@livekit/agents 1.0.16 → 1.0.18

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 (99) hide show
  1. package/dist/inference/llm.cjs +35 -13
  2. package/dist/inference/llm.cjs.map +1 -1
  3. package/dist/inference/llm.d.cts +10 -5
  4. package/dist/inference/llm.d.ts +10 -5
  5. package/dist/inference/llm.d.ts.map +1 -1
  6. package/dist/inference/llm.js +35 -13
  7. package/dist/inference/llm.js.map +1 -1
  8. package/dist/llm/chat_context.d.cts +1 -1
  9. package/dist/llm/chat_context.d.ts +1 -1
  10. package/dist/llm/llm.cjs.map +1 -1
  11. package/dist/llm/llm.d.cts +1 -1
  12. package/dist/llm/llm.d.ts +1 -1
  13. package/dist/llm/llm.d.ts.map +1 -1
  14. package/dist/llm/llm.js.map +1 -1
  15. package/dist/llm/provider_format/google.cjs.map +1 -1
  16. package/dist/llm/provider_format/google.d.cts +1 -1
  17. package/dist/llm/provider_format/google.d.ts +1 -1
  18. package/dist/llm/provider_format/google.d.ts.map +1 -1
  19. package/dist/llm/provider_format/google.js.map +1 -1
  20. package/dist/llm/provider_format/index.d.cts +1 -1
  21. package/dist/llm/provider_format/index.d.ts +1 -1
  22. package/dist/llm/provider_format/index.d.ts.map +1 -1
  23. package/dist/llm/realtime.cjs.map +1 -1
  24. package/dist/llm/realtime.d.cts +4 -0
  25. package/dist/llm/realtime.d.ts +4 -0
  26. package/dist/llm/realtime.d.ts.map +1 -1
  27. package/dist/llm/realtime.js.map +1 -1
  28. package/dist/llm/utils.cjs +2 -2
  29. package/dist/llm/utils.cjs.map +1 -1
  30. package/dist/llm/utils.d.cts +1 -1
  31. package/dist/llm/utils.d.ts +1 -1
  32. package/dist/llm/utils.d.ts.map +1 -1
  33. package/dist/llm/utils.js +2 -2
  34. package/dist/llm/utils.js.map +1 -1
  35. package/dist/llm/zod-utils.cjs +6 -3
  36. package/dist/llm/zod-utils.cjs.map +1 -1
  37. package/dist/llm/zod-utils.d.cts +1 -1
  38. package/dist/llm/zod-utils.d.ts +1 -1
  39. package/dist/llm/zod-utils.d.ts.map +1 -1
  40. package/dist/llm/zod-utils.js +6 -3
  41. package/dist/llm/zod-utils.js.map +1 -1
  42. package/dist/llm/zod-utils.test.cjs +83 -0
  43. package/dist/llm/zod-utils.test.cjs.map +1 -1
  44. package/dist/llm/zod-utils.test.js +83 -0
  45. package/dist/llm/zod-utils.test.js.map +1 -1
  46. package/dist/stt/stt.cjs +0 -1
  47. package/dist/stt/stt.cjs.map +1 -1
  48. package/dist/stt/stt.d.ts.map +1 -1
  49. package/dist/stt/stt.js +0 -1
  50. package/dist/stt/stt.js.map +1 -1
  51. package/dist/tts/tts.cjs +2 -4
  52. package/dist/tts/tts.cjs.map +1 -1
  53. package/dist/tts/tts.d.ts.map +1 -1
  54. package/dist/tts/tts.js +3 -5
  55. package/dist/tts/tts.js.map +1 -1
  56. package/dist/utils.cjs.map +1 -1
  57. package/dist/utils.d.cts +7 -0
  58. package/dist/utils.d.ts +7 -0
  59. package/dist/utils.d.ts.map +1 -1
  60. package/dist/utils.js.map +1 -1
  61. package/dist/voice/agent_activity.cjs +69 -20
  62. package/dist/voice/agent_activity.cjs.map +1 -1
  63. package/dist/voice/agent_activity.d.ts.map +1 -1
  64. package/dist/voice/agent_activity.js +69 -20
  65. package/dist/voice/agent_activity.js.map +1 -1
  66. package/dist/voice/agent_session.cjs +40 -1
  67. package/dist/voice/agent_session.cjs.map +1 -1
  68. package/dist/voice/agent_session.d.cts +5 -0
  69. package/dist/voice/agent_session.d.ts +5 -0
  70. package/dist/voice/agent_session.d.ts.map +1 -1
  71. package/dist/voice/agent_session.js +40 -1
  72. package/dist/voice/agent_session.js.map +1 -1
  73. package/dist/voice/interruption_detection.test.cjs +114 -0
  74. package/dist/voice/interruption_detection.test.cjs.map +1 -0
  75. package/dist/voice/interruption_detection.test.js +113 -0
  76. package/dist/voice/interruption_detection.test.js.map +1 -0
  77. package/dist/voice/room_io/room_io.cjs +3 -0
  78. package/dist/voice/room_io/room_io.cjs.map +1 -1
  79. package/dist/voice/room_io/room_io.d.cts +1 -0
  80. package/dist/voice/room_io/room_io.d.ts +1 -0
  81. package/dist/voice/room_io/room_io.d.ts.map +1 -1
  82. package/dist/voice/room_io/room_io.js +3 -0
  83. package/dist/voice/room_io/room_io.js.map +1 -1
  84. package/package.json +3 -3
  85. package/src/inference/llm.ts +53 -21
  86. package/src/llm/__snapshots__/zod-utils.test.ts.snap +218 -0
  87. package/src/llm/llm.ts +1 -1
  88. package/src/llm/provider_format/google.ts +4 -4
  89. package/src/llm/realtime.ts +8 -1
  90. package/src/llm/utils.ts +7 -2
  91. package/src/llm/zod-utils.test.ts +101 -0
  92. package/src/llm/zod-utils.ts +12 -3
  93. package/src/stt/stt.ts +2 -1
  94. package/src/tts/tts.ts +7 -5
  95. package/src/utils.ts +17 -0
  96. package/src/voice/agent_activity.ts +96 -24
  97. package/src/voice/agent_session.ts +54 -0
  98. package/src/voice/interruption_detection.test.ts +151 -0
  99. package/src/voice/room_io/room_io.ts +4 -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 { 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 { getJobContext } from '../job.js';\nimport { 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 type { TTS, TTSError } from '../tts/tts.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 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 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 { RoomIO, type RoomInputOptions, type RoomOutputOptions } from './room_io/index.js';\nimport type { UnknownUserData } from './run_context.js';\nimport type { SpeechHandle } from './speech_handle.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}\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} 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};\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\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 } = opts;\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\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 set userData(value: UserData) {\n this._userData = value;\n }\n\n async start({\n agent,\n room,\n inputOptions,\n outputOptions,\n }: {\n agent: Agent;\n room: Room;\n inputOptions?: Partial<RoomInputOptions>;\n outputOptions?: Partial<RoomOutputOptions>;\n }): Promise<void> {\n if (this.started) {\n return;\n }\n\n this.agent = agent;\n this._updateAgentState('initializing');\n\n const tasks: Promise<void>[] = [];\n // Check for existing input/output configuration and warn if needed\n if (this.input.audio && inputOptions?.audioEnabled !== false) {\n this.logger.warn('RoomIO audio input is enabled but input.audio is already set, ignoring..');\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 const ctx = getJobContext();\n if (ctx && ctx.room === room && !room.isConnected) {\n this.logger.debug('Auto-connecting to room via job context');\n tasks.push(ctx.connect());\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._updateAgentState('listening');\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 return this.activity.say(text, options);\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 if (this.activity.draining) {\n if (!this.nextActivity) {\n throw new Error('AgentSession is closing, cannot use generateReply()');\n }\n return this.nextActivity.generateReply({ userMessage, ...options });\n }\n\n return this.activity.generateReply({ userMessage, ...options });\n }\n\n private async updateActivity(agent: Agent): Promise<void> {\n // TODO(AJS-129): add lock to agent activity core lifecycle\n this.nextActivity = new AgentActivity(agent, this);\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 await this.activity.start();\n\n if (this._input.audio) {\n this.activity.attachAudioInput(this._input.audio.stream);\n }\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 /** @internal */\n _closeSoon({\n reason,\n drain = false,\n error = null,\n }: {\n reason: CloseReason;\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 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 _updateAgentState(state: AgentState) {\n if (this._agentState === state) {\n return;\n }\n\n const oldState = this._agentState;\n this._agentState = state;\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 const oldState = this.userState;\n this.userState = state;\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 async closeImpl(\n reason: CloseReason,\n error: RealtimeModelError | LLMError | TTSError | STTError | null = null,\n drain: boolean = false,\n ): Promise<void> {\n if (!this.started) {\n return;\n }\n\n if (this.activity) {\n if (!drain) {\n try {\n this.activity.interrupt();\n } catch (error) {\n // uninterruptible speech [copied from python]\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 this.activity.detachAudioInput();\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 this.started = false;\n\n this.emit(AgentSessionEventTypes.Close, createCloseEvent(reason, error));\n\n this.userState = 'listening';\n this._agentState = 'initializing';\n\n this.logger.info({ reason, error }, 'AgentSession closed');\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAKA,yBAA6B;AAE7B,uBAOO;AACP,iBAA8B;AAC9B,0BAAyC;AAGzC,iBAAoB;AAMpB,4BAA8B;AAE9B,oBAkBO;AACP,gBAAwC;AACxC,qBAAsE;AAetE,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;AACxB;AA0BO,MAAM,qBAEF,gCAA+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,aAAS,gBAAI;AAAA,EAEb;AAAA,EACA;AAAA,EACA,cAA0B;AAAA,EAE1B;AAAA,EACA;AAAA,EAEA,cAAoC;AAAA,EAE5C,YAAY,MAAqC;AAC/C,UAAM;AAEN,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,eAAe;AAAA,IACjB,IAAI;AAEJ,SAAK,MAAM;AAEX,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,iBAAAA,IAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,iBAAAC,IAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,iBAAAC,IAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,SAAK,gBAAgB;AACrB,SAAK,YAAY;AAGjB,SAAK,SAAS,IAAI,qBAAW,KAAK,mBAAmB;AACrD,SAAK,UAAU,IAAI,sBAAY,KAAK,sBAAsB,KAAK,mBAAmB;AAGlF,SAAK,WAAW,gCAAY,MAAM;AAClC,SAAK,UAAU,EAAE,GAAG,qBAAqB,GAAG,aAAa;AAAA,EAC3D;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,EAEA,IAAI,SAAS,OAAiB;AAC5B,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,MAAM,MAAM;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAKkB;AAChB,QAAI,KAAK,SAAS;AAChB;AAAA,IACF;AAEA,SAAK,QAAQ;AACb,SAAK,kBAAkB,cAAc;AAErC,UAAM,QAAyB,CAAC;AAEhC,QAAI,KAAK,MAAM,UAAS,6CAAc,kBAAiB,OAAO;AAC5D,WAAK,OAAO,KAAK,0EAA0E;AAAA,IAC7F;AAEA,QAAI,KAAK,OAAO,UAAS,+CAAe,kBAAiB,OAAO;AAC9D,WAAK,OAAO;AAAA,QACV;AAAA,MACF;AAAA,IACF;AAEA,QAAI,KAAK,OAAO,kBAAiB,+CAAe,0BAAyB,OAAO;AAC9E,WAAK,OAAO;AAAA,QACV;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS,IAAI,sBAAO;AAAA,MACvB,cAAc;AAAA,MACd;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC;AACD,SAAK,OAAO,MAAM;AAElB,UAAM,UAAM,0BAAc;AAC1B,QAAI,OAAO,IAAI,SAAS,QAAQ,CAAC,KAAK,aAAa;AACjD,WAAK,OAAO,MAAM,yCAAyC;AAC3D,YAAM,KAAK,IAAI,QAAQ,CAAC;AAAA,IAC1B;AAEA,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,kBAAkB,WAAW;AAAA,EACpC;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,WAAO,KAAK,SAAS,IAAI,MAAM,OAAO;AAAA,EACxC;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,gCAAY;AAAA,MACd,MAAM;AAAA,MACN,SAAS,QAAQ;AAAA,IACnB,CAAC,IACD;AAEJ,QAAI,KAAK,SAAS,UAAU;AAC1B,UAAI,CAAC,KAAK,cAAc;AACtB,cAAM,IAAI,MAAM,qDAAqD;AAAA,MACvE;AACA,aAAO,KAAK,aAAa,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,IACpE;AAEA,WAAO,KAAK,SAAS,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,eAAe,OAA6B;AAExD,SAAK,eAAe,IAAI,oCAAc,OAAO,IAAI;AAEjD,QAAI,KAAK,UAAU;AACjB,YAAM,KAAK,SAAS,MAAM;AAC1B,YAAM,KAAK,SAAS,MAAM;AAAA,IAC5B;AAEA,SAAK,WAAW,KAAK;AACrB,SAAK,eAAe;AAEpB,UAAM,KAAK,SAAS,MAAM;AAE1B,QAAI,KAAK,OAAO,OAAO;AACrB,WAAK,SAAS,iBAAiB,KAAK,OAAO,MAAM,MAAM;AAAA,IACzD;AAAA,EACF;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,0BAAY,cAAc;AAAA,EACjD;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;AAEA,SAAK,OAAO,MAAM,OAAO,oDAAoD;AAE7E,SAAK,eAAe,YAAY;AAC9B,YAAM,KAAK,UAAU,0BAAY,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,qCAAuB,2BAAuB,gDAAiC,IAAI,CAAC;AAAA,EAChG;AAAA;AAAA,EAGA,kBAAkB,OAAmB;AACnC,QAAI,KAAK,gBAAgB,OAAO;AAC9B;AAAA,IACF;AAEA,UAAM,WAAW,KAAK;AACtB,SAAK,cAAc;AACnB,SAAK;AAAA,MACH,qCAAuB;AAAA,UACvB,4CAA6B,UAAU,KAAK;AAAA,IAC9C;AAAA,EACF;AAAA;AAAA,EAGA,iBAAiB,OAAkB,mBAA4B;AAC7D,QAAI,KAAK,cAAc,OAAO;AAC5B;AAAA,IACF;AAEA,UAAM,WAAW,KAAK;AACtB,SAAK,YAAY;AACjB,SAAK;AAAA,MACH,qCAAuB;AAAA,UACvB,2CAA4B,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,EAErC,MAAc,UACZ,QACA,QAAoE,MACpE,QAAiB,OACF;AAzcnB;AA0cI,QAAI,CAAC,KAAK,SAAS;AACjB;AAAA,IACF;AAEA,QAAI,KAAK,UAAU;AACjB,UAAI,CAAC,OAAO;AACV,YAAI;AACF,eAAK,SAAS,UAAU;AAAA,QAC1B,SAASC,QAAO;AAAA,QAIhB;AAAA,MACF;AACA,YAAM,KAAK,SAAS,MAAM;AAE1B,cAAM,UAAK,SAAS,kBAAd,mBAA6B;AACnC,WAAK,SAAS,iBAAiB;AAAA,IACjC;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,SAAK,UAAU;AAEf,SAAK,KAAK,qCAAuB,WAAO,gCAAiB,QAAQ,KAAK,CAAC;AAEvE,SAAK,YAAY;AACjB,SAAK,cAAc;AAEnB,SAAK,OAAO,KAAK,EAAE,QAAQ,MAAM,GAAG,qBAAqB;AAAA,EAC3D;AACF;","names":["InferenceSTT","InferenceLLM","InferenceTTS","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 { 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 { getJobContext } from '../job.js';\nimport { 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 type { TTS, TTSError } from '../tts/tts.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 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 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 { RoomIO, type RoomInputOptions, type RoomOutputOptions } from './room_io/index.js';\nimport type { UnknownUserData } from './run_context.js';\nimport type { SpeechHandle } from './speech_handle.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}\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} 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};\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 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 } = opts;\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.on(AgentSessionEventTypes.UserInputTranscribed, this._onUserInputTranscribed.bind(this));\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 set userData(value: UserData) {\n this._userData = value;\n }\n\n async start({\n agent,\n room,\n inputOptions,\n outputOptions,\n }: {\n agent: Agent;\n room: Room;\n inputOptions?: Partial<RoomInputOptions>;\n outputOptions?: Partial<RoomOutputOptions>;\n }): Promise<void> {\n if (this.started) {\n return;\n }\n\n this.agent = agent;\n this._updateAgentState('initializing');\n\n const tasks: Promise<void>[] = [];\n // Check for existing input/output configuration and warn if needed\n if (this.input.audio && inputOptions?.audioEnabled !== false) {\n this.logger.warn('RoomIO audio input is enabled but input.audio is already set, ignoring..');\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 const ctx = getJobContext();\n if (ctx && ctx.room === room && !room.isConnected) {\n this.logger.debug('Auto-connecting to room via job context');\n tasks.push(ctx.connect());\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._updateAgentState('listening');\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 return this.activity.say(text, options);\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 if (this.activity.draining) {\n if (!this.nextActivity) {\n throw new Error('AgentSession is closing, cannot use generateReply()');\n }\n return this.nextActivity.generateReply({ userMessage, ...options });\n }\n\n return this.activity.generateReply({ userMessage, ...options });\n }\n\n private async updateActivity(agent: Agent): Promise<void> {\n // TODO(AJS-129): add lock to agent activity core lifecycle\n this.nextActivity = new AgentActivity(agent, this);\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 await this.activity.start();\n\n if (this._input.audio) {\n this.activity.attachAudioInput(this._input.audio.stream);\n }\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 /** @internal */\n _closeSoon({\n reason,\n drain = false,\n error = null,\n }: {\n reason: CloseReason;\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 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 _updateAgentState(state: AgentState) {\n if (this._agentState === state) {\n return;\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 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: CloseReason,\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\n if (this.activity) {\n if (!drain) {\n try {\n this.activity.interrupt();\n } catch (error) {\n // uninterruptible speech [copied from python]\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 this.activity.detachAudioInput();\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 this.started = false;\n\n this.emit(AgentSessionEventTypes.Close, createCloseEvent(reason, error));\n\n this.userState = 'listening';\n this._agentState = 'initializing';\n\n this.logger.info({ reason, error }, 'AgentSession closed');\n }\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAKA,yBAA6B;AAE7B,uBAOO;AACP,iBAA8B;AAC9B,0BAAyC;AAGzC,iBAAoB;AAMpB,4BAA8B;AAE9B,oBAkBO;AACP,gBAAwC;AACxC,qBAAsE;AAgBtE,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;AACnB;AA0BO,MAAM,qBAEF,gCAA+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,aAAS,gBAAI;AAAA,EAEb;AAAA,EACA;AAAA,EACA,cAA0B;AAAA,EAE1B;AAAA,EACA;AAAA,EAEA,cAAoC;AAAA,EACpC,gBAAuC;AAAA,EAE/C,YAAY,MAAqC;AAC/C,UAAM;AAEN,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,eAAe;AAAA,IACjB,IAAI;AAEJ,SAAK,MAAM;AAEX,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,iBAAAA,IAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,iBAAAC,IAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,MAAM,iBAAAC,IAAa,gBAAgB,GAAG;AAAA,IAC7C,OAAO;AACL,WAAK,MAAM;AAAA,IACb;AAEA,SAAK,gBAAgB;AACrB,SAAK,YAAY;AAGjB,SAAK,SAAS,IAAI,qBAAW,KAAK,mBAAmB;AACrD,SAAK,UAAU,IAAI,sBAAY,KAAK,sBAAsB,KAAK,mBAAmB;AAGlF,SAAK,WAAW,gCAAY,MAAM;AAClC,SAAK,UAAU,EAAE,GAAG,qBAAqB,GAAG,aAAa;AAEzD,SAAK,GAAG,qCAAuB,sBAAsB,KAAK,wBAAwB,KAAK,IAAI,CAAC;AAAA,EAC9F;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,EAEA,IAAI,SAAS,OAAiB;AAC5B,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,MAAM,MAAM;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAKkB;AAChB,QAAI,KAAK,SAAS;AAChB;AAAA,IACF;AAEA,SAAK,QAAQ;AACb,SAAK,kBAAkB,cAAc;AAErC,UAAM,QAAyB,CAAC;AAEhC,QAAI,KAAK,MAAM,UAAS,6CAAc,kBAAiB,OAAO;AAC5D,WAAK,OAAO,KAAK,0EAA0E;AAAA,IAC7F;AAEA,QAAI,KAAK,OAAO,UAAS,+CAAe,kBAAiB,OAAO;AAC9D,WAAK,OAAO;AAAA,QACV;AAAA,MACF;AAAA,IACF;AAEA,QAAI,KAAK,OAAO,kBAAiB,+CAAe,0BAAyB,OAAO;AAC9E,WAAK,OAAO;AAAA,QACV;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS,IAAI,sBAAO;AAAA,MACvB,cAAc;AAAA,MACd;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC;AACD,SAAK,OAAO,MAAM;AAElB,UAAM,UAAM,0BAAc;AAC1B,QAAI,OAAO,IAAI,SAAS,QAAQ,CAAC,KAAK,aAAa;AACjD,WAAK,OAAO,MAAM,yCAAyC;AAC3D,YAAM,KAAK,IAAI,QAAQ,CAAC;AAAA,IAC1B;AAEA,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,kBAAkB,WAAW;AAAA,EACpC;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,WAAO,KAAK,SAAS,IAAI,MAAM,OAAO;AAAA,EACxC;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,gCAAY;AAAA,MACd,MAAM;AAAA,MACN,SAAS,QAAQ;AAAA,IACnB,CAAC,IACD;AAEJ,QAAI,KAAK,SAAS,UAAU;AAC1B,UAAI,CAAC,KAAK,cAAc;AACtB,cAAM,IAAI,MAAM,qDAAqD;AAAA,MACvE;AACA,aAAO,KAAK,aAAa,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,IACpE;AAEA,WAAO,KAAK,SAAS,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,eAAe,OAA6B;AAExD,SAAK,eAAe,IAAI,oCAAc,OAAO,IAAI;AAEjD,QAAI,KAAK,UAAU;AACjB,YAAM,KAAK,SAAS,MAAM;AAC1B,YAAM,KAAK,SAAS,MAAM;AAAA,IAC5B;AAEA,SAAK,WAAW,KAAK;AACrB,SAAK,eAAe;AAEpB,UAAM,KAAK,SAAS,MAAM;AAE1B,QAAI,KAAK,OAAO,OAAO;AACrB,WAAK,SAAS,iBAAiB,KAAK,OAAO,MAAM,MAAM;AAAA,IACzD;AAAA,EACF;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,0BAAY,cAAc;AAAA,EACjD;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;AAEA,SAAK,OAAO,MAAM,OAAO,oDAAoD;AAE7E,SAAK,eAAe,YAAY;AAC9B,YAAM,KAAK,UAAU,0BAAY,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,qCAAuB,2BAAuB,gDAAiC,IAAI,CAAC;AAAA,EAChG;AAAA;AAAA,EAGA,kBAAkB,OAAmB;AACnC,QAAI,KAAK,gBAAgB,OAAO;AAC9B;AAAA,IACF;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,qCAAuB;AAAA,UACvB,4CAA6B,UAAU,KAAK;AAAA,IAC9C;AAAA,EACF;AAAA;AAAA,EAGA,iBAAiB,OAAkB,mBAA4B;AAC7D,QAAI,KAAK,cAAc,OAAO;AAC5B;AAAA,IACF;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,qCAAuB;AAAA,UACvB,2CAA4B,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;AA7fnB;AA8fI,QAAI,CAAC,KAAK,SAAS;AACjB;AAAA,IACF;AAEA,SAAK,qBAAqB;AAE1B,QAAI,KAAK,UAAU;AACjB,UAAI,CAAC,OAAO;AACV,YAAI;AACF,eAAK,SAAS,UAAU;AAAA,QAC1B,SAASC,QAAO;AAAA,QAIhB;AAAA,MACF;AACA,YAAM,KAAK,SAAS,MAAM;AAE1B,cAAM,UAAK,SAAS,kBAAd,mBAA6B;AACnC,WAAK,SAAS,iBAAiB;AAAA,IACjC;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,SAAK,UAAU;AAEf,SAAK,KAAK,qCAAuB,WAAO,gCAAiB,QAAQ,KAAK,CAAC;AAEvE,SAAK,YAAY;AACjB,SAAK,cAAc;AAEnB,SAAK,OAAO,KAAK,EAAE,QAAQ,MAAM,GAAG,qBAAqB;AAAA,EAC3D;AACF;","names":["InferenceSTT","InferenceLLM","InferenceTTS","error"]}
@@ -26,6 +26,7 @@ export interface VoiceOptions {
26
26
  maxEndpointingDelay: number;
27
27
  maxToolSteps: number;
28
28
  preemptiveGeneration: boolean;
29
+ userAwayTimeout?: number | null;
29
30
  }
30
31
  export type TurnDetectionMode = 'stt' | 'vad' | 'realtime_llm' | 'manual' | _TurnDetector;
31
32
  export type AgentSessionCallbacks = {
@@ -69,6 +70,7 @@ export declare class AgentSession<UserData = UnknownUserData> extends AgentSessi
69
70
  private _input;
70
71
  private _output;
71
72
  private closingTask;
73
+ private userAwayTimer;
72
74
  constructor(opts: AgentSessionOptions<UserData>);
73
75
  get input(): AgentInput;
74
76
  get output(): AgentOutput;
@@ -118,6 +120,9 @@ export declare class AgentSession<UserData = UnknownUserData> extends AgentSessi
118
120
  private onAudioInputChanged;
119
121
  private onAudioOutputChanged;
120
122
  private onTextOutputChanged;
123
+ private _setUserAwayTimer;
124
+ private _cancelUserAwayTimer;
125
+ private _onUserInputTranscribed;
121
126
  private closeImpl;
122
127
  }
123
128
  export {};
@@ -26,6 +26,7 @@ export interface VoiceOptions {
26
26
  maxEndpointingDelay: number;
27
27
  maxToolSteps: number;
28
28
  preemptiveGeneration: boolean;
29
+ userAwayTimeout?: number | null;
29
30
  }
30
31
  export type TurnDetectionMode = 'stt' | 'vad' | 'realtime_llm' | 'manual' | _TurnDetector;
31
32
  export type AgentSessionCallbacks = {
@@ -69,6 +70,7 @@ export declare class AgentSession<UserData = UnknownUserData> extends AgentSessi
69
70
  private _input;
70
71
  private _output;
71
72
  private closingTask;
73
+ private userAwayTimer;
72
74
  constructor(opts: AgentSessionOptions<UserData>);
73
75
  get input(): AgentInput;
74
76
  get output(): AgentOutput;
@@ -118,6 +120,9 @@ export declare class AgentSession<UserData = UnknownUserData> extends AgentSessi
118
120
  private onAudioInputChanged;
119
121
  private onAudioOutputChanged;
120
122
  private onTextOutputChanged;
123
+ private _setUserAwayTimer;
124
+ private _cancelUserAwayTimer;
125
+ private _onUserInputTranscribed;
121
126
  private closeImpl;
122
127
  }
123
128
  export {};
@@ -1 +1 @@
1
- {"version":3,"file":"agent_session.d.ts","sourceRoot":"","sources":["../../src/voice/agent_session.ts"],"names":[],"mappings":";AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,IAAI,EAAE,MAAM,mBAAmB,CAAC;AAC1D,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAEhF,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AACtD,OAAO,EAIL,KAAK,SAAS,EACd,KAAK,cAAc,EACnB,KAAK,cAAc,EACpB,MAAM,uBAAuB,CAAC;AAE/B,OAAO,EAAE,WAAW,EAAE,WAAW,EAAE,MAAM,wBAAwB,CAAC;AAClE,OAAO,KAAK,EAAE,GAAG,EAAE,aAAa,EAAE,kBAAkB,EAAE,UAAU,EAAE,MAAM,iBAAiB,CAAC;AAC1F,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAE9C,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAC3C,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAC9C,OAAO,KAAK,EAAE,GAAG,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AACnD,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,WAAW,CAAC;AACrC,OAAO,KAAK,EAAE,KAAK,EAAE,MAAM,YAAY,CAAC;AAExC,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,wBAAwB,CAAC;AAC5D,OAAO,EACL,sBAAsB,EACtB,KAAK,UAAU,EACf,KAAK,sBAAsB,EAC3B,KAAK,UAAU,EACf,WAAW,EACX,KAAK,0BAA0B,EAC/B,KAAK,UAAU,EACf,KAAK,0BAA0B,EAC/B,KAAK,qBAAqB,EAC1B,KAAK,kBAAkB,EACvB,KAAK,yBAAyB,EAC9B,KAAK,SAAS,EACd,KAAK,qBAAqB,EAK3B,MAAM,aAAa,CAAC;AACrB,OAAO,EAAE,UAAU,EAAE,WAAW,EAAE,MAAM,SAAS,CAAC;AAClD,OAAO,EAAU,KAAK,gBAAgB,EAAE,KAAK,iBAAiB,EAAE,MAAM,oBAAoB,CAAC;AAC3F,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,kBAAkB,CAAC;AACxD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAEvD,MAAM,WAAW,YAAY;IAC3B,kBAAkB,EAAE,OAAO,CAAC;IAC5B,6BAA6B,EAAE,OAAO,CAAC;IACvC,uBAAuB,EAAE,MAAM,CAAC;IAChC,oBAAoB,EAAE,MAAM,CAAC;IAC7B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,YAAY,EAAE,MAAM,CAAC;IACrB,oBAAoB,EAAE,OAAO,CAAC;CAC/B;AAaD,MAAM,MAAM,iBAAiB,GAAG,KAAK,GAAG,KAAK,GAAG,cAAc,GAAG,QAAQ,GAAG,aAAa,CAAC;AAE1F,MAAM,MAAM,qBAAqB,GAAG;IAClC,CAAC,sBAAsB,CAAC,oBAAoB,CAAC,EAAE,CAAC,EAAE,EAAE,yBAAyB,KAAK,IAAI,CAAC;IACvF,CAAC,sBAAsB,CAAC,iBAAiB,CAAC,EAAE,CAAC,EAAE,EAAE,sBAAsB,KAAK,IAAI,CAAC;IACjF,CAAC,sBAAsB,CAAC,gBAAgB,CAAC,EAAE,CAAC,EAAE,EAAE,qBAAqB,KAAK,IAAI,CAAC;IAC/E,CAAC,sBAAsB,CAAC,qBAAqB,CAAC,EAAE,CAAC,EAAE,EAAE,0BAA0B,KAAK,IAAI,CAAC;IACzF,CAAC,sBAAsB,CAAC,qBAAqB,CAAC,EAAE,CAAC,EAAE,EAAE,0BAA0B,KAAK,IAAI,CAAC;IACzF,CAAC,sBAAsB,CAAC,gBAAgB,CAAC,EAAE,CAAC,EAAE,EAAE,qBAAqB,KAAK,IAAI,CAAC;IAC/E,CAAC,sBAAsB,CAAC,aAAa,CAAC,EAAE,CAAC,EAAE,EAAE,kBAAkB,KAAK,IAAI,CAAC;IACzE,CAAC,sBAAsB,CAAC,KAAK,CAAC,EAAE,CAAC,EAAE,EAAE,UAAU,KAAK,IAAI,CAAC;IACzD,CAAC,sBAAsB,CAAC,KAAK,CAAC,EAAE,CAAC,EAAE,EAAE,UAAU,KAAK,IAAI,CAAC;CAC1D,CAAC;AAEF,MAAM,MAAM,mBAAmB,CAAC,QAAQ,GAAG,eAAe,IAAI;IAC5D,aAAa,CAAC,EAAE,iBAAiB,CAAC;IAClC,GAAG,CAAC,EAAE,GAAG,GAAG,cAAc,CAAC;IAC3B,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,GAAG,aAAa,GAAG,SAAS,CAAC;IACtC,GAAG,CAAC,EAAE,GAAG,GAAG,cAAc,CAAC;IAC3B,QAAQ,CAAC,EAAE,QAAQ,CAAC;IACpB,YAAY,CAAC,EAAE,OAAO,CAAC,YAAY,CAAC,CAAC;CACtC,CAAC;2CAImC,aAAa,qBAAqB,CAAC;AAFxE,qBAAa,YAAY,CACvB,QAAQ,GAAG,eAAe,CAC1B,SAAQ,iBAA+D;IACvE,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,GAAG,aAAa,CAAC;IAC1B,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,aAAa,CAAC,EAAE,iBAAiB,CAAC;IAElC,QAAQ,CAAC,OAAO,EAAE,YAAY,CAAC;IAE/B,OAAO,CAAC,KAAK,CAAC,CAAQ;IACtB,OAAO,CAAC,QAAQ,CAAC,CAAgB;IACjC,OAAO,CAAC,YAAY,CAAC,CAAgB;IACrC,OAAO,CAAC,OAAO,CAAS;IACxB,OAAO,CAAC,SAAS,CAA0B;IAE3C,OAAO,CAAC,MAAM,CAAC,CAAS;IACxB,OAAO,CAAC,MAAM,CAAS;IAEvB,OAAO,CAAC,QAAQ,CAAc;IAC9B,OAAO,CAAC,SAAS,CAAuB;IACxC,OAAO,CAAC,WAAW,CAA8B;IAEjD,OAAO,CAAC,MAAM,CAAa;IAC3B,OAAO,CAAC,OAAO,CAAc;IAE7B,OAAO,CAAC,WAAW,CAA8B;gBAErC,IAAI,EAAE,mBAAmB,CAAC,QAAQ,CAAC;IA6C/C,IAAI,KAAK,IAAI,UAAU,CAEtB;IAED,IAAI,MAAM,IAAI,WAAW,CAExB;IAED,IAAI,QAAQ,IAAI,QAAQ,CAMvB;IAED,IAAI,OAAO,IAAI,WAAW,CAEzB;IAED,IAAI,QAAQ,CAAC,KAAK,EAAE,QAAQ,EAE3B;IAEK,KAAK,CAAC,EACV,KAAK,EACL,IAAI,EACJ,YAAY,EACZ,aAAa,GACd,EAAE;QACD,KAAK,EAAE,KAAK,CAAC;QACb,IAAI,EAAE,IAAI,CAAC;QACX,YAAY,CAAC,EAAE,OAAO,CAAC,gBAAgB,CAAC,CAAC;QACzC,aAAa,CAAC,EAAE,OAAO,CAAC,iBAAiB,CAAC,CAAC;KAC5C,GAAG,OAAO,CAAC,IAAI,CAAC;IAyDjB,WAAW,CAAC,KAAK,EAAE,KAAK,GAAG,IAAI;IAQ/B,cAAc;IAQd,aAAa;IAOb,GAAG,CACD,IAAI,EAAE,MAAM,GAAG,cAAc,CAAC,MAAM,CAAC,EACrC,OAAO,CAAC,EAAE;QACR,KAAK,CAAC,EAAE,cAAc,CAAC,UAAU,CAAC,CAAC;QACnC,kBAAkB,CAAC,EAAE,OAAO,CAAC;QAC7B,YAAY,CAAC,EAAE,OAAO,CAAC;KACxB,GACA,YAAY;IAQf,SAAS;IAOT,aAAa,CAAC,OAAO,CAAC,EAAE;QACtB,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,YAAY,CAAC,EAAE,MAAM,CAAC;QACtB,UAAU,CAAC,EAAE,UAAU,CAAC;QACxB,kBAAkB,CAAC,EAAE,OAAO,CAAC;KAC9B,GAAG,YAAY;YAsBF,cAAc;IAmB5B,IAAI,OAAO,IAAI,WAAW,CAEzB;IAED,IAAI,UAAU,IAAI,UAAU,CAE3B;IAED,IAAI,YAAY,IAAI,KAAK,CAMxB;IAEK,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IAI5B,gBAAgB;IAChB,UAAU,CAAC,EACT,MAAM,EACN,KAAa,EACb,KAAY,GACb,EAAE;QACD,MAAM,EAAE,WAAW,CAAC;QACpB,KAAK,CAAC,EAAE,OAAO,CAAC;QAChB,KAAK,CAAC,EAAE,kBAAkB,GAAG,QAAQ,GAAG,QAAQ,GAAG,QAAQ,GAAG,IAAI,CAAC;KACpE,GAAG,IAAI;IAOR,gBAAgB;IAChB,QAAQ,CAAC,KAAK,EAAE,kBAAkB,GAAG,QAAQ,GAAG,QAAQ,GAAG,QAAQ,GAAG,IAAI;IAc1E,gBAAgB;IAChB,sBAAsB,CAAC,IAAI,EAAE,WAAW,GAAG,IAAI;IAK/C,gBAAgB;IAChB,iBAAiB,CAAC,KAAK,EAAE,UAAU;IAanC,gBAAgB;IAChB,gBAAgB,CAAC,KAAK,EAAE,SAAS,EAAE,iBAAiB,CAAC,EAAE,MAAM;IAc7D,OAAO,CAAC,mBAAmB;IAU3B,OAAO,CAAC,oBAAoB;IAE5B,OAAO,CAAC,mBAAmB;YAEb,SAAS;CA6CxB"}
1
+ {"version":3,"file":"agent_session.d.ts","sourceRoot":"","sources":["../../src/voice/agent_session.ts"],"names":[],"mappings":";AAGA,OAAO,KAAK,EAAE,UAAU,EAAE,IAAI,EAAE,MAAM,mBAAmB,CAAC;AAC1D,OAAO,KAAK,EAAE,iBAAiB,IAAI,YAAY,EAAE,MAAM,wBAAwB,CAAC;AAEhF,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,iBAAiB,CAAC;AACtD,OAAO,EAIL,KAAK,SAAS,EACd,KAAK,cAAc,EACnB,KAAK,cAAc,EACpB,MAAM,uBAAuB,CAAC;AAE/B,OAAO,EAAE,WAAW,EAAE,WAAW,EAAE,MAAM,wBAAwB,CAAC;AAClE,OAAO,KAAK,EAAE,GAAG,EAAE,aAAa,EAAE,kBAAkB,EAAE,UAAU,EAAE,MAAM,iBAAiB,CAAC;AAC1F,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAE9C,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,iBAAiB,CAAC;AAC3C,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAC9C,OAAO,KAAK,EAAE,GAAG,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AACnD,OAAO,KAAK,EAAE,GAAG,EAAE,MAAM,WAAW,CAAC;AACrC,OAAO,KAAK,EAAE,KAAK,EAAE,MAAM,YAAY,CAAC;AAExC,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,wBAAwB,CAAC;AAC5D,OAAO,EACL,sBAAsB,EACtB,KAAK,UAAU,EACf,KAAK,sBAAsB,EAC3B,KAAK,UAAU,EACf,WAAW,EACX,KAAK,0BAA0B,EAC/B,KAAK,UAAU,EACf,KAAK,0BAA0B,EAC/B,KAAK,qBAAqB,EAC1B,KAAK,kBAAkB,EACvB,KAAK,yBAAyB,EAC9B,KAAK,SAAS,EACd,KAAK,qBAAqB,EAK3B,MAAM,aAAa,CAAC;AACrB,OAAO,EAAE,UAAU,EAAE,WAAW,EAAE,MAAM,SAAS,CAAC;AAClD,OAAO,EAAU,KAAK,gBAAgB,EAAE,KAAK,iBAAiB,EAAE,MAAM,oBAAoB,CAAC;AAC3F,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,kBAAkB,CAAC;AACxD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AAEvD,MAAM,WAAW,YAAY;IAC3B,kBAAkB,EAAE,OAAO,CAAC;IAC5B,6BAA6B,EAAE,OAAO,CAAC;IACvC,uBAAuB,EAAE,MAAM,CAAC;IAChC,oBAAoB,EAAE,MAAM,CAAC;IAC7B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,mBAAmB,EAAE,MAAM,CAAC;IAC5B,YAAY,EAAE,MAAM,CAAC;IACrB,oBAAoB,EAAE,OAAO,CAAC;IAC9B,eAAe,CAAC,EAAE,MAAM,GAAG,IAAI,CAAC;CACjC;AAcD,MAAM,MAAM,iBAAiB,GAAG,KAAK,GAAG,KAAK,GAAG,cAAc,GAAG,QAAQ,GAAG,aAAa,CAAC;AAE1F,MAAM,MAAM,qBAAqB,GAAG;IAClC,CAAC,sBAAsB,CAAC,oBAAoB,CAAC,EAAE,CAAC,EAAE,EAAE,yBAAyB,KAAK,IAAI,CAAC;IACvF,CAAC,sBAAsB,CAAC,iBAAiB,CAAC,EAAE,CAAC,EAAE,EAAE,sBAAsB,KAAK,IAAI,CAAC;IACjF,CAAC,sBAAsB,CAAC,gBAAgB,CAAC,EAAE,CAAC,EAAE,EAAE,qBAAqB,KAAK,IAAI,CAAC;IAC/E,CAAC,sBAAsB,CAAC,qBAAqB,CAAC,EAAE,CAAC,EAAE,EAAE,0BAA0B,KAAK,IAAI,CAAC;IACzF,CAAC,sBAAsB,CAAC,qBAAqB,CAAC,EAAE,CAAC,EAAE,EAAE,0BAA0B,KAAK,IAAI,CAAC;IACzF,CAAC,sBAAsB,CAAC,gBAAgB,CAAC,EAAE,CAAC,EAAE,EAAE,qBAAqB,KAAK,IAAI,CAAC;IAC/E,CAAC,sBAAsB,CAAC,aAAa,CAAC,EAAE,CAAC,EAAE,EAAE,kBAAkB,KAAK,IAAI,CAAC;IACzE,CAAC,sBAAsB,CAAC,KAAK,CAAC,EAAE,CAAC,EAAE,EAAE,UAAU,KAAK,IAAI,CAAC;IACzD,CAAC,sBAAsB,CAAC,KAAK,CAAC,EAAE,CAAC,EAAE,EAAE,UAAU,KAAK,IAAI,CAAC;CAC1D,CAAC;AAEF,MAAM,MAAM,mBAAmB,CAAC,QAAQ,GAAG,eAAe,IAAI;IAC5D,aAAa,CAAC,EAAE,iBAAiB,CAAC;IAClC,GAAG,CAAC,EAAE,GAAG,GAAG,cAAc,CAAC;IAC3B,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,GAAG,aAAa,GAAG,SAAS,CAAC;IACtC,GAAG,CAAC,EAAE,GAAG,GAAG,cAAc,CAAC;IAC3B,QAAQ,CAAC,EAAE,QAAQ,CAAC;IACpB,YAAY,CAAC,EAAE,OAAO,CAAC,YAAY,CAAC,CAAC;CACtC,CAAC;2CAImC,aAAa,qBAAqB,CAAC;AAFxE,qBAAa,YAAY,CACvB,QAAQ,GAAG,eAAe,CAC1B,SAAQ,iBAA+D;IACvE,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,GAAG,CAAC,EAAE,GAAG,GAAG,aAAa,CAAC;IAC1B,GAAG,CAAC,EAAE,GAAG,CAAC;IACV,aAAa,CAAC,EAAE,iBAAiB,CAAC;IAElC,QAAQ,CAAC,OAAO,EAAE,YAAY,CAAC;IAE/B,OAAO,CAAC,KAAK,CAAC,CAAQ;IACtB,OAAO,CAAC,QAAQ,CAAC,CAAgB;IACjC,OAAO,CAAC,YAAY,CAAC,CAAgB;IACrC,OAAO,CAAC,OAAO,CAAS;IACxB,OAAO,CAAC,SAAS,CAA0B;IAE3C,OAAO,CAAC,MAAM,CAAC,CAAS;IACxB,OAAO,CAAC,MAAM,CAAS;IAEvB,OAAO,CAAC,QAAQ,CAAc;IAC9B,OAAO,CAAC,SAAS,CAAuB;IACxC,OAAO,CAAC,WAAW,CAA8B;IAEjD,OAAO,CAAC,MAAM,CAAa;IAC3B,OAAO,CAAC,OAAO,CAAc;IAE7B,OAAO,CAAC,WAAW,CAA8B;IACjD,OAAO,CAAC,aAAa,CAA+B;gBAExC,IAAI,EAAE,mBAAmB,CAAC,QAAQ,CAAC;IA+C/C,IAAI,KAAK,IAAI,UAAU,CAEtB;IAED,IAAI,MAAM,IAAI,WAAW,CAExB;IAED,IAAI,QAAQ,IAAI,QAAQ,CAMvB;IAED,IAAI,OAAO,IAAI,WAAW,CAEzB;IAED,IAAI,QAAQ,CAAC,KAAK,EAAE,QAAQ,EAE3B;IAEK,KAAK,CAAC,EACV,KAAK,EACL,IAAI,EACJ,YAAY,EACZ,aAAa,GACd,EAAE;QACD,KAAK,EAAE,KAAK,CAAC;QACb,IAAI,EAAE,IAAI,CAAC;QACX,YAAY,CAAC,EAAE,OAAO,CAAC,gBAAgB,CAAC,CAAC;QACzC,aAAa,CAAC,EAAE,OAAO,CAAC,iBAAiB,CAAC,CAAC;KAC5C,GAAG,OAAO,CAAC,IAAI,CAAC;IAyDjB,WAAW,CAAC,KAAK,EAAE,KAAK,GAAG,IAAI;IAQ/B,cAAc;IAQd,aAAa;IAOb,GAAG,CACD,IAAI,EAAE,MAAM,GAAG,cAAc,CAAC,MAAM,CAAC,EACrC,OAAO,CAAC,EAAE;QACR,KAAK,CAAC,EAAE,cAAc,CAAC,UAAU,CAAC,CAAC;QACnC,kBAAkB,CAAC,EAAE,OAAO,CAAC;QAC7B,YAAY,CAAC,EAAE,OAAO,CAAC;KACxB,GACA,YAAY;IAQf,SAAS;IAOT,aAAa,CAAC,OAAO,CAAC,EAAE;QACtB,SAAS,CAAC,EAAE,MAAM,CAAC;QACnB,YAAY,CAAC,EAAE,MAAM,CAAC;QACtB,UAAU,CAAC,EAAE,UAAU,CAAC;QACxB,kBAAkB,CAAC,EAAE,OAAO,CAAC;KAC9B,GAAG,YAAY;YAsBF,cAAc;IAmB5B,IAAI,OAAO,IAAI,WAAW,CAEzB;IAED,IAAI,UAAU,IAAI,UAAU,CAE3B;IAED,IAAI,YAAY,IAAI,KAAK,CAMxB;IAEK,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;IAI5B,gBAAgB;IAChB,UAAU,CAAC,EACT,MAAM,EACN,KAAa,EACb,KAAY,GACb,EAAE;QACD,MAAM,EAAE,WAAW,CAAC;QACpB,KAAK,CAAC,EAAE,OAAO,CAAC;QAChB,KAAK,CAAC,EAAE,kBAAkB,GAAG,QAAQ,GAAG,QAAQ,GAAG,QAAQ,GAAG,IAAI,CAAC;KACpE,GAAG,IAAI;IAOR,gBAAgB;IAChB,QAAQ,CAAC,KAAK,EAAE,kBAAkB,GAAG,QAAQ,GAAG,QAAQ,GAAG,QAAQ,GAAG,IAAI;IAc1E,gBAAgB;IAChB,sBAAsB,CAAC,IAAI,EAAE,WAAW,GAAG,IAAI;IAK/C,gBAAgB;IAChB,iBAAiB,CAAC,KAAK,EAAE,UAAU;IAqBnC,gBAAgB;IAChB,gBAAgB,CAAC,KAAK,EAAE,SAAS,EAAE,iBAAiB,CAAC,EAAE,MAAM;IAsB7D,OAAO,CAAC,mBAAmB;IAU3B,OAAO,CAAC,oBAAoB;IAE5B,OAAO,CAAC,mBAAmB;IAE3B,OAAO,CAAC,iBAAiB;IAiBzB,OAAO,CAAC,oBAAoB;IAO5B,OAAO,CAAC,uBAAuB;YAOjB,SAAS;CA+CxB"}
@@ -26,7 +26,8 @@ const defaultVoiceOptions = {
26
26
  minEndpointingDelay: 500,
27
27
  maxEndpointingDelay: 6e3,
28
28
  maxToolSteps: 3,
29
- preemptiveGeneration: false
29
+ preemptiveGeneration: false,
30
+ userAwayTimeout: 15
30
31
  };
31
32
  class AgentSession extends EventEmitter {
32
33
  vad;
@@ -48,6 +49,7 @@ class AgentSession extends EventEmitter {
48
49
  _input;
49
50
  _output;
50
51
  closingTask = null;
52
+ userAwayTimer = null;
51
53
  constructor(opts) {
52
54
  super();
53
55
  const {
@@ -81,6 +83,7 @@ class AgentSession extends EventEmitter {
81
83
  this._output = new AgentOutput(this.onAudioOutputChanged, this.onTextOutputChanged);
82
84
  this._chatCtx = ChatContext.empty();
83
85
  this.options = { ...defaultVoiceOptions, ...voiceOptions };
86
+ this.on(AgentSessionEventTypes.UserInputTranscribed, this._onUserInputTranscribed.bind(this));
84
87
  }
85
88
  get input() {
86
89
  return this._input;
@@ -257,6 +260,11 @@ class AgentSession extends EventEmitter {
257
260
  }
258
261
  const oldState = this._agentState;
259
262
  this._agentState = state;
263
+ if (state === "listening" && this.userState === "listening") {
264
+ this._setUserAwayTimer();
265
+ } else {
266
+ this._cancelUserAwayTimer();
267
+ }
260
268
  this.emit(
261
269
  AgentSessionEventTypes.AgentStateChanged,
262
270
  createAgentStateChangedEvent(oldState, state)
@@ -269,6 +277,11 @@ class AgentSession extends EventEmitter {
269
277
  }
270
278
  const oldState = this.userState;
271
279
  this.userState = state;
280
+ if (state === "listening" && this._agentState === "listening") {
281
+ this._setUserAwayTimer();
282
+ } else {
283
+ this._cancelUserAwayTimer();
284
+ }
272
285
  this.emit(
273
286
  AgentSessionEventTypes.UserStateChanged,
274
287
  createUserStateChangedEvent(oldState, state)
@@ -287,11 +300,37 @@ class AgentSession extends EventEmitter {
287
300
  }
288
301
  onTextOutputChanged() {
289
302
  }
303
+ _setUserAwayTimer() {
304
+ this._cancelUserAwayTimer();
305
+ if (this.options.userAwayTimeout === null || this.options.userAwayTimeout === void 0) {
306
+ return;
307
+ }
308
+ if (this.roomIO && !this.roomIO.isParticipantAvailable) {
309
+ return;
310
+ }
311
+ this.userAwayTimer = setTimeout(() => {
312
+ this.logger.debug("User away timeout triggered");
313
+ this._updateUserState("away");
314
+ }, this.options.userAwayTimeout * 1e3);
315
+ }
316
+ _cancelUserAwayTimer() {
317
+ if (this.userAwayTimer !== null) {
318
+ clearTimeout(this.userAwayTimer);
319
+ this.userAwayTimer = null;
320
+ }
321
+ }
322
+ _onUserInputTranscribed(ev) {
323
+ if (this.userState === "away" && ev.isFinal) {
324
+ this.logger.debug("User returned from away state due to speech input");
325
+ this._updateUserState("listening");
326
+ }
327
+ }
290
328
  async closeImpl(reason, error = null, drain = false) {
291
329
  var _a, _b, _c;
292
330
  if (!this.started) {
293
331
  return;
294
332
  }
333
+ this._cancelUserAwayTimer();
295
334
  if (this.activity) {
296
335
  if (!drain) {
297
336
  try {
@@ -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 { 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 { getJobContext } from '../job.js';\nimport { 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 type { TTS, TTSError } from '../tts/tts.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 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 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 { RoomIO, type RoomInputOptions, type RoomOutputOptions } from './room_io/index.js';\nimport type { UnknownUserData } from './run_context.js';\nimport type { SpeechHandle } from './speech_handle.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}\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} 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};\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\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 } = opts;\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\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 set userData(value: UserData) {\n this._userData = value;\n }\n\n async start({\n agent,\n room,\n inputOptions,\n outputOptions,\n }: {\n agent: Agent;\n room: Room;\n inputOptions?: Partial<RoomInputOptions>;\n outputOptions?: Partial<RoomOutputOptions>;\n }): Promise<void> {\n if (this.started) {\n return;\n }\n\n this.agent = agent;\n this._updateAgentState('initializing');\n\n const tasks: Promise<void>[] = [];\n // Check for existing input/output configuration and warn if needed\n if (this.input.audio && inputOptions?.audioEnabled !== false) {\n this.logger.warn('RoomIO audio input is enabled but input.audio is already set, ignoring..');\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 const ctx = getJobContext();\n if (ctx && ctx.room === room && !room.isConnected) {\n this.logger.debug('Auto-connecting to room via job context');\n tasks.push(ctx.connect());\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._updateAgentState('listening');\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 return this.activity.say(text, options);\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 if (this.activity.draining) {\n if (!this.nextActivity) {\n throw new Error('AgentSession is closing, cannot use generateReply()');\n }\n return this.nextActivity.generateReply({ userMessage, ...options });\n }\n\n return this.activity.generateReply({ userMessage, ...options });\n }\n\n private async updateActivity(agent: Agent): Promise<void> {\n // TODO(AJS-129): add lock to agent activity core lifecycle\n this.nextActivity = new AgentActivity(agent, this);\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 await this.activity.start();\n\n if (this._input.audio) {\n this.activity.attachAudioInput(this._input.audio.stream);\n }\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 /** @internal */\n _closeSoon({\n reason,\n drain = false,\n error = null,\n }: {\n reason: CloseReason;\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 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 _updateAgentState(state: AgentState) {\n if (this._agentState === state) {\n return;\n }\n\n const oldState = this._agentState;\n this._agentState = state;\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 const oldState = this.userState;\n this.userState = state;\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 async closeImpl(\n reason: CloseReason,\n error: RealtimeModelError | LLMError | TTSError | STTError | null = null,\n drain: boolean = false,\n ): Promise<void> {\n if (!this.started) {\n return;\n }\n\n if (this.activity) {\n if (!drain) {\n try {\n this.activity.interrupt();\n } catch (error) {\n // uninterruptible speech [copied from python]\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 this.activity.detachAudioInput();\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 this.started = false;\n\n this.emit(AgentSessionEventTypes.Close, createCloseEvent(reason, error));\n\n this.userState = 'listening';\n this._agentState = 'initializing';\n\n this.logger.info({ reason, error }, 'AgentSession closed');\n }\n}\n"],"mappings":"AAKA,SAAS,oBAAoB;AAE7B;AAAA,EACE,OAAO;AAAA,EACP,OAAO;AAAA,EACP,OAAO;AAAA,OAIF;AACP,SAAS,qBAAqB;AAC9B,SAAS,aAAa,mBAAmB;AAGzC,SAAS,WAAW;AAMpB,SAAS,qBAAqB;AAE9B;AAAA,EACE;AAAA,EAIA;AAAA,EASA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,YAAY,mBAAmB;AACxC,SAAS,cAA6D;AAetE,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;AACxB;AA0BO,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,EAE5C,YAAY,MAAqC;AAC/C,UAAM;AAEN,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,eAAe;AAAA,IACjB,IAAI;AAEJ,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;AAAA,EAC3D;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,EAEA,IAAI,SAAS,OAAiB;AAC5B,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,MAAM,MAAM;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAKkB;AAChB,QAAI,KAAK,SAAS;AAChB;AAAA,IACF;AAEA,SAAK,QAAQ;AACb,SAAK,kBAAkB,cAAc;AAErC,UAAM,QAAyB,CAAC;AAEhC,QAAI,KAAK,MAAM,UAAS,6CAAc,kBAAiB,OAAO;AAC5D,WAAK,OAAO,KAAK,0EAA0E;AAAA,IAC7F;AAEA,QAAI,KAAK,OAAO,UAAS,+CAAe,kBAAiB,OAAO;AAC9D,WAAK,OAAO;AAAA,QACV;AAAA,MACF;AAAA,IACF;AAEA,QAAI,KAAK,OAAO,kBAAiB,+CAAe,0BAAyB,OAAO;AAC9E,WAAK,OAAO;AAAA,QACV;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS,IAAI,OAAO;AAAA,MACvB,cAAc;AAAA,MACd;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC;AACD,SAAK,OAAO,MAAM;AAElB,UAAM,MAAM,cAAc;AAC1B,QAAI,OAAO,IAAI,SAAS,QAAQ,CAAC,KAAK,aAAa;AACjD,WAAK,OAAO,MAAM,yCAAyC;AAC3D,YAAM,KAAK,IAAI,QAAQ,CAAC;AAAA,IAC1B;AAEA,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,kBAAkB,WAAW;AAAA,EACpC;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,WAAO,KAAK,SAAS,IAAI,MAAM,OAAO;AAAA,EACxC;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,QAAI,KAAK,SAAS,UAAU;AAC1B,UAAI,CAAC,KAAK,cAAc;AACtB,cAAM,IAAI,MAAM,qDAAqD;AAAA,MACvE;AACA,aAAO,KAAK,aAAa,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,IACpE;AAEA,WAAO,KAAK,SAAS,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,eAAe,OAA6B;AAExD,SAAK,eAAe,IAAI,cAAc,OAAO,IAAI;AAEjD,QAAI,KAAK,UAAU;AACjB,YAAM,KAAK,SAAS,MAAM;AAC1B,YAAM,KAAK,SAAS,MAAM;AAAA,IAC5B;AAEA,SAAK,WAAW,KAAK;AACrB,SAAK,eAAe;AAEpB,UAAM,KAAK,SAAS,MAAM;AAE1B,QAAI,KAAK,OAAO,OAAO;AACrB,WAAK,SAAS,iBAAiB,KAAK,OAAO,MAAM,MAAM;AAAA,IACzD;AAAA,EACF;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;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;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,kBAAkB,OAAmB;AACnC,QAAI,KAAK,gBAAgB,OAAO;AAC9B;AAAA,IACF;AAEA,UAAM,WAAW,KAAK;AACtB,SAAK,cAAc;AACnB,SAAK;AAAA,MACH,uBAAuB;AAAA,MACvB,6BAA6B,UAAU,KAAK;AAAA,IAC9C;AAAA,EACF;AAAA;AAAA,EAGA,iBAAiB,OAAkB,mBAA4B;AAC7D,QAAI,KAAK,cAAc,OAAO;AAC5B;AAAA,IACF;AAEA,UAAM,WAAW,KAAK;AACtB,SAAK,YAAY;AACjB,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,EAErC,MAAc,UACZ,QACA,QAAoE,MACpE,QAAiB,OACF;AAzcnB;AA0cI,QAAI,CAAC,KAAK,SAAS;AACjB;AAAA,IACF;AAEA,QAAI,KAAK,UAAU;AACjB,UAAI,CAAC,OAAO;AACV,YAAI;AACF,eAAK,SAAS,UAAU;AAAA,QAC1B,SAASA,QAAO;AAAA,QAIhB;AAAA,MACF;AACA,YAAM,KAAK,SAAS,MAAM;AAE1B,cAAM,UAAK,SAAS,kBAAd,mBAA6B;AACnC,WAAK,SAAS,iBAAiB;AAAA,IACjC;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,SAAK,UAAU;AAEf,SAAK,KAAK,uBAAuB,OAAO,iBAAiB,QAAQ,KAAK,CAAC;AAEvE,SAAK,YAAY;AACjB,SAAK,cAAc;AAEnB,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 { 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 { getJobContext } from '../job.js';\nimport { 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 type { TTS, TTSError } from '../tts/tts.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 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 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 { RoomIO, type RoomInputOptions, type RoomOutputOptions } from './room_io/index.js';\nimport type { UnknownUserData } from './run_context.js';\nimport type { SpeechHandle } from './speech_handle.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}\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} 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};\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 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 } = opts;\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.on(AgentSessionEventTypes.UserInputTranscribed, this._onUserInputTranscribed.bind(this));\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 set userData(value: UserData) {\n this._userData = value;\n }\n\n async start({\n agent,\n room,\n inputOptions,\n outputOptions,\n }: {\n agent: Agent;\n room: Room;\n inputOptions?: Partial<RoomInputOptions>;\n outputOptions?: Partial<RoomOutputOptions>;\n }): Promise<void> {\n if (this.started) {\n return;\n }\n\n this.agent = agent;\n this._updateAgentState('initializing');\n\n const tasks: Promise<void>[] = [];\n // Check for existing input/output configuration and warn if needed\n if (this.input.audio && inputOptions?.audioEnabled !== false) {\n this.logger.warn('RoomIO audio input is enabled but input.audio is already set, ignoring..');\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 const ctx = getJobContext();\n if (ctx && ctx.room === room && !room.isConnected) {\n this.logger.debug('Auto-connecting to room via job context');\n tasks.push(ctx.connect());\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._updateAgentState('listening');\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 return this.activity.say(text, options);\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 if (this.activity.draining) {\n if (!this.nextActivity) {\n throw new Error('AgentSession is closing, cannot use generateReply()');\n }\n return this.nextActivity.generateReply({ userMessage, ...options });\n }\n\n return this.activity.generateReply({ userMessage, ...options });\n }\n\n private async updateActivity(agent: Agent): Promise<void> {\n // TODO(AJS-129): add lock to agent activity core lifecycle\n this.nextActivity = new AgentActivity(agent, this);\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 await this.activity.start();\n\n if (this._input.audio) {\n this.activity.attachAudioInput(this._input.audio.stream);\n }\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 /** @internal */\n _closeSoon({\n reason,\n drain = false,\n error = null,\n }: {\n reason: CloseReason;\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 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 _updateAgentState(state: AgentState) {\n if (this._agentState === state) {\n return;\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 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: CloseReason,\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\n if (this.activity) {\n if (!drain) {\n try {\n this.activity.interrupt();\n } catch (error) {\n // uninterruptible speech [copied from python]\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 this.activity.detachAudioInput();\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 this.started = false;\n\n this.emit(AgentSessionEventTypes.Close, createCloseEvent(reason, error));\n\n this.userState = 'listening';\n this._agentState = 'initializing';\n\n this.logger.info({ reason, error }, 'AgentSession closed');\n }\n}\n"],"mappings":"AAKA,SAAS,oBAAoB;AAE7B;AAAA,EACE,OAAO;AAAA,EACP,OAAO;AAAA,EACP,OAAO;AAAA,OAIF;AACP,SAAS,qBAAqB;AAC9B,SAAS,aAAa,mBAAmB;AAGzC,SAAS,WAAW;AAMpB,SAAS,qBAAqB;AAE9B;AAAA,EACE;AAAA,EAIA;AAAA,EASA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,YAAY,mBAAmB;AACxC,SAAS,cAA6D;AAgBtE,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;AACnB;AA0BO,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,EAE/C,YAAY,MAAqC;AAC/C,UAAM;AAEN,UAAM;AAAA,MACJ;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA;AAAA,MACA,eAAe;AAAA,IACjB,IAAI;AAEJ,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,GAAG,uBAAuB,sBAAsB,KAAK,wBAAwB,KAAK,IAAI,CAAC;AAAA,EAC9F;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,EAEA,IAAI,SAAS,OAAiB;AAC5B,SAAK,YAAY;AAAA,EACnB;AAAA,EAEA,MAAM,MAAM;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAKkB;AAChB,QAAI,KAAK,SAAS;AAChB;AAAA,IACF;AAEA,SAAK,QAAQ;AACb,SAAK,kBAAkB,cAAc;AAErC,UAAM,QAAyB,CAAC;AAEhC,QAAI,KAAK,MAAM,UAAS,6CAAc,kBAAiB,OAAO;AAC5D,WAAK,OAAO,KAAK,0EAA0E;AAAA,IAC7F;AAEA,QAAI,KAAK,OAAO,UAAS,+CAAe,kBAAiB,OAAO;AAC9D,WAAK,OAAO;AAAA,QACV;AAAA,MACF;AAAA,IACF;AAEA,QAAI,KAAK,OAAO,kBAAiB,+CAAe,0BAAyB,OAAO;AAC9E,WAAK,OAAO;AAAA,QACV;AAAA,MACF;AAAA,IACF;AAEA,SAAK,SAAS,IAAI,OAAO;AAAA,MACvB,cAAc;AAAA,MACd;AAAA,MACA;AAAA,MACA;AAAA,IACF,CAAC;AACD,SAAK,OAAO,MAAM;AAElB,UAAM,MAAM,cAAc;AAC1B,QAAI,OAAO,IAAI,SAAS,QAAQ,CAAC,KAAK,aAAa;AACjD,WAAK,OAAO,MAAM,yCAAyC;AAC3D,YAAM,KAAK,IAAI,QAAQ,CAAC;AAAA,IAC1B;AAEA,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,kBAAkB,WAAW;AAAA,EACpC;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,WAAO,KAAK,SAAS,IAAI,MAAM,OAAO;AAAA,EACxC;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,QAAI,KAAK,SAAS,UAAU;AAC1B,UAAI,CAAC,KAAK,cAAc;AACtB,cAAM,IAAI,MAAM,qDAAqD;AAAA,MACvE;AACA,aAAO,KAAK,aAAa,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,IACpE;AAEA,WAAO,KAAK,SAAS,cAAc,EAAE,aAAa,GAAG,QAAQ,CAAC;AAAA,EAChE;AAAA,EAEA,MAAc,eAAe,OAA6B;AAExD,SAAK,eAAe,IAAI,cAAc,OAAO,IAAI;AAEjD,QAAI,KAAK,UAAU;AACjB,YAAM,KAAK,SAAS,MAAM;AAC1B,YAAM,KAAK,SAAS,MAAM;AAAA,IAC5B;AAEA,SAAK,WAAW,KAAK;AACrB,SAAK,eAAe;AAEpB,UAAM,KAAK,SAAS,MAAM;AAE1B,QAAI,KAAK,OAAO,OAAO;AACrB,WAAK,SAAS,iBAAiB,KAAK,OAAO,MAAM,MAAM;AAAA,IACzD;AAAA,EACF;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;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;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,kBAAkB,OAAmB;AACnC,QAAI,KAAK,gBAAgB,OAAO;AAC9B;AAAA,IACF;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,mBAA4B;AAC7D,QAAI,KAAK,cAAc,OAAO;AAC5B;AAAA,IACF;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;AA7fnB;AA8fI,QAAI,CAAC,KAAK,SAAS;AACjB;AAAA,IACF;AAEA,SAAK,qBAAqB;AAE1B,QAAI,KAAK,UAAU;AACjB,UAAI,CAAC,OAAO;AACV,YAAI;AACF,eAAK,SAAS,UAAU;AAAA,QAC1B,SAASA,QAAO;AAAA,QAIhB;AAAA,MACF;AACA,YAAM,KAAK,SAAS,MAAM;AAE1B,cAAM,UAAK,SAAS,kBAAd,mBAA6B;AACnC,WAAK,SAAS,iBAAiB;AAAA,IACjC;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,SAAK,UAAU;AAEf,SAAK,KAAK,uBAAuB,OAAO,iBAAiB,QAAQ,KAAK,CAAC;AAEvE,SAAK,YAAY;AACjB,SAAK,cAAc;AAEnB,SAAK,OAAO,KAAK,EAAE,QAAQ,MAAM,GAAG,qBAAqB;AAAA,EAC3D;AACF;","names":["error"]}
@@ -0,0 +1,114 @@
1
+ "use strict";
2
+ var import_vitest = require("vitest");
3
+ var import_word = require("../tokenize/basic/word.cjs");
4
+ (0, import_vitest.describe)("Interruption Detection - Word Counting", () => {
5
+ (0, import_vitest.describe)("Word Splitting Behavior", () => {
6
+ (0, import_vitest.it)("should count empty string as 0 words", () => {
7
+ const text = "";
8
+ const wordCount = (0, import_word.splitWords)(text, true).length;
9
+ (0, import_vitest.expect)(wordCount).toBe(0);
10
+ });
11
+ (0, import_vitest.it)("should count single word correctly", () => {
12
+ const text = "hello";
13
+ const wordCount = (0, import_word.splitWords)(text, true).length;
14
+ (0, import_vitest.expect)(wordCount).toBe(1);
15
+ });
16
+ (0, import_vitest.it)("should count two words correctly", () => {
17
+ const text = "hello world";
18
+ const wordCount = (0, import_word.splitWords)(text, true).length;
19
+ (0, import_vitest.expect)(wordCount).toBe(2);
20
+ });
21
+ (0, import_vitest.it)("should count multiple words correctly", () => {
22
+ const text = "hello this is a full sentence";
23
+ const wordCount = (0, import_word.splitWords)(text, true).length;
24
+ (0, import_vitest.expect)(wordCount).toBe(6);
25
+ });
26
+ (0, import_vitest.it)("should handle punctuation correctly", () => {
27
+ const text = "hello, world!";
28
+ const wordCount = (0, import_word.splitWords)(text, true).length;
29
+ (0, import_vitest.expect)(wordCount).toBe(2);
30
+ });
31
+ (0, import_vitest.it)("should handle multiple spaces between words", () => {
32
+ const text = "hello world";
33
+ const wordCount = (0, import_word.splitWords)(text, true).length;
34
+ (0, import_vitest.expect)(wordCount).toBe(2);
35
+ });
36
+ (0, import_vitest.it)("should count whitespace-only string as 0 words", () => {
37
+ const text = " ";
38
+ const wordCount = (0, import_word.splitWords)(text, true).length;
39
+ (0, import_vitest.expect)(wordCount).toBe(0);
40
+ });
41
+ (0, import_vitest.it)("should handle leading and trailing whitespace", () => {
42
+ const text = " hello world ";
43
+ const wordCount = (0, import_word.splitWords)(text, true).length;
44
+ (0, import_vitest.expect)(wordCount).toBe(2);
45
+ });
46
+ });
47
+ (0, import_vitest.describe)("Integration: Full Interruption Check Logic", () => {
48
+ (0, import_vitest.it)("should block interruption for empty transcript with threshold 2", () => {
49
+ const text = "";
50
+ const minInterruptionWords = 2;
51
+ const normalizedText = text ?? "";
52
+ const wordCount = (0, import_word.splitWords)(normalizedText, true).length;
53
+ const shouldBlock = wordCount < minInterruptionWords;
54
+ (0, import_vitest.expect)(normalizedText).toBe("");
55
+ (0, import_vitest.expect)(wordCount).toBe(0);
56
+ (0, import_vitest.expect)(shouldBlock).toBe(true);
57
+ });
58
+ (0, import_vitest.it)("should block interruption for undefined transcript with threshold 2", () => {
59
+ const text = void 0;
60
+ const minInterruptionWords = 2;
61
+ const normalizedText = text ?? "";
62
+ const wordCount = (0, import_word.splitWords)(normalizedText, true).length;
63
+ const shouldBlock = wordCount < minInterruptionWords;
64
+ (0, import_vitest.expect)(normalizedText).toBe("");
65
+ (0, import_vitest.expect)(wordCount).toBe(0);
66
+ (0, import_vitest.expect)(shouldBlock).toBe(true);
67
+ });
68
+ (0, import_vitest.it)("should block interruption for single word with threshold 2", () => {
69
+ const text = "hello";
70
+ const minInterruptionWords = 2;
71
+ const normalizedText = text ?? "";
72
+ const wordCount = (0, import_word.splitWords)(normalizedText, true).length;
73
+ const shouldBlock = wordCount < minInterruptionWords;
74
+ (0, import_vitest.expect)(normalizedText).toBe("hello");
75
+ (0, import_vitest.expect)(wordCount).toBe(1);
76
+ (0, import_vitest.expect)(shouldBlock).toBe(true);
77
+ });
78
+ (0, import_vitest.it)("should allow interruption when word count exactly meets threshold", () => {
79
+ const text = "hello world";
80
+ const minInterruptionWords = 2;
81
+ const normalizedText = text ?? "";
82
+ const wordCount = (0, import_word.splitWords)(normalizedText, true).length;
83
+ const shouldBlock = wordCount < minInterruptionWords;
84
+ (0, import_vitest.expect)(normalizedText).toBe("hello world");
85
+ (0, import_vitest.expect)(wordCount).toBe(2);
86
+ (0, import_vitest.expect)(shouldBlock).toBe(false);
87
+ });
88
+ (0, import_vitest.it)("should allow interruption when word count exceeds threshold", () => {
89
+ const text = "hello this is a full sentence";
90
+ const minInterruptionWords = 2;
91
+ const normalizedText = text ?? "";
92
+ const wordCount = (0, import_word.splitWords)(normalizedText, true).length;
93
+ const shouldBlock = wordCount < minInterruptionWords;
94
+ (0, import_vitest.expect)(normalizedText).toBe("hello this is a full sentence");
95
+ (0, import_vitest.expect)(wordCount).toBe(6);
96
+ (0, import_vitest.expect)(shouldBlock).toBe(false);
97
+ });
98
+ (0, import_vitest.it)("should apply consistent word counting logic in both methods", () => {
99
+ const transcripts = ["", "hello", "hello world", "this is a longer sentence"];
100
+ const threshold = 2;
101
+ transcripts.forEach((transcript) => {
102
+ const text1 = transcript;
103
+ const normalizedText1 = text1 ?? "";
104
+ const wordCount1 = (0, import_word.splitWords)(normalizedText1, true).length;
105
+ const shouldBlock1 = wordCount1 < threshold;
106
+ const wordCount2 = (0, import_word.splitWords)(transcript, true).length;
107
+ const shouldBlock2 = wordCount2 < threshold;
108
+ (0, import_vitest.expect)(wordCount1).toBe(wordCount2);
109
+ (0, import_vitest.expect)(shouldBlock1).toBe(shouldBlock2);
110
+ });
111
+ });
112
+ });
113
+ });
114
+ //# sourceMappingURL=interruption_detection.test.cjs.map
@@ -0,0 +1 @@
1
+ {"version":3,"sources":["../../src/voice/interruption_detection.test.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\n\n/**\n * Unit tests for interruption detection logic in AgentActivity.\n *\n * Tests the refactored minInterruptionWords check which ensures:\n * - Consistent word count filtering across all speech scenarios\n * - Proper handling of empty strings, undefined, and short speech\n * - Interruptions allowed only when word count meets or exceeds minInterruptionWords threshold\n */\nimport { describe, expect, it } from 'vitest';\nimport { splitWords } from '../tokenize/basic/word.js';\n\ndescribe('Interruption Detection - Word Counting', () => {\n describe('Word Splitting Behavior', () => {\n it('should count empty string as 0 words', () => {\n const text = '';\n const wordCount = splitWords(text, true).length;\n expect(wordCount).toBe(0);\n });\n\n it('should count single word correctly', () => {\n const text = 'hello';\n const wordCount = splitWords(text, true).length;\n expect(wordCount).toBe(1);\n });\n\n it('should count two words correctly', () => {\n const text = 'hello world';\n const wordCount = splitWords(text, true).length;\n expect(wordCount).toBe(2);\n });\n\n it('should count multiple words correctly', () => {\n const text = 'hello this is a full sentence';\n const wordCount = splitWords(text, true).length;\n expect(wordCount).toBe(6);\n });\n\n it('should handle punctuation correctly', () => {\n const text = 'hello, world!';\n const wordCount = splitWords(text, true).length;\n expect(wordCount).toBe(2);\n });\n\n it('should handle multiple spaces between words', () => {\n const text = 'hello world';\n const wordCount = splitWords(text, true).length;\n expect(wordCount).toBe(2);\n });\n\n it('should count whitespace-only string as 0 words', () => {\n const text = ' ';\n const wordCount = splitWords(text, true).length;\n expect(wordCount).toBe(0);\n });\n\n it('should handle leading and trailing whitespace', () => {\n const text = ' hello world ';\n const wordCount = splitWords(text, true).length;\n expect(wordCount).toBe(2);\n });\n });\n\n describe('Integration: Full Interruption Check Logic', () => {\n it('should block interruption for empty transcript with threshold 2', () => {\n const text = '';\n const minInterruptionWords = 2;\n\n const normalizedText = text ?? '';\n const wordCount = splitWords(normalizedText, true).length;\n const shouldBlock = wordCount < minInterruptionWords;\n\n expect(normalizedText).toBe('');\n expect(wordCount).toBe(0);\n expect(shouldBlock).toBe(true);\n });\n\n it('should block interruption for undefined transcript with threshold 2', () => {\n const text: string | undefined = undefined;\n const minInterruptionWords = 2;\n\n const normalizedText = text ?? '';\n const wordCount = splitWords(normalizedText, true).length;\n const shouldBlock = wordCount < minInterruptionWords;\n\n expect(normalizedText).toBe('');\n expect(wordCount).toBe(0);\n expect(shouldBlock).toBe(true);\n });\n\n it('should block interruption for single word with threshold 2', () => {\n const text = 'hello';\n const minInterruptionWords = 2;\n\n const normalizedText = text ?? '';\n const wordCount = splitWords(normalizedText, true).length;\n const shouldBlock = wordCount < minInterruptionWords;\n\n expect(normalizedText).toBe('hello');\n expect(wordCount).toBe(1);\n expect(shouldBlock).toBe(true);\n });\n\n it('should allow interruption when word count exactly meets threshold', () => {\n const text = 'hello world';\n const minInterruptionWords = 2;\n\n const normalizedText = text ?? '';\n const wordCount = splitWords(normalizedText, true).length;\n const shouldBlock = wordCount < minInterruptionWords;\n\n expect(normalizedText).toBe('hello world');\n expect(wordCount).toBe(2);\n expect(shouldBlock).toBe(false);\n });\n\n it('should allow interruption when word count exceeds threshold', () => {\n const text = 'hello this is a full sentence';\n const minInterruptionWords = 2;\n\n const normalizedText = text ?? '';\n const wordCount = splitWords(normalizedText, true).length;\n const shouldBlock = wordCount < minInterruptionWords;\n\n expect(normalizedText).toBe('hello this is a full sentence');\n expect(wordCount).toBe(6);\n expect(shouldBlock).toBe(false);\n });\n\n it('should apply consistent word counting logic in both methods', () => {\n const transcripts = ['', 'hello', 'hello world', 'this is a longer sentence'];\n const threshold = 2;\n\n transcripts.forEach((transcript) => {\n const text1 = transcript;\n const normalizedText1 = text1 ?? '';\n const wordCount1 = splitWords(normalizedText1, true).length;\n const shouldBlock1 = wordCount1 < threshold;\n\n const wordCount2 = splitWords(transcript, true).length;\n const shouldBlock2 = wordCount2 < threshold;\n\n expect(wordCount1).toBe(wordCount2);\n expect(shouldBlock1).toBe(shouldBlock2);\n });\n });\n });\n});\n"],"mappings":";AAYA,oBAAqC;AACrC,kBAA2B;AAAA,IAE3B,wBAAS,0CAA0C,MAAM;AACvD,8BAAS,2BAA2B,MAAM;AACxC,0BAAG,wCAAwC,MAAM;AAC/C,YAAM,OAAO;AACb,YAAM,gBAAY,wBAAW,MAAM,IAAI,EAAE;AACzC,gCAAO,SAAS,EAAE,KAAK,CAAC;AAAA,IAC1B,CAAC;AAED,0BAAG,sCAAsC,MAAM;AAC7C,YAAM,OAAO;AACb,YAAM,gBAAY,wBAAW,MAAM,IAAI,EAAE;AACzC,gCAAO,SAAS,EAAE,KAAK,CAAC;AAAA,IAC1B,CAAC;AAED,0BAAG,oCAAoC,MAAM;AAC3C,YAAM,OAAO;AACb,YAAM,gBAAY,wBAAW,MAAM,IAAI,EAAE;AACzC,gCAAO,SAAS,EAAE,KAAK,CAAC;AAAA,IAC1B,CAAC;AAED,0BAAG,yCAAyC,MAAM;AAChD,YAAM,OAAO;AACb,YAAM,gBAAY,wBAAW,MAAM,IAAI,EAAE;AACzC,gCAAO,SAAS,EAAE,KAAK,CAAC;AAAA,IAC1B,CAAC;AAED,0BAAG,uCAAuC,MAAM;AAC9C,YAAM,OAAO;AACb,YAAM,gBAAY,wBAAW,MAAM,IAAI,EAAE;AACzC,gCAAO,SAAS,EAAE,KAAK,CAAC;AAAA,IAC1B,CAAC;AAED,0BAAG,+CAA+C,MAAM;AACtD,YAAM,OAAO;AACb,YAAM,gBAAY,wBAAW,MAAM,IAAI,EAAE;AACzC,gCAAO,SAAS,EAAE,KAAK,CAAC;AAAA,IAC1B,CAAC;AAED,0BAAG,kDAAkD,MAAM;AACzD,YAAM,OAAO;AACb,YAAM,gBAAY,wBAAW,MAAM,IAAI,EAAE;AACzC,gCAAO,SAAS,EAAE,KAAK,CAAC;AAAA,IAC1B,CAAC;AAED,0BAAG,iDAAiD,MAAM;AACxD,YAAM,OAAO;AACb,YAAM,gBAAY,wBAAW,MAAM,IAAI,EAAE;AACzC,gCAAO,SAAS,EAAE,KAAK,CAAC;AAAA,IAC1B,CAAC;AAAA,EACH,CAAC;AAED,8BAAS,8CAA8C,MAAM;AAC3D,0BAAG,mEAAmE,MAAM;AAC1E,YAAM,OAAO;AACb,YAAM,uBAAuB;AAE7B,YAAM,iBAAiB,QAAQ;AAC/B,YAAM,gBAAY,wBAAW,gBAAgB,IAAI,EAAE;AACnD,YAAM,cAAc,YAAY;AAEhC,gCAAO,cAAc,EAAE,KAAK,EAAE;AAC9B,gCAAO,SAAS,EAAE,KAAK,CAAC;AACxB,gCAAO,WAAW,EAAE,KAAK,IAAI;AAAA,IAC/B,CAAC;AAED,0BAAG,uEAAuE,MAAM;AAC9E,YAAM,OAA2B;AACjC,YAAM,uBAAuB;AAE7B,YAAM,iBAAiB,QAAQ;AAC/B,YAAM,gBAAY,wBAAW,gBAAgB,IAAI,EAAE;AACnD,YAAM,cAAc,YAAY;AAEhC,gCAAO,cAAc,EAAE,KAAK,EAAE;AAC9B,gCAAO,SAAS,EAAE,KAAK,CAAC;AACxB,gCAAO,WAAW,EAAE,KAAK,IAAI;AAAA,IAC/B,CAAC;AAED,0BAAG,8DAA8D,MAAM;AACrE,YAAM,OAAO;AACb,YAAM,uBAAuB;AAE7B,YAAM,iBAAiB,QAAQ;AAC/B,YAAM,gBAAY,wBAAW,gBAAgB,IAAI,EAAE;AACnD,YAAM,cAAc,YAAY;AAEhC,gCAAO,cAAc,EAAE,KAAK,OAAO;AACnC,gCAAO,SAAS,EAAE,KAAK,CAAC;AACxB,gCAAO,WAAW,EAAE,KAAK,IAAI;AAAA,IAC/B,CAAC;AAED,0BAAG,qEAAqE,MAAM;AAC5E,YAAM,OAAO;AACb,YAAM,uBAAuB;AAE7B,YAAM,iBAAiB,QAAQ;AAC/B,YAAM,gBAAY,wBAAW,gBAAgB,IAAI,EAAE;AACnD,YAAM,cAAc,YAAY;AAEhC,gCAAO,cAAc,EAAE,KAAK,aAAa;AACzC,gCAAO,SAAS,EAAE,KAAK,CAAC;AACxB,gCAAO,WAAW,EAAE,KAAK,KAAK;AAAA,IAChC,CAAC;AAED,0BAAG,+DAA+D,MAAM;AACtE,YAAM,OAAO;AACb,YAAM,uBAAuB;AAE7B,YAAM,iBAAiB,QAAQ;AAC/B,YAAM,gBAAY,wBAAW,gBAAgB,IAAI,EAAE;AACnD,YAAM,cAAc,YAAY;AAEhC,gCAAO,cAAc,EAAE,KAAK,+BAA+B;AAC3D,gCAAO,SAAS,EAAE,KAAK,CAAC;AACxB,gCAAO,WAAW,EAAE,KAAK,KAAK;AAAA,IAChC,CAAC;AAED,0BAAG,+DAA+D,MAAM;AACtE,YAAM,cAAc,CAAC,IAAI,SAAS,eAAe,2BAA2B;AAC5E,YAAM,YAAY;AAElB,kBAAY,QAAQ,CAAC,eAAe;AAClC,cAAM,QAAQ;AACd,cAAM,kBAAkB,SAAS;AACjC,cAAM,iBAAa,wBAAW,iBAAiB,IAAI,EAAE;AACrD,cAAM,eAAe,aAAa;AAElC,cAAM,iBAAa,wBAAW,YAAY,IAAI,EAAE;AAChD,cAAM,eAAe,aAAa;AAElC,kCAAO,UAAU,EAAE,KAAK,UAAU;AAClC,kCAAO,YAAY,EAAE,KAAK,YAAY;AAAA,MACxC,CAAC;AAAA,IACH,CAAC;AAAA,EACH,CAAC;AACH,CAAC;","names":[]}
@@ -0,0 +1,113 @@
1
+ import { describe, expect, it } from "vitest";
2
+ import { splitWords } from "../tokenize/basic/word.js";
3
+ describe("Interruption Detection - Word Counting", () => {
4
+ describe("Word Splitting Behavior", () => {
5
+ it("should count empty string as 0 words", () => {
6
+ const text = "";
7
+ const wordCount = splitWords(text, true).length;
8
+ expect(wordCount).toBe(0);
9
+ });
10
+ it("should count single word correctly", () => {
11
+ const text = "hello";
12
+ const wordCount = splitWords(text, true).length;
13
+ expect(wordCount).toBe(1);
14
+ });
15
+ it("should count two words correctly", () => {
16
+ const text = "hello world";
17
+ const wordCount = splitWords(text, true).length;
18
+ expect(wordCount).toBe(2);
19
+ });
20
+ it("should count multiple words correctly", () => {
21
+ const text = "hello this is a full sentence";
22
+ const wordCount = splitWords(text, true).length;
23
+ expect(wordCount).toBe(6);
24
+ });
25
+ it("should handle punctuation correctly", () => {
26
+ const text = "hello, world!";
27
+ const wordCount = splitWords(text, true).length;
28
+ expect(wordCount).toBe(2);
29
+ });
30
+ it("should handle multiple spaces between words", () => {
31
+ const text = "hello world";
32
+ const wordCount = splitWords(text, true).length;
33
+ expect(wordCount).toBe(2);
34
+ });
35
+ it("should count whitespace-only string as 0 words", () => {
36
+ const text = " ";
37
+ const wordCount = splitWords(text, true).length;
38
+ expect(wordCount).toBe(0);
39
+ });
40
+ it("should handle leading and trailing whitespace", () => {
41
+ const text = " hello world ";
42
+ const wordCount = splitWords(text, true).length;
43
+ expect(wordCount).toBe(2);
44
+ });
45
+ });
46
+ describe("Integration: Full Interruption Check Logic", () => {
47
+ it("should block interruption for empty transcript with threshold 2", () => {
48
+ const text = "";
49
+ const minInterruptionWords = 2;
50
+ const normalizedText = text ?? "";
51
+ const wordCount = splitWords(normalizedText, true).length;
52
+ const shouldBlock = wordCount < minInterruptionWords;
53
+ expect(normalizedText).toBe("");
54
+ expect(wordCount).toBe(0);
55
+ expect(shouldBlock).toBe(true);
56
+ });
57
+ it("should block interruption for undefined transcript with threshold 2", () => {
58
+ const text = void 0;
59
+ const minInterruptionWords = 2;
60
+ const normalizedText = text ?? "";
61
+ const wordCount = splitWords(normalizedText, true).length;
62
+ const shouldBlock = wordCount < minInterruptionWords;
63
+ expect(normalizedText).toBe("");
64
+ expect(wordCount).toBe(0);
65
+ expect(shouldBlock).toBe(true);
66
+ });
67
+ it("should block interruption for single word with threshold 2", () => {
68
+ const text = "hello";
69
+ const minInterruptionWords = 2;
70
+ const normalizedText = text ?? "";
71
+ const wordCount = splitWords(normalizedText, true).length;
72
+ const shouldBlock = wordCount < minInterruptionWords;
73
+ expect(normalizedText).toBe("hello");
74
+ expect(wordCount).toBe(1);
75
+ expect(shouldBlock).toBe(true);
76
+ });
77
+ it("should allow interruption when word count exactly meets threshold", () => {
78
+ const text = "hello world";
79
+ const minInterruptionWords = 2;
80
+ const normalizedText = text ?? "";
81
+ const wordCount = splitWords(normalizedText, true).length;
82
+ const shouldBlock = wordCount < minInterruptionWords;
83
+ expect(normalizedText).toBe("hello world");
84
+ expect(wordCount).toBe(2);
85
+ expect(shouldBlock).toBe(false);
86
+ });
87
+ it("should allow interruption when word count exceeds threshold", () => {
88
+ const text = "hello this is a full sentence";
89
+ const minInterruptionWords = 2;
90
+ const normalizedText = text ?? "";
91
+ const wordCount = splitWords(normalizedText, true).length;
92
+ const shouldBlock = wordCount < minInterruptionWords;
93
+ expect(normalizedText).toBe("hello this is a full sentence");
94
+ expect(wordCount).toBe(6);
95
+ expect(shouldBlock).toBe(false);
96
+ });
97
+ it("should apply consistent word counting logic in both methods", () => {
98
+ const transcripts = ["", "hello", "hello world", "this is a longer sentence"];
99
+ const threshold = 2;
100
+ transcripts.forEach((transcript) => {
101
+ const text1 = transcript;
102
+ const normalizedText1 = text1 ?? "";
103
+ const wordCount1 = splitWords(normalizedText1, true).length;
104
+ const shouldBlock1 = wordCount1 < threshold;
105
+ const wordCount2 = splitWords(transcript, true).length;
106
+ const shouldBlock2 = wordCount2 < threshold;
107
+ expect(wordCount1).toBe(wordCount2);
108
+ expect(shouldBlock1).toBe(shouldBlock2);
109
+ });
110
+ });
111
+ });
112
+ });
113
+ //# sourceMappingURL=interruption_detection.test.js.map