@livekit/agents 0.0.0-20260120144724

package/dist/voice/audio_recognition.cjs.map

@@ -0,0 +1 @@
+{"version":3,"sources":["../../src/voice/audio_recognition.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport { AudioFrame } from '@livekit/rtc-node';\nimport type { Context, Span } from '@opentelemetry/api';\nimport type { WritableStreamDefaultWriter } from 'node:stream/web';\nimport { ReadableStream } from 'node:stream/web';\nimport type { AdaptiveInterruptionDetector } from '../inference/interruption/AdaptiveInterruptionDetector.js';\nimport {\n  InterruptionStreamBase,\n  InterruptionStreamSentinel,\n} from '../inference/interruption/InterruptionStream.js';\nimport type { InterruptionEvent } from '../inference/interruption/interruption.js';\nimport { type ChatContext } from '../llm/chat_context.js';\nimport { log } from '../log.js';\nimport { DeferredReadableStream, isStreamReaderReleaseError } from '../stream/deferred_stream.js';\nimport { IdentityTransform } from '../stream/identity_transform.js';\nimport { mergeReadableStreams } from '../stream/merge_readable_streams.js';\nimport { type SpeechEvent, SpeechEventType } from '../stt/stt.js';\nimport { traceTypes, tracer } from '../telemetry/index.js';\nimport { Task, delay } from '../utils.js';\nimport { type VAD, type VADEvent, VADEventType } from '../vad.js';\nimport type { TurnDetectionMode } from './agent_session.js';\nimport type { STTNode } from './io.js';\n\nexport interface EndOfTurnInfo {\n  newTranscript: string;\n  transcriptConfidence: number;\n  transcriptionDelay: number;\n  endOfUtteranceDelay: number;\n  startedSpeakingAt: number | undefined;\n  stoppedSpeakingAt: number | undefined;\n}\n\nexport interface PreemptiveGenerationInfo {\n  newTranscript: string;\n  transcriptConfidence: number;\n}\n\nexport interface RecognitionHooks {\n  onStartOfSpeech: (ev: VADEvent) => void;\n  onVADInferenceDone: (ev: VADEvent) => void;\n  onEndOfSpeech: (ev: VADEvent) => void;\n  onInterimTranscript: (ev: SpeechEvent) => void;\n  onFinalTranscript: (ev: SpeechEvent) => void;\n  onEndOfTurn: (info: EndOfTurnInfo) => Promise<boolean>;\n  onPreemptiveGeneration: (info: PreemptiveGenerationInfo) => void;\n  onInterruption: (ev: InterruptionEvent) => void;\n\n  retrieveChatCtx: () => ChatContext;\n}\n\nexport interface _TurnDetector {\n  unlikelyThreshold: (language?: string) => Promise<number | undefined>;\n  supportsLanguage: (language?: string) => Promise<boolean>;\n  predictEndOfTurn(chatCtx: ChatContext): Promise<number>;\n}\n\nexport interface AudioRecognitionOptions {\n  recognitionHooks: RecognitionHooks;\n  stt?: STTNode;\n  vad?: VAD;\n  interruptionDetector?: AdaptiveInterruptionDetector;\n  turnDetector?: _TurnDetector;\n  turnDetectionMode?: Exclude<TurnDetectionMode, _TurnDetector>;\n  minEndpointingDelay: number;\n  maxEndpointingDelay: number;\n  rootSpanContext?: Context;\n}\n\nexport class AudioRecognition {\n  private hooks: RecognitionHooks;\n  private stt?: STTNode;\n  private vad?: VAD;\n  private turnDetector?: _TurnDetector;\n  private turnDetectionMode?: Exclude<TurnDetectionMode, _TurnDetector>;\n  private minEndpointingDelay: number;\n  private maxEndpointingDelay: number;\n  private lastLanguage?: string;\n  private rootSpanContext?: Context;\n\n  private deferredInputStream: DeferredReadableStream<AudioFrame>;\n  private logger = log();\n  private lastFinalTranscriptTime = 0;\n  private audioTranscript = '';\n  private audioInterimTranscript = '';\n  private audioPreflightTranscript = '';\n  private finalTranscriptConfidence: number[] = [];\n  private lastSpeakingTime: number | undefined;\n  private speechStartTime: number | undefined;\n  private userTurnCommitted = false;\n  private speaking = false;\n  private sampleRate?: number;\n\n  private userTurnSpan?: Span;\n\n  private vadInputStream: ReadableStream<AudioFrame>;\n  private sttInputStream: ReadableStream<AudioFrame>;\n  private interruptionInputStream: ReadableStream<AudioFrame>;\n  private silenceAudioTransform = new IdentityTransform<AudioFrame>();\n  private silenceAudioWriter: WritableStreamDefaultWriter<AudioFrame>;\n\n  // all cancellable tasks\n  private bounceEOUTask?: Task<void>;\n  private commitUserTurnTask?: Task<void>;\n  private vadTask?: Task<void>;\n  private sttTask?: Task<void>;\n  private interruptionTask?: Task<void>;\n\n  // interruption detection\n  private interruptionDetector?: AdaptiveInterruptionDetector;\n  private interruptionStream?: InterruptionStreamBase;\n  private interruptionEnabled = false;\n  private agentSpeaking = false;\n\n  constructor(opts: AudioRecognitionOptions) {\n    this.hooks = opts.recognitionHooks;\n    this.stt = opts.stt;\n    this.vad = opts.vad;\n    this.interruptionDetector = opts.interruptionDetector;\n    this.turnDetector = opts.turnDetector;\n    this.turnDetectionMode = opts.turnDetectionMode;\n    this.minEndpointingDelay = opts.minEndpointingDelay;\n    this.maxEndpointingDelay = opts.maxEndpointingDelay;\n    this.lastLanguage = undefined;\n    this.rootSpanContext = opts.rootSpanContext;\n\n    // Interruption detection is only enabled if both detector and VAD are provided\n    this.interruptionEnabled = this.interruptionDetector !== undefined && this.vad !== undefined;\n\n    this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n    const [vadInputStream, rest] = this.deferredInputStream.stream.tee();\n    const [sttInputStream, interruptionInputStream] = rest.tee();\n    this.vadInputStream = vadInputStream;\n    this.sttInputStream = mergeReadableStreams(sttInputStream, this.silenceAudioTransform.readable);\n    this.interruptionInputStream = interruptionInputStream;\n    this.silenceAudioWriter = this.silenceAudioTransform.writable.getWriter();\n  }\n\n  /**\n   * Current transcript of the user's speech, including interim transcript if available.\n   */\n  get currentTranscript(): string {\n    if (this.audioInterimTranscript) {\n      return `${this.audioTranscript} ${this.audioInterimTranscript}`.trim();\n    }\n    return this.audioTranscript;\n  }\n\n  async start() {\n    this.vadTask = Task.from(({ signal }) => this.createVadTask(this.vad, signal));\n    this.vadTask.result.catch((err) => {\n      this.logger.error(`Error running VAD task: ${err}`);\n    });\n\n    this.sttTask = Task.from(({ signal }) => this.createSttTask(this.stt, signal));\n    this.sttTask.result.catch((err) => {\n      this.logger.error(`Error running STT task: ${err}`);\n    });\n\n    if (this.interruptionEnabled && this.interruptionDetector) {\n      this.interruptionTask = Task.from(({ signal }) =>\n        this.createInterruptionTask(this.interruptionDetector!, signal),\n      );\n      this.interruptionTask.result.catch((err) => {\n        this.logger.error(`Error running interruption task: ${err}`);\n      });\n    }\n  }\n\n  private async onSTTEvent(ev: SpeechEvent) {\n    if (\n      this.turnDetectionMode === 'manual' &&\n      this.userTurnCommitted &&\n      (this.bounceEOUTask === undefined ||\n        this.bounceEOUTask.done ||\n        ev.type == SpeechEventType.INTERIM_TRANSCRIPT)\n    ) {\n      // ignore stt event if user turn already committed and EOU task is done\n      // or it's an interim transcript\n      this.logger.debug(\n        {\n          userTurnCommitted: this.userTurnCommitted,\n          eouTaskDone: this.bounceEOUTask?.done,\n          evType: ev.type,\n          turnDetectionMode: this.turnDetectionMode,\n        },\n        'ignoring stt event',\n      );\n      return;\n    }\n\n    switch (ev.type) {\n      case SpeechEventType.FINAL_TRANSCRIPT:\n        this.hooks.onFinalTranscript(ev);\n        const transcript = ev.alternatives?.[0]?.text;\n        const confidence = ev.alternatives?.[0]?.confidence ?? 0;\n        this.lastLanguage = ev.alternatives?.[0]?.language;\n\n        if (!transcript) {\n          // stt final transcript received but no transcript\n          return;\n        }\n\n        this.logger.debug(\n          {\n            user_transcript: transcript,\n            language: this.lastLanguage,\n          },\n          'received user transcript',\n        );\n\n        this.lastFinalTranscriptTime = Date.now();\n        this.audioTranscript += ` ${transcript}`;\n        this.audioTranscript = this.audioTranscript.trimStart();\n        this.finalTranscriptConfidence.push(confidence);\n        const transcriptChanged = this.audioTranscript !== this.audioPreflightTranscript;\n        this.audioInterimTranscript = '';\n        this.audioPreflightTranscript = '';\n\n        if (!this.vad || this.lastSpeakingTime === undefined) {\n          // vad disabled, use stt timestamp\n          // TODO: this would screw up transcription latency metrics\n          // but we'll live with it for now.\n          // the correct way is to ensure STT fires SpeechEventType.END_OF_SPEECH\n          // and using that timestamp for lastSpeakingTime\n          this.lastSpeakingTime = Date.now();\n        }\n\n        if (this.vadBaseTurnDetection || this.userTurnCommitted) {\n          if (transcriptChanged) {\n            this.logger.debug(\n              { transcript: this.audioTranscript },\n              'triggering preemptive generation (FINAL_TRANSCRIPT)',\n            );\n            this.hooks.onPreemptiveGeneration({\n              newTranscript: this.audioTranscript,\n              transcriptConfidence:\n                this.finalTranscriptConfidence.length > 0\n                  ? this.finalTranscriptConfidence.reduce((a, b) => a + b, 0) /\n                    this.finalTranscriptConfidence.length\n                  : 0,\n            });\n          }\n\n          if (!this.speaking) {\n            const chatCtx = this.hooks.retrieveChatCtx();\n            this.logger.debug('running EOU detection on stt FINAL_TRANSCRIPT');\n            this.runEOUDetection(chatCtx);\n          }\n        }\n        break;\n      case SpeechEventType.PREFLIGHT_TRANSCRIPT:\n        this.hooks.onInterimTranscript(ev);\n        const preflightTranscript = ev.alternatives?.[0]?.text ?? '';\n        const preflightConfidence = ev.alternatives?.[0]?.confidence ?? 0;\n        const preflightLanguage = ev.alternatives?.[0]?.language;\n\n        const MIN_LANGUAGE_DETECTION_LENGTH = 5;\n        if (\n          !this.lastLanguage ||\n          (preflightLanguage && preflightTranscript.length > MIN_LANGUAGE_DETECTION_LENGTH)\n        ) {\n          this.lastLanguage = preflightLanguage;\n        }\n\n        if (!preflightTranscript) {\n          return;\n        }\n\n        this.logger.debug(\n          {\n            user_transcript: preflightTranscript,\n            language: this.lastLanguage,\n          },\n          'received user preflight transcript',\n        );\n\n        // still need to increment it as it's used for turn detection,\n        this.lastFinalTranscriptTime = Date.now();\n        // preflight transcript includes all pre-committed transcripts (including final transcript from the previous STT run)\n        this.audioPreflightTranscript =\n          `${this.audioTranscript} ${preflightTranscript}`.trimStart();\n        this.audioInterimTranscript = preflightTranscript;\n\n        if (!this.vad || this.lastSpeakingTime === undefined) {\n          // vad disabled, use stt timestamp\n          this.lastSpeakingTime = Date.now();\n        }\n\n        if (this.turnDetectionMode !== 'manual' || this.userTurnCommitted) {\n          const confidenceVals = [...this.finalTranscriptConfidence, preflightConfidence];\n          this.logger.debug(\n            {\n              transcript:\n                this.audioPreflightTranscript.length > 100\n                  ? this.audioPreflightTranscript.slice(0, 100) + '...'\n                  : this.audioPreflightTranscript,\n            },\n            'triggering preemptive generation (PREFLIGHT_TRANSCRIPT)',\n          );\n          this.hooks.onPreemptiveGeneration({\n            newTranscript: this.audioPreflightTranscript,\n            transcriptConfidence:\n              confidenceVals.length > 0\n                ? confidenceVals.reduce((a, b) => a + b, 0) / confidenceVals.length\n                : 0,\n          });\n        }\n        break;\n      case SpeechEventType.INTERIM_TRANSCRIPT:\n        this.logger.debug({ transcript: ev.alternatives?.[0]?.text }, 'interim transcript');\n        this.hooks.onInterimTranscript(ev);\n        this.audioInterimTranscript = ev.alternatives?.[0]?.text ?? '';\n        break;\n      case SpeechEventType.START_OF_SPEECH:\n        if (this.turnDetectionMode !== 'stt') break;\n        this.hooks.onStartOfSpeech({\n          type: VADEventType.START_OF_SPEECH,\n          samplesIndex: 0,\n          timestamp: Date.now(),\n          speechDuration: 0,\n          silenceDuration: 0,\n          frames: [],\n          probability: 0,\n          inferenceDuration: 0,\n          speaking: true,\n          rawAccumulatedSilence: 0,\n          rawAccumulatedSpeech: 0,\n        });\n        this.speaking = true;\n        this.lastSpeakingTime = Date.now();\n\n        this.bounceEOUTask?.cancel();\n        break;\n      case SpeechEventType.END_OF_SPEECH:\n        if (this.turnDetectionMode !== 'stt') break;\n        this.hooks.onEndOfSpeech({\n          type: VADEventType.END_OF_SPEECH,\n          samplesIndex: 0,\n          timestamp: Date.now(),\n          speechDuration: 0,\n          silenceDuration: 0,\n          frames: [],\n          probability: 0,\n          inferenceDuration: 0,\n          speaking: false,\n          rawAccumulatedSilence: 0,\n          rawAccumulatedSpeech: 0,\n        });\n        this.speaking = false;\n        this.userTurnCommitted = true;\n        this.lastSpeakingTime = Date.now();\n\n        if (!this.speaking) {\n          const chatCtx = this.hooks.retrieveChatCtx();\n          this.logger.debug('running EOU detection on stt END_OF_SPEECH');\n          this.runEOUDetection(chatCtx);\n        }\n    }\n  }\n\n  private runEOUDetection(chatCtx: ChatContext) {\n    this.logger.debug(\n      {\n        stt: this.stt,\n        audioTranscript: this.audioTranscript,\n        turnDetectionMode: this.turnDetectionMode,\n      },\n      'running EOU detection',\n    );\n\n    if (this.stt && !this.audioTranscript && this.turnDetectionMode !== 'manual') {\n      // stt enabled but no transcript yet\n      this.logger.debug('skipping EOU detection');\n      return;\n    }\n\n    chatCtx = chatCtx.copy();\n    chatCtx.addMessage({ role: 'user', content: this.audioTranscript });\n\n    const turnDetector =\n      // disable EOU model if manual turn detection enabled\n      this.audioTranscript && this.turnDetectionMode !== 'manual' ? this.turnDetector : undefined;\n\n    const bounceEOUTask =\n      (\n        lastSpeakingTime: number | undefined,\n        lastFinalTranscriptTime: number,\n        speechStartTime: number | undefined,\n      ) =>\n      async (controller: AbortController) => {\n        let endpointingDelay = this.minEndpointingDelay;\n\n        if (turnDetector) {\n          await tracer.startActiveSpan(\n            async (span) => {\n              this.logger.debug('Running turn detector model');\n\n              let endOfTurnProbability = 0.0;\n              let unlikelyThreshold: number | undefined;\n\n              if (!(await turnDetector.supportsLanguage(this.lastLanguage))) {\n                this.logger.debug(`Turn detector does not support language ${this.lastLanguage}`);\n              } else {\n                try {\n                  endOfTurnProbability = await turnDetector.predictEndOfTurn(chatCtx);\n                  unlikelyThreshold = await turnDetector.unlikelyThreshold(this.lastLanguage);\n\n                  this.logger.debug(\n                    { endOfTurnProbability, unlikelyThreshold, language: this.lastLanguage },\n                    'end of turn probability',\n                  );\n\n                  if (unlikelyThreshold && endOfTurnProbability < unlikelyThreshold) {\n                    endpointingDelay = this.maxEndpointingDelay;\n                  }\n                } catch (error) {\n                  this.logger.error(error, 'Error predicting end of turn');\n                }\n              }\n\n              span.setAttribute(\n                traceTypes.ATTR_CHAT_CTX,\n                JSON.stringify(chatCtx.toJSON({ excludeTimestamp: false })),\n              );\n              span.setAttribute(traceTypes.ATTR_EOU_PROBABILITY, endOfTurnProbability);\n              span.setAttribute(traceTypes.ATTR_EOU_UNLIKELY_THRESHOLD, unlikelyThreshold ?? 0);\n              span.setAttribute(traceTypes.ATTR_EOU_DELAY, endpointingDelay);\n              span.setAttribute(traceTypes.ATTR_EOU_LANGUAGE, this.lastLanguage ?? '');\n            },\n            {\n              name: 'eou_detection',\n              context: this.rootSpanContext,\n            },\n          );\n        }\n\n        let extraSleep = endpointingDelay;\n        if (lastSpeakingTime !== undefined) {\n          extraSleep += lastSpeakingTime - Date.now();\n        }\n\n        if (extraSleep > 0) {\n          // add delay to see if there's a potential upcoming EOU task that cancels this one\n          await delay(Math.max(extraSleep, 0), { signal: controller.signal });\n        }\n\n        this.logger.debug({ transcript: this.audioTranscript }, 'end of user turn');\n\n        const confidenceAvg =\n          this.finalTranscriptConfidence.length > 0\n            ? this.finalTranscriptConfidence.reduce((a, b) => a + b, 0) /\n              this.finalTranscriptConfidence.length\n            : 0;\n\n        let startedSpeakingAt: number | undefined;\n        let stoppedSpeakingAt: number | undefined;\n        let transcriptionDelay: number | undefined;\n        let endOfUtteranceDelay: number | undefined;\n\n        // sometimes, we can't calculate the metrics because VAD was unreliable.\n        // in this case, we just ignore the calculation, it's better than providing likely wrong values\n        if (\n          lastFinalTranscriptTime !== 0 &&\n          lastSpeakingTime !== undefined &&\n          speechStartTime !== undefined\n        ) {\n          startedSpeakingAt = speechStartTime;\n          stoppedSpeakingAt = lastSpeakingTime;\n          transcriptionDelay = Math.max(lastFinalTranscriptTime - lastSpeakingTime, 0);\n          endOfUtteranceDelay = Date.now() - lastSpeakingTime;\n        }\n\n        const committed = await this.hooks.onEndOfTurn({\n          newTranscript: this.audioTranscript,\n          transcriptConfidence: confidenceAvg,\n          transcriptionDelay: transcriptionDelay ?? 0,\n          endOfUtteranceDelay: endOfUtteranceDelay ?? 0,\n          startedSpeakingAt,\n          stoppedSpeakingAt,\n        });\n\n        if (committed) {\n          this._endUserTurnSpan({\n            transcript: this.audioTranscript,\n            confidence: confidenceAvg,\n            transcriptionDelay: transcriptionDelay ?? 0,\n            endOfUtteranceDelay: endOfUtteranceDelay ?? 0,\n          });\n\n          // clear the transcript if the user turn was committed\n          this.audioTranscript = '';\n          this.finalTranscriptConfidence = [];\n          this.lastSpeakingTime = undefined;\n          this.lastFinalTranscriptTime = 0;\n          this.speechStartTime = undefined;\n        }\n\n        this.userTurnCommitted = false;\n      };\n\n    // cancel any existing EOU task\n    this.bounceEOUTask?.cancel();\n    // copy the values before awaiting (the values can change)\n    this.bounceEOUTask = Task.from(\n      bounceEOUTask(this.lastSpeakingTime, this.lastFinalTranscriptTime, this.speechStartTime),\n    );\n\n    this.bounceEOUTask.result\n      .then(() => {\n        this.logger.debug('EOU detection task completed');\n      })\n      .catch((err: unknown) => {\n        if (err instanceof Error && err.message.includes('This operation was aborted')) {\n          // ignore aborted errors\n          return;\n        }\n        this.logger.error(err, 'Error in EOU detection task:');\n      });\n  }\n\n  private async createSttTask(stt: STTNode | undefined, signal: AbortSignal) {\n    if (!stt) return;\n\n    this.logger.debug('createSttTask: create stt stream from stt node');\n\n    const sttStream = await stt(this.sttInputStream, {});\n\n    if (signal.aborted || sttStream === null) return;\n\n    if (sttStream instanceof ReadableStream) {\n      const reader = sttStream.getReader();\n\n      signal.addEventListener('abort', async () => {\n        try {\n          reader.releaseLock();\n          await sttStream?.cancel();\n        } catch (e) {\n          this.logger.debug('createSttTask: error during abort handler:', e);\n        }\n      });\n\n      try {\n        while (true) {\n          if (signal.aborted) break;\n\n          const { done, value: ev } = await reader.read();\n          if (done) break;\n\n          if (typeof ev === 'string') {\n            throw new Error('STT node must yield SpeechEvent');\n          } else {\n            await this.onSTTEvent(ev);\n          }\n        }\n      } catch (e) {\n        if (isStreamReaderReleaseError(e)) {\n          return;\n        }\n        this.logger.error({ error: e }, 'createSttTask: error reading sttStream');\n      } finally {\n        reader.releaseLock();\n        try {\n          await sttStream.cancel();\n        } catch (e) {\n          this.logger.debug(\n            'createSttTask: error cancelling sttStream (may already be cancelled):',\n            e,\n          );\n        }\n      }\n    }\n  }\n\n  private async createVadTask(vad: VAD | undefined, signal: AbortSignal) {\n    if (!vad) return;\n\n    const vadStream = vad.stream();\n    vadStream.updateInputStream(this.vadInputStream);\n\n    const abortHandler = () => {\n      vadStream.detachInputStream();\n      vadStream.close();\n      signal.removeEventListener('abort', abortHandler);\n    };\n    signal.addEventListener('abort', abortHandler);\n\n    try {\n      for await (const ev of vadStream) {\n        if (signal.aborted) break;\n\n        switch (ev.type) {\n          case VADEventType.START_OF_SPEECH:\n            this.logger.debug('VAD task: START_OF_SPEECH');\n            this.hooks.onStartOfSpeech(ev);\n            this.speaking = true;\n\n            if (!this.userTurnSpan) {\n              this.userTurnSpan = tracer.startSpan({\n                name: 'user_turn',\n                context: this.rootSpanContext,\n              });\n            }\n\n            // Capture sample rate from the first VAD event if not already set\n            if (ev.frames.length > 0 && ev.frames[0]) {\n              this.sampleRate = ev.frames[0].sampleRate;\n            }\n\n            // If agent is speaking, user speech is overlap - trigger interruption detection\n            if (this.agentSpeaking && this.interruptionEnabled) {\n              this.onStartOfOverlapSpeech(ev.speechDuration, this.userTurnSpan);\n            }\n\n            this.bounceEOUTask?.cancel();\n            break;\n          case VADEventType.INFERENCE_DONE:\n            this.hooks.onVADInferenceDone(ev);\n            // for metrics, get the \"earliest\" signal of speech as possible\n            if (ev.rawAccumulatedSpeech > 0.0) {\n              this.lastSpeakingTime = Date.now();\n\n              if (this.speechStartTime === undefined) {\n                this.speechStartTime = Date.now();\n              }\n            }\n            break;\n          case VADEventType.END_OF_SPEECH:\n            this.logger.debug('VAD task: END_OF_SPEECH');\n            this.hooks.onEndOfSpeech(ev);\n\n            // when VAD fires END_OF_SPEECH, it already waited for the silence_duration\n            this.speaking = false;\n\n            // If we were in overlap speech (agent speaking + user speaking), end it\n            if (this.agentSpeaking && this.interruptionEnabled) {\n              this.onEndOfOverlapSpeech();\n            }\n\n            if (\n              this.vadBaseTurnDetection ||\n              (this.turnDetectionMode === 'stt' && this.userTurnCommitted)\n            ) {\n              const chatCtx = this.hooks.retrieveChatCtx();\n              this.runEOUDetection(chatCtx);\n            }\n            break;\n        }\n      }\n    } catch (e) {\n      this.logger.error(e, 'Error in VAD task');\n    } finally {\n      this.logger.debug('VAD task closed');\n    }\n  }\n\n  private async createInterruptionTask(\n    interruptionDetector: AdaptiveInterruptionDetector,\n    signal: AbortSignal,\n  ) {\n    // Create the interruption stream from the detector\n    this.interruptionStream = interruptionDetector.createStream();\n\n    // Forward audio frames to the interruption stream\n    const reader = this.interruptionInputStream.getReader();\n\n    const forwardTask = (async () => {\n      try {\n        while (!signal.aborted) {\n          const { done, value: frame } = await reader.read();\n          if (done) break;\n          await this.interruptionStream?.pushFrame(frame);\n        }\n      } catch (e) {\n        if (!signal.aborted) {\n          this.logger.error(e, 'Error forwarding audio to interruption stream');\n        }\n      } finally {\n        reader.releaseLock();\n      }\n    })();\n\n    // Read interruption events from the stream\n    const eventStream = this.interruptionStream.stream;\n    const eventReader = eventStream.getReader();\n\n    const abortHandler = () => {\n      eventReader.releaseLock();\n      this.interruptionStream?.close();\n      signal.removeEventListener('abort', abortHandler);\n    };\n    signal.addEventListener('abort', abortHandler);\n\n    try {\n      while (!signal.aborted) {\n        const { done, value: ev } = await eventReader.read();\n        if (done) break;\n\n        this.logger.debug({ type: ev.type, probability: ev.probability }, 'Interruption event');\n        this.hooks.onInterruption(ev);\n      }\n    } catch (e) {\n      if (!signal.aborted) {\n        this.logger.error(e, 'Error in interruption task');\n      }\n    } finally {\n      this.logger.debug('Interruption task closed');\n      await forwardTask;\n    }\n  }\n\n  /**\n   * Called when the agent starts speaking.\n   * Enables interruption detection by sending the agent-speech-started sentinel.\n   */\n  onStartOfAgentSpeech(): void {\n    this.agentSpeaking = true;\n\n    if (!this.interruptionEnabled || !this.interruptionStream) {\n      return;\n    }\n\n    this.interruptionStream.pushFrame(InterruptionStreamSentinel.speechStarted());\n  }\n\n  /**\n   * Called when the agent stops speaking.\n   * Disables interruption detection by sending the agent-speech-ended sentinel.\n   */\n  onEndOfAgentSpeech(): void {\n    if (!this.interruptionEnabled || !this.interruptionStream) {\n      this.agentSpeaking = false;\n      return;\n    }\n\n    this.interruptionStream.pushFrame(InterruptionStreamSentinel.speechEnded());\n\n    if (this.agentSpeaking) {\n      // No interruption was detected, end the overlap inference (idempotent)\n      this.onEndOfOverlapSpeech();\n    }\n\n    this.agentSpeaking = false;\n  }\n\n  /**\n   * Called when user starts speaking while agent is speaking (overlap speech).\n   * This triggers the interruption detection inference.\n   */\n  onStartOfOverlapSpeech(speechDuration: number, userSpeakingSpan?: Span): void {\n    if (!this.interruptionEnabled || !this.interruptionStream) {\n      return;\n    }\n\n    if (this.agentSpeaking && userSpeakingSpan) {\n      this.interruptionStream.pushFrame(\n        InterruptionStreamSentinel.overlapSpeechStarted(speechDuration, userSpeakingSpan),\n      );\n    }\n  }\n\n  /**\n   * Called when user stops speaking during overlap.\n   * This ends the interruption detection inference for this overlap period.\n   */\n  onEndOfOverlapSpeech(): void {\n    if (!this.interruptionEnabled || !this.interruptionStream) {\n      return;\n    }\n\n    this.interruptionStream.pushFrame(InterruptionStreamSentinel.overlapSpeechEnded());\n  }\n\n  setInputAudioStream(audioStream: ReadableStream<AudioFrame>) {\n    this.deferredInputStream.setSource(audioStream);\n  }\n\n  detachInputAudioStream() {\n    this.deferredInputStream.detachSource();\n  }\n\n  clearUserTurn() {\n    this.audioTranscript = '';\n    this.audioInterimTranscript = '';\n    this.audioPreflightTranscript = '';\n    this.finalTranscriptConfidence = [];\n    this.userTurnCommitted = false;\n\n    this.sttTask?.cancelAndWait().finally(() => {\n      this.sttTask = Task.from(({ signal }) => this.createSttTask(this.stt, signal));\n      this.sttTask.result.catch((err) => {\n        this.logger.error(`Error running STT task: ${err}`);\n      });\n    });\n  }\n\n  commitUserTurn(audioDetached: boolean) {\n    const commitUserTurnTask =\n      (delayDuration: number = 500) =>\n      async (controller: AbortController) => {\n        if (Date.now() - this.lastFinalTranscriptTime > delayDuration) {\n          // flush the stt by pushing silence\n          if (audioDetached && this.sampleRate !== undefined) {\n            const numSamples = Math.floor(this.sampleRate * 0.5);\n            const silence = new Int16Array(numSamples * 2);\n            const silenceFrame = new AudioFrame(silence, this.sampleRate, 1, numSamples);\n            this.silenceAudioWriter.write(silenceFrame);\n          }\n\n          // wait for the final transcript to be available\n          await delay(delayDuration, { signal: controller.signal });\n        }\n\n        if (this.audioInterimTranscript) {\n          // append interim transcript in case the final transcript is not ready\n          this.audioTranscript = `${this.audioTranscript} ${this.audioInterimTranscript}`.trim();\n        }\n        this.audioInterimTranscript = '';\n\n        const chatCtx = this.hooks.retrieveChatCtx();\n        this.logger.debug('running EOU detection on commitUserTurn');\n        this.runEOUDetection(chatCtx);\n        this.userTurnCommitted = true;\n      };\n\n    // cancel any existing commit user turn task\n    this.commitUserTurnTask?.cancel();\n    this.commitUserTurnTask = Task.from(commitUserTurnTask());\n\n    this.commitUserTurnTask.result\n      .then(() => {\n        this.logger.debug('User turn committed');\n      })\n      .catch((err: unknown) => {\n        this.logger.error(err, 'Error in user turn commit task:');\n      });\n  }\n\n  async close() {\n    this.detachInputAudioStream();\n    this.silenceAudioWriter.releaseLock();\n    await this.commitUserTurnTask?.cancelAndWait();\n    await this.sttTask?.cancelAndWait();\n    await this.vadTask?.cancelAndWait();\n    await this.bounceEOUTask?.cancelAndWait();\n    await this.interruptionTask?.cancelAndWait();\n    await this.interruptionStream?.close();\n  }\n\n  private _endUserTurnSpan({\n    transcript,\n    confidence,\n    transcriptionDelay,\n    endOfUtteranceDelay,\n  }: {\n    transcript: string;\n    confidence: number;\n    transcriptionDelay: number;\n    endOfUtteranceDelay: number;\n  }): void {\n    if (this.userTurnSpan) {\n      this.userTurnSpan.setAttributes({\n        [traceTypes.ATTR_USER_TRANSCRIPT]: transcript,\n        [traceTypes.ATTR_TRANSCRIPT_CONFIDENCE]: confidence,\n        [traceTypes.ATTR_TRANSCRIPTION_DELAY]: transcriptionDelay,\n        [traceTypes.ATTR_END_OF_TURN_DELAY]: endOfUtteranceDelay,\n      });\n      this.userTurnSpan.end();\n      this.userTurnSpan = undefined;\n    }\n  }\n\n  private get vadBaseTurnDetection() {\n    return ['vad', undefined].includes(this.turnDetectionMode);\n  }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAGA,sBAA2B;AAG3B,iBAA+B;AAE/B,gCAGO;AAEP,0BAAiC;AACjC,iBAAoB;AACpB,6BAAmE;AACnE,gCAAkC;AAClC,oCAAqC;AACrC,iBAAkD;AAClD,uBAAmC;AACnC,mBAA4B;AAC5B,iBAAsD;AAiD/C,MAAM,iBAAiB;AAAA,EACpB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAEA;AAAA,EACA,aAAS,gBAAI;AAAA,EACb,0BAA0B;AAAA,EAC1B,kBAAkB;AAAA,EAClB,yBAAyB;AAAA,EACzB,2BAA2B;AAAA,EAC3B,4BAAsC,CAAC;AAAA,EACvC;AAAA,EACA;AAAA,EACA,oBAAoB;AAAA,EACpB,WAAW;AAAA,EACX;AAAA,EAEA;AAAA,EAEA;AAAA,EACA;AAAA,EACA;AAAA,EACA,wBAAwB,IAAI,4CAA8B;AAAA,EAC1D;AAAA;AAAA,EAGA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAGA;AAAA,EACA;AAAA,EACA,sBAAsB;AAAA,EACtB,gBAAgB;AAAA,EAExB,YAAY,MAA+B;AACzC,SAAK,QAAQ,KAAK;AAClB,SAAK,MAAM,KAAK;AAChB,SAAK,MAAM,KAAK;AAChB,SAAK,uBAAuB,KAAK;AACjC,SAAK,eAAe,KAAK;AACzB,SAAK,oBAAoB,KAAK;AAC9B,SAAK,sBAAsB,KAAK;AAChC,SAAK,sBAAsB,KAAK;AAChC,SAAK,eAAe;AACpB,SAAK,kBAAkB,KAAK;AAG5B,SAAK,sBAAsB,KAAK,yBAAyB,UAAa,KAAK,QAAQ;AAEnF,SAAK,sBAAsB,IAAI,8CAAmC;AAClE,UAAM,CAAC,gBAAgB,IAAI,IAAI,KAAK,oBAAoB,OAAO,IAAI;AACnE,UAAM,CAAC,gBAAgB,uBAAuB,IAAI,KAAK,IAAI;AAC3D,SAAK,iBAAiB;AACtB,SAAK,qBAAiB,oDAAqB,gBAAgB,KAAK,sBAAsB,QAAQ;AAC9F,SAAK,0BAA0B;AAC/B,SAAK,qBAAqB,KAAK,sBAAsB,SAAS,UAAU;AAAA,EAC1E;AAAA;AAAA;AAAA;AAAA,EAKA,IAAI,oBAA4B;AAC9B,QAAI,KAAK,wBAAwB;AAC/B,aAAO,GAAG,KAAK,eAAe,IAAI,KAAK,sBAAsB,GAAG,KAAK;AAAA,IACvE;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,QAAQ;AACZ,SAAK,UAAU,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,cAAc,KAAK,KAAK,MAAM,CAAC;AAC7E,SAAK,QAAQ,OAAO,MAAM,CAAC,QAAQ;AACjC,WAAK,OAAO,MAAM,2BAA2B,GAAG,EAAE;AAAA,IACpD,CAAC;AAED,SAAK,UAAU,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,cAAc,KAAK,KAAK,MAAM,CAAC;AAC7E,SAAK,QAAQ,OAAO,MAAM,CAAC,QAAQ;AACjC,WAAK,OAAO,MAAM,2BAA2B,GAAG,EAAE;AAAA,IACpD,CAAC;AAED,QAAI,KAAK,uBAAuB,KAAK,sBAAsB;AACzD,WAAK,mBAAmB,kBAAK;AAAA,QAAK,CAAC,EAAE,OAAO,MAC1C,KAAK,uBAAuB,KAAK,sBAAuB,MAAM;AAAA,MAChE;AACA,WAAK,iBAAiB,OAAO,MAAM,CAAC,QAAQ;AAC1C,aAAK,OAAO,MAAM,oCAAoC,GAAG,EAAE;AAAA,MAC7D,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,MAAc,WAAW,IAAiB;AA1K5C;AA2KI,QACE,KAAK,sBAAsB,YAC3B,KAAK,sBACJ,KAAK,kBAAkB,UACtB,KAAK,cAAc,QACnB,GAAG,QAAQ,2BAAgB,qBAC7B;AAGA,WAAK,OAAO;AAAA,QACV;AAAA,UACE,mBAAmB,KAAK;AAAA,UACxB,cAAa,UAAK,kBAAL,mBAAoB;AAAA,UACjC,QAAQ,GAAG;AAAA,UACX,mBAAmB,KAAK;AAAA,QAC1B;AAAA,QACA;AAAA,MACF;AACA;AAAA,IACF;AAEA,YAAQ,GAAG,MAAM;AAAA,MACf,KAAK,2BAAgB;AACnB,aAAK,MAAM,kBAAkB,EAAE;AAC/B,cAAM,cAAa,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB;AACzC,cAAM,eAAa,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB,eAAc;AACvD,aAAK,gBAAe,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB;AAE1C,YAAI,CAAC,YAAY;AAEf;AAAA,QACF;AAEA,aAAK,OAAO;AAAA,UACV;AAAA,YACE,iBAAiB;AAAA,YACjB,UAAU,KAAK;AAAA,UACjB;AAAA,UACA;AAAA,QACF;AAEA,aAAK,0BAA0B,KAAK,IAAI;AACxC,aAAK,mBAAmB,IAAI,UAAU;AACtC,aAAK,kBAAkB,KAAK,gBAAgB,UAAU;AACtD,aAAK,0BAA0B,KAAK,UAAU;AAC9C,cAAM,oBAAoB,KAAK,oBAAoB,KAAK;AACxD,aAAK,yBAAyB;AAC9B,aAAK,2BAA2B;AAEhC,YAAI,CAAC,KAAK,OAAO,KAAK,qBAAqB,QAAW;AAMpD,eAAK,mBAAmB,KAAK,IAAI;AAAA,QACnC;AAEA,YAAI,KAAK,wBAAwB,KAAK,mBAAmB;AACvD,cAAI,mBAAmB;AACrB,iBAAK,OAAO;AAAA,cACV,EAAE,YAAY,KAAK,gBAAgB;AAAA,cACnC;AAAA,YACF;AACA,iBAAK,MAAM,uBAAuB;AAAA,cAChC,eAAe,KAAK;AAAA,cACpB,sBACE,KAAK,0BAA0B,SAAS,IACpC,KAAK,0BAA0B,OAAO,CAAC,GAAG,MAAM,IAAI,GAAG,CAAC,IACxD,KAAK,0BAA0B,SAC/B;AAAA,YACR,CAAC;AAAA,UACH;AAEA,cAAI,CAAC,KAAK,UAAU;AAClB,kBAAM,UAAU,KAAK,MAAM,gBAAgB;AAC3C,iBAAK,OAAO,MAAM,+CAA+C;AACjE,iBAAK,gBAAgB,OAAO;AAAA,UAC9B;AAAA,QACF;AACA;AAAA,MACF,KAAK,2BAAgB;AACnB,aAAK,MAAM,oBAAoB,EAAE;AACjC,cAAM,wBAAsB,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB,SAAQ;AAC1D,cAAM,wBAAsB,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB,eAAc;AAChE,cAAM,qBAAoB,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB;AAEhD,cAAM,gCAAgC;AACtC,YACE,CAAC,KAAK,gBACL,qBAAqB,oBAAoB,SAAS,+BACnD;AACA,eAAK,eAAe;AAAA,QACtB;AAEA,YAAI,CAAC,qBAAqB;AACxB;AAAA,QACF;AAEA,aAAK,OAAO;AAAA,UACV;AAAA,YACE,iBAAiB;AAAA,YACjB,UAAU,KAAK;AAAA,UACjB;AAAA,UACA;AAAA,QACF;AAGA,aAAK,0BAA0B,KAAK,IAAI;AAExC,aAAK,2BACH,GAAG,KAAK,eAAe,IAAI,mBAAmB,GAAG,UAAU;AAC7D,aAAK,yBAAyB;AAE9B,YAAI,CAAC,KAAK,OAAO,KAAK,qBAAqB,QAAW;AAEpD,eAAK,mBAAmB,KAAK,IAAI;AAAA,QACnC;AAEA,YAAI,KAAK,sBAAsB,YAAY,KAAK,mBAAmB;AACjE,gBAAM,iBAAiB,CAAC,GAAG,KAAK,2BAA2B,mBAAmB;AAC9E,eAAK,OAAO;AAAA,YACV;AAAA,cACE,YACE,KAAK,yBAAyB,SAAS,MACnC,KAAK,yBAAyB,MAAM,GAAG,GAAG,IAAI,QAC9C,KAAK;AAAA,YACb;AAAA,YACA;AAAA,UACF;AACA,eAAK,MAAM,uBAAuB;AAAA,YAChC,eAAe,KAAK;AAAA,YACpB,sBACE,eAAe,SAAS,IACpB,eAAe,OAAO,CAAC,GAAG,MAAM,IAAI,GAAG,CAAC,IAAI,eAAe,SAC3D;AAAA,UACR,CAAC;AAAA,QACH;AACA;AAAA,MACF,KAAK,2BAAgB;AACnB,aAAK,OAAO,MAAM,EAAE,aAAY,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB,KAAK,GAAG,oBAAoB;AAClF,aAAK,MAAM,oBAAoB,EAAE;AACjC,aAAK,2BAAyB,cAAG,iBAAH,mBAAkB,OAAlB,mBAAsB,SAAQ;AAC5D;AAAA,MACF,KAAK,2BAAgB;AACnB,YAAI,KAAK,sBAAsB,MAAO;AACtC,aAAK,MAAM,gBAAgB;AAAA,UACzB,MAAM,wBAAa;AAAA,UACnB,cAAc;AAAA,UACd,WAAW,KAAK,IAAI;AAAA,UACpB,gBAAgB;AAAA,UAChB,iBAAiB;AAAA,UACjB,QAAQ,CAAC;AAAA,UACT,aAAa;AAAA,UACb,mBAAmB;AAAA,UACnB,UAAU;AAAA,UACV,uBAAuB;AAAA,UACvB,sBAAsB;AAAA,QACxB,CAAC;AACD,aAAK,WAAW;AAChB,aAAK,mBAAmB,KAAK,IAAI;AAEjC,mBAAK,kBAAL,mBAAoB;AACpB;AAAA,MACF,KAAK,2BAAgB;AACnB,YAAI,KAAK,sBAAsB,MAAO;AACtC,aAAK,MAAM,cAAc;AAAA,UACvB,MAAM,wBAAa;AAAA,UACnB,cAAc;AAAA,UACd,WAAW,KAAK,IAAI;AAAA,UACpB,gBAAgB;AAAA,UAChB,iBAAiB;AAAA,UACjB,QAAQ,CAAC;AAAA,UACT,aAAa;AAAA,UACb,mBAAmB;AAAA,UACnB,UAAU;AAAA,UACV,uBAAuB;AAAA,UACvB,sBAAsB;AAAA,QACxB,CAAC;AACD,aAAK,WAAW;AAChB,aAAK,oBAAoB;AACzB,aAAK,mBAAmB,KAAK,IAAI;AAEjC,YAAI,CAAC,KAAK,UAAU;AAClB,gBAAM,UAAU,KAAK,MAAM,gBAAgB;AAC3C,eAAK,OAAO,MAAM,4CAA4C;AAC9D,eAAK,gBAAgB,OAAO;AAAA,QAC9B;AAAA,IACJ;AAAA,EACF;AAAA,EAEQ,gBAAgB,SAAsB;AA1WhD;AA2WI,SAAK,OAAO;AAAA,MACV;AAAA,QACE,KAAK,KAAK;AAAA,QACV,iBAAiB,KAAK;AAAA,QACtB,mBAAmB,KAAK;AAAA,MAC1B;AAAA,MACA;AAAA,IACF;AAEA,QAAI,KAAK,OAAO,CAAC,KAAK,mBAAmB,KAAK,sBAAsB,UAAU;AAE5E,WAAK,OAAO,MAAM,wBAAwB;AAC1C;AAAA,IACF;AAEA,cAAU,QAAQ,KAAK;AACvB,YAAQ,WAAW,EAAE,MAAM,QAAQ,SAAS,KAAK,gBAAgB,CAAC;AAElE,UAAM;AAAA;AAAA,MAEJ,KAAK,mBAAmB,KAAK,sBAAsB,WAAW,KAAK,eAAe;AAAA;AAEpF,UAAM,gBACJ,CACE,kBACA,yBACA,oBAEF,OAAO,eAAgC;AACrC,UAAI,mBAAmB,KAAK;AAE5B,UAAI,cAAc;AAChB,cAAM,wBAAO;AAAA,UACX,OAAO,SAAS;AACd,iBAAK,OAAO,MAAM,6BAA6B;AAE/C,gBAAI,uBAAuB;AAC3B,gBAAI;AAEJ,gBAAI,CAAE,MAAM,aAAa,iBAAiB,KAAK,YAAY,GAAI;AAC7D,mBAAK,OAAO,MAAM,2CAA2C,KAAK,YAAY,EAAE;AAAA,YAClF,OAAO;AACL,kBAAI;AACF,uCAAuB,MAAM,aAAa,iBAAiB,OAAO;AAClE,oCAAoB,MAAM,aAAa,kBAAkB,KAAK,YAAY;AAE1E,qBAAK,OAAO;AAAA,kBACV,EAAE,sBAAsB,mBAAmB,UAAU,KAAK,aAAa;AAAA,kBACvE;AAAA,gBACF;AAEA,oBAAI,qBAAqB,uBAAuB,mBAAmB;AACjE,qCAAmB,KAAK;AAAA,gBAC1B;AAAA,cACF,SAAS,OAAO;AACd,qBAAK,OAAO,MAAM,OAAO,8BAA8B;AAAA,cACzD;AAAA,YACF;AAEA,iBAAK;AAAA,cACH,4BAAW;AAAA,cACX,KAAK,UAAU,QAAQ,OAAO,EAAE,kBAAkB,MAAM,CAAC,CAAC;AAAA,YAC5D;AACA,iBAAK,aAAa,4BAAW,sBAAsB,oBAAoB;AACvE,iBAAK,aAAa,4BAAW,6BAA6B,qBAAqB,CAAC;AAChF,iBAAK,aAAa,4BAAW,gBAAgB,gBAAgB;AAC7D,iBAAK,aAAa,4BAAW,mBAAmB,KAAK,gBAAgB,EAAE;AAAA,UACzE;AAAA,UACA;AAAA,YACE,MAAM;AAAA,YACN,SAAS,KAAK;AAAA,UAChB;AAAA,QACF;AAAA,MACF;AAEA,UAAI,aAAa;AACjB,UAAI,qBAAqB,QAAW;AAClC,sBAAc,mBAAmB,KAAK,IAAI;AAAA,MAC5C;AAEA,UAAI,aAAa,GAAG;AAElB,kBAAM,oBAAM,KAAK,IAAI,YAAY,CAAC,GAAG,EAAE,QAAQ,WAAW,OAAO,CAAC;AAAA,MACpE;AAEA,WAAK,OAAO,MAAM,EAAE,YAAY,KAAK,gBAAgB,GAAG,kBAAkB;AAE1E,YAAM,gBACJ,KAAK,0BAA0B,SAAS,IACpC,KAAK,0BAA0B,OAAO,CAAC,GAAG,MAAM,IAAI,GAAG,CAAC,IACxD,KAAK,0BAA0B,SAC/B;AAEN,UAAI;AACJ,UAAI;AACJ,UAAI;AACJ,UAAI;AAIJ,UACE,4BAA4B,KAC5B,qBAAqB,UACrB,oBAAoB,QACpB;AACA,4BAAoB;AACpB,4BAAoB;AACpB,6BAAqB,KAAK,IAAI,0BAA0B,kBAAkB,CAAC;AAC3E,8BAAsB,KAAK,IAAI,IAAI;AAAA,MACrC;AAEA,YAAM,YAAY,MAAM,KAAK,MAAM,YAAY;AAAA,QAC7C,eAAe,KAAK;AAAA,QACpB,sBAAsB;AAAA,QACtB,oBAAoB,sBAAsB;AAAA,QAC1C,qBAAqB,uBAAuB;AAAA,QAC5C;AAAA,QACA;AAAA,MACF,CAAC;AAED,UAAI,WAAW;AACb,aAAK,iBAAiB;AAAA,UACpB,YAAY,KAAK;AAAA,UACjB,YAAY;AAAA,UACZ,oBAAoB,sBAAsB;AAAA,UAC1C,qBAAqB,uBAAuB;AAAA,QAC9C,CAAC;AAGD,aAAK,kBAAkB;AACvB,aAAK,4BAA4B,CAAC;AAClC,aAAK,mBAAmB;AACxB,aAAK,0BAA0B;AAC/B,aAAK,kBAAkB;AAAA,MACzB;AAEA,WAAK,oBAAoB;AAAA,IAC3B;AAGF,eAAK,kBAAL,mBAAoB;AAEpB,SAAK,gBAAgB,kBAAK;AAAA,MACxB,cAAc,KAAK,kBAAkB,KAAK,yBAAyB,KAAK,eAAe;AAAA,IACzF;AAEA,SAAK,cAAc,OAChB,KAAK,MAAM;AACV,WAAK,OAAO,MAAM,8BAA8B;AAAA,IAClD,CAAC,EACA,MAAM,CAAC,QAAiB;AACvB,UAAI,eAAe,SAAS,IAAI,QAAQ,SAAS,4BAA4B,GAAG;AAE9E;AAAA,MACF;AACA,WAAK,OAAO,MAAM,KAAK,8BAA8B;AAAA,IACvD,CAAC;AAAA,EACL;AAAA,EAEA,MAAc,cAAc,KAA0B,QAAqB;AACzE,QAAI,CAAC,IAAK;AAEV,SAAK,OAAO,MAAM,gDAAgD;AAElE,UAAM,YAAY,MAAM,IAAI,KAAK,gBAAgB,CAAC,CAAC;AAEnD,QAAI,OAAO,WAAW,cAAc,KAAM;AAE1C,QAAI,qBAAqB,2BAAgB;AACvC,YAAM,SAAS,UAAU,UAAU;AAEnC,aAAO,iBAAiB,SAAS,YAAY;AAC3C,YAAI;AACF,iBAAO,YAAY;AACnB,iBAAM,uCAAW;AAAA,QACnB,SAAS,GAAG;AACV,eAAK,OAAO,MAAM,8CAA8C,CAAC;AAAA,QACnE;AAAA,MACF,CAAC;AAED,UAAI;AACF,eAAO,MAAM;AACX,cAAI,OAAO,QAAS;AAEpB,gBAAM,EAAE,MAAM,OAAO,GAAG,IAAI,MAAM,OAAO,KAAK;AAC9C,cAAI,KAAM;AAEV,cAAI,OAAO,OAAO,UAAU;AAC1B,kBAAM,IAAI,MAAM,iCAAiC;AAAA,UACnD,OAAO;AACL,kBAAM,KAAK,WAAW,EAAE;AAAA,UAC1B;AAAA,QACF;AAAA,MACF,SAAS,GAAG;AACV,gBAAI,mDAA2B,CAAC,GAAG;AACjC;AAAA,QACF;AACA,aAAK,OAAO,MAAM,EAAE,OAAO,EAAE,GAAG,wCAAwC;AAAA,MAC1E,UAAE;AACA,eAAO,YAAY;AACnB,YAAI;AACF,gBAAM,UAAU,OAAO;AAAA,QACzB,SAAS,GAAG;AACV,eAAK,OAAO;AAAA,YACV;AAAA,YACA;AAAA,UACF;AAAA,QACF;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAAA,EAEA,MAAc,cAAc,KAAsB,QAAqB;AA/jBzE;AAgkBI,QAAI,CAAC,IAAK;AAEV,UAAM,YAAY,IAAI,OAAO;AAC7B,cAAU,kBAAkB,KAAK,cAAc;AAE/C,UAAM,eAAe,MAAM;AACzB,gBAAU,kBAAkB;AAC5B,gBAAU,MAAM;AAChB,aAAO,oBAAoB,SAAS,YAAY;AAAA,IAClD;AACA,WAAO,iBAAiB,SAAS,YAAY;AAE7C,QAAI;AACF,uBAAiB,MAAM,WAAW;AAChC,YAAI,OAAO,QAAS;AAEpB,gBAAQ,GAAG,MAAM;AAAA,UACf,KAAK,wBAAa;AAChB,iBAAK,OAAO,MAAM,2BAA2B;AAC7C,iBAAK,MAAM,gBAAgB,EAAE;AAC7B,iBAAK,WAAW;AAEhB,gBAAI,CAAC,KAAK,cAAc;AACtB,mBAAK,eAAe,wBAAO,UAAU;AAAA,gBACnC,MAAM;AAAA,gBACN,SAAS,KAAK;AAAA,cAChB,CAAC;AAAA,YACH;AAGA,gBAAI,GAAG,OAAO,SAAS,KAAK,GAAG,OAAO,CAAC,GAAG;AACxC,mBAAK,aAAa,GAAG,OAAO,CAAC,EAAE;AAAA,YACjC;AAGA,gBAAI,KAAK,iBAAiB,KAAK,qBAAqB;AAClD,mBAAK,uBAAuB,GAAG,gBAAgB,KAAK,YAAY;AAAA,YAClE;AAEA,uBAAK,kBAAL,mBAAoB;AACpB;AAAA,UACF,KAAK,wBAAa;AAChB,iBAAK,MAAM,mBAAmB,EAAE;AAEhC,gBAAI,GAAG,uBAAuB,GAAK;AACjC,mBAAK,mBAAmB,KAAK,IAAI;AAEjC,kBAAI,KAAK,oBAAoB,QAAW;AACtC,qBAAK,kBAAkB,KAAK,IAAI;AAAA,cAClC;AAAA,YACF;AACA;AAAA,UACF,KAAK,wBAAa;AAChB,iBAAK,OAAO,MAAM,yBAAyB;AAC3C,iBAAK,MAAM,cAAc,EAAE;AAG3B,iBAAK,WAAW;AAGhB,gBAAI,KAAK,iBAAiB,KAAK,qBAAqB;AAClD,mBAAK,qBAAqB;AAAA,YAC5B;AAEA,gBACE,KAAK,wBACJ,KAAK,sBAAsB,SAAS,KAAK,mBAC1C;AACA,oBAAM,UAAU,KAAK,MAAM,gBAAgB;AAC3C,mBAAK,gBAAgB,OAAO;AAAA,YAC9B;AACA;AAAA,QACJ;AAAA,MACF;AAAA,IACF,SAAS,GAAG;AACV,WAAK,OAAO,MAAM,GAAG,mBAAmB;AAAA,IAC1C,UAAE;AACA,WAAK,OAAO,MAAM,iBAAiB;AAAA,IACrC;AAAA,EACF;AAAA,EAEA,MAAc,uBACZ,sBACA,QACA;AAEA,SAAK,qBAAqB,qBAAqB,aAAa;AAG5D,UAAM,SAAS,KAAK,wBAAwB,UAAU;AAEtD,UAAM,eAAe,YAAY;AA3pBrC;AA4pBM,UAAI;AACF,eAAO,CAAC,OAAO,SAAS;AACtB,gBAAM,EAAE,MAAM,OAAO,MAAM,IAAI,MAAM,OAAO,KAAK;AACjD,cAAI,KAAM;AACV,kBAAM,UAAK,uBAAL,mBAAyB,UAAU;AAAA,QAC3C;AAAA,MACF,SAAS,GAAG;AACV,YAAI,CAAC,OAAO,SAAS;AACnB,eAAK,OAAO,MAAM,GAAG,+CAA+C;AAAA,QACtE;AAAA,MACF,UAAE;AACA,eAAO,YAAY;AAAA,MACrB;AAAA,IACF,GAAG;AAGH,UAAM,cAAc,KAAK,mBAAmB;AAC5C,UAAM,cAAc,YAAY,UAAU;AAE1C,UAAM,eAAe,MAAM;AA/qB/B;AAgrBM,kBAAY,YAAY;AACxB,iBAAK,uBAAL,mBAAyB;AACzB,aAAO,oBAAoB,SAAS,YAAY;AAAA,IAClD;AACA,WAAO,iBAAiB,SAAS,YAAY;AAE7C,QAAI;AACF,aAAO,CAAC,OAAO,SAAS;AACtB,cAAM,EAAE,MAAM,OAAO,GAAG,IAAI,MAAM,YAAY,KAAK;AACnD,YAAI,KAAM;AAEV,aAAK,OAAO,MAAM,EAAE,MAAM,GAAG,MAAM,aAAa,GAAG,YAAY,GAAG,oBAAoB;AACtF,aAAK,MAAM,eAAe,EAAE;AAAA,MAC9B;AAAA,IACF,SAAS,GAAG;AACV,UAAI,CAAC,OAAO,SAAS;AACnB,aAAK,OAAO,MAAM,GAAG,4BAA4B;AAAA,MACnD;AAAA,IACF,UAAE;AACA,WAAK,OAAO,MAAM,0BAA0B;AAC5C,YAAM;AAAA,IACR;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,uBAA6B;AAC3B,SAAK,gBAAgB;AAErB,QAAI,CAAC,KAAK,uBAAuB,CAAC,KAAK,oBAAoB;AACzD;AAAA,IACF;AAEA,SAAK,mBAAmB,UAAU,qDAA2B,cAAc,CAAC;AAAA,EAC9E;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,qBAA2B;AACzB,QAAI,CAAC,KAAK,uBAAuB,CAAC,KAAK,oBAAoB;AACzD,WAAK,gBAAgB;AACrB;AAAA,IACF;AAEA,SAAK,mBAAmB,UAAU,qDAA2B,YAAY,CAAC;AAE1E,QAAI,KAAK,eAAe;AAEtB,WAAK,qBAAqB;AAAA,IAC5B;AAEA,SAAK,gBAAgB;AAAA,EACvB;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,uBAAuB,gBAAwB,kBAA+B;AAC5E,QAAI,CAAC,KAAK,uBAAuB,CAAC,KAAK,oBAAoB;AACzD;AAAA,IACF;AAEA,QAAI,KAAK,iBAAiB,kBAAkB;AAC1C,WAAK,mBAAmB;AAAA,QACtB,qDAA2B,qBAAqB,gBAAgB,gBAAgB;AAAA,MAClF;AAAA,IACF;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,uBAA6B;AAC3B,QAAI,CAAC,KAAK,uBAAuB,CAAC,KAAK,oBAAoB;AACzD;AAAA,IACF;AAEA,SAAK,mBAAmB,UAAU,qDAA2B,mBAAmB,CAAC;AAAA,EACnF;AAAA,EAEA,oBAAoB,aAAyC;AAC3D,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,yBAAyB;AACvB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA,EAEA,gBAAgB;AA9wBlB;AA+wBI,SAAK,kBAAkB;AACvB,SAAK,yBAAyB;AAC9B,SAAK,2BAA2B;AAChC,SAAK,4BAA4B,CAAC;AAClC,SAAK,oBAAoB;AAEzB,eAAK,YAAL,mBAAc,gBAAgB,QAAQ,MAAM;AAC1C,WAAK,UAAU,kBAAK,KAAK,CAAC,EAAE,OAAO,MAAM,KAAK,cAAc,KAAK,KAAK,MAAM,CAAC;AAC7E,WAAK,QAAQ,OAAO,MAAM,CAAC,QAAQ;AACjC,aAAK,OAAO,MAAM,2BAA2B,GAAG,EAAE;AAAA,MACpD,CAAC;AAAA,IACH;AAAA,EACF;AAAA,EAEA,eAAe,eAAwB;AA7xBzC;AA8xBI,UAAM,qBACJ,CAAC,gBAAwB,QACzB,OAAO,eAAgC;AACrC,UAAI,KAAK,IAAI,IAAI,KAAK,0BAA0B,eAAe;AAE7D,YAAI,iBAAiB,KAAK,eAAe,QAAW;AAClD,gBAAM,aAAa,KAAK,MAAM,KAAK,aAAa,GAAG;AACnD,gBAAM,UAAU,IAAI,WAAW,aAAa,CAAC;AAC7C,gBAAM,eAAe,IAAI,2BAAW,SAAS,KAAK,YAAY,GAAG,UAAU;AAC3E,eAAK,mBAAmB,MAAM,YAAY;AAAA,QAC5C;AAGA,kBAAM,oBAAM,eAAe,EAAE,QAAQ,WAAW,OAAO,CAAC;AAAA,MAC1D;AAEA,UAAI,KAAK,wBAAwB;AAE/B,aAAK,kBAAkB,GAAG,KAAK,eAAe,IAAI,KAAK,sBAAsB,GAAG,KAAK;AAAA,MACvF;AACA,WAAK,yBAAyB;AAE9B,YAAM,UAAU,KAAK,MAAM,gBAAgB;AAC3C,WAAK,OAAO,MAAM,yCAAyC;AAC3D,WAAK,gBAAgB,OAAO;AAC5B,WAAK,oBAAoB;AAAA,IAC3B;AAGF,eAAK,uBAAL,mBAAyB;AACzB,SAAK,qBAAqB,kBAAK,KAAK,mBAAmB,CAAC;AAExD,SAAK,mBAAmB,OACrB,KAAK,MAAM;AACV,WAAK,OAAO,MAAM,qBAAqB;AAAA,IACzC,CAAC,EACA,MAAM,CAAC,QAAiB;AACvB,WAAK,OAAO,MAAM,KAAK,iCAAiC;AAAA,IAC1D,CAAC;AAAA,EACL;AAAA,EAEA,MAAM,QAAQ;AAv0BhB;AAw0BI,SAAK,uBAAuB;AAC5B,SAAK,mBAAmB,YAAY;AACpC,YAAM,UAAK,uBAAL,mBAAyB;AAC/B,YAAM,UAAK,YAAL,mBAAc;AACpB,YAAM,UAAK,YAAL,mBAAc;AACpB,YAAM,UAAK,kBAAL,mBAAoB;AAC1B,YAAM,UAAK,qBAAL,mBAAuB;AAC7B,YAAM,UAAK,uBAAL,mBAAyB;AAAA,EACjC;AAAA,EAEQ,iBAAiB;AAAA,IACvB;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAKS;AACP,QAAI,KAAK,cAAc;AACrB,WAAK,aAAa,cAAc;AAAA,QAC9B,CAAC,4BAAW,oBAAoB,GAAG;AAAA,QACnC,CAAC,4BAAW,0BAA0B,GAAG;AAAA,QACzC,CAAC,4BAAW,wBAAwB,GAAG;AAAA,QACvC,CAAC,4BAAW,sBAAsB,GAAG;AAAA,MACvC,CAAC;AACD,WAAK,aAAa,IAAI;AACtB,WAAK,eAAe;AAAA,IACtB;AAAA,EACF;AAAA,EAEA,IAAY,uBAAuB;AACjC,WAAO,CAAC,OAAO,MAAS,EAAE,SAAS,KAAK,iBAAiB;AAAA,EAC3D;AACF;","names":[]}

package/dist/voice/audio_recognition.d.cts

@@ -0,0 +1,127 @@
+/// <reference types="node" resolution-mode="require"/>
+import { AudioFrame } from '@livekit/rtc-node';
+import type { Context, Span } from '@opentelemetry/api';
+import { ReadableStream } from 'node:stream/web';
+import type { AdaptiveInterruptionDetector } from '../inference/interruption/AdaptiveInterruptionDetector.js';
+import type { InterruptionEvent } from '../inference/interruption/interruption.js';
+import { type ChatContext } from '../llm/chat_context.js';
+import { type SpeechEvent } from '../stt/stt.js';
+import { type VAD, type VADEvent } from '../vad.js';
+import type { TurnDetectionMode } from './agent_session.js';
+import type { STTNode } from './io.js';
+export interface EndOfTurnInfo {
+    newTranscript: string;
+    transcriptConfidence: number;
+    transcriptionDelay: number;
+    endOfUtteranceDelay: number;
+    startedSpeakingAt: number | undefined;
+    stoppedSpeakingAt: number | undefined;
+}
+export interface PreemptiveGenerationInfo {
+    newTranscript: string;
+    transcriptConfidence: number;
+}
+export interface RecognitionHooks {
+    onStartOfSpeech: (ev: VADEvent) => void;
+    onVADInferenceDone: (ev: VADEvent) => void;
+    onEndOfSpeech: (ev: VADEvent) => void;
+    onInterimTranscript: (ev: SpeechEvent) => void;
+    onFinalTranscript: (ev: SpeechEvent) => void;
+    onEndOfTurn: (info: EndOfTurnInfo) => Promise<boolean>;
+    onPreemptiveGeneration: (info: PreemptiveGenerationInfo) => void;
+    onInterruption: (ev: InterruptionEvent) => void;
+    retrieveChatCtx: () => ChatContext;
+}
+export interface _TurnDetector {
+    unlikelyThreshold: (language?: string) => Promise<number | undefined>;
+    supportsLanguage: (language?: string) => Promise<boolean>;
+    predictEndOfTurn(chatCtx: ChatContext): Promise<number>;
+}
+export interface AudioRecognitionOptions {
+    recognitionHooks: RecognitionHooks;
+    stt?: STTNode;
+    vad?: VAD;
+    interruptionDetector?: AdaptiveInterruptionDetector;
+    turnDetector?: _TurnDetector;
+    turnDetectionMode?: Exclude<TurnDetectionMode, _TurnDetector>;
+    minEndpointingDelay: number;
+    maxEndpointingDelay: number;
+    rootSpanContext?: Context;
+}
+export declare class AudioRecognition {
+    private hooks;
+    private stt?;
+    private vad?;
+    private turnDetector?;
+    private turnDetectionMode?;
+    private minEndpointingDelay;
+    private maxEndpointingDelay;
+    private lastLanguage?;
+    private rootSpanContext?;
+    private deferredInputStream;
+    private logger;
+    private lastFinalTranscriptTime;
+    private audioTranscript;
+    private audioInterimTranscript;
+    private audioPreflightTranscript;
+    private finalTranscriptConfidence;
+    private lastSpeakingTime;
+    private speechStartTime;
+    private userTurnCommitted;
+    private speaking;
+    private sampleRate?;
+    private userTurnSpan?;
+    private vadInputStream;
+    private sttInputStream;
+    private interruptionInputStream;
+    private silenceAudioTransform;
+    private silenceAudioWriter;
+    private bounceEOUTask?;
+    private commitUserTurnTask?;
+    private vadTask?;
+    private sttTask?;
+    private interruptionTask?;
+    private interruptionDetector?;
+    private interruptionStream?;
+    private interruptionEnabled;
+    private agentSpeaking;
+    constructor(opts: AudioRecognitionOptions);
+    /**
+     * Current transcript of the user's speech, including interim transcript if available.
+     */
+    get currentTranscript(): string;
+    start(): Promise<void>;
+    private onSTTEvent;
+    private runEOUDetection;
+    private createSttTask;
+    private createVadTask;
+    private createInterruptionTask;
+    /**
+     * Called when the agent starts speaking.
+     * Enables interruption detection by sending the agent-speech-started sentinel.
+     */
+    onStartOfAgentSpeech(): void;
+    /**
+     * Called when the agent stops speaking.
+     * Disables interruption detection by sending the agent-speech-ended sentinel.
+     */
+    onEndOfAgentSpeech(): void;
+    /**
+     * Called when user starts speaking while agent is speaking (overlap speech).
+     * This triggers the interruption detection inference.
+     */
+    onStartOfOverlapSpeech(speechDuration: number, userSpeakingSpan?: Span): void;
+    /**
+     * Called when user stops speaking during overlap.
+     * This ends the interruption detection inference for this overlap period.
+     */
+    onEndOfOverlapSpeech(): void;
+    setInputAudioStream(audioStream: ReadableStream<AudioFrame>): void;
+    detachInputAudioStream(): void;
+    clearUserTurn(): void;
+    commitUserTurn(audioDetached: boolean): void;
+    close(): Promise<void>;
+    private _endUserTurnSpan;
+    private get vadBaseTurnDetection();
+}
+//# sourceMappingURL=audio_recognition.d.ts.map

package/dist/voice/audio_recognition.d.ts

@@ -0,0 +1,127 @@
+/// <reference types="node" resolution-mode="require"/>
+import { AudioFrame } from '@livekit/rtc-node';
+import type { Context, Span } from '@opentelemetry/api';
+import { ReadableStream } from 'node:stream/web';
+import type { AdaptiveInterruptionDetector } from '../inference/interruption/AdaptiveInterruptionDetector.js';
+import type { InterruptionEvent } from '../inference/interruption/interruption.js';
+import { type ChatContext } from '../llm/chat_context.js';
+import { type SpeechEvent } from '../stt/stt.js';
+import { type VAD, type VADEvent } from '../vad.js';
+import type { TurnDetectionMode } from './agent_session.js';
+import type { STTNode } from './io.js';
+export interface EndOfTurnInfo {
+    newTranscript: string;
+    transcriptConfidence: number;
+    transcriptionDelay: number;
+    endOfUtteranceDelay: number;
+    startedSpeakingAt: number | undefined;
+    stoppedSpeakingAt: number | undefined;
+}
+export interface PreemptiveGenerationInfo {
+    newTranscript: string;
+    transcriptConfidence: number;
+}
+export interface RecognitionHooks {
+    onStartOfSpeech: (ev: VADEvent) => void;
+    onVADInferenceDone: (ev: VADEvent) => void;
+    onEndOfSpeech: (ev: VADEvent) => void;
+    onInterimTranscript: (ev: SpeechEvent) => void;
+    onFinalTranscript: (ev: SpeechEvent) => void;
+    onEndOfTurn: (info: EndOfTurnInfo) => Promise<boolean>;
+    onPreemptiveGeneration: (info: PreemptiveGenerationInfo) => void;
+    onInterruption: (ev: InterruptionEvent) => void;
+    retrieveChatCtx: () => ChatContext;
+}
+export interface _TurnDetector {
+    unlikelyThreshold: (language?: string) => Promise<number | undefined>;
+    supportsLanguage: (language?: string) => Promise<boolean>;
+    predictEndOfTurn(chatCtx: ChatContext): Promise<number>;
+}
+export interface AudioRecognitionOptions {
+    recognitionHooks: RecognitionHooks;
+    stt?: STTNode;
+    vad?: VAD;
+    interruptionDetector?: AdaptiveInterruptionDetector;
+    turnDetector?: _TurnDetector;
+    turnDetectionMode?: Exclude<TurnDetectionMode, _TurnDetector>;
+    minEndpointingDelay: number;
+    maxEndpointingDelay: number;
+    rootSpanContext?: Context;
+}
+export declare class AudioRecognition {
+    private hooks;
+    private stt?;
+    private vad?;
+    private turnDetector?;
+    private turnDetectionMode?;
+    private minEndpointingDelay;
+    private maxEndpointingDelay;
+    private lastLanguage?;
+    private rootSpanContext?;
+    private deferredInputStream;
+    private logger;
+    private lastFinalTranscriptTime;
+    private audioTranscript;
+    private audioInterimTranscript;
+    private audioPreflightTranscript;
+    private finalTranscriptConfidence;
+    private lastSpeakingTime;
+    private speechStartTime;
+    private userTurnCommitted;
+    private speaking;
+    private sampleRate?;
+    private userTurnSpan?;
+    private vadInputStream;
+    private sttInputStream;
+    private interruptionInputStream;
+    private silenceAudioTransform;
+    private silenceAudioWriter;
+    private bounceEOUTask?;
+    private commitUserTurnTask?;
+    private vadTask?;
+    private sttTask?;
+    private interruptionTask?;
+    private interruptionDetector?;
+    private interruptionStream?;
+    private interruptionEnabled;
+    private agentSpeaking;
+    constructor(opts: AudioRecognitionOptions);
+    /**
+     * Current transcript of the user's speech, including interim transcript if available.
+     */
+    get currentTranscript(): string;
+    start(): Promise<void>;
+    private onSTTEvent;
+    private runEOUDetection;
+    private createSttTask;
+    private createVadTask;
+    private createInterruptionTask;
+    /**
+     * Called when the agent starts speaking.
+     * Enables interruption detection by sending the agent-speech-started sentinel.
+     */
+    onStartOfAgentSpeech(): void;
+    /**
+     * Called when the agent stops speaking.
+     * Disables interruption detection by sending the agent-speech-ended sentinel.
+     */
+    onEndOfAgentSpeech(): void;
+    /**
+     * Called when user starts speaking while agent is speaking (overlap speech).
+     * This triggers the interruption detection inference.
+     */
+    onStartOfOverlapSpeech(speechDuration: number, userSpeakingSpan?: Span): void;
+    /**
+     * Called when user stops speaking during overlap.
+     * This ends the interruption detection inference for this overlap period.
+     */
+    onEndOfOverlapSpeech(): void;
+    setInputAudioStream(audioStream: ReadableStream<AudioFrame>): void;
+    detachInputAudioStream(): void;
+    clearUserTurn(): void;
+    commitUserTurn(audioDetached: boolean): void;
+    close(): Promise<void>;
+    private _endUserTurnSpan;
+    private get vadBaseTurnDetection();
+}
+//# sourceMappingURL=audio_recognition.d.ts.map

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

@@ -0,0 +1 @@
+{"version":3,"file":"audio_recognition.d.ts","sourceRoot":"","sources":["../../src/voice/audio_recognition.ts"],"names":[],"mappings":";AAGA,OAAO,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AAC/C,OAAO,KAAK,EAAE,OAAO,EAAE,IAAI,EAAE,MAAM,oBAAoB,CAAC;AAExD,OAAO,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AACjD,OAAO,KAAK,EAAE,4BAA4B,EAAE,MAAM,2DAA2D,CAAC;AAK9G,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,2CAA2C,CAAC;AACnF,OAAO,EAAE,KAAK,WAAW,EAAE,MAAM,wBAAwB,CAAC;AAK1D,OAAO,EAAE,KAAK,WAAW,EAAmB,MAAM,eAAe,CAAC;AAGlE,OAAO,EAAE,KAAK,GAAG,EAAE,KAAK,QAAQ,EAAgB,MAAM,WAAW,CAAC;AAClE,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,oBAAoB,CAAC;AAC5D,OAAO,KAAK,EAAE,OAAO,EAAE,MAAM,SAAS,CAAC;AAEvC,MAAM,WAAW,aAAa;IAC5B,aAAa,EAAE,MAAM,CAAC;IACtB,oBAAoB,EAAE,MAAM,CAAC;IAC7B,kBAAkB,EAAE,MAAM,CAAC;IAC3B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,iBAAiB,EAAE,MAAM,GAAG,SAAS,CAAC;IACtC,iBAAiB,EAAE,MAAM,GAAG,SAAS,CAAC;CACvC;AAED,MAAM,WAAW,wBAAwB;IACvC,aAAa,EAAE,MAAM,CAAC;IACtB,oBAAoB,EAAE,MAAM,CAAC;CAC9B;AAED,MAAM,WAAW,gBAAgB;IAC/B,eAAe,EAAE,CAAC,EAAE,EAAE,QAAQ,KAAK,IAAI,CAAC;IACxC,kBAAkB,EAAE,CAAC,EAAE,EAAE,QAAQ,KAAK,IAAI,CAAC;IAC3C,aAAa,EAAE,CAAC,EAAE,EAAE,QAAQ,KAAK,IAAI,CAAC;IACtC,mBAAmB,EAAE,CAAC,EAAE,EAAE,WAAW,KAAK,IAAI,CAAC;IAC/C,iBAAiB,EAAE,CAAC,EAAE,EAAE,WAAW,KAAK,IAAI,CAAC;IAC7C,WAAW,EAAE,CAAC,IAAI,EAAE,aAAa,KAAK,OAAO,CAAC,OAAO,CAAC,CAAC;IACvD,sBAAsB,EAAE,CAAC,IAAI,EAAE,wBAAwB,KAAK,IAAI,CAAC;IACjE,cAAc,EAAE,CAAC,EAAE,EAAE,iBAAiB,KAAK,IAAI,CAAC;IAEhD,eAAe,EAAE,MAAM,WAAW,CAAC;CACpC;AAED,MAAM,WAAW,aAAa;IAC5B,iBAAiB,EAAE,CAAC,QAAQ,CAAC,EAAE,MAAM,KAAK,OAAO,CAAC,MAAM,GAAG,SAAS,CAAC,CAAC;IACtE,gBAAgB,EAAE,CAAC,QAAQ,CAAC,EAAE,MAAM,KAAK,OAAO,CAAC,OAAO,CAAC,CAAC;IAC1D,gBAAgB,CAAC,OAAO,EAAE,WAAW,GAAG,OAAO,CAAC,MAAM,CAAC,CAAC;CACzD;AAED,MAAM,WAAW,uBAAuB;IACtC,gBAAgB,EAAE,gBAAgB,CAAC;IACnC,GAAG,CAAC,EAAE,OAAO,CAAC;IACd,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,oBAAoB,CAAC,EAAE,4BAA4B,CAAC;IACpD,YAAY,CAAC,EAAE,aAAa,CAAC;IAC7B,iBAAiB,CAAC,EAAE,OAAO,CAAC,iBAAiB,EAAE,aAAa,CAAC,CAAC;IAC9D,mBAAmB,EAAE,MAAM,CAAC;IAC5B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,eAAe,CAAC,EAAE,OAAO,CAAC;CAC3B;AAED,qBAAa,gBAAgB;IAC3B,OAAO,CAAC,KAAK,CAAmB;IAChC,OAAO,CAAC,GAAG,CAAC,CAAU;IACtB,OAAO,CAAC,GAAG,CAAC,CAAM;IAClB,OAAO,CAAC,YAAY,CAAC,CAAgB;IACrC,OAAO,CAAC,iBAAiB,CAAC,CAA4C;IACtE,OAAO,CAAC,mBAAmB,CAAS;IACpC,OAAO,CAAC,mBAAmB,CAAS;IACpC,OAAO,CAAC,YAAY,CAAC,CAAS;IAC9B,OAAO,CAAC,eAAe,CAAC,CAAU;IAElC,OAAO,CAAC,mBAAmB,CAAqC;IAChE,OAAO,CAAC,MAAM,CAAS;IACvB,OAAO,CAAC,uBAAuB,CAAK;IACpC,OAAO,CAAC,eAAe,CAAM;IAC7B,OAAO,CAAC,sBAAsB,CAAM;IACpC,OAAO,CAAC,wBAAwB,CAAM;IACtC,OAAO,CAAC,yBAAyB,CAAgB;IACjD,OAAO,CAAC,gBAAgB,CAAqB;IAC7C,OAAO,CAAC,eAAe,CAAqB;IAC5C,OAAO,CAAC,iBAAiB,CAAS;IAClC,OAAO,CAAC,QAAQ,CAAS;IACzB,OAAO,CAAC,UAAU,CAAC,CAAS;IAE5B,OAAO,CAAC,YAAY,CAAC,CAAO;IAE5B,OAAO,CAAC,cAAc,CAA6B;IACnD,OAAO,CAAC,cAAc,CAA6B;IACnD,OAAO,CAAC,uBAAuB,CAA6B;IAC5D,OAAO,CAAC,qBAAqB,CAAuC;IACpE,OAAO,CAAC,kBAAkB,CAA0C;IAGpE,OAAO,CAAC,aAAa,CAAC,CAAa;IACnC,OAAO,CAAC,kBAAkB,CAAC,CAAa;IACxC,OAAO,CAAC,OAAO,CAAC,CAAa;IAC7B,OAAO,CAAC,OAAO,CAAC,CAAa;IAC7B,OAAO,CAAC,gBAAgB,CAAC,CAAa;IAGtC,OAAO,CAAC,oBAAoB,CAAC,CAA+B;IAC5D,OAAO,CAAC,kBAAkB,CAAC,CAAyB;IACpD,OAAO,CAAC,mBAAmB,CAAS;IACpC,OAAO,CAAC,aAAa,CAAS;gBAElB,IAAI,EAAE,uBAAuB;IAwBzC;;OAEG;IACH,IAAI,iBAAiB,IAAI,MAAM,CAK9B;IAEK,KAAK;YAqBG,UAAU;IAgMxB,OAAO,CAAC,eAAe;YAgKT,aAAa;YAqDb,aAAa;YAkFb,sBAAsB;IAuDpC;;;OAGG;IACH,oBAAoB,IAAI,IAAI;IAU5B;;;OAGG;IACH,kBAAkB,IAAI,IAAI;IAgB1B;;;OAGG;IACH,sBAAsB,CAAC,cAAc,EAAE,MAAM,EAAE,gBAAgB,CAAC,EAAE,IAAI,GAAG,IAAI;IAY7E;;;OAGG;IACH,oBAAoB,IAAI,IAAI;IAQ5B,mBAAmB,CAAC,WAAW,EAAE,cAAc,CAAC,UAAU,CAAC;IAI3D,sBAAsB;IAItB,aAAa;IAeb,cAAc,CAAC,aAAa,EAAE,OAAO;IA0C/B,KAAK;IAWX,OAAO,CAAC,gBAAgB;IAuBxB,OAAO,KAAK,oBAAoB,GAE/B;CACF"}