends TTSModels> extends BaseSynthesizeSt
     };
 
     const sendClientEvent = async (event: TtsClientEvent) => {
+      // Don't send events to a closed WebSocket or aborted controller
+      if (this.abortController.signal.aborted || closing) return;
+
       const validatedEvent = await ttsClientEventSchema.parseAsync(event);
       if (!ws || ws.readyState !== WebSocket.OPEN) {
         this.#logger.warn('Trying to send client TTS event to a closed WebSocket');
@@ -321,7 +324,7 @@ export class SynthesizeStream<TModel extends TTSModels> extends BaseSynthesizeSt
       return new Promise<void>((resolve, reject) => {
         this.abortController.signal.addEventListener('abort', () => {
           resourceCleanup();
-          reject(new Error('WebSocket connection aborted'));
+          resolve(); // Abort is triggered by close(), which is a normal shutdown, not an error
         });
 
         ws.on('message', async (data) => {
@@ -420,8 +423,6 @@ export class SynthesizeStream<TModel extends TTSModels> extends BaseSynthesizeSt
         createWsListenerTask(ws),
         createRecvTask(),
       ]);
-    } catch (e) {
-      this.#logger.error({ error: e }, 'Error in SynthesizeStream');
     } finally {
       resourceCleanup();
     }

package/src/stt/stt.ts

@@ -257,7 +257,18 @@ export abstract class SpeechStream implements AsyncIterableIterator<SpeechEvent>
 
   protected async monitorMetrics() {
     for await (const event of this.queue) {
-      this.output.put(event);
+      if (!this.output.closed) {
+        try {
+          this.output.put(event);
+        } catch (e) {
+          if (e instanceof Error && e.message.includes('Queue is closed')) {
+            this.logger.warn(
+              { err: e },
+              'Queue closed during transcript processing (expected during disconnect)',
+            );
+          }
+        }
+      }
       if (event.type !== SpeechEventType.RECOGNITION_USAGE) continue;
       const metrics: STTMetrics = {
         type: 'stt_metrics',
@@ -270,7 +281,9 @@ export abstract class SpeechStream implements AsyncIterableIterator<SpeechEvent>
       };
       this.#stt.emit('metrics_collected', metrics);
     }
-    this.output.close();
+    if (!this.output.closed) {
+      this.output.close();
+    }
   }
 
   protected abstract run(): Promise<void>;
@@ -336,9 +349,9 @@ export abstract class SpeechStream implements AsyncIterableIterator<SpeechEvent>
 
   /** Close both the input and output of the STT stream */
   close() {
-    this.input.close();
-    this.queue.close();
-    this.output.close();
+    if (!this.input.closed) this.input.close();
+    if (!this.queue.closed) this.queue.close();
+    if (!this.output.closed) this.output.close();
     this.closed = true;
   }
 

package/src/tts/tts.ts

@@ -307,12 +307,11 @@ export abstract class SynthesizeStream
     }
     this.#metricsText += text;
 
-    if (this.input.closed) {
-      throw new Error('Input is closed');
-    }
-    if (this.closed) {
-      throw new Error('Stream is closed');
+    if (this.input.closed || this.closed) {
+      // Stream was aborted/closed, silently skip
+      return;
     }
+
     this.input.put(text);
   }
 
@@ -322,24 +321,24 @@ export abstract class SynthesizeStream
       this.#metricsPendingTexts.push(this.#metricsText);
       this.#metricsText = '';
     }
-    if (this.input.closed) {
-      throw new Error('Input is closed');
-    }
-    if (this.closed) {
-      throw new Error('Stream is closed');
+
+    if (this.input.closed || this.closed) {
+      // Stream was aborted/closed, silently skip
+      return;
     }
+
     this.input.put(SynthesizeStream.FLUSH_SENTINEL);
   }
 
   /** Mark the input as ended and forbid additional pushes */
   endInput() {
     this.flush();
-    if (this.input.closed) {
-      throw new Error('Input is closed');
-    }
-    if (this.closed) {
-      throw new Error('Stream is closed');
+
+    if (this.input.closed || this.closed) {
+      // Stream was aborted/closed, silently skip
+      return;
     }
+
     this.input.close();
   }