@livekit/agents 1.0.44 → 1.0.46

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (157) hide show
  1. package/dist/ipc/supervised_proc.cjs +1 -1
  2. package/dist/ipc/supervised_proc.cjs.map +1 -1
  3. package/dist/ipc/supervised_proc.js +1 -1
  4. package/dist/ipc/supervised_proc.js.map +1 -1
  5. package/dist/llm/llm.cjs +1 -1
  6. package/dist/llm/llm.cjs.map +1 -1
  7. package/dist/llm/llm.js +1 -1
  8. package/dist/llm/llm.js.map +1 -1
  9. package/dist/log.cjs +13 -9
  10. package/dist/log.cjs.map +1 -1
  11. package/dist/log.d.cts +1 -1
  12. package/dist/log.d.ts +1 -1
  13. package/dist/log.d.ts.map +1 -1
  14. package/dist/log.js +13 -9
  15. package/dist/log.js.map +1 -1
  16. package/dist/stream/index.cjs +3 -0
  17. package/dist/stream/index.cjs.map +1 -1
  18. package/dist/stream/index.d.cts +1 -0
  19. package/dist/stream/index.d.ts +1 -0
  20. package/dist/stream/index.d.ts.map +1 -1
  21. package/dist/stream/index.js +2 -0
  22. package/dist/stream/index.js.map +1 -1
  23. package/dist/stream/multi_input_stream.cjs +139 -0
  24. package/dist/stream/multi_input_stream.cjs.map +1 -0
  25. package/dist/stream/multi_input_stream.d.cts +55 -0
  26. package/dist/stream/multi_input_stream.d.ts +55 -0
  27. package/dist/stream/multi_input_stream.d.ts.map +1 -0
  28. package/dist/stream/multi_input_stream.js +115 -0
  29. package/dist/stream/multi_input_stream.js.map +1 -0
  30. package/dist/stream/multi_input_stream.test.cjs +340 -0
  31. package/dist/stream/multi_input_stream.test.cjs.map +1 -0
  32. package/dist/stream/multi_input_stream.test.js +339 -0
  33. package/dist/stream/multi_input_stream.test.js.map +1 -0
  34. package/dist/stt/stt.cjs +2 -2
  35. package/dist/stt/stt.cjs.map +1 -1
  36. package/dist/stt/stt.js +2 -2
  37. package/dist/stt/stt.js.map +1 -1
  38. package/dist/telemetry/trace_types.cjs +42 -0
  39. package/dist/telemetry/trace_types.cjs.map +1 -1
  40. package/dist/telemetry/trace_types.d.cts +14 -0
  41. package/dist/telemetry/trace_types.d.ts +14 -0
  42. package/dist/telemetry/trace_types.d.ts.map +1 -1
  43. package/dist/telemetry/trace_types.js +28 -0
  44. package/dist/telemetry/trace_types.js.map +1 -1
  45. package/dist/tts/fallback_adapter.cjs +466 -0
  46. package/dist/tts/fallback_adapter.cjs.map +1 -0
  47. package/dist/tts/fallback_adapter.d.cts +110 -0
  48. package/dist/tts/fallback_adapter.d.ts +110 -0
  49. package/dist/tts/fallback_adapter.d.ts.map +1 -0
  50. package/dist/tts/fallback_adapter.js +442 -0
  51. package/dist/tts/fallback_adapter.js.map +1 -0
  52. package/dist/tts/index.cjs +3 -0
  53. package/dist/tts/index.cjs.map +1 -1
  54. package/dist/tts/index.d.cts +1 -0
  55. package/dist/tts/index.d.ts +1 -0
  56. package/dist/tts/index.d.ts.map +1 -1
  57. package/dist/tts/index.js +2 -0
  58. package/dist/tts/index.js.map +1 -1
  59. package/dist/tts/tts.cjs +2 -2
  60. package/dist/tts/tts.cjs.map +1 -1
  61. package/dist/tts/tts.js +2 -2
  62. package/dist/tts/tts.js.map +1 -1
  63. package/dist/utils.cjs +13 -0
  64. package/dist/utils.cjs.map +1 -1
  65. package/dist/utils.d.cts +1 -0
  66. package/dist/utils.d.ts +1 -0
  67. package/dist/utils.d.ts.map +1 -1
  68. package/dist/utils.js +13 -0
  69. package/dist/utils.js.map +1 -1
  70. package/dist/vad.cjs +11 -10
  71. package/dist/vad.cjs.map +1 -1
  72. package/dist/vad.d.cts +5 -3
  73. package/dist/vad.d.ts +5 -3
  74. package/dist/vad.d.ts.map +1 -1
  75. package/dist/vad.js +11 -10
  76. package/dist/vad.js.map +1 -1
  77. package/dist/voice/agent_activity.cjs +35 -10
  78. package/dist/voice/agent_activity.cjs.map +1 -1
  79. package/dist/voice/agent_activity.d.cts +1 -0
  80. package/dist/voice/agent_activity.d.ts +1 -0
  81. package/dist/voice/agent_activity.d.ts.map +1 -1
  82. package/dist/voice/agent_activity.js +35 -10
  83. package/dist/voice/agent_activity.js.map +1 -1
  84. package/dist/voice/agent_session.cjs +19 -7
  85. package/dist/voice/agent_session.cjs.map +1 -1
  86. package/dist/voice/agent_session.d.cts +3 -2
  87. package/dist/voice/agent_session.d.ts +3 -2
  88. package/dist/voice/agent_session.d.ts.map +1 -1
  89. package/dist/voice/agent_session.js +19 -7
  90. package/dist/voice/agent_session.js.map +1 -1
  91. package/dist/voice/audio_recognition.cjs +85 -36
  92. package/dist/voice/audio_recognition.cjs.map +1 -1
  93. package/dist/voice/audio_recognition.d.cts +22 -1
  94. package/dist/voice/audio_recognition.d.ts +22 -1
  95. package/dist/voice/audio_recognition.d.ts.map +1 -1
  96. package/dist/voice/audio_recognition.js +89 -36
  97. package/dist/voice/audio_recognition.js.map +1 -1
  98. package/dist/voice/audio_recognition_span.test.cjs +233 -0
  99. package/dist/voice/audio_recognition_span.test.cjs.map +1 -0
  100. package/dist/voice/audio_recognition_span.test.js +232 -0
  101. package/dist/voice/audio_recognition_span.test.js.map +1 -0
  102. package/dist/voice/io.cjs +6 -3
  103. package/dist/voice/io.cjs.map +1 -1
  104. package/dist/voice/io.d.cts +3 -2
  105. package/dist/voice/io.d.ts +3 -2
  106. package/dist/voice/io.d.ts.map +1 -1
  107. package/dist/voice/io.js +6 -3
  108. package/dist/voice/io.js.map +1 -1
  109. package/dist/voice/recorder_io/recorder_io.cjs +3 -1
  110. package/dist/voice/recorder_io/recorder_io.cjs.map +1 -1
  111. package/dist/voice/recorder_io/recorder_io.d.ts.map +1 -1
  112. package/dist/voice/recorder_io/recorder_io.js +3 -1
  113. package/dist/voice/recorder_io/recorder_io.js.map +1 -1
  114. package/dist/voice/room_io/_input.cjs +23 -20
  115. package/dist/voice/room_io/_input.cjs.map +1 -1
  116. package/dist/voice/room_io/_input.d.cts +2 -2
  117. package/dist/voice/room_io/_input.d.ts +2 -2
  118. package/dist/voice/room_io/_input.d.ts.map +1 -1
  119. package/dist/voice/room_io/_input.js +13 -9
  120. package/dist/voice/room_io/_input.js.map +1 -1
  121. package/dist/voice/room_io/room_io.cjs +9 -0
  122. package/dist/voice/room_io/room_io.cjs.map +1 -1
  123. package/dist/voice/room_io/room_io.d.cts +3 -1
  124. package/dist/voice/room_io/room_io.d.ts +3 -1
  125. package/dist/voice/room_io/room_io.d.ts.map +1 -1
  126. package/dist/voice/room_io/room_io.js +9 -0
  127. package/dist/voice/room_io/room_io.js.map +1 -1
  128. package/dist/voice/utils.cjs +47 -0
  129. package/dist/voice/utils.cjs.map +1 -0
  130. package/dist/voice/utils.d.cts +4 -0
  131. package/dist/voice/utils.d.ts +4 -0
  132. package/dist/voice/utils.d.ts.map +1 -0
  133. package/dist/voice/utils.js +23 -0
  134. package/dist/voice/utils.js.map +1 -0
  135. package/package.json +1 -1
  136. package/src/ipc/supervised_proc.ts +1 -1
  137. package/src/llm/llm.ts +1 -1
  138. package/src/log.ts +22 -11
  139. package/src/stream/index.ts +1 -0
  140. package/src/stream/multi_input_stream.test.ts +540 -0
  141. package/src/stream/multi_input_stream.ts +172 -0
  142. package/src/stt/stt.ts +2 -2
  143. package/src/telemetry/trace_types.ts +18 -0
  144. package/src/tts/fallback_adapter.ts +579 -0
  145. package/src/tts/index.ts +1 -0
  146. package/src/tts/tts.ts +2 -2
  147. package/src/utils.ts +16 -0
  148. package/src/vad.ts +12 -11
  149. package/src/voice/agent_activity.ts +25 -0
  150. package/src/voice/agent_session.ts +17 -11
  151. package/src/voice/audio_recognition.ts +114 -38
  152. package/src/voice/audio_recognition_span.test.ts +261 -0
  153. package/src/voice/io.ts +7 -4
  154. package/src/voice/recorder_io/recorder_io.ts +2 -1
  155. package/src/voice/room_io/_input.ts +16 -10
  156. package/src/voice/room_io/room_io.ts +12 -0
  157. package/src/voice/utils.ts +29 -0
@@ -1 +1 @@
1
- {"version":3,"sources":["../../src/voice/agent_session.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame, Room } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport type { Context, Span } from '@opentelemetry/api';\nimport { ROOT_CONTEXT, context as otelContext, trace } from '@opentelemetry/api';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport {\n LLM as InferenceLLM,\n STT as InferenceSTT,\n TTS as InferenceTTS,\n type LLMModels,\n type STTModelString,\n type TTSModelString,\n} from '../inference/index.js';\nimport { type JobContext, getJobContext } from '../job.js';\nimport type { FunctionCall, FunctionCallOutput } from '../llm/chat_context.js';\nimport { AgentHandoffItem, ChatContext, ChatMessage } from '../llm/chat_context.js';\nimport type { LLM, RealtimeModel, RealtimeModelError, ToolChoice } from '../llm/index.js';\nimport type { LLMError } from '../llm/llm.js';\nimport { log } from '../log.js';\nimport type { STT } from '../stt/index.js';\nimport type { STTError } from '../stt/stt.js';\nimport { traceTypes, tracer } from '../telemetry/index.js';\nimport type { TTS, TTSError } from '../tts/tts.js';\nimport {\n DEFAULT_API_CONNECT_OPTIONS,\n DEFAULT_SESSION_CONNECT_OPTIONS,\n type ResolvedSessionConnectOptions,\n type SessionConnectOptions,\n} from '../types.js';\nimport type { VAD } from '../vad.js';\nimport type { Agent } from './agent.js';\nimport { AgentActivity } from './agent_activity.js';\nimport type { _TurnDetector } from './audio_recognition.js';\nimport {\n type AgentEvent,\n AgentSessionEventTypes,\n type AgentState,\n type AgentStateChangedEvent,\n type CloseEvent,\n CloseReason,\n type ConversationItemAddedEvent,\n type ErrorEvent,\n type FunctionToolsExecutedEvent,\n type MetricsCollectedEvent,\n type ShutdownReason,\n type SpeechCreatedEvent,\n type UserInputTranscribedEvent,\n type UserState,\n type UserStateChangedEvent,\n createAgentStateChangedEvent,\n createCloseEvent,\n createConversationItemAddedEvent,\n createUserStateChangedEvent,\n} from './events.js';\nimport { AgentInput, AgentOutput } from './io.js';\nimport { RecorderIO } from './recorder_io/index.js';\nimport { RoomIO, type RoomInputOptions, type RoomOutputOptions } from './room_io/index.js';\nimport type { UnknownUserData } from './run_context.js';\nimport type { SpeechHandle } from './speech_handle.js';\nimport { RunResult } from './testing/run_result.js';\n\nexport interface VoiceOptions {\n allowInterruptions: boolean;\n discardAudioIfUninterruptible: boolean;\n minInterruptionDuration: number;\n minInterruptionWords: number;\n minEndpointingDelay: number;\n maxEndpointingDelay: number;\n maxToolSteps: number;\n preemptiveGeneration: boolean;\n userAwayTimeout?: number | null;\n useTtsAlignedTranscript: boolean;\n}\n\nconst defaultVoiceOptions: VoiceOptions = {\n allowInterruptions: true,\n discardAudioIfUninterruptible: true,\n minInterruptionDuration: 500,\n minInterruptionWords: 0,\n minEndpointingDelay: 500,\n maxEndpointingDelay: 6000,\n maxToolSteps: 3,\n preemptiveGeneration: false,\n userAwayTimeout: 15.0,\n useTtsAlignedTranscript: true,\n} as const;\n\nexport type TurnDetectionMode = 'stt' | 'vad' | 'realtime_llm' | 'manual' | _TurnDetector;\n\nexport type AgentSessionCallbacks = {\n [AgentSessionEventTypes.UserInputTranscribed]: (ev: UserInputTranscribedEvent) => void;\n [AgentSessionEventTypes.AgentStateChanged]: (ev: AgentStateChangedEvent) => void;\n [AgentSessionEventTypes.UserStateChanged]: (ev: UserStateChangedEvent) => void;\n [AgentSessionEventTypes.ConversationItemAdded]: (ev: ConversationItemAddedEvent) => void;\n [AgentSessionEventTypes.FunctionToolsExecuted]: (ev: FunctionToolsExecutedEvent) => void;\n [AgentSessionEventTypes.MetricsCollected]: (ev: MetricsCollectedEvent) => void;\n [AgentSessionEventTypes.SpeechCreated]: (ev: SpeechCreatedEvent) => void;\n [AgentSessionEventTypes.Error]: (ev: ErrorEvent) => void;\n [AgentSessionEventTypes.Close]: (ev: CloseEvent) => void;\n};\n\nexport type AgentSessionOptions<UserData = UnknownUserData> = {\n turnDetection?: TurnDetectionMode;\n stt?: STT | STTModelString;\n vad?: VAD;\n llm?: LLM | RealtimeModel | LLMModels;\n tts?: TTS | TTSModelString;\n userData?: UserData;\n voiceOptions?: Partial<VoiceOptions>;\n connOptions?: SessionConnectOptions;\n};\n\nexport class AgentSession<\n UserData = UnknownUserData,\n> extends (EventEmitter as new () => TypedEmitter<AgentSessionCallbacks>) {\n vad?: VAD;\n stt?: STT;\n llm?: LLM | RealtimeModel;\n tts?: TTS;\n turnDetection?: TurnDetectionMode;\n\n readonly options: VoiceOptions;\n\n private agent?: Agent;\n private activity?: AgentActivity;\n private nextActivity?: AgentActivity;\n private started = false;\n private userState: UserState = 'listening';\n\n private roomIO?: RoomIO;\n private logger = log();\n\n private _chatCtx: ChatContext;\n private _userData: UserData | undefined;\n private _agentState: AgentState = 'initializing';\n\n private _input: AgentInput;\n private _output: AgentOutput;\n\n private closingTask: Promise<void> | null = null;\n private userAwayTimer: NodeJS.Timeout | null = null;\n\n // Connection options for STT, LLM, and TTS\n private _connOptions: ResolvedSessionConnectOptions;\n\n // Unrecoverable error counts, reset after agent speaking\n private llmErrorCounts = 0;\n private ttsErrorCounts = 0;\n\n private sessionSpan?: Span;\n private userSpeakingSpan?: Span;\n private agentSpeakingSpan?: Span;\n\n /** @internal */\n _recorderIO?: RecorderIO;\n\n /** @internal */\n rootSpanContext?: Context;\n\n /** @internal */\n _recordedEvents: AgentEvent[] = [];\n\n /** @internal */\n _enableRecording = false;\n\n /** @internal - Timestamp when the session started (milliseconds) */\n _startedAt?: number;\n\n /** @internal - Current run state for testing */\n _globalRunState?: RunResult;\n\n constructor(opts: AgentSessionOptions<UserData>) {\n super();\n\n const {\n vad,\n stt,\n llm,\n tts,\n turnDetection,\n userData,\n voiceOptions = defaultVoiceOptions,\n connOptions,\n } = opts;\n\n // Merge user-provided connOptions with defaults\n this._connOptions = {\n sttConnOptions: { ...DEFAULT_API_CONNECT_OPTIONS, ...connOptions?.sttConnOptions },\n llmConnOptions: { ...DEFAULT_API_CONNECT_OPTIONS, ...connOptions?.llmConnOptions },\n ttsConnOptions: { ...DEFAULT_API_CONNECT_OPTIONS, ...connOptions?.ttsConnOptions },\n maxUnrecoverableErrors:\n connOptions?.maxUnrecoverableErrors ??\n DEFAULT_SESSION_CONNECT_OPTIONS.maxUnrecoverableErrors,\n };\n\n this.vad = vad;\n\n if (typeof stt === 'string') {\n this.stt = InferenceSTT.fromModelString(stt);\n } else {\n this.stt = stt;\n }\n\n if (typeof llm === 'string') {\n this.llm = InferenceLLM.fromModelString(llm);\n } else {\n this.llm = llm;\n }\n\n if (typeof tts === 'string') {\n this.tts = InferenceTTS.fromModelString(tts);\n } else {\n this.tts = tts;\n }\n\n this.turnDetection = turnDetection;\n this._userData = userData;\n\n // configurable IO\n this._input = new AgentInput(this.onAudioInputChanged);\n this._output = new AgentOutput(this.onAudioOutputChanged, this.onTextOutputChanged);\n\n // This is the \"global\" chat context, it holds the entire conversation history\n this._chatCtx = ChatContext.empty();\n this.options = { ...defaultVoiceOptions, ...voiceOptions };\n\n this._onUserInputTranscribed = this._onUserInputTranscribed.bind(this);\n this.on(AgentSessionEventTypes.UserInputTranscribed, this._onUserInputTranscribed);\n }\n\n emit<K extends keyof AgentSessionCallbacks>(\n event: K,\n ...args: Parameters<AgentSessionCallbacks[K]>\n ): boolean {\n const eventData = args[0] as AgentEvent;\n this._recordedEvents.push(eventData);\n return super.emit(event, ...args);\n }\n\n get input(): AgentInput {\n return this._input;\n }\n\n get output(): AgentOutput {\n return this._output;\n }\n\n get userData(): UserData {\n if (this._userData === undefined) {\n throw new Error('Voice agent userData is not set');\n }\n\n return this._userData;\n }\n\n get history(): ChatContext {\n return this._chatCtx;\n }\n\n /** Connection options for STT, LLM, and TTS. */\n get connOptions(): ResolvedSessionConnectOptions {\n return this._connOptions;\n }\n\n get useTtsAlignedTranscript(): boolean {\n return this.options.useTtsAlignedTranscript;\n }\n\n set userData(value: UserData) {\n this._userData = value;\n }\n\n private async _startImpl({\n agent,\n room,\n inputOptions,\n outputOptions,\n span,\n }: {\n agent: Agent;\n room?: Room;\n inputOptions?: Partial<RoomInputOptions>;\n outputOptions?: Partial<RoomOutputOptions>;\n span: Span;\n }): Promise<void> {\n span.setAttribute(traceTypes.ATTR_AGENT_LABEL, agent.id);\n\n this.agent = agent;\n this._updateAgentState('initializing');\n\n const tasks: Promise<void>[] = [];\n\n if (room && !this.roomIO) {\n // Check for existing input/output configuration and warn if needed\n if (this.input.audio && inputOptions?.audioEnabled !== false) {\n this.logger.warn(\n 'RoomIO audio input is enabled but input.audio is already set, ignoring..',\n );\n }\n\n if (this.output.audio && outputOptions?.audioEnabled !== false) {\n this.logger.warn(\n 'RoomIO audio output is enabled but output.audio is already set, ignoring..',\n );\n }\n\n if (this.output.transcription && outputOptions?.transcriptionEnabled !== false) {\n this.logger.warn(\n 'RoomIO transcription output is enabled but output.transcription is already set, ignoring..',\n );\n }\n\n this.roomIO = new RoomIO({\n agentSession: this,\n room,\n inputOptions,\n outputOptions,\n });\n this.roomIO.start();\n }\n\n let ctx: JobContext | undefined = undefined;\n try {\n ctx = getJobContext();\n } catch {\n // JobContext is not available in evals\n }\n\n if (ctx) {\n if (room && ctx.room === room && !room.isConnected) {\n this.logger.debug('Auto-connecting to room via job context');\n tasks.push(ctx.connect());\n }\n\n if (ctx._primaryAgentSession === undefined) {\n ctx._primaryAgentSession = this;\n } else if (this._enableRecording) {\n throw new Error(\n 'Only one `AgentSession` can be the primary at a time. If you want to ignore primary designation, use `session.start({ record: false })`.',\n );\n }\n\n if (this.input.audio && this.output.audio && this._enableRecording) {\n this._recorderIO = new RecorderIO({ agentSession: this });\n this.input.audio = this._recorderIO.recordInput(this.input.audio);\n this.output.audio = this._recorderIO.recordOutput(this.output.audio);\n\n // Start recording to session directory\n const sessionDir = ctx.sessionDirectory;\n if (sessionDir) {\n tasks.push(this._recorderIO.start(`${sessionDir}/audio.ogg`));\n }\n }\n }\n\n // TODO(AJS-265): add shutdown callback to job context\n tasks.push(this.updateActivity(this.agent));\n\n await Promise.allSettled(tasks);\n\n // Log used IO configuration\n this.logger.debug(\n `using audio io: ${this.input.audio ? '`' + this.input.audio.constructor.name + '`' : '(none)'} -> \\`AgentSession\\` -> ${this.output.audio ? '`' + this.output.audio.constructor.name + '`' : '(none)'}`,\n );\n\n this.logger.debug(\n `using transcript io: \\`AgentSession\\` -> ${this.output.transcription ? '`' + this.output.transcription.constructor.name + '`' : '(none)'}`,\n );\n\n this.started = true;\n this._startedAt = Date.now();\n this._updateAgentState('listening');\n }\n\n async start({\n agent,\n room,\n inputOptions,\n outputOptions,\n record,\n }: {\n agent: Agent;\n room?: Room;\n inputOptions?: Partial<RoomInputOptions>;\n outputOptions?: Partial<RoomOutputOptions>;\n record?: boolean;\n }): Promise<void> {\n if (this.started) {\n return;\n }\n\n let ctx: JobContext | undefined = undefined;\n try {\n ctx = getJobContext();\n\n if (record === undefined) {\n record = ctx.job.enableRecording;\n }\n\n this._enableRecording = record;\n\n if (this._enableRecording) {\n ctx.initRecording();\n }\n } catch (error) {\n // JobContext is not available in evals\n this.logger.warn('JobContext is not available');\n }\n\n this.sessionSpan = tracer.startSpan({\n name: 'agent_session',\n context: ROOT_CONTEXT,\n });\n\n this.rootSpanContext = trace.setSpan(ROOT_CONTEXT, this.sessionSpan);\n\n await this._startImpl({\n agent,\n room,\n inputOptions,\n outputOptions,\n span: this.sessionSpan,\n });\n }\n\n updateAgent(agent: Agent): void {\n this.agent = agent;\n\n if (this.started) {\n this.updateActivity(agent);\n }\n }\n\n commitUserTurn() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n this.activity.commitUserTurn();\n }\n\n clearUserTurn() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n this.activity.clearUserTurn();\n }\n\n say(\n text: string | ReadableStream<string>,\n options?: {\n audio?: ReadableStream<AudioFrame>;\n allowInterruptions?: boolean;\n addToChatCtx?: boolean;\n },\n ): SpeechHandle {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n const doSay = (activity: AgentActivity) => {\n return activity.say(text, options);\n };\n\n // attach to the session span if called outside of the AgentSession\n const activeSpan = trace.getActiveSpan();\n if (!activeSpan && this.rootSpanContext) {\n return otelContext.with(this.rootSpanContext, () => doSay(this.activity!));\n }\n\n return doSay(this.activity);\n }\n\n interrupt() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n return this.activity.interrupt();\n }\n\n generateReply(options?: {\n userInput?: string;\n instructions?: string;\n toolChoice?: ToolChoice;\n allowInterruptions?: boolean;\n }): SpeechHandle {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n const userMessage = options?.userInput\n ? new ChatMessage({\n role: 'user',\n content: options.userInput,\n })\n : undefined;\n\n const doGenerateReply = (activity: AgentActivity, nextActivity?: AgentActivity) => {\n if (activity.draining) {\n if (!nextActivity) {\n throw new Error('AgentSession is closing, cannot use generateReply()');\n }\n return nextActivity.generateReply({ userMessage, ...options });\n }\n return activity.generateReply({ userMessage, ...options });\n };\n\n // attach to the session span if called outside of the AgentSession\n const activeSpan = trace.getActiveSpan();\n let handle: SpeechHandle;\n if (!activeSpan && this.rootSpanContext) {\n handle = otelContext.with(this.rootSpanContext, () =>\n doGenerateReply(this.activity!, this.nextActivity),\n );\n } else {\n handle = doGenerateReply(this.activity!, this.nextActivity);\n }\n\n if (this._globalRunState) {\n this._globalRunState._watchHandle(handle);\n }\n\n return handle;\n }\n\n /**\n * Run a test with user input and return a result for assertions.\n *\n * This method is primarily used for testing agent behavior without\n * requiring a real room connection.\n *\n * @example\n * ```typescript\n * const result = await session.run({ userInput: 'Hello' });\n * result.expect.nextEvent().isMessage({ role: 'assistant' });\n * result.expect.noMoreEvents();\n * ```\n *\n * @param options - Run options including user input\n * @returns A RunResult that resolves when the agent finishes responding\n *\n * TODO: Add outputType parameter for typed outputs (parity with Python)\n */\n run(options: { userInput: string }): RunResult {\n if (this._globalRunState && !this._globalRunState.done()) {\n throw new Error('nested runs are not supported');\n }\n\n const runState = new RunResult({ userInput: options.userInput });\n this._globalRunState = runState;\n this.generateReply({ userInput: options.userInput });\n\n return runState;\n }\n\n private async updateActivity(agent: Agent): Promise<void> {\n const runWithContext = async () => {\n // TODO(AJS-129): add lock to agent activity core lifecycle\n this.nextActivity = new AgentActivity(agent, this);\n\n const previousActivity = this.activity;\n\n if (this.activity) {\n await this.activity.drain();\n await this.activity.close();\n }\n\n this.activity = this.nextActivity;\n this.nextActivity = undefined;\n\n this._chatCtx.insert(\n new AgentHandoffItem({\n oldAgentId: previousActivity?.agent.id,\n newAgentId: agent.id,\n }),\n );\n this.logger.debug(\n { previousAgentId: previousActivity?.agent.id, newAgentId: agent.id },\n 'Agent handoff inserted into chat context',\n );\n\n await this.activity.start();\n\n if (this._input.audio) {\n this.activity.attachAudioInput(this._input.audio.stream);\n }\n };\n\n // Run within session span context if available\n if (this.rootSpanContext) {\n return otelContext.with(this.rootSpanContext, runWithContext);\n }\n\n return runWithContext();\n }\n\n get chatCtx(): ChatContext {\n return this._chatCtx.copy();\n }\n\n get agentState(): AgentState {\n return this._agentState;\n }\n\n get currentAgent(): Agent {\n if (!this.agent) {\n throw new Error('AgentSession is not running');\n }\n\n return this.agent;\n }\n\n async close(): Promise<void> {\n await this.closeImpl(CloseReason.USER_INITIATED);\n }\n\n shutdown(options?: { drain?: boolean; reason?: ShutdownReason }): void {\n const { drain = true, reason = CloseReason.USER_INITIATED } = options ?? {};\n\n this._closeSoon({\n reason,\n drain,\n });\n }\n\n /** @internal */\n _closeSoon({\n reason,\n drain = false,\n error = null,\n }: {\n reason: ShutdownReason;\n drain?: boolean;\n error?: RealtimeModelError | STTError | TTSError | LLMError | null;\n }): void {\n if (this.closingTask) {\n return;\n }\n this.closeImpl(reason, error, drain);\n }\n\n /** @internal */\n _onError(error: RealtimeModelError | STTError | TTSError | LLMError): void {\n if (this.closingTask || error.recoverable) {\n return;\n }\n\n // Track error counts per type to implement max_unrecoverable_errors logic\n if (error.type === 'llm_error') {\n this.llmErrorCounts += 1;\n if (this.llmErrorCounts <= this._connOptions.maxUnrecoverableErrors) {\n return;\n }\n } else if (error.type === 'tts_error') {\n this.ttsErrorCounts += 1;\n if (this.ttsErrorCounts <= this._connOptions.maxUnrecoverableErrors) {\n return;\n }\n }\n\n this.logger.error(error, 'AgentSession is closing due to unrecoverable error');\n\n this.closingTask = (async () => {\n await this.closeImpl(CloseReason.ERROR, error);\n })().then(() => {\n this.closingTask = null;\n });\n }\n\n /** @internal */\n _conversationItemAdded(item: ChatMessage): void {\n this._chatCtx.insert(item);\n this.emit(AgentSessionEventTypes.ConversationItemAdded, createConversationItemAddedEvent(item));\n }\n\n /** @internal */\n _toolItemsAdded(items: (FunctionCall | FunctionCallOutput)[]): void {\n this._chatCtx.insert(items);\n }\n\n /** @internal */\n _updateAgentState(state: AgentState, options?: { startTime?: number; otelContext?: Context }) {\n if (this._agentState === state) {\n return;\n }\n\n if (state === 'speaking') {\n // Reset error counts when agent starts speaking\n this.llmErrorCounts = 0;\n this.ttsErrorCounts = 0;\n\n if (this.agentSpeakingSpan === undefined) {\n this.agentSpeakingSpan = tracer.startSpan({\n name: 'agent_speaking',\n context: options?.otelContext ?? this.rootSpanContext,\n startTime: options?.startTime,\n });\n\n // TODO(brian): PR4 - Set participant attributes if roomIO.room.localParticipant is available\n // (Ref: Python agent_session.py line 1161-1164)\n }\n } else if (this.agentSpeakingSpan !== undefined) {\n // TODO(brian): PR4 - Set ATTR_END_TIME attribute if available\n this.agentSpeakingSpan.end();\n this.agentSpeakingSpan = undefined;\n }\n\n const oldState = this._agentState;\n this._agentState = state;\n\n // Handle user away timer based on state changes\n if (state === 'listening' && this.userState === 'listening') {\n this._setUserAwayTimer();\n } else {\n this._cancelUserAwayTimer();\n }\n\n this.emit(\n AgentSessionEventTypes.AgentStateChanged,\n createAgentStateChangedEvent(oldState, state),\n );\n }\n\n /** @internal */\n _updateUserState(state: UserState, lastSpeakingTime?: number) {\n if (this.userState === state) {\n return;\n }\n\n if (state === 'speaking' && this.userSpeakingSpan === undefined) {\n this.userSpeakingSpan = tracer.startSpan({\n name: 'user_speaking',\n context: this.rootSpanContext,\n startTime: lastSpeakingTime,\n });\n\n // TODO(brian): PR4 - Set participant attributes if roomIO.linkedParticipant is available\n // (Ref: Python agent_session.py line 1192-1195)\n } else if (this.userSpeakingSpan !== undefined) {\n this.userSpeakingSpan.end(lastSpeakingTime);\n this.userSpeakingSpan = undefined;\n }\n\n const oldState = this.userState;\n this.userState = state;\n\n // Handle user away timer based on state changes\n if (state === 'listening' && this._agentState === 'listening') {\n this._setUserAwayTimer();\n } else {\n this._cancelUserAwayTimer();\n }\n\n this.emit(\n AgentSessionEventTypes.UserStateChanged,\n createUserStateChangedEvent(oldState, state),\n );\n }\n\n // -- User changed input/output streams/sinks --\n private onAudioInputChanged(): void {\n if (!this.started) {\n return;\n }\n\n if (this.activity && this._input.audio) {\n this.activity.attachAudioInput(this._input.audio.stream);\n }\n }\n\n private onAudioOutputChanged(): void {}\n\n private onTextOutputChanged(): void {}\n\n private _setUserAwayTimer(): void {\n this._cancelUserAwayTimer();\n\n if (this.options.userAwayTimeout === null || this.options.userAwayTimeout === undefined) {\n return;\n }\n\n if (this.roomIO && !this.roomIO.isParticipantAvailable) {\n return;\n }\n\n this.userAwayTimer = setTimeout(() => {\n this.logger.debug('User away timeout triggered');\n this._updateUserState('away');\n }, this.options.userAwayTimeout * 1000);\n }\n\n private _cancelUserAwayTimer(): void {\n if (this.userAwayTimer !== null) {\n clearTimeout(this.userAwayTimer);\n this.userAwayTimer = null;\n }\n }\n\n private _onUserInputTranscribed(ev: UserInputTranscribedEvent): void {\n if (this.userState === 'away' && ev.isFinal) {\n this.logger.debug('User returned from away state due to speech input');\n this._updateUserState('listening');\n }\n }\n\n private async closeImpl(\n reason: ShutdownReason,\n error: RealtimeModelError | LLMError | TTSError | STTError | null = null,\n drain: boolean = false,\n ): Promise<void> {\n if (this.rootSpanContext) {\n return otelContext.with(this.rootSpanContext, async () => {\n await this.closeImplInner(reason, error, drain);\n });\n }\n\n return this.closeImplInner(reason, error, drain);\n }\n\n private async closeImplInner(\n reason: ShutdownReason,\n error: RealtimeModelError | LLMError | TTSError | STTError | null = null,\n drain: boolean = false,\n ): Promise<void> {\n if (!this.started) {\n return;\n }\n\n this._cancelUserAwayTimer();\n this.off(AgentSessionEventTypes.UserInputTranscribed, this._onUserInputTranscribed);\n\n if (this.activity) {\n if (!drain) {\n try {\n this.activity.interrupt();\n } catch (error) {\n // TODO(shubhra): force interrupt or wait for it to finish?\n // it might be an audio played from the error callback\n }\n }\n await this.activity.drain();\n // wait any uninterruptible speech to finish\n await this.activity.currentSpeech?.waitForPlayout();\n try {\n this.activity.detachAudioInput();\n } catch (error) {\n // Ignore detach errors during cleanup - source may not have been set\n }\n }\n\n // Close recorder before detaching inputs/outputs (keep reference for session report)\n if (this._recorderIO) {\n await this._recorderIO.close();\n }\n\n // detach the inputs and outputs\n this.input.audio = null;\n this.output.audio = null;\n this.output.transcription = null;\n\n await this.roomIO?.close();\n this.roomIO = undefined;\n\n await this.activity?.close();\n this.activity = undefined;\n\n if (this.sessionSpan) {\n this.sessionSpan.end();\n this.sessionSpan = undefined;\n }\n\n if (this.userSpeakingSpan) {\n this.userSpeakingSpan.end();\n this.userSpeakingSpan = undefined;\n }\n\n if (this.agentSpeakingSpan) {\n this.agentSpeakingSpan.end();\n this.agentSpeakingSpan = undefined;\n }\n\n this.started = false;\n\n this.emit(AgentSessionEventTypes.Close, createCloseEvent(reason, error));\n\n this.userState = 'listening';\n this._agentState = 'initializing';\n this.rootSpanContext = undefined;\n this.llmErrorCounts = 0;\n this.ttsErrorCounts = 0;\n\n this.logger.info({ reason, error }, 'AgentSession closed');\n }\n}\n"],"mappings":"AAMA,SAAS,cAAc,WAAW,aAAa,aAAa;AAC5D,SAAS,oBAAoB;AAE7B;AAAA,EACE,OAAO;AAAA,EACP,OAAO;AAAA,EACP,OAAO;AAAA,OAIF;AACP,SAA0B,qBAAqB;AAE/C,SAAS,kBAAkB,aAAa,mBAAmB;AAG3D,SAAS,WAAW;AAGpB,SAAS,YAAY,cAAc;AAEnC;AAAA,EACE;AAAA,EACA;AAAA,OAGK;AAGP,SAAS,qBAAqB;AAE9B;AAAA,EAEE;AAAA,EAIA;AAAA,EAUA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,YAAY,mBAAmB;AACxC,SAAS,kBAAkB;AAC3B,SAAS,cAA6D;AAGtE,SAAS,iBAAiB;AAe1B,MAAM,sBAAoC;AAAA,EACxC,oBAAoB;AAAA,EACpB,+BAA+B;AAAA,EAC/B,yBAAyB;AAAA,EACzB,sBAAsB;AAAA,EACtB,qBAAqB;AAAA,EACrB,qBAAqB;AAAA,EACrB,cAAc;AAAA,EACd,sBAAsB;AAAA,EACtB,iBAAiB;AAAA,EACjB,yBAAyB;AAC3B;AA2BO,MAAM,qBAEF,aAA+D;AAAA,EACxE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAES;AAAA,EAED;AAAA,EACA;AAAA,EACA;AAAA,EACA,UAAU;AAAA,EACV,YAAuB;AAAA,EAEvB;AAAA,EACA,SAAS,IAAI;AAAA,EAEb;AAAA,EACA;AAAA,EACA,cAA0B;AAAA,EAE1B;AAAA,EACA;AAAA,EAEA,cAAoC;AAAA,EACpC,gBAAuC;AAAA;AAAA,EAGvC;AAAA;AAAA,EAGA,iBAAiB;AAAA,EACjB,iBAAiB;AAAA,EAEjB;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAGR;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA,kBAAgC,CAAC;AAAA;AAAA,EAGjC,mBAAmB;AAAA;AAAA,EAGnB;AAAA;AAAA,EAGA;AAAA,EAEA,YAAY,MAAqC;AAC/C,UAAM;AAEN,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,eAAe;AAAA,MACf;AAAA,IACF,IAAI;AAGJ,SAAK,eAAe;AAAA,MAClB,gBAAgB,EAAE,GAAG,6BAA6B,GAAG,2CAAa,eAAe;AAAA,MACjF,gBAAgB,EAAE,GAAG,6BAA6B,GAAG,2CAAa,eAAe;AAAA,MACjF,gBAAgB,EAAE,GAAG,6BAA6B,GAAG,2CAAa,eAAe;AAAA,MACjF,yBACE,2CAAa,2BACb,gCAAgC;AAAA,IACpC;AAEA,SAAK,MAAM;AAEX,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,aAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,aAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,aAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,SAAK,gBAAgB;AACrB,SAAK,YAAY;AAGjB,SAAK,SAAS,IAAI,WAAW,KAAK,mBAAmB;AACrD,SAAK,UAAU,IAAI,YAAY,KAAK,sBAAsB,KAAK,mBAAmB;AAGlF,SAAK,WAAW,YAAY,MAAM;AAClC,SAAK,UAAU,EAAE,GAAG,qBAAqB,GAAG,aAAa;AAEzD,SAAK,0BAA0B,KAAK,wBAAwB,KAAK,IAAI;AACrE,SAAK,GAAG,uBAAuB,sBAAsB,KAAK,uBAAuB;AAAA,EACnF;AAAA,EAEA,KACE,UACG,MACM;AACT,UAAM,YAAY,KAAK,CAAC;AACxB,SAAK,gBAAgB,KAAK,SAAS;AACnC,WAAO,MAAM,KAAK,OAAO,GAAG,IAAI;AAAA,EAClC;AAAA,EAEA,IAAI,QAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAsB;AACxB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,WAAqB;AACvB,QAAI,KAAK,cAAc,QAAW;AAChC,YAAM,IAAI,MAAM,iCAAiC;AAAA,IACnD;AAEA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,cAA6C;AAC/C,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,0BAAmC;AACrC,WAAO,KAAK,QAAQ;AAAA,EACtB;AAAA,EAEA,IAAI,SAAS,OAAiB;AAC5B,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,MAAc,WAAW;AAAA,IACvB;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAMkB;AAChB,SAAK,aAAa,WAAW,kBAAkB,MAAM,EAAE;AAEvD,SAAK,QAAQ;AACb,SAAK,kBAAkB,cAAc;AAErC,UAAM,QAAyB,CAAC;AAEhC,QAAI,QAAQ,CAAC,KAAK,QAAQ;AAExB,UAAI,KAAK,MAAM,UAAS,6CAAc,kBAAiB,OAAO;AAC5D,aAAK,OAAO;AAAA,UACV;AAAA,QACF;AAAA,MACF;AAEA,UAAI,KAAK,OAAO,UAAS,+CAAe,kBAAiB,OAAO;AAC9D,aAAK,OAAO;AAAA,UACV;AAAA,QACF;AAAA,MACF;AAEA,UAAI,KAAK,OAAO,kBAAiB,+CAAe,0BAAyB,OAAO;AAC9E,aAAK,OAAO;AAAA,UACV;AAAA,QACF;AAAA,MACF;AAEA,WAAK,SAAS,IAAI,OAAO;AAAA,QACvB,cAAc;AAAA,QACd;AAAA,QACA;AAAA,QACA;AAAA,MACF,CAAC;AACD,WAAK,OAAO,MAAM;AAAA,IACpB;AAEA,QAAI,MAA8B;AAClC,QAAI;AACF,YAAM,cAAc;AAAA,IACtB,QAAQ;AAAA,IAER;AAEA,QAAI,KAAK;AACP,UAAI,QAAQ,IAAI,SAAS,QAAQ,CAAC,KAAK,aAAa;AAClD,aAAK,OAAO,MAAM,yCAAyC;AAC3D,cAAM,KAAK,IAAI,QAAQ,CAAC;AAAA,MAC1B;AAEA,UAAI,IAAI,yBAAyB,QAAW;AAC1C,YAAI,uBAAuB;AAAA,MAC7B,WAAW,KAAK,kBAAkB;AAChC,cAAM,IAAI;AAAA,UACR;AAAA,QACF;AAAA,MACF;AAEA,UAAI,KAAK,MAAM,SAAS,KAAK,OAAO,SAAS,KAAK,kBAAkB;AAClE,aAAK,cAAc,IAAI,WAAW,EAAE,cAAc,KAAK,CAAC;AACxD,aAAK,MAAM,QAAQ,KAAK,YAAY,YAAY,KAAK,MAAM,KAAK;AAChE,aAAK,OAAO,QAAQ,KAAK,YAAY,aAAa,KAAK,OAAO,KAAK;AAGnE,cAAM,aAAa,IAAI;AACvB,YAAI,YAAY;AACd,gBAAM,KAAK,KAAK,YAAY,MAAM,GAAG,UAAU,YAAY,CAAC;AAAA,QAC9D;AAAA,MACF;AAAA,IACF;AAGA,UAAM,KAAK,KAAK,eAAe,KAAK,KAAK,CAAC;AAE1C,UAAM,QAAQ,WAAW,KAAK;AAG9B,SAAK,OAAO;AAAA,MACV,mBAAmB,KAAK,MAAM,QAAQ,MAAM,KAAK,MAAM,MAAM,YAAY,OAAO,MAAM,QAAQ,2BAA2B,KAAK,OAAO,QAAQ,MAAM,KAAK,OAAO,MAAM,YAAY,OAAO,MAAM,QAAQ;AAAA,IACxM;AAEA,SAAK,OAAO;AAAA,MACV,4CAA4C,KAAK,OAAO,gBAAgB,MAAM,KAAK,OAAO,cAAc,YAAY,OAAO,MAAM,QAAQ;AAAA,IAC3I;AAEA,SAAK,UAAU;AACf,SAAK,aAAa,KAAK,IAAI;AAC3B,SAAK,kBAAkB,WAAW;AAAA,EACpC;AAAA,EAEA,MAAM,MAAM;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAMkB;AAChB,QAAI,KAAK,SAAS;AAChB;AAAA,IACF;AAEA,QAAI,MAA8B;AAClC,QAAI;AACF,YAAM,cAAc;AAEpB,UAAI,WAAW,QAAW;AACxB,iBAAS,IAAI,IAAI;AAAA,MACnB;AAEA,WAAK,mBAAmB;AAExB,UAAI,KAAK,kBAAkB;AACzB,YAAI,cAAc;AAAA,MACpB;AAAA,IACF,SAAS,OAAO;AAEd,WAAK,OAAO,KAAK,6BAA6B;AAAA,IAChD;AAEA,SAAK,cAAc,OAAO,UAAU;AAAA,MAClC,MAAM;AAAA,MACN,SAAS;AAAA,IACX,CAAC;AAED,SAAK,kBAAkB,MAAM,QAAQ,cAAc,KAAK,WAAW;AAEnE,UAAM,KAAK,WAAW;AAAA,MACpB;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,MAAM,KAAK;AAAA,IACb,CAAC;AAAA,EACH;AAAA,EAEA,YAAY,OAAoB;AAC9B,SAAK,QAAQ;AAEb,QAAI,KAAK,SAAS;AAChB,WAAK,eAAe,KAAK;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,iBAAiB;AACf,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,SAAK,SAAS,eAAe;AAAA,EAC/B;AAAA,EAEA,gBAAgB;AACd,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AACA,SAAK,SAAS,cAAc;AAAA,EAC9B;AAAA,EAEA,IACE,MACA,SAKc;AACd,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,UAAM,QAAQ,CAAC,aAA4B;AACzC,aAAO,SAAS,IAAI,MAAM,OAAO;AAAA,IACnC;AAGA,UAAM,aAAa,MAAM,cAAc;AACvC,QAAI,CAAC,cAAc,KAAK,iBAAiB;AACvC,aAAO,YAAY,KAAK,KAAK,iBAAiB,MAAM,MAAM,KAAK,QAAS,CAAC;AAAA,IAC3E;AAEA,WAAO,MAAM,KAAK,QAAQ;AAAA,EAC5B;AAAA,EAEA,YAAY;AACV,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AACA,WAAO,KAAK,SAAS,UAAU;AAAA,EACjC;AAAA,EAEA,cAAc,SAKG;AACf,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,UAAM,eAAc,mCAAS,aACzB,IAAI,YAAY;AAAA,MACd,MAAM;AAAA,MACN,SAAS,QAAQ;AAAA,IACnB,CAAC,IACD;AAEJ,UAAM,kBAAkB,CAAC,UAAyB,iBAAiC;AACjF,UAAI,SAAS,UAAU;AACrB,YAAI,CAAC,cAAc;AACjB,gBAAM,IAAI,MAAM,qDAAqD;AAAA,QACvE;AACA,eAAO,aAAa,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,MAC/D;AACA,aAAO,SAAS,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,IAC3D;AAGA,UAAM,aAAa,MAAM,cAAc;AACvC,QAAI;AACJ,QAAI,CAAC,cAAc,KAAK,iBAAiB;AACvC,eAAS,YAAY;AAAA,QAAK,KAAK;AAAA,QAAiB,MAC9C,gBAAgB,KAAK,UAAW,KAAK,YAAY;AAAA,MACnD;AAAA,IACF,OAAO;AACL,eAAS,gBAAgB,KAAK,UAAW,KAAK,YAAY;AAAA,IAC5D;AAEA,QAAI,KAAK,iBAAiB;AACxB,WAAK,gBAAgB,aAAa,MAAM;AAAA,IAC1C;AAEA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAoBA,IAAI,SAA2C;AAC7C,QAAI,KAAK,mBAAmB,CAAC,KAAK,gBAAgB,KAAK,GAAG;AACxD,YAAM,IAAI,MAAM,+BAA+B;AAAA,IACjD;AAEA,UAAM,WAAW,IAAI,UAAU,EAAE,WAAW,QAAQ,UAAU,CAAC;AAC/D,SAAK,kBAAkB;AACvB,SAAK,cAAc,EAAE,WAAW,QAAQ,UAAU,CAAC;AAEnD,WAAO;AAAA,EACT;AAAA,EAEA,MAAc,eAAe,OAA6B;AACxD,UAAM,iBAAiB,YAAY;AAEjC,WAAK,eAAe,IAAI,cAAc,OAAO,IAAI;AAEjD,YAAM,mBAAmB,KAAK;AAE9B,UAAI,KAAK,UAAU;AACjB,cAAM,KAAK,SAAS,MAAM;AAC1B,cAAM,KAAK,SAAS,MAAM;AAAA,MAC5B;AAEA,WAAK,WAAW,KAAK;AACrB,WAAK,eAAe;AAEpB,WAAK,SAAS;AAAA,QACZ,IAAI,iBAAiB;AAAA,UACnB,YAAY,qDAAkB,MAAM;AAAA,UACpC,YAAY,MAAM;AAAA,QACpB,CAAC;AAAA,MACH;AACA,WAAK,OAAO;AAAA,QACV,EAAE,iBAAiB,qDAAkB,MAAM,IAAI,YAAY,MAAM,GAAG;AAAA,QACpE;AAAA,MACF;AAEA,YAAM,KAAK,SAAS,MAAM;AAE1B,UAAI,KAAK,OAAO,OAAO;AACrB,aAAK,SAAS,iBAAiB,KAAK,OAAO,MAAM,MAAM;AAAA,MACzD;AAAA,IACF;AAGA,QAAI,KAAK,iBAAiB;AACxB,aAAO,YAAY,KAAK,KAAK,iBAAiB,cAAc;AAAA,IAC9D;AAEA,WAAO,eAAe;AAAA,EACxB;AAAA,EAEA,IAAI,UAAuB;AACzB,WAAO,KAAK,SAAS,KAAK;AAAA,EAC5B;AAAA,EAEA,IAAI,aAAyB;AAC3B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAsB;AACxB,QAAI,CAAC,KAAK,OAAO;AACf,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,QAAuB;AAC3B,UAAM,KAAK,UAAU,YAAY,cAAc;AAAA,EACjD;AAAA,EAEA,SAAS,SAA8D;AACrE,UAAM,EAAE,QAAQ,MAAM,SAAS,YAAY,eAAe,IAAI,WAAW,CAAC;AAE1E,SAAK,WAAW;AAAA,MACd;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA;AAAA,EAGA,WAAW;AAAA,IACT;AAAA,IACA,QAAQ;AAAA,IACR,QAAQ;AAAA,EACV,GAIS;AACP,QAAI,KAAK,aAAa;AACpB;AAAA,IACF;AACA,SAAK,UAAU,QAAQ,OAAO,KAAK;AAAA,EACrC;AAAA;AAAA,EAGA,SAAS,OAAkE;AACzE,QAAI,KAAK,eAAe,MAAM,aAAa;AACzC;AAAA,IACF;AAGA,QAAI,MAAM,SAAS,aAAa;AAC9B,WAAK,kBAAkB;AACvB,UAAI,KAAK,kBAAkB,KAAK,aAAa,wBAAwB;AACnE;AAAA,MACF;AAAA,IACF,WAAW,MAAM,SAAS,aAAa;AACrC,WAAK,kBAAkB;AACvB,UAAI,KAAK,kBAAkB,KAAK,aAAa,wBAAwB;AACnE;AAAA,MACF;AAAA,IACF;AAEA,SAAK,OAAO,MAAM,OAAO,oDAAoD;AAE7E,SAAK,eAAe,YAAY;AAC9B,YAAM,KAAK,UAAU,YAAY,OAAO,KAAK;AAAA,IAC/C,GAAG,EAAE,KAAK,MAAM;AACd,WAAK,cAAc;AAAA,IACrB,CAAC;AAAA,EACH;AAAA;AAAA,EAGA,uBAAuB,MAAyB;AAC9C,SAAK,SAAS,OAAO,IAAI;AACzB,SAAK,KAAK,uBAAuB,uBAAuB,iCAAiC,IAAI,CAAC;AAAA,EAChG;AAAA;AAAA,EAGA,gBAAgB,OAAoD;AAClE,SAAK,SAAS,OAAO,KAAK;AAAA,EAC5B;AAAA;AAAA,EAGA,kBAAkB,OAAmB,SAAyD;AAC5F,QAAI,KAAK,gBAAgB,OAAO;AAC9B;AAAA,IACF;AAEA,QAAI,UAAU,YAAY;AAExB,WAAK,iBAAiB;AACtB,WAAK,iBAAiB;AAEtB,UAAI,KAAK,sBAAsB,QAAW;AACxC,aAAK,oBAAoB,OAAO,UAAU;AAAA,UACxC,MAAM;AAAA,UACN,UAAS,mCAAS,gBAAe,KAAK;AAAA,UACtC,WAAW,mCAAS;AAAA,QACtB,CAAC;AAAA,MAIH;AAAA,IACF,WAAW,KAAK,sBAAsB,QAAW;AAE/C,WAAK,kBAAkB,IAAI;AAC3B,WAAK,oBAAoB;AAAA,IAC3B;AAEA,UAAM,WAAW,KAAK;AACtB,SAAK,cAAc;AAGnB,QAAI,UAAU,eAAe,KAAK,cAAc,aAAa;AAC3D,WAAK,kBAAkB;AAAA,IACzB,OAAO;AACL,WAAK,qBAAqB;AAAA,IAC5B;AAEA,SAAK;AAAA,MACH,uBAAuB;AAAA,MACvB,6BAA6B,UAAU,KAAK;AAAA,IAC9C;AAAA,EACF;AAAA;AAAA,EAGA,iBAAiB,OAAkB,kBAA2B;AAC5D,QAAI,KAAK,cAAc,OAAO;AAC5B;AAAA,IACF;AAEA,QAAI,UAAU,cAAc,KAAK,qBAAqB,QAAW;AAC/D,WAAK,mBAAmB,OAAO,UAAU;AAAA,QACvC,MAAM;AAAA,QACN,SAAS,KAAK;AAAA,QACd,WAAW;AAAA,MACb,CAAC;AAAA,IAIH,WAAW,KAAK,qBAAqB,QAAW;AAC9C,WAAK,iBAAiB,IAAI,gBAAgB;AAC1C,WAAK,mBAAmB;AAAA,IAC1B;AAEA,UAAM,WAAW,KAAK;AACtB,SAAK,YAAY;AAGjB,QAAI,UAAU,eAAe,KAAK,gBAAgB,aAAa;AAC7D,WAAK,kBAAkB;AAAA,IACzB,OAAO;AACL,WAAK,qBAAqB;AAAA,IAC5B;AAEA,SAAK;AAAA,MACH,uBAAuB;AAAA,MACvB,4BAA4B,UAAU,KAAK;AAAA,IAC7C;AAAA,EACF;AAAA;AAAA,EAGQ,sBAA4B;AAClC,QAAI,CAAC,KAAK,SAAS;AACjB;AAAA,IACF;AAEA,QAAI,KAAK,YAAY,KAAK,OAAO,OAAO;AACtC,WAAK,SAAS,iBAAiB,KAAK,OAAO,MAAM,MAAM;AAAA,IACzD;AAAA,EACF;AAAA,EAEQ,uBAA6B;AAAA,EAAC;AAAA,EAE9B,sBAA4B;AAAA,EAAC;AAAA,EAE7B,oBAA0B;AAChC,SAAK,qBAAqB;AAE1B,QAAI,KAAK,QAAQ,oBAAoB,QAAQ,KAAK,QAAQ,oBAAoB,QAAW;AACvF;AAAA,IACF;AAEA,QAAI,KAAK,UAAU,CAAC,KAAK,OAAO,wBAAwB;AACtD;AAAA,IACF;AAEA,SAAK,gBAAgB,WAAW,MAAM;AACpC,WAAK,OAAO,MAAM,6BAA6B;AAC/C,WAAK,iBAAiB,MAAM;AAAA,IAC9B,GAAG,KAAK,QAAQ,kBAAkB,GAAI;AAAA,EACxC;AAAA,EAEQ,uBAA6B;AACnC,QAAI,KAAK,kBAAkB,MAAM;AAC/B,mBAAa,KAAK,aAAa;AAC/B,WAAK,gBAAgB;AAAA,IACvB;AAAA,EACF;AAAA,EAEQ,wBAAwB,IAAqC;AACnE,QAAI,KAAK,cAAc,UAAU,GAAG,SAAS;AAC3C,WAAK,OAAO,MAAM,mDAAmD;AACrE,WAAK,iBAAiB,WAAW;AAAA,IACnC;AAAA,EACF;AAAA,EAEA,MAAc,UACZ,QACA,QAAoE,MACpE,QAAiB,OACF;AACf,QAAI,KAAK,iBAAiB;AACxB,aAAO,YAAY,KAAK,KAAK,iBAAiB,YAAY;AACxD,cAAM,KAAK,eAAe,QAAQ,OAAO,KAAK;AAAA,MAChD,CAAC;AAAA,IACH;AAEA,WAAO,KAAK,eAAe,QAAQ,OAAO,KAAK;AAAA,EACjD;AAAA,EAEA,MAAc,eACZ,QACA,QAAoE,MACpE,QAAiB,OACF;AA3zBnB;AA4zBI,QAAI,CAAC,KAAK,SAAS;AACjB;AAAA,IACF;AAEA,SAAK,qBAAqB;AAC1B,SAAK,IAAI,uBAAuB,sBAAsB,KAAK,uBAAuB;AAElF,QAAI,KAAK,UAAU;AACjB,UAAI,CAAC,OAAO;AACV,YAAI;AACF,eAAK,SAAS,UAAU;AAAA,QAC1B,SAASA,QAAO;AAAA,QAGhB;AAAA,MACF;AACA,YAAM,KAAK,SAAS,MAAM;AAE1B,cAAM,UAAK,SAAS,kBAAd,mBAA6B;AACnC,UAAI;AACF,aAAK,SAAS,iBAAiB;AAAA,MACjC,SAASA,QAAO;AAAA,MAEhB;AAAA,IACF;AAGA,QAAI,KAAK,aAAa;AACpB,YAAM,KAAK,YAAY,MAAM;AAAA,IAC/B;AAGA,SAAK,MAAM,QAAQ;AACnB,SAAK,OAAO,QAAQ;AACpB,SAAK,OAAO,gBAAgB;AAE5B,YAAM,UAAK,WAAL,mBAAa;AACnB,SAAK,SAAS;AAEd,YAAM,UAAK,aAAL,mBAAe;AACrB,SAAK,WAAW;AAEhB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,IAAI;AACrB,WAAK,cAAc;AAAA,IACrB;AAEA,QAAI,KAAK,kBAAkB;AACzB,WAAK,iBAAiB,IAAI;AAC1B,WAAK,mBAAmB;AAAA,IAC1B;AAEA,QAAI,KAAK,mBAAmB;AAC1B,WAAK,kBAAkB,IAAI;AAC3B,WAAK,oBAAoB;AAAA,IAC3B;AAEA,SAAK,UAAU;AAEf,SAAK,KAAK,uBAAuB,OAAO,iBAAiB,QAAQ,KAAK,CAAC;AAEvE,SAAK,YAAY;AACjB,SAAK,cAAc;AACnB,SAAK,kBAAkB;AACvB,SAAK,iBAAiB;AACtB,SAAK,iBAAiB;AAEtB,SAAK,OAAO,KAAK,EAAE,QAAQ,MAAM,GAAG,qBAAqB;AAAA,EAC3D;AACF;","names":["error"]}
1
+ {"version":3,"sources":["../../src/voice/agent_session.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2024 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame, Room } from '@livekit/rtc-node';\nimport type { TypedEventEmitter as TypedEmitter } from '@livekit/typed-emitter';\nimport type { Context, Span } from '@opentelemetry/api';\nimport { ROOT_CONTEXT, context as otelContext, trace } from '@opentelemetry/api';\nimport { EventEmitter } from 'node:events';\nimport type { ReadableStream } from 'node:stream/web';\nimport {\n LLM as InferenceLLM,\n STT as InferenceSTT,\n TTS as InferenceTTS,\n type LLMModels,\n type STTModelString,\n type TTSModelString,\n} from '../inference/index.js';\nimport { type JobContext, getJobContext } from '../job.js';\nimport type { FunctionCall, FunctionCallOutput } from '../llm/chat_context.js';\nimport { AgentHandoffItem, ChatContext, ChatMessage } from '../llm/chat_context.js';\nimport type { LLM, RealtimeModel, RealtimeModelError, ToolChoice } from '../llm/index.js';\nimport type { LLMError } from '../llm/llm.js';\nimport { log } from '../log.js';\nimport type { STT } from '../stt/index.js';\nimport type { STTError } from '../stt/stt.js';\nimport { traceTypes, tracer } from '../telemetry/index.js';\nimport type { TTS, TTSError } from '../tts/tts.js';\nimport {\n DEFAULT_API_CONNECT_OPTIONS,\n DEFAULT_SESSION_CONNECT_OPTIONS,\n type ResolvedSessionConnectOptions,\n type SessionConnectOptions,\n} from '../types.js';\nimport type { VAD } from '../vad.js';\nimport type { Agent } from './agent.js';\nimport { AgentActivity } from './agent_activity.js';\nimport type { _TurnDetector } from './audio_recognition.js';\nimport {\n type AgentEvent,\n AgentSessionEventTypes,\n type AgentState,\n type AgentStateChangedEvent,\n type CloseEvent,\n CloseReason,\n type ConversationItemAddedEvent,\n type ErrorEvent,\n type FunctionToolsExecutedEvent,\n type MetricsCollectedEvent,\n type ShutdownReason,\n type SpeechCreatedEvent,\n type UserInputTranscribedEvent,\n type UserState,\n type UserStateChangedEvent,\n createAgentStateChangedEvent,\n createCloseEvent,\n createConversationItemAddedEvent,\n createUserStateChangedEvent,\n} from './events.js';\nimport { AgentInput, AgentOutput } from './io.js';\nimport { RecorderIO } from './recorder_io/index.js';\nimport { RoomIO, type RoomInputOptions, type RoomOutputOptions } from './room_io/index.js';\nimport type { UnknownUserData } from './run_context.js';\nimport type { SpeechHandle } from './speech_handle.js';\nimport { RunResult } from './testing/run_result.js';\nimport { setParticipantSpanAttributes } from './utils.js';\n\nexport interface VoiceOptions {\n allowInterruptions: boolean;\n discardAudioIfUninterruptible: boolean;\n minInterruptionDuration: number;\n minInterruptionWords: number;\n minEndpointingDelay: number;\n maxEndpointingDelay: number;\n maxToolSteps: number;\n preemptiveGeneration: boolean;\n userAwayTimeout?: number | null;\n useTtsAlignedTranscript: boolean;\n}\n\nconst defaultVoiceOptions: VoiceOptions = {\n allowInterruptions: true,\n discardAudioIfUninterruptible: true,\n minInterruptionDuration: 500,\n minInterruptionWords: 0,\n minEndpointingDelay: 500,\n maxEndpointingDelay: 6000,\n maxToolSteps: 3,\n preemptiveGeneration: false,\n userAwayTimeout: 15.0,\n useTtsAlignedTranscript: true,\n} as const;\n\nexport type TurnDetectionMode = 'stt' | 'vad' | 'realtime_llm' | 'manual' | _TurnDetector;\n\nexport type AgentSessionCallbacks = {\n [AgentSessionEventTypes.UserInputTranscribed]: (ev: UserInputTranscribedEvent) => void;\n [AgentSessionEventTypes.AgentStateChanged]: (ev: AgentStateChangedEvent) => void;\n [AgentSessionEventTypes.UserStateChanged]: (ev: UserStateChangedEvent) => void;\n [AgentSessionEventTypes.ConversationItemAdded]: (ev: ConversationItemAddedEvent) => void;\n [AgentSessionEventTypes.FunctionToolsExecuted]: (ev: FunctionToolsExecutedEvent) => void;\n [AgentSessionEventTypes.MetricsCollected]: (ev: MetricsCollectedEvent) => void;\n [AgentSessionEventTypes.SpeechCreated]: (ev: SpeechCreatedEvent) => void;\n [AgentSessionEventTypes.Error]: (ev: ErrorEvent) => void;\n [AgentSessionEventTypes.Close]: (ev: CloseEvent) => void;\n};\n\nexport type AgentSessionOptions<UserData = UnknownUserData> = {\n turnDetection?: TurnDetectionMode;\n stt?: STT | STTModelString;\n vad?: VAD;\n llm?: LLM | RealtimeModel | LLMModels;\n tts?: TTS | TTSModelString;\n userData?: UserData;\n voiceOptions?: Partial<VoiceOptions>;\n connOptions?: SessionConnectOptions;\n};\n\nexport class AgentSession<\n UserData = UnknownUserData,\n> extends (EventEmitter as new () => TypedEmitter<AgentSessionCallbacks>) {\n vad?: VAD;\n stt?: STT;\n llm?: LLM | RealtimeModel;\n tts?: TTS;\n turnDetection?: TurnDetectionMode;\n\n readonly options: VoiceOptions;\n\n private agent?: Agent;\n private activity?: AgentActivity;\n private nextActivity?: AgentActivity;\n private started = false;\n private userState: UserState = 'listening';\n\n /** @internal */\n _roomIO?: RoomIO;\n private logger = log();\n\n private _chatCtx: ChatContext;\n private _userData: UserData | undefined;\n private _agentState: AgentState = 'initializing';\n\n private _input: AgentInput;\n private _output: AgentOutput;\n\n private closingTask: Promise<void> | null = null;\n private userAwayTimer: NodeJS.Timeout | null = null;\n\n // Connection options for STT, LLM, and TTS\n private _connOptions: ResolvedSessionConnectOptions;\n\n // Unrecoverable error counts, reset after agent speaking\n private llmErrorCounts = 0;\n private ttsErrorCounts = 0;\n\n private sessionSpan?: Span;\n private userSpeakingSpan?: Span;\n private agentSpeakingSpan?: Span;\n\n /** @internal */\n _recorderIO?: RecorderIO;\n\n /** @internal */\n rootSpanContext?: Context;\n\n /** @internal */\n _recordedEvents: AgentEvent[] = [];\n\n /** @internal */\n _enableRecording = false;\n\n /** @internal - Timestamp when the session started (milliseconds) */\n _startedAt?: number;\n\n /** @internal - Current run state for testing */\n _globalRunState?: RunResult;\n\n constructor(opts: AgentSessionOptions<UserData>) {\n super();\n\n const {\n vad,\n stt,\n llm,\n tts,\n turnDetection,\n userData,\n voiceOptions = defaultVoiceOptions,\n connOptions,\n } = opts;\n\n // Merge user-provided connOptions with defaults\n this._connOptions = {\n sttConnOptions: { ...DEFAULT_API_CONNECT_OPTIONS, ...connOptions?.sttConnOptions },\n llmConnOptions: { ...DEFAULT_API_CONNECT_OPTIONS, ...connOptions?.llmConnOptions },\n ttsConnOptions: { ...DEFAULT_API_CONNECT_OPTIONS, ...connOptions?.ttsConnOptions },\n maxUnrecoverableErrors:\n connOptions?.maxUnrecoverableErrors ??\n DEFAULT_SESSION_CONNECT_OPTIONS.maxUnrecoverableErrors,\n };\n\n this.vad = vad;\n\n if (typeof stt === 'string') {\n this.stt = InferenceSTT.fromModelString(stt);\n } else {\n this.stt = stt;\n }\n\n if (typeof llm === 'string') {\n this.llm = InferenceLLM.fromModelString(llm);\n } else {\n this.llm = llm;\n }\n\n if (typeof tts === 'string') {\n this.tts = InferenceTTS.fromModelString(tts);\n } else {\n this.tts = tts;\n }\n\n this.turnDetection = turnDetection;\n this._userData = userData;\n\n // configurable IO\n this._input = new AgentInput(this.onAudioInputChanged);\n this._output = new AgentOutput(this.onAudioOutputChanged, this.onTextOutputChanged);\n\n // This is the \"global\" chat context, it holds the entire conversation history\n this._chatCtx = ChatContext.empty();\n this.options = { ...defaultVoiceOptions, ...voiceOptions };\n\n this._onUserInputTranscribed = this._onUserInputTranscribed.bind(this);\n this.on(AgentSessionEventTypes.UserInputTranscribed, this._onUserInputTranscribed);\n }\n\n emit<K extends keyof AgentSessionCallbacks>(\n event: K,\n ...args: Parameters<AgentSessionCallbacks[K]>\n ): boolean {\n const eventData = args[0] as AgentEvent;\n this._recordedEvents.push(eventData);\n return super.emit(event, ...args);\n }\n\n get input(): AgentInput {\n return this._input;\n }\n\n get output(): AgentOutput {\n return this._output;\n }\n\n get userData(): UserData {\n if (this._userData === undefined) {\n throw new Error('Voice agent userData is not set');\n }\n\n return this._userData;\n }\n\n get history(): ChatContext {\n return this._chatCtx;\n }\n\n /** Connection options for STT, LLM, and TTS. */\n get connOptions(): ResolvedSessionConnectOptions {\n return this._connOptions;\n }\n\n get useTtsAlignedTranscript(): boolean {\n return this.options.useTtsAlignedTranscript;\n }\n\n set userData(value: UserData) {\n this._userData = value;\n }\n\n private async _startImpl({\n agent,\n room,\n inputOptions,\n outputOptions,\n span,\n }: {\n agent: Agent;\n room?: Room;\n inputOptions?: Partial<RoomInputOptions>;\n outputOptions?: Partial<RoomOutputOptions>;\n span: Span;\n }): Promise<void> {\n span.setAttribute(traceTypes.ATTR_AGENT_LABEL, agent.id);\n\n this.agent = agent;\n this._updateAgentState('initializing');\n\n const tasks: Promise<void>[] = [];\n\n if (room && !this._roomIO) {\n // Check for existing input/output configuration and warn if needed\n if (this.input.audio && inputOptions?.audioEnabled !== false) {\n this.logger.warn(\n 'RoomIO audio input is enabled but input.audio is already set, ignoring..',\n );\n }\n\n if (this.output.audio && outputOptions?.audioEnabled !== false) {\n this.logger.warn(\n 'RoomIO audio output is enabled but output.audio is already set, ignoring..',\n );\n }\n\n if (this.output.transcription && outputOptions?.transcriptionEnabled !== false) {\n this.logger.warn(\n 'RoomIO transcription output is enabled but output.transcription is already set, ignoring..',\n );\n }\n\n this._roomIO = new RoomIO({\n agentSession: this,\n room,\n inputOptions,\n outputOptions,\n });\n this._roomIO.start();\n }\n\n let ctx: JobContext | undefined = undefined;\n try {\n ctx = getJobContext();\n } catch {\n // JobContext is not available in evals\n }\n\n if (ctx) {\n if (room && ctx.room === room && !room.isConnected) {\n this.logger.debug('Auto-connecting to room via job context');\n tasks.push(ctx.connect());\n }\n\n if (ctx._primaryAgentSession === undefined) {\n ctx._primaryAgentSession = this;\n } else if (this._enableRecording) {\n throw new Error(\n 'Only one `AgentSession` can be the primary at a time. If you want to ignore primary designation, use `session.start({ record: false })`.',\n );\n }\n\n if (this.input.audio && this.output.audio && this._enableRecording) {\n this._recorderIO = new RecorderIO({ agentSession: this });\n this.input.audio = this._recorderIO.recordInput(this.input.audio);\n this.output.audio = this._recorderIO.recordOutput(this.output.audio);\n\n // Start recording to session directory\n const sessionDir = ctx.sessionDirectory;\n if (sessionDir) {\n tasks.push(this._recorderIO.start(`${sessionDir}/audio.ogg`));\n }\n }\n }\n\n // TODO(AJS-265): add shutdown callback to job context\n tasks.push(this.updateActivity(this.agent));\n\n await Promise.allSettled(tasks);\n\n // Log used IO configuration\n this.logger.debug(\n `using audio io: ${this.input.audio ? '`' + this.input.audio.constructor.name + '`' : '(none)'} -> \\`AgentSession\\` -> ${this.output.audio ? '`' + this.output.audio.constructor.name + '`' : '(none)'}`,\n );\n\n this.logger.debug(\n `using transcript io: \\`AgentSession\\` -> ${this.output.transcription ? '`' + this.output.transcription.constructor.name + '`' : '(none)'}`,\n );\n\n this.started = true;\n this._startedAt = Date.now();\n this._updateAgentState('listening');\n }\n\n async start({\n agent,\n room,\n inputOptions,\n outputOptions,\n record,\n }: {\n agent: Agent;\n room?: Room;\n inputOptions?: Partial<RoomInputOptions>;\n outputOptions?: Partial<RoomOutputOptions>;\n record?: boolean;\n }): Promise<void> {\n if (this.started) {\n return;\n }\n\n let ctx: JobContext | undefined = undefined;\n try {\n ctx = getJobContext();\n\n if (record === undefined) {\n record = ctx.job.enableRecording;\n }\n\n this._enableRecording = record;\n\n if (this._enableRecording) {\n ctx.initRecording();\n }\n } catch (error) {\n // JobContext is not available in evals\n this.logger.warn('JobContext is not available');\n }\n\n this.sessionSpan = tracer.startSpan({\n name: 'agent_session',\n context: ROOT_CONTEXT,\n });\n\n this.rootSpanContext = trace.setSpan(ROOT_CONTEXT, this.sessionSpan);\n\n await this._startImpl({\n agent,\n room,\n inputOptions,\n outputOptions,\n span: this.sessionSpan,\n });\n }\n\n updateAgent(agent: Agent): void {\n this.agent = agent;\n\n if (this.started) {\n this.updateActivity(agent);\n }\n }\n\n commitUserTurn() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n this.activity.commitUserTurn();\n }\n\n clearUserTurn() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n this.activity.clearUserTurn();\n }\n\n say(\n text: string | ReadableStream<string>,\n options?: {\n audio?: ReadableStream<AudioFrame>;\n allowInterruptions?: boolean;\n addToChatCtx?: boolean;\n },\n ): SpeechHandle {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n const doSay = (activity: AgentActivity) => {\n return activity.say(text, options);\n };\n\n // attach to the session span if called outside of the AgentSession\n const activeSpan = trace.getActiveSpan();\n if (!activeSpan && this.rootSpanContext) {\n return otelContext.with(this.rootSpanContext, () => doSay(this.activity!));\n }\n\n return doSay(this.activity);\n }\n\n interrupt() {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n return this.activity.interrupt();\n }\n\n generateReply(options?: {\n userInput?: string;\n instructions?: string;\n toolChoice?: ToolChoice;\n allowInterruptions?: boolean;\n }): SpeechHandle {\n if (!this.activity) {\n throw new Error('AgentSession is not running');\n }\n\n const userMessage = options?.userInput\n ? new ChatMessage({\n role: 'user',\n content: options.userInput,\n })\n : undefined;\n\n const doGenerateReply = (activity: AgentActivity, nextActivity?: AgentActivity) => {\n if (activity.draining) {\n if (!nextActivity) {\n throw new Error('AgentSession is closing, cannot use generateReply()');\n }\n return nextActivity.generateReply({ userMessage, ...options });\n }\n return activity.generateReply({ userMessage, ...options });\n };\n\n // attach to the session span if called outside of the AgentSession\n const activeSpan = trace.getActiveSpan();\n let handle: SpeechHandle;\n if (!activeSpan && this.rootSpanContext) {\n handle = otelContext.with(this.rootSpanContext, () =>\n doGenerateReply(this.activity!, this.nextActivity),\n );\n } else {\n handle = doGenerateReply(this.activity!, this.nextActivity);\n }\n\n if (this._globalRunState) {\n this._globalRunState._watchHandle(handle);\n }\n\n return handle;\n }\n\n /**\n * Run a test with user input and return a result for assertions.\n *\n * This method is primarily used for testing agent behavior without\n * requiring a real room connection.\n *\n * @example\n * ```typescript\n * const result = await session.run({ userInput: 'Hello' });\n * result.expect.nextEvent().isMessage({ role: 'assistant' });\n * result.expect.noMoreEvents();\n * ```\n *\n * @param options - Run options including user input\n * @returns A RunResult that resolves when the agent finishes responding\n *\n * TODO: Add outputType parameter for typed outputs (parity with Python)\n */\n run(options: { userInput: string }): RunResult {\n if (this._globalRunState && !this._globalRunState.done()) {\n throw new Error('nested runs are not supported');\n }\n\n const runState = new RunResult({ userInput: options.userInput });\n this._globalRunState = runState;\n this.generateReply({ userInput: options.userInput });\n\n return runState;\n }\n\n private async updateActivity(agent: Agent): Promise<void> {\n const runWithContext = async () => {\n // TODO(AJS-129): add lock to agent activity core lifecycle\n this.nextActivity = new AgentActivity(agent, this);\n\n const previousActivity = this.activity;\n\n if (this.activity) {\n await this.activity.drain();\n await this.activity.close();\n }\n\n this.activity = this.nextActivity;\n this.nextActivity = undefined;\n\n this._chatCtx.insert(\n new AgentHandoffItem({\n oldAgentId: previousActivity?.agent.id,\n newAgentId: agent.id,\n }),\n );\n this.logger.debug(\n { previousAgentId: previousActivity?.agent.id, newAgentId: agent.id },\n 'Agent handoff inserted into chat context',\n );\n\n await this.activity.start();\n\n if (this._input.audio) {\n this.activity.attachAudioInput(this._input.audio.stream);\n }\n };\n\n // Run within session span context if available\n if (this.rootSpanContext) {\n return otelContext.with(this.rootSpanContext, runWithContext);\n }\n\n return runWithContext();\n }\n\n get chatCtx(): ChatContext {\n return this._chatCtx.copy();\n }\n\n get agentState(): AgentState {\n return this._agentState;\n }\n\n get currentAgent(): Agent {\n if (!this.agent) {\n throw new Error('AgentSession is not running');\n }\n\n return this.agent;\n }\n\n async close(): Promise<void> {\n await this.closeImpl(CloseReason.USER_INITIATED);\n }\n\n shutdown(options?: { drain?: boolean; reason?: ShutdownReason }): void {\n const { drain = true, reason = CloseReason.USER_INITIATED } = options ?? {};\n\n this._closeSoon({\n reason,\n drain,\n });\n }\n\n /** @internal */\n _closeSoon({\n reason,\n drain = false,\n error = null,\n }: {\n reason: ShutdownReason;\n drain?: boolean;\n error?: RealtimeModelError | STTError | TTSError | LLMError | null;\n }): void {\n if (this.closingTask) {\n return;\n }\n this.closeImpl(reason, error, drain);\n }\n\n /** @internal */\n _onError(error: RealtimeModelError | STTError | TTSError | LLMError): void {\n if (this.closingTask || error.recoverable) {\n return;\n }\n\n // Track error counts per type to implement max_unrecoverable_errors logic\n if (error.type === 'llm_error') {\n this.llmErrorCounts += 1;\n if (this.llmErrorCounts <= this._connOptions.maxUnrecoverableErrors) {\n return;\n }\n } else if (error.type === 'tts_error') {\n this.ttsErrorCounts += 1;\n if (this.ttsErrorCounts <= this._connOptions.maxUnrecoverableErrors) {\n return;\n }\n }\n\n this.logger.error(error, 'AgentSession is closing due to unrecoverable error');\n\n this.closingTask = (async () => {\n await this.closeImpl(CloseReason.ERROR, error);\n })().then(() => {\n this.closingTask = null;\n });\n }\n\n /** @internal */\n _conversationItemAdded(item: ChatMessage): void {\n this._chatCtx.insert(item);\n this.emit(AgentSessionEventTypes.ConversationItemAdded, createConversationItemAddedEvent(item));\n }\n\n /** @internal */\n _toolItemsAdded(items: (FunctionCall | FunctionCallOutput)[]): void {\n this._chatCtx.insert(items);\n }\n\n /** @internal */\n _updateAgentState(state: AgentState, options?: { startTime?: number; otelContext?: Context }) {\n if (this._agentState === state) {\n return;\n }\n\n if (state === 'speaking') {\n // Reset error counts when agent starts speaking\n this.llmErrorCounts = 0;\n this.ttsErrorCounts = 0;\n\n if (this.agentSpeakingSpan === undefined) {\n this.agentSpeakingSpan = tracer.startSpan({\n name: 'agent_speaking',\n context: options?.otelContext ?? this.rootSpanContext,\n startTime: options?.startTime,\n });\n\n const localParticipant = this._roomIO?.localParticipant;\n if (localParticipant) {\n setParticipantSpanAttributes(this.agentSpeakingSpan, localParticipant);\n }\n }\n } else if (this.agentSpeakingSpan !== undefined) {\n // TODO(brian): PR4 - Set ATTR_END_TIME attribute if available\n this.agentSpeakingSpan.end();\n this.agentSpeakingSpan = undefined;\n }\n\n const oldState = this._agentState;\n this._agentState = state;\n\n // Handle user away timer based on state changes\n if (state === 'listening' && this.userState === 'listening') {\n this._setUserAwayTimer();\n } else {\n this._cancelUserAwayTimer();\n }\n\n this.emit(\n AgentSessionEventTypes.AgentStateChanged,\n createAgentStateChangedEvent(oldState, state),\n );\n }\n\n /** @internal */\n _updateUserState(state: UserState, lastSpeakingTime?: number) {\n if (this.userState === state) {\n return;\n }\n\n if (state === 'speaking' && this.userSpeakingSpan === undefined) {\n this.userSpeakingSpan = tracer.startSpan({\n name: 'user_speaking',\n context: this.rootSpanContext,\n startTime: lastSpeakingTime,\n });\n\n const linked = this._roomIO?.linkedParticipant;\n if (linked) {\n setParticipantSpanAttributes(this.userSpeakingSpan, linked);\n }\n } else if (this.userSpeakingSpan !== undefined) {\n this.userSpeakingSpan.end(lastSpeakingTime);\n this.userSpeakingSpan = undefined;\n }\n\n const oldState = this.userState;\n this.userState = state;\n\n // Handle user away timer based on state changes\n if (state === 'listening' && this._agentState === 'listening') {\n this._setUserAwayTimer();\n } else {\n this._cancelUserAwayTimer();\n }\n\n this.emit(\n AgentSessionEventTypes.UserStateChanged,\n createUserStateChangedEvent(oldState, state),\n );\n }\n\n // -- User changed input/output streams/sinks --\n private onAudioInputChanged(): void {\n if (!this.started) {\n return;\n }\n\n if (this.activity && this._input.audio) {\n this.activity.attachAudioInput(this._input.audio.stream);\n }\n }\n\n private onAudioOutputChanged(): void {}\n\n private onTextOutputChanged(): void {}\n\n private _setUserAwayTimer(): void {\n this._cancelUserAwayTimer();\n\n if (this.options.userAwayTimeout === null || this.options.userAwayTimeout === undefined) {\n return;\n }\n\n if (this._roomIO && !this._roomIO.isParticipantAvailable) {\n return;\n }\n\n this.userAwayTimer = setTimeout(() => {\n this.logger.debug('User away timeout triggered');\n this._updateUserState('away');\n }, this.options.userAwayTimeout * 1000);\n }\n\n private _cancelUserAwayTimer(): void {\n if (this.userAwayTimer !== null) {\n clearTimeout(this.userAwayTimer);\n this.userAwayTimer = null;\n }\n }\n\n private _onUserInputTranscribed(ev: UserInputTranscribedEvent): void {\n if (this.userState === 'away' && ev.isFinal) {\n this.logger.debug('User returned from away state due to speech input');\n this._updateUserState('listening');\n }\n }\n\n private async closeImpl(\n reason: ShutdownReason,\n error: RealtimeModelError | LLMError | TTSError | STTError | null = null,\n drain: boolean = false,\n ): Promise<void> {\n if (this.rootSpanContext) {\n return otelContext.with(this.rootSpanContext, async () => {\n await this.closeImplInner(reason, error, drain);\n });\n }\n\n return this.closeImplInner(reason, error, drain);\n }\n\n private async closeImplInner(\n reason: ShutdownReason,\n error: RealtimeModelError | LLMError | TTSError | STTError | null = null,\n drain: boolean = false,\n ): Promise<void> {\n if (!this.started) {\n return;\n }\n\n this._cancelUserAwayTimer();\n this.off(AgentSessionEventTypes.UserInputTranscribed, this._onUserInputTranscribed);\n\n if (this.activity) {\n if (!drain) {\n try {\n this.activity.interrupt();\n } catch (error) {\n // TODO(shubhra): force interrupt or wait for it to finish?\n // it might be an audio played from the error callback\n }\n }\n await this.activity.drain();\n // wait any uninterruptible speech to finish\n await this.activity.currentSpeech?.waitForPlayout();\n try {\n this.activity.detachAudioInput();\n } catch (error) {\n // Ignore detach errors during cleanup - source may not have been set\n }\n }\n\n // Close recorder before detaching inputs/outputs (keep reference for session report)\n if (this._recorderIO) {\n await this._recorderIO.close();\n }\n\n // detach the inputs and outputs\n this.input.audio = null;\n this.output.audio = null;\n this.output.transcription = null;\n\n await this._roomIO?.close();\n this._roomIO = undefined;\n\n await this.activity?.close();\n this.activity = undefined;\n\n if (this.sessionSpan) {\n this.sessionSpan.end();\n this.sessionSpan = undefined;\n }\n\n if (this.userSpeakingSpan) {\n this.userSpeakingSpan.end();\n this.userSpeakingSpan = undefined;\n }\n\n if (this.agentSpeakingSpan) {\n this.agentSpeakingSpan.end();\n this.agentSpeakingSpan = undefined;\n }\n\n this.started = false;\n\n this.emit(AgentSessionEventTypes.Close, createCloseEvent(reason, error));\n\n this.userState = 'listening';\n this._agentState = 'initializing';\n this.rootSpanContext = undefined;\n this.llmErrorCounts = 0;\n this.ttsErrorCounts = 0;\n\n this.logger.info({ reason, error }, 'AgentSession closed');\n }\n}\n"],"mappings":"AAMA,SAAS,cAAc,WAAW,aAAa,aAAa;AAC5D,SAAS,oBAAoB;AAE7B;AAAA,EACE,OAAO;AAAA,EACP,OAAO;AAAA,EACP,OAAO;AAAA,OAIF;AACP,SAA0B,qBAAqB;AAE/C,SAAS,kBAAkB,aAAa,mBAAmB;AAG3D,SAAS,WAAW;AAGpB,SAAS,YAAY,cAAc;AAEnC;AAAA,EACE;AAAA,EACA;AAAA,OAGK;AAGP,SAAS,qBAAqB;AAE9B;AAAA,EAEE;AAAA,EAIA;AAAA,EAUA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,YAAY,mBAAmB;AACxC,SAAS,kBAAkB;AAC3B,SAAS,cAA6D;AAGtE,SAAS,iBAAiB;AAC1B,SAAS,oCAAoC;AAe7C,MAAM,sBAAoC;AAAA,EACxC,oBAAoB;AAAA,EACpB,+BAA+B;AAAA,EAC/B,yBAAyB;AAAA,EACzB,sBAAsB;AAAA,EACtB,qBAAqB;AAAA,EACrB,qBAAqB;AAAA,EACrB,cAAc;AAAA,EACd,sBAAsB;AAAA,EACtB,iBAAiB;AAAA,EACjB,yBAAyB;AAC3B;AA2BO,MAAM,qBAEF,aAA+D;AAAA,EACxE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAES;AAAA,EAED;AAAA,EACA;AAAA,EACA;AAAA,EACA,UAAU;AAAA,EACV,YAAuB;AAAA;AAAA,EAG/B;AAAA,EACQ,SAAS,IAAI;AAAA,EAEb;AAAA,EACA;AAAA,EACA,cAA0B;AAAA,EAE1B;AAAA,EACA;AAAA,EAEA,cAAoC;AAAA,EACpC,gBAAuC;AAAA;AAAA,EAGvC;AAAA;AAAA,EAGA,iBAAiB;AAAA,EACjB,iBAAiB;AAAA,EAEjB;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAGR;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA,kBAAgC,CAAC;AAAA;AAAA,EAGjC,mBAAmB;AAAA;AAAA,EAGnB;AAAA;AAAA,EAGA;AAAA,EAEA,YAAY,MAAqC;AAC/C,UAAM;AAEN,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,eAAe;AAAA,MACf;AAAA,IACF,IAAI;AAGJ,SAAK,eAAe;AAAA,MAClB,gBAAgB,EAAE,GAAG,6BAA6B,GAAG,2CAAa,eAAe;AAAA,MACjF,gBAAgB,EAAE,GAAG,6BAA6B,GAAG,2CAAa,eAAe;AAAA,MACjF,gBAAgB,EAAE,GAAG,6BAA6B,GAAG,2CAAa,eAAe;AAAA,MACjF,yBACE,2CAAa,2BACb,gCAAgC;AAAA,IACpC;AAEA,SAAK,MAAM;AAEX,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,aAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,aAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,aAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,SAAK,gBAAgB;AACrB,SAAK,YAAY;AAGjB,SAAK,SAAS,IAAI,WAAW,KAAK,mBAAmB;AACrD,SAAK,UAAU,IAAI,YAAY,KAAK,sBAAsB,KAAK,mBAAmB;AAGlF,SAAK,WAAW,YAAY,MAAM;AAClC,SAAK,UAAU,EAAE,GAAG,qBAAqB,GAAG,aAAa;AAEzD,SAAK,0BAA0B,KAAK,wBAAwB,KAAK,IAAI;AACrE,SAAK,GAAG,uBAAuB,sBAAsB,KAAK,uBAAuB;AAAA,EACnF;AAAA,EAEA,KACE,UACG,MACM;AACT,UAAM,YAAY,KAAK,CAAC;AACxB,SAAK,gBAAgB,KAAK,SAAS;AACnC,WAAO,MAAM,KAAK,OAAO,GAAG,IAAI;AAAA,EAClC;AAAA,EAEA,IAAI,QAAoB;AACtB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,SAAsB;AACxB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,WAAqB;AACvB,QAAI,KAAK,cAAc,QAAW;AAChC,YAAM,IAAI,MAAM,iCAAiC;AAAA,IACnD;AAEA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA;AAAA,EAGA,IAAI,cAA6C;AAC/C,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,0BAAmC;AACrC,WAAO,KAAK,QAAQ;AAAA,EACtB;AAAA,EAEA,IAAI,SAAS,OAAiB;AAC5B,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,MAAc,WAAW;AAAA,IACvB;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAMkB;AAChB,SAAK,aAAa,WAAW,kBAAkB,MAAM,EAAE;AAEvD,SAAK,QAAQ;AACb,SAAK,kBAAkB,cAAc;AAErC,UAAM,QAAyB,CAAC;AAEhC,QAAI,QAAQ,CAAC,KAAK,SAAS;AAEzB,UAAI,KAAK,MAAM,UAAS,6CAAc,kBAAiB,OAAO;AAC5D,aAAK,OAAO;AAAA,UACV;AAAA,QACF;AAAA,MACF;AAEA,UAAI,KAAK,OAAO,UAAS,+CAAe,kBAAiB,OAAO;AAC9D,aAAK,OAAO;AAAA,UACV;AAAA,QACF;AAAA,MACF;AAEA,UAAI,KAAK,OAAO,kBAAiB,+CAAe,0BAAyB,OAAO;AAC9E,aAAK,OAAO;AAAA,UACV;AAAA,QACF;AAAA,MACF;AAEA,WAAK,UAAU,IAAI,OAAO;AAAA,QACxB,cAAc;AAAA,QACd;AAAA,QACA;AAAA,QACA;AAAA,MACF,CAAC;AACD,WAAK,QAAQ,MAAM;AAAA,IACrB;AAEA,QAAI,MAA8B;AAClC,QAAI;AACF,YAAM,cAAc;AAAA,IACtB,QAAQ;AAAA,IAER;AAEA,QAAI,KAAK;AACP,UAAI,QAAQ,IAAI,SAAS,QAAQ,CAAC,KAAK,aAAa;AAClD,aAAK,OAAO,MAAM,yCAAyC;AAC3D,cAAM,KAAK,IAAI,QAAQ,CAAC;AAAA,MAC1B;AAEA,UAAI,IAAI,yBAAyB,QAAW;AAC1C,YAAI,uBAAuB;AAAA,MAC7B,WAAW,KAAK,kBAAkB;AAChC,cAAM,IAAI;AAAA,UACR;AAAA,QACF;AAAA,MACF;AAEA,UAAI,KAAK,MAAM,SAAS,KAAK,OAAO,SAAS,KAAK,kBAAkB;AAClE,aAAK,cAAc,IAAI,WAAW,EAAE,cAAc,KAAK,CAAC;AACxD,aAAK,MAAM,QAAQ,KAAK,YAAY,YAAY,KAAK,MAAM,KAAK;AAChE,aAAK,OAAO,QAAQ,KAAK,YAAY,aAAa,KAAK,OAAO,KAAK;AAGnE,cAAM,aAAa,IAAI;AACvB,YAAI,YAAY;AACd,gBAAM,KAAK,KAAK,YAAY,MAAM,GAAG,UAAU,YAAY,CAAC;AAAA,QAC9D;AAAA,MACF;AAAA,IACF;AAGA,UAAM,KAAK,KAAK,eAAe,KAAK,KAAK,CAAC;AAE1C,UAAM,QAAQ,WAAW,KAAK;AAG9B,SAAK,OAAO;AAAA,MACV,mBAAmB,KAAK,MAAM,QAAQ,MAAM,KAAK,MAAM,MAAM,YAAY,OAAO,MAAM,QAAQ,2BAA2B,KAAK,OAAO,QAAQ,MAAM,KAAK,OAAO,MAAM,YAAY,OAAO,MAAM,QAAQ;AAAA,IACxM;AAEA,SAAK,OAAO;AAAA,MACV,4CAA4C,KAAK,OAAO,gBAAgB,MAAM,KAAK,OAAO,cAAc,YAAY,OAAO,MAAM,QAAQ;AAAA,IAC3I;AAEA,SAAK,UAAU;AACf,SAAK,aAAa,KAAK,IAAI;AAC3B,SAAK,kBAAkB,WAAW;AAAA,EACpC;AAAA,EAEA,MAAM,MAAM;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAMkB;AAChB,QAAI,KAAK,SAAS;AAChB;AAAA,IACF;AAEA,QAAI,MAA8B;AAClC,QAAI;AACF,YAAM,cAAc;AAEpB,UAAI,WAAW,QAAW;AACxB,iBAAS,IAAI,IAAI;AAAA,MACnB;AAEA,WAAK,mBAAmB;AAExB,UAAI,KAAK,kBAAkB;AACzB,YAAI,cAAc;AAAA,MACpB;AAAA,IACF,SAAS,OAAO;AAEd,WAAK,OAAO,KAAK,6BAA6B;AAAA,IAChD;AAEA,SAAK,cAAc,OAAO,UAAU;AAAA,MAClC,MAAM;AAAA,MACN,SAAS;AAAA,IACX,CAAC;AAED,SAAK,kBAAkB,MAAM,QAAQ,cAAc,KAAK,WAAW;AAEnE,UAAM,KAAK,WAAW;AAAA,MACpB;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,MAAM,KAAK;AAAA,IACb,CAAC;AAAA,EACH;AAAA,EAEA,YAAY,OAAoB;AAC9B,SAAK,QAAQ;AAEb,QAAI,KAAK,SAAS;AAChB,WAAK,eAAe,KAAK;AAAA,IAC3B;AAAA,EACF;AAAA,EAEA,iBAAiB;AACf,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,SAAK,SAAS,eAAe;AAAA,EAC/B;AAAA,EAEA,gBAAgB;AACd,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AACA,SAAK,SAAS,cAAc;AAAA,EAC9B;AAAA,EAEA,IACE,MACA,SAKc;AACd,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,UAAM,QAAQ,CAAC,aAA4B;AACzC,aAAO,SAAS,IAAI,MAAM,OAAO;AAAA,IACnC;AAGA,UAAM,aAAa,MAAM,cAAc;AACvC,QAAI,CAAC,cAAc,KAAK,iBAAiB;AACvC,aAAO,YAAY,KAAK,KAAK,iBAAiB,MAAM,MAAM,KAAK,QAAS,CAAC;AAAA,IAC3E;AAEA,WAAO,MAAM,KAAK,QAAQ;AAAA,EAC5B;AAAA,EAEA,YAAY;AACV,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AACA,WAAO,KAAK,SAAS,UAAU;AAAA,EACjC;AAAA,EAEA,cAAc,SAKG;AACf,QAAI,CAAC,KAAK,UAAU;AAClB,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,UAAM,eAAc,mCAAS,aACzB,IAAI,YAAY;AAAA,MACd,MAAM;AAAA,MACN,SAAS,QAAQ;AAAA,IACnB,CAAC,IACD;AAEJ,UAAM,kBAAkB,CAAC,UAAyB,iBAAiC;AACjF,UAAI,SAAS,UAAU;AACrB,YAAI,CAAC,cAAc;AACjB,gBAAM,IAAI,MAAM,qDAAqD;AAAA,QACvE;AACA,eAAO,aAAa,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,MAC/D;AACA,aAAO,SAAS,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,IAC3D;AAGA,UAAM,aAAa,MAAM,cAAc;AACvC,QAAI;AACJ,QAAI,CAAC,cAAc,KAAK,iBAAiB;AACvC,eAAS,YAAY;AAAA,QAAK,KAAK;AAAA,QAAiB,MAC9C,gBAAgB,KAAK,UAAW,KAAK,YAAY;AAAA,MACnD;AAAA,IACF,OAAO;AACL,eAAS,gBAAgB,KAAK,UAAW,KAAK,YAAY;AAAA,IAC5D;AAEA,QAAI,KAAK,iBAAiB;AACxB,WAAK,gBAAgB,aAAa,MAAM;AAAA,IAC1C;AAEA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAoBA,IAAI,SAA2C;AAC7C,QAAI,KAAK,mBAAmB,CAAC,KAAK,gBAAgB,KAAK,GAAG;AACxD,YAAM,IAAI,MAAM,+BAA+B;AAAA,IACjD;AAEA,UAAM,WAAW,IAAI,UAAU,EAAE,WAAW,QAAQ,UAAU,CAAC;AAC/D,SAAK,kBAAkB;AACvB,SAAK,cAAc,EAAE,WAAW,QAAQ,UAAU,CAAC;AAEnD,WAAO;AAAA,EACT;AAAA,EAEA,MAAc,eAAe,OAA6B;AACxD,UAAM,iBAAiB,YAAY;AAEjC,WAAK,eAAe,IAAI,cAAc,OAAO,IAAI;AAEjD,YAAM,mBAAmB,KAAK;AAE9B,UAAI,KAAK,UAAU;AACjB,cAAM,KAAK,SAAS,MAAM;AAC1B,cAAM,KAAK,SAAS,MAAM;AAAA,MAC5B;AAEA,WAAK,WAAW,KAAK;AACrB,WAAK,eAAe;AAEpB,WAAK,SAAS;AAAA,QACZ,IAAI,iBAAiB;AAAA,UACnB,YAAY,qDAAkB,MAAM;AAAA,UACpC,YAAY,MAAM;AAAA,QACpB,CAAC;AAAA,MACH;AACA,WAAK,OAAO;AAAA,QACV,EAAE,iBAAiB,qDAAkB,MAAM,IAAI,YAAY,MAAM,GAAG;AAAA,QACpE;AAAA,MACF;AAEA,YAAM,KAAK,SAAS,MAAM;AAE1B,UAAI,KAAK,OAAO,OAAO;AACrB,aAAK,SAAS,iBAAiB,KAAK,OAAO,MAAM,MAAM;AAAA,MACzD;AAAA,IACF;AAGA,QAAI,KAAK,iBAAiB;AACxB,aAAO,YAAY,KAAK,KAAK,iBAAiB,cAAc;AAAA,IAC9D;AAEA,WAAO,eAAe;AAAA,EACxB;AAAA,EAEA,IAAI,UAAuB;AACzB,WAAO,KAAK,SAAS,KAAK;AAAA,EAC5B;AAAA,EAEA,IAAI,aAAyB;AAC3B,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAsB;AACxB,QAAI,CAAC,KAAK,OAAO;AACf,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,QAAuB;AAC3B,UAAM,KAAK,UAAU,YAAY,cAAc;AAAA,EACjD;AAAA,EAEA,SAAS,SAA8D;AACrE,UAAM,EAAE,QAAQ,MAAM,SAAS,YAAY,eAAe,IAAI,WAAW,CAAC;AAE1E,SAAK,WAAW;AAAA,MACd;AAAA,MACA;AAAA,IACF,CAAC;AAAA,EACH;AAAA;AAAA,EAGA,WAAW;AAAA,IACT;AAAA,IACA,QAAQ;AAAA,IACR,QAAQ;AAAA,EACV,GAIS;AACP,QAAI,KAAK,aAAa;AACpB;AAAA,IACF;AACA,SAAK,UAAU,QAAQ,OAAO,KAAK;AAAA,EACrC;AAAA;AAAA,EAGA,SAAS,OAAkE;AACzE,QAAI,KAAK,eAAe,MAAM,aAAa;AACzC;AAAA,IACF;AAGA,QAAI,MAAM,SAAS,aAAa;AAC9B,WAAK,kBAAkB;AACvB,UAAI,KAAK,kBAAkB,KAAK,aAAa,wBAAwB;AACnE;AAAA,MACF;AAAA,IACF,WAAW,MAAM,SAAS,aAAa;AACrC,WAAK,kBAAkB;AACvB,UAAI,KAAK,kBAAkB,KAAK,aAAa,wBAAwB;AACnE;AAAA,MACF;AAAA,IACF;AAEA,SAAK,OAAO,MAAM,OAAO,oDAAoD;AAE7E,SAAK,eAAe,YAAY;AAC9B,YAAM,KAAK,UAAU,YAAY,OAAO,KAAK;AAAA,IAC/C,GAAG,EAAE,KAAK,MAAM;AACd,WAAK,cAAc;AAAA,IACrB,CAAC;AAAA,EACH;AAAA;AAAA,EAGA,uBAAuB,MAAyB;AAC9C,SAAK,SAAS,OAAO,IAAI;AACzB,SAAK,KAAK,uBAAuB,uBAAuB,iCAAiC,IAAI,CAAC;AAAA,EAChG;AAAA;AAAA,EAGA,gBAAgB,OAAoD;AAClE,SAAK,SAAS,OAAO,KAAK;AAAA,EAC5B;AAAA;AAAA,EAGA,kBAAkB,OAAmB,SAAyD;AA/qBhG;AAgrBI,QAAI,KAAK,gBAAgB,OAAO;AAC9B;AAAA,IACF;AAEA,QAAI,UAAU,YAAY;AAExB,WAAK,iBAAiB;AACtB,WAAK,iBAAiB;AAEtB,UAAI,KAAK,sBAAsB,QAAW;AACxC,aAAK,oBAAoB,OAAO,UAAU;AAAA,UACxC,MAAM;AAAA,UACN,UAAS,mCAAS,gBAAe,KAAK;AAAA,UACtC,WAAW,mCAAS;AAAA,QACtB,CAAC;AAED,cAAM,oBAAmB,UAAK,YAAL,mBAAc;AACvC,YAAI,kBAAkB;AACpB,uCAA6B,KAAK,mBAAmB,gBAAgB;AAAA,QACvE;AAAA,MACF;AAAA,IACF,WAAW,KAAK,sBAAsB,QAAW;AAE/C,WAAK,kBAAkB,IAAI;AAC3B,WAAK,oBAAoB;AAAA,IAC3B;AAEA,UAAM,WAAW,KAAK;AACtB,SAAK,cAAc;AAGnB,QAAI,UAAU,eAAe,KAAK,cAAc,aAAa;AAC3D,WAAK,kBAAkB;AAAA,IACzB,OAAO;AACL,WAAK,qBAAqB;AAAA,IAC5B;AAEA,SAAK;AAAA,MACH,uBAAuB;AAAA,MACvB,6BAA6B,UAAU,KAAK;AAAA,IAC9C;AAAA,EACF;AAAA;AAAA,EAGA,iBAAiB,OAAkB,kBAA2B;AA5tBhE;AA6tBI,QAAI,KAAK,cAAc,OAAO;AAC5B;AAAA,IACF;AAEA,QAAI,UAAU,cAAc,KAAK,qBAAqB,QAAW;AAC/D,WAAK,mBAAmB,OAAO,UAAU;AAAA,QACvC,MAAM;AAAA,QACN,SAAS,KAAK;AAAA,QACd,WAAW;AAAA,MACb,CAAC;AAED,YAAM,UAAS,UAAK,YAAL,mBAAc;AAC7B,UAAI,QAAQ;AACV,qCAA6B,KAAK,kBAAkB,MAAM;AAAA,MAC5D;AAAA,IACF,WAAW,KAAK,qBAAqB,QAAW;AAC9C,WAAK,iBAAiB,IAAI,gBAAgB;AAC1C,WAAK,mBAAmB;AAAA,IAC1B;AAEA,UAAM,WAAW,KAAK;AACtB,SAAK,YAAY;AAGjB,QAAI,UAAU,eAAe,KAAK,gBAAgB,aAAa;AAC7D,WAAK,kBAAkB;AAAA,IACzB,OAAO;AACL,WAAK,qBAAqB;AAAA,IAC5B;AAEA,SAAK;AAAA,MACH,uBAAuB;AAAA,MACvB,4BAA4B,UAAU,KAAK;AAAA,IAC7C;AAAA,EACF;AAAA;AAAA,EAGQ,sBAA4B;AAClC,QAAI,CAAC,KAAK,SAAS;AACjB;AAAA,IACF;AAEA,QAAI,KAAK,YAAY,KAAK,OAAO,OAAO;AACtC,WAAK,SAAS,iBAAiB,KAAK,OAAO,MAAM,MAAM;AAAA,IACzD;AAAA,EACF;AAAA,EAEQ,uBAA6B;AAAA,EAAC;AAAA,EAE9B,sBAA4B;AAAA,EAAC;AAAA,EAE7B,oBAA0B;AAChC,SAAK,qBAAqB;AAE1B,QAAI,KAAK,QAAQ,oBAAoB,QAAQ,KAAK,QAAQ,oBAAoB,QAAW;AACvF;AAAA,IACF;AAEA,QAAI,KAAK,WAAW,CAAC,KAAK,QAAQ,wBAAwB;AACxD;AAAA,IACF;AAEA,SAAK,gBAAgB,WAAW,MAAM;AACpC,WAAK,OAAO,MAAM,6BAA6B;AAC/C,WAAK,iBAAiB,MAAM;AAAA,IAC9B,GAAG,KAAK,QAAQ,kBAAkB,GAAI;AAAA,EACxC;AAAA,EAEQ,uBAA6B;AACnC,QAAI,KAAK,kBAAkB,MAAM;AAC/B,mBAAa,KAAK,aAAa;AAC/B,WAAK,gBAAgB;AAAA,IACvB;AAAA,EACF;AAAA,EAEQ,wBAAwB,IAAqC;AACnE,QAAI,KAAK,cAAc,UAAU,GAAG,SAAS;AAC3C,WAAK,OAAO,MAAM,mDAAmD;AACrE,WAAK,iBAAiB,WAAW;AAAA,IACnC;AAAA,EACF;AAAA,EAEA,MAAc,UACZ,QACA,QAAoE,MACpE,QAAiB,OACF;AACf,QAAI,KAAK,iBAAiB;AACxB,aAAO,YAAY,KAAK,KAAK,iBAAiB,YAAY;AACxD,cAAM,KAAK,eAAe,QAAQ,OAAO,KAAK;AAAA,MAChD,CAAC;AAAA,IACH;AAEA,WAAO,KAAK,eAAe,QAAQ,OAAO,KAAK;AAAA,EACjD;AAAA,EAEA,MAAc,eACZ,QACA,QAAoE,MACpE,QAAiB,OACF;AAj0BnB;AAk0BI,QAAI,CAAC,KAAK,SAAS;AACjB;AAAA,IACF;AAEA,SAAK,qBAAqB;AAC1B,SAAK,IAAI,uBAAuB,sBAAsB,KAAK,uBAAuB;AAElF,QAAI,KAAK,UAAU;AACjB,UAAI,CAAC,OAAO;AACV,YAAI;AACF,eAAK,SAAS,UAAU;AAAA,QAC1B,SAASA,QAAO;AAAA,QAGhB;AAAA,MACF;AACA,YAAM,KAAK,SAAS,MAAM;AAE1B,cAAM,UAAK,SAAS,kBAAd,mBAA6B;AACnC,UAAI;AACF,aAAK,SAAS,iBAAiB;AAAA,MACjC,SAASA,QAAO;AAAA,MAEhB;AAAA,IACF;AAGA,QAAI,KAAK,aAAa;AACpB,YAAM,KAAK,YAAY,MAAM;AAAA,IAC/B;AAGA,SAAK,MAAM,QAAQ;AACnB,SAAK,OAAO,QAAQ;AACpB,SAAK,OAAO,gBAAgB;AAE5B,YAAM,UAAK,YAAL,mBAAc;AACpB,SAAK,UAAU;AAEf,YAAM,UAAK,aAAL,mBAAe;AACrB,SAAK,WAAW;AAEhB,QAAI,KAAK,aAAa;AACpB,WAAK,YAAY,IAAI;AACrB,WAAK,cAAc;AAAA,IACrB;AAEA,QAAI,KAAK,kBAAkB;AACzB,WAAK,iBAAiB,IAAI;AAC1B,WAAK,mBAAmB;AAAA,IAC1B;AAEA,QAAI,KAAK,mBAAmB;AAC1B,WAAK,kBAAkB,IAAI;AAC3B,WAAK,oBAAoB;AAAA,IAC3B;AAEA,SAAK,UAAU;AAEf,SAAK,KAAK,uBAAuB,OAAO,iBAAiB,QAAQ,KAAK,CAAC;AAEvE,SAAK,YAAY;AACjB,SAAK,cAAc;AACnB,SAAK,kBAAkB;AACvB,SAAK,iBAAiB;AACtB,SAAK,iBAAiB;AAEtB,SAAK,OAAO,KAAK,EAAE,QAAQ,MAAM,GAAG,qBAAqB;AAAA,EAC3D;AACF;","names":["error"]}
@@ -22,6 +22,7 @@ __export(audio_recognition_exports, {
22
22
  });
23
23
  module.exports = __toCommonJS(audio_recognition_exports);
24
24
  var import_rtc_node = require("@livekit/rtc-node");
25
+ var import_api = require("@opentelemetry/api");
25
26
  var import_web = require("node:stream/web");
26
27
  var import_chat_context = require("../llm/chat_context.cjs");
27
28
  var import_log = require("../log.cjs");
@@ -32,6 +33,7 @@ var import_stt = require("../stt/stt.cjs");
32
33
  var import_telemetry = require("../telemetry/index.cjs");
33
34
  var import_utils = require("../utils.cjs");
34
35
  var import_vad = require("../vad.cjs");
36
+ var import_utils2 = require("./utils.cjs");
35
37
  class AudioRecognition {
36
38
  hooks;
37
39
  stt;
@@ -42,6 +44,9 @@ class AudioRecognition {
42
44
  maxEndpointingDelay;
43
45
  lastLanguage;
44
46
  rootSpanContext;
47
+ sttModel;
48
+ sttProvider;
49
+ getLinkedParticipant;
45
50
  deferredInputStream;
46
51
  logger = (0, import_log.log)();
47
52
  lastFinalTranscriptTime = 0;
@@ -74,6 +79,9 @@ class AudioRecognition {
74
79
  this.maxEndpointingDelay = opts.maxEndpointingDelay;
75
80
  this.lastLanguage = void 0;
76
81
  this.rootSpanContext = opts.rootSpanContext;
82
+ this.sttModel = opts.sttModel;
83
+ this.sttProvider = opts.sttProvider;
84
+ this.getLinkedParticipant = opts.getLinkedParticipant;
77
85
  this.deferredInputStream = new import_deferred_stream.DeferredReadableStream();
78
86
  const [vadInputStream, sttInputStream] = this.deferredInputStream.stream.tee();
79
87
  this.vadInputStream = vadInputStream;
@@ -99,6 +107,32 @@ class AudioRecognition {
99
107
  this.logger.error(`Error running STT task: ${err}`);
100
108
  });
101
109
  }
110
+ ensureUserTurnSpan(startTime) {
111
+ var _a;
112
+ if (this.userTurnSpan && this.userTurnSpan.isRecording()) {
113
+ return this.userTurnSpan;
114
+ }
115
+ this.userTurnSpan = import_telemetry.tracer.startSpan({
116
+ name: "user_turn",
117
+ context: this.rootSpanContext,
118
+ startTime
119
+ });
120
+ const participant = (_a = this.getLinkedParticipant) == null ? void 0 : _a.call(this);
121
+ if (participant) {
122
+ (0, import_utils2.setParticipantSpanAttributes)(this.userTurnSpan, participant);
123
+ }
124
+ if (this.sttModel) {
125
+ this.userTurnSpan.setAttribute(import_telemetry.traceTypes.ATTR_GEN_AI_REQUEST_MODEL, this.sttModel);
126
+ }
127
+ if (this.sttProvider) {
128
+ this.userTurnSpan.setAttribute(import_telemetry.traceTypes.ATTR_GEN_AI_PROVIDER_NAME, this.sttProvider);
129
+ }
130
+ return this.userTurnSpan;
131
+ }
132
+ userTurnContext(span) {
133
+ const base = this.rootSpanContext ?? import_api.ROOT_CONTEXT;
134
+ return import_api.trace.setSpan(base, span);
135
+ }
102
136
  async onSTTEvent(ev) {
103
137
  var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q, _r;
104
138
  if (this.turnDetectionMode === "manual" && this.userTurnCommitted && (this.bounceEOUTask === void 0 || this.bounceEOUTask.done || ev.type == import_stt.SpeechEventType.INTERIM_TRANSCRIPT)) {
@@ -203,38 +237,50 @@ class AudioRecognition {
203
237
  break;
204
238
  case import_stt.SpeechEventType.START_OF_SPEECH:
205
239
  if (this.turnDetectionMode !== "stt") break;
206
- this.hooks.onStartOfSpeech({
207
- type: import_vad.VADEventType.START_OF_SPEECH,
208
- samplesIndex: 0,
209
- timestamp: Date.now(),
210
- speechDuration: 0,
211
- silenceDuration: 0,
212
- frames: [],
213
- probability: 0,
214
- inferenceDuration: 0,
215
- speaking: true,
216
- rawAccumulatedSilence: 0,
217
- rawAccumulatedSpeech: 0
218
- });
240
+ {
241
+ const span = this.ensureUserTurnSpan(Date.now());
242
+ const ctx = this.userTurnContext(span);
243
+ import_api.context.with(ctx, () => {
244
+ this.hooks.onStartOfSpeech({
245
+ type: import_vad.VADEventType.START_OF_SPEECH,
246
+ samplesIndex: 0,
247
+ timestamp: Date.now(),
248
+ speechDuration: 0,
249
+ silenceDuration: 0,
250
+ frames: [],
251
+ probability: 0,
252
+ inferenceDuration: 0,
253
+ speaking: true,
254
+ rawAccumulatedSilence: 0,
255
+ rawAccumulatedSpeech: 0
256
+ });
257
+ });
258
+ }
219
259
  this.speaking = true;
220
260
  this.lastSpeakingTime = Date.now();
221
261
  (_r = this.bounceEOUTask) == null ? void 0 : _r.cancel();
222
262
  break;
223
263
  case import_stt.SpeechEventType.END_OF_SPEECH:
224
264
  if (this.turnDetectionMode !== "stt") break;
225
- this.hooks.onEndOfSpeech({
226
- type: import_vad.VADEventType.END_OF_SPEECH,
227
- samplesIndex: 0,
228
- timestamp: Date.now(),
229
- speechDuration: 0,
230
- silenceDuration: 0,
231
- frames: [],
232
- probability: 0,
233
- inferenceDuration: 0,
234
- speaking: false,
235
- rawAccumulatedSilence: 0,
236
- rawAccumulatedSpeech: 0
237
- });
265
+ {
266
+ const span = this.ensureUserTurnSpan();
267
+ const ctx = this.userTurnContext(span);
268
+ import_api.context.with(ctx, () => {
269
+ this.hooks.onEndOfSpeech({
270
+ type: import_vad.VADEventType.END_OF_SPEECH,
271
+ samplesIndex: 0,
272
+ timestamp: Date.now(),
273
+ speechDuration: 0,
274
+ silenceDuration: 0,
275
+ frames: [],
276
+ probability: 0,
277
+ inferenceDuration: 0,
278
+ speaking: false,
279
+ rawAccumulatedSilence: 0,
280
+ rawAccumulatedSpeech: 0
281
+ });
282
+ });
283
+ }
238
284
  this.speaking = false;
239
285
  this.userTurnCommitted = true;
240
286
  this.lastSpeakingTime = Date.now();
@@ -267,6 +313,8 @@ class AudioRecognition {
267
313
  );
268
314
  const bounceEOUTask = (lastSpeakingTime, lastFinalTranscriptTime, speechStartTime) => async (controller) => {
269
315
  let endpointingDelay = this.minEndpointingDelay;
316
+ const userTurnSpan = this.ensureUserTurnSpan();
317
+ const userTurnCtx = this.userTurnContext(userTurnSpan);
270
318
  if (turnDetector) {
271
319
  await import_telemetry.tracer.startActiveSpan(
272
320
  async (span) => {
@@ -301,7 +349,7 @@ class AudioRecognition {
301
349
  },
302
350
  {
303
351
  name: "eou_detection",
304
- context: this.rootSpanContext
352
+ context: userTurnCtx
305
353
  }
306
354
  );
307
355
  }
@@ -421,16 +469,13 @@ class AudioRecognition {
421
469
  switch (ev.type) {
422
470
  case import_vad.VADEventType.START_OF_SPEECH:
423
471
  this.logger.debug("VAD task: START_OF_SPEECH");
424
- this.hooks.onStartOfSpeech(ev);
425
- this.speaking = true;
426
- if (!this.userTurnSpan) {
472
+ {
427
473
  const startTime = Date.now() - ev.speechDuration;
428
- this.userTurnSpan = import_telemetry.tracer.startSpan({
429
- name: "user_turn",
430
- context: this.rootSpanContext,
431
- startTime
432
- });
474
+ const span = this.ensureUserTurnSpan(startTime);
475
+ const ctx = this.userTurnContext(span);
476
+ import_api.context.with(ctx, () => this.hooks.onStartOfSpeech(ev));
433
477
  }
478
+ this.speaking = true;
434
479
  if (ev.frames.length > 0 && ev.frames[0]) {
435
480
  this.sampleRate = ev.frames[0].sampleRate;
436
481
  }
@@ -447,7 +492,11 @@ class AudioRecognition {
447
492
  break;
448
493
  case import_vad.VADEventType.END_OF_SPEECH:
449
494
  this.logger.debug("VAD task: END_OF_SPEECH");
450
- this.hooks.onEndOfSpeech(ev);
495
+ {
496
+ const span = this.ensureUserTurnSpan();
497
+ const ctx = this.userTurnContext(span);
498
+ import_api.context.with(ctx, () => this.hooks.onEndOfSpeech(ev));
499
+ }
451
500
  this.speaking = false;
452
501
  if (this.vadBaseTurnDetection || this.turnDetectionMode === "stt" && this.userTurnCommitted) {
453
502
  const chatCtx = this.hooks.retrieveChatCtx();
@@ -1 +1 @@
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 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 /** The new transcript text from the user's speech. */\n newTranscript: string;\n /** Confidence score of the transcript (0-1). */\n transcriptConfidence: number;\n /** Delay from speech stop to final transcription in milliseconds. */\n transcriptionDelay: number;\n /** Delay from speech stop to end of utterance detection in milliseconds. */\n endOfUtteranceDelay: number;\n /** Timestamp when user started speaking (milliseconds since epoch). */\n startedSpeakingAt: number | undefined;\n /** Timestamp when user stopped speaking (milliseconds since epoch). */\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\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 /** Hooks for recognition events. */\n recognitionHooks: RecognitionHooks;\n /** Speech-to-text node. */\n stt?: STTNode;\n /** Voice activity detection. */\n vad?: VAD;\n /** Turn detector for end-of-turn prediction. */\n turnDetector?: _TurnDetector;\n /** Turn detection mode. */\n turnDetectionMode?: Exclude<TurnDetectionMode, _TurnDetector>;\n /** Minimum endpointing delay in milliseconds. */\n minEndpointingDelay: number;\n /** Maximum endpointing delay in milliseconds. */\n maxEndpointingDelay: number;\n /** Root span context for tracing. */\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 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\n constructor(opts: AudioRecognitionOptions) {\n this.hooks = opts.recognitionHooks;\n this.stt = opts.stt;\n this.vad = opts.vad;\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 this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n const [vadInputStream, sttInputStream] = this.deferredInputStream.stream.tee();\n this.vadInputStream = vadInputStream;\n this.sttInputStream = mergeReadableStreams(sttInputStream, this.silenceAudioTransform.readable);\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\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 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.hooks.onFinalTranscript(ev);\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 const startTime = Date.now() - ev.speechDuration;\n this.userTurnSpan = tracer.startSpan({\n name: 'user_turn',\n context: this.rootSpanContext,\n startTime,\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 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 (\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 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 }\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;AAC/B,0BAAiC;AACjC,iBAAoB;AACpB,6BAAmE;AACnE,gCAAkC;AAClC,oCAAqC;AACrC,iBAAkD;AAClD,uBAAmC;AACnC,mBAA4B;AAC5B,iBAAsD;AA6D/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,wBAAwB,IAAI,4CAA8B;AAAA,EAC1D;AAAA;AAAA,EAGA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAER,YAAY,MAA+B;AACzC,SAAK,QAAQ,KAAK;AAClB,SAAK,MAAM,KAAK;AAChB,SAAK,MAAM,KAAK;AAChB,SAAK,eAAe,KAAK;AACzB,SAAK,oBAAoB,KAAK;AAC9B,SAAK,sBAAsB,KAAK;AAChC,SAAK,sBAAsB,KAAK;AAChC,SAAK,eAAe;AACpB,SAAK,kBAAkB,KAAK;AAE5B,SAAK,sBAAsB,IAAI,8CAAmC;AAClE,UAAM,CAAC,gBAAgB,cAAc,IAAI,KAAK,oBAAoB,OAAO,IAAI;AAC7E,SAAK,iBAAiB;AACtB,SAAK,qBAAiB,oDAAqB,gBAAgB,KAAK,sBAAsB,QAAQ;AAC9F,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;AAAA,EACH;AAAA,EAEA,MAAc,WAAW,IAAiB;AAzJ5C;AA0JI,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,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,MAAM,kBAAkB,EAAE;AAE/B,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;AA1VhD;AA2VI,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/iBzE;AAgjBI,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,oBAAM,YAAY,KAAK,IAAI,IAAI,GAAG;AAClC,mBAAK,eAAe,wBAAO,UAAU;AAAA,gBACnC,MAAM;AAAA,gBACN,SAAS,KAAK;AAAA,gBACd;AAAA,cACF,CAAC;AAAA,YACH;AAGA,gBAAI,GAAG,OAAO,SAAS,KAAK,GAAG,OAAO,CAAC,GAAG;AACxC,mBAAK,aAAa,GAAG,OAAO,CAAC,EAAE;AAAA,YACjC;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;AAEhB,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,oBAAoB,aAAyC;AAC3D,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,yBAAyB;AACvB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA,EAEA,gBAAgB;AAjoBlB;AAkoBI,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;AAhpBzC;AAipBI,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;AA1rBhB;AA2rBI,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;AAAA,EAC5B;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":[]}
1
+ {"version":3,"sources":["../../src/voice/audio_recognition.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { ParticipantKind } from '@livekit/rtc-node';\nimport { AudioFrame } from '@livekit/rtc-node';\nimport {\n type Context,\n ROOT_CONTEXT,\n type Span,\n context as otelContext,\n trace,\n} from '@opentelemetry/api';\nimport type { WritableStreamDefaultWriter } from 'node:stream/web';\nimport { ReadableStream } from 'node:stream/web';\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';\nimport { setParticipantSpanAttributes } from './utils.js';\n\nexport interface EndOfTurnInfo {\n /** The new transcript text from the user's speech. */\n newTranscript: string;\n /** Confidence score of the transcript (0-1). */\n transcriptConfidence: number;\n /** Delay from speech stop to final transcription in milliseconds. */\n transcriptionDelay: number;\n /** Delay from speech stop to end of utterance detection in milliseconds. */\n endOfUtteranceDelay: number;\n /** Timestamp when user started speaking (milliseconds since epoch). */\n startedSpeakingAt: number | undefined;\n /** Timestamp when user stopped speaking (milliseconds since epoch). */\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\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 /** Hooks for recognition events. */\n recognitionHooks: RecognitionHooks;\n /** Speech-to-text node. */\n stt?: STTNode;\n /** Voice activity detection. */\n vad?: VAD;\n /** Turn detector for end-of-turn prediction. */\n turnDetector?: _TurnDetector;\n /** Turn detection mode. */\n turnDetectionMode?: Exclude<TurnDetectionMode, _TurnDetector>;\n /** Minimum endpointing delay in milliseconds. */\n minEndpointingDelay: number;\n /** Maximum endpointing delay in milliseconds. */\n maxEndpointingDelay: number;\n /** Root span context for tracing. */\n rootSpanContext?: Context;\n /** STT model name for tracing */\n sttModel?: string;\n /** STT provider name for tracing */\n sttProvider?: string;\n /** Getter for linked participant for span attribution */\n getLinkedParticipant?: () => ParticipantLike | undefined;\n}\n\n/**\n * Minimal participant shape for span attribution.\n * Compatible with both `LocalParticipant` and `RemoteParticipant` from `@livekit/rtc-node`.\n */\nexport interface ParticipantLike {\n sid: string | undefined;\n identity: string;\n kind: ParticipantKind;\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 private sttModel?: string;\n private sttProvider?: string;\n private getLinkedParticipant?: () => ParticipantLike | undefined;\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 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\n constructor(opts: AudioRecognitionOptions) {\n this.hooks = opts.recognitionHooks;\n this.stt = opts.stt;\n this.vad = opts.vad;\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 this.sttModel = opts.sttModel;\n this.sttProvider = opts.sttProvider;\n this.getLinkedParticipant = opts.getLinkedParticipant;\n\n this.deferredInputStream = new DeferredReadableStream<AudioFrame>();\n const [vadInputStream, sttInputStream] = this.deferredInputStream.stream.tee();\n this.vadInputStream = vadInputStream;\n this.sttInputStream = mergeReadableStreams(sttInputStream, this.silenceAudioTransform.readable);\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\n private ensureUserTurnSpan(startTime?: number): Span {\n if (this.userTurnSpan && this.userTurnSpan.isRecording()) {\n return this.userTurnSpan;\n }\n\n this.userTurnSpan = tracer.startSpan({\n name: 'user_turn',\n context: this.rootSpanContext,\n startTime,\n });\n\n const participant = this.getLinkedParticipant?.();\n if (participant) {\n setParticipantSpanAttributes(this.userTurnSpan, participant);\n }\n\n if (this.sttModel) {\n this.userTurnSpan.setAttribute(traceTypes.ATTR_GEN_AI_REQUEST_MODEL, this.sttModel);\n }\n if (this.sttProvider) {\n this.userTurnSpan.setAttribute(traceTypes.ATTR_GEN_AI_PROVIDER_NAME, this.sttProvider);\n }\n\n return this.userTurnSpan;\n }\n\n private userTurnContext(span: Span): Context {\n const base = this.rootSpanContext ?? ROOT_CONTEXT;\n return trace.setSpan(base, span);\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 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.hooks.onFinalTranscript(ev);\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 {\n const span = this.ensureUserTurnSpan(Date.now());\n const ctx = this.userTurnContext(span);\n otelContext.with(ctx, () => {\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 });\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 {\n const span = this.ensureUserTurnSpan();\n const ctx = this.userTurnContext(span);\n otelContext.with(ctx, () => {\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 });\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 const userTurnSpan = this.ensureUserTurnSpan();\n const userTurnCtx = this.userTurnContext(userTurnSpan);\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: userTurnCtx,\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 {\n const startTime = Date.now() - ev.speechDuration;\n const span = this.ensureUserTurnSpan(startTime);\n const ctx = this.userTurnContext(span);\n otelContext.with(ctx, () => this.hooks.onStartOfSpeech(ev));\n }\n this.speaking = true;\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 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 {\n const span = this.ensureUserTurnSpan();\n const ctx = this.userTurnContext(span);\n otelContext.with(ctx, () => this.hooks.onEndOfSpeech(ev));\n }\n\n // when VAD fires END_OF_SPEECH, it already waited for the silence_duration\n this.speaking = false;\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 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 }\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;AAIA,sBAA2B;AAC3B,iBAMO;AAEP,iBAA+B;AAC/B,0BAAiC;AACjC,iBAAoB;AACpB,6BAAmE;AACnE,gCAAkC;AAClC,oCAAqC;AACrC,iBAAkD;AAClD,uBAAmC;AACnC,mBAA4B;AAC5B,iBAAsD;AAGtD,IAAAA,gBAA6C;AA2EtC,MAAM,iBAAiB;AAAA,EACpB;AAAA,EACA;AAAA,EACA;AAAA,EACA;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,wBAAwB,IAAI,4CAA8B;AAAA,EAC1D;AAAA;AAAA,EAGA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAER,YAAY,MAA+B;AACzC,SAAK,QAAQ,KAAK;AAClB,SAAK,MAAM,KAAK;AAChB,SAAK,MAAM,KAAK;AAChB,SAAK,eAAe,KAAK;AACzB,SAAK,oBAAoB,KAAK;AAC9B,SAAK,sBAAsB,KAAK;AAChC,SAAK,sBAAsB,KAAK;AAChC,SAAK,eAAe;AACpB,SAAK,kBAAkB,KAAK;AAC5B,SAAK,WAAW,KAAK;AACrB,SAAK,cAAc,KAAK;AACxB,SAAK,uBAAuB,KAAK;AAEjC,SAAK,sBAAsB,IAAI,8CAAmC;AAClE,UAAM,CAAC,gBAAgB,cAAc,IAAI,KAAK,oBAAoB,OAAO,IAAI;AAC7E,SAAK,iBAAiB;AACtB,SAAK,qBAAiB,oDAAqB,gBAAgB,KAAK,sBAAsB,QAAQ;AAC9F,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;AAAA,EACH;AAAA,EAEQ,mBAAmB,WAA0B;AAvLvD;AAwLI,QAAI,KAAK,gBAAgB,KAAK,aAAa,YAAY,GAAG;AACxD,aAAO,KAAK;AAAA,IACd;AAEA,SAAK,eAAe,wBAAO,UAAU;AAAA,MACnC,MAAM;AAAA,MACN,SAAS,KAAK;AAAA,MACd;AAAA,IACF,CAAC;AAED,UAAM,eAAc,UAAK,yBAAL;AACpB,QAAI,aAAa;AACf,sDAA6B,KAAK,cAAc,WAAW;AAAA,IAC7D;AAEA,QAAI,KAAK,UAAU;AACjB,WAAK,aAAa,aAAa,4BAAW,2BAA2B,KAAK,QAAQ;AAAA,IACpF;AACA,QAAI,KAAK,aAAa;AACpB,WAAK,aAAa,aAAa,4BAAW,2BAA2B,KAAK,WAAW;AAAA,IACvF;AAEA,WAAO,KAAK;AAAA,EACd;AAAA,EAEQ,gBAAgB,MAAqB;AAC3C,UAAM,OAAO,KAAK,mBAAmB;AACrC,WAAO,iBAAM,QAAQ,MAAM,IAAI;AAAA,EACjC;AAAA,EAEA,MAAc,WAAW,IAAiB;AAtN5C;AAuNI,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,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,MAAM,kBAAkB,EAAE;AAE/B,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;AACE,gBAAM,OAAO,KAAK,mBAAmB,KAAK,IAAI,CAAC;AAC/C,gBAAM,MAAM,KAAK,gBAAgB,IAAI;AACrC,qBAAAC,QAAY,KAAK,KAAK,MAAM;AAC1B,iBAAK,MAAM,gBAAgB;AAAA,cACzB,MAAM,wBAAa;AAAA,cACnB,cAAc;AAAA,cACd,WAAW,KAAK,IAAI;AAAA,cACpB,gBAAgB;AAAA,cAChB,iBAAiB;AAAA,cACjB,QAAQ,CAAC;AAAA,cACT,aAAa;AAAA,cACb,mBAAmB;AAAA,cACnB,UAAU;AAAA,cACV,uBAAuB;AAAA,cACvB,sBAAsB;AAAA,YACxB,CAAC;AAAA,UACH,CAAC;AAAA,QACH;AACA,aAAK,WAAW;AAChB,aAAK,mBAAmB,KAAK,IAAI;AAEjC,mBAAK,kBAAL,mBAAoB;AACpB;AAAA,MACF,KAAK,2BAAgB;AACnB,YAAI,KAAK,sBAAsB,MAAO;AACtC;AACE,gBAAM,OAAO,KAAK,mBAAmB;AACrC,gBAAM,MAAM,KAAK,gBAAgB,IAAI;AACrC,qBAAAA,QAAY,KAAK,KAAK,MAAM;AAC1B,iBAAK,MAAM,cAAc;AAAA,cACvB,MAAM,wBAAa;AAAA,cACnB,cAAc;AAAA,cACd,WAAW,KAAK,IAAI;AAAA,cACpB,gBAAgB;AAAA,cAChB,iBAAiB;AAAA,cACjB,QAAQ,CAAC;AAAA,cACT,aAAa;AAAA,cACb,mBAAmB;AAAA,cACnB,UAAU;AAAA,cACV,uBAAuB;AAAA,cACvB,sBAAsB;AAAA,YACxB,CAAC;AAAA,UACH,CAAC;AAAA,QACH;AACA,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;AAnahD;AAoaI,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,YAAM,eAAe,KAAK,mBAAmB;AAC7C,YAAM,cAAc,KAAK,gBAAgB,YAAY;AAErD,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;AAAA,UACX;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;AA3nBzE;AA4nBI,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;AACE,oBAAM,YAAY,KAAK,IAAI,IAAI,GAAG;AAClC,oBAAM,OAAO,KAAK,mBAAmB,SAAS;AAC9C,oBAAM,MAAM,KAAK,gBAAgB,IAAI;AACrC,yBAAAA,QAAY,KAAK,KAAK,MAAM,KAAK,MAAM,gBAAgB,EAAE,CAAC;AAAA,YAC5D;AACA,iBAAK,WAAW;AAGhB,gBAAI,GAAG,OAAO,SAAS,KAAK,GAAG,OAAO,CAAC,GAAG;AACxC,mBAAK,aAAa,GAAG,OAAO,CAAC,EAAE;AAAA,YACjC;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;AACE,oBAAM,OAAO,KAAK,mBAAmB;AACrC,oBAAM,MAAM,KAAK,gBAAgB,IAAI;AACrC,yBAAAA,QAAY,KAAK,KAAK,MAAM,KAAK,MAAM,cAAc,EAAE,CAAC;AAAA,YAC1D;AAGA,iBAAK,WAAW;AAEhB,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,oBAAoB,aAAyC;AAC3D,SAAK,oBAAoB,UAAU,WAAW;AAAA,EAChD;AAAA,EAEA,yBAAyB;AACvB,SAAK,oBAAoB,aAAa;AAAA,EACxC;AAAA,EAEA,gBAAgB;AA7sBlB;AA8sBI,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;AA5tBzC;AA6tBI,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;AAtwBhB;AAuwBI,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;AAAA,EAC5B;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":["import_utils","otelContext"]}