@livekit/agents 1.0.38 → 1.0.39

package/dist/voice/agent.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../../src/voice/agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { AsyncLocalStorage } from 'node:async_hooks';\nimport { 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 { ReadonlyChatContext } from '../llm/chat_context.js';\nimport type { ChatMessage, FunctionCall, RealtimeModel } from '../llm/index.js';\nimport {\n  type ChatChunk,\n  ChatContext,\n  LLM,\n  type ToolChoice,\n  type ToolContext,\n} from '../llm/index.js';\nimport type { STT, SpeechEvent } from '../stt/index.js';\nimport { StreamAdapter as STTStreamAdapter } from '../stt/index.js';\nimport { SentenceTokenizer as BasicSentenceTokenizer } from '../tokenize/basic/index.js';\nimport type { TTS } from '../tts/index.js';\nimport { SynthesizeStream, StreamAdapter as TTSStreamAdapter } from '../tts/index.js';\nimport type { VAD } from '../vad.js';\nimport type { AgentActivity } from './agent_activity.js';\nimport type { AgentSession, TurnDetectionMode } from './agent_session.js';\n\nexport const asyncLocalStorage = new AsyncLocalStorage<{ functionCall?: FunctionCall }>();\nexport const STOP_RESPONSE_SYMBOL = Symbol('StopResponse');\n\nexport class StopResponse extends Error {\n  constructor() {\n    super();\n    this.name = 'StopResponse';\n\n    Object.defineProperty(this, STOP_RESPONSE_SYMBOL, {\n      value: true,\n    });\n  }\n}\n\nexport function isStopResponse(value: unknown): value is StopResponse {\n  return (\n    value !== undefined &&\n    value !== null &&\n    typeof value === 'object' &&\n    STOP_RESPONSE_SYMBOL in value\n  );\n}\n\nexport interface ModelSettings {\n  /** The tool choice to use when calling the LLM. */\n  toolChoice?: ToolChoice;\n}\n\nexport interface AgentOptions<UserData> {\n  id?: string;\n  instructions: string;\n  chatCtx?: ChatContext;\n  tools?: ToolContext<UserData>;\n  turnDetection?: TurnDetectionMode;\n  stt?: STT | STTModelString;\n  vad?: VAD;\n  llm?: LLM | RealtimeModel | LLMModels;\n  tts?: TTS | TTSModelString;\n  allowInterruptions?: boolean;\n  minConsecutiveSpeechDelay?: number;\n}\n\nexport class Agent<UserData = any> {\n  private _id: string;\n  private turnDetection?: TurnDetectionMode;\n  private _stt?: STT;\n  private _vad?: VAD;\n  private _llm?: LLM | RealtimeModel;\n  private _tts?: TTS;\n\n  /** @internal */\n  _agentActivity?: AgentActivity;\n\n  /** @internal */\n  _chatCtx: ChatContext;\n\n  /** @internal */\n  _instructions: string;\n\n  /** @internal */\n  _tools?: ToolContext<UserData>;\n\n  constructor({\n    id,\n    instructions,\n    chatCtx,\n    tools,\n    turnDetection,\n    stt,\n    vad,\n    llm,\n    tts,\n  }: AgentOptions<UserData>) {\n    if (id) {\n      this._id = id;\n    } else {\n      // Convert class name to snake_case\n      const className = this.constructor.name;\n      if (className === 'Agent') {\n        this._id = 'default_agent';\n      } else {\n        this._id = className\n          .replace(/([A-Z])/g, '_$1')\n          .toLowerCase()\n          .replace(/^_/, '');\n      }\n    }\n\n    this._instructions = instructions;\n    this._tools = { ...tools };\n    this._chatCtx = chatCtx\n      ? chatCtx.copy({\n          toolCtx: this._tools,\n        })\n      : ChatContext.empty();\n\n    this.turnDetection = turnDetection;\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._agentActivity = undefined;\n  }\n\n  get vad(): VAD | undefined {\n    return this._vad;\n  }\n\n  get stt(): STT | undefined {\n    return this._stt;\n  }\n\n  get llm(): LLM | RealtimeModel | undefined {\n    return this._llm;\n  }\n\n  get tts(): TTS | undefined {\n    return this._tts;\n  }\n\n  get chatCtx(): ReadonlyChatContext {\n    return new ReadonlyChatContext(this._chatCtx.items);\n  }\n\n  get id(): string {\n    return this._id;\n  }\n\n  get instructions(): string {\n    return this._instructions;\n  }\n\n  get toolCtx(): ToolContext<UserData> {\n    return { ...this._tools };\n  }\n\n  get session(): AgentSession<UserData> {\n    return this.getActivityOrThrow().agentSession as AgentSession<UserData>;\n  }\n\n  async onEnter(): Promise<void> {}\n\n  async onExit(): Promise<void> {}\n\n  async transcriptionNode(\n    text: ReadableStream<string>,\n    modelSettings: ModelSettings,\n  ): Promise<ReadableStream<string> | null> {\n    return Agent.default.transcriptionNode(this, text, modelSettings);\n  }\n\n  async onUserTurnCompleted(_chatCtx: ChatContext, _newMessage: ChatMessage): Promise<void> {}\n\n  async sttNode(\n    audio: ReadableStream<AudioFrame>,\n    modelSettings: ModelSettings,\n  ): Promise<ReadableStream<SpeechEvent | string> | null> {\n    return Agent.default.sttNode(this, audio, modelSettings);\n  }\n\n  async llmNode(\n    chatCtx: ChatContext,\n    toolCtx: ToolContext,\n    modelSettings: ModelSettings,\n  ): Promise<ReadableStream<ChatChunk | string> | null> {\n    return Agent.default.llmNode(this, chatCtx, toolCtx, modelSettings);\n  }\n\n  async ttsNode(\n    text: ReadableStream<string>,\n    modelSettings: ModelSettings,\n  ): Promise<ReadableStream<AudioFrame> | null> {\n    return Agent.default.ttsNode(this, text, modelSettings);\n  }\n\n  async realtimeAudioOutputNode(\n    audio: ReadableStream<AudioFrame>,\n    modelSettings: ModelSettings,\n  ): Promise<ReadableStream<AudioFrame> | null> {\n    return Agent.default.realtimeAudioOutputNode(this, audio, modelSettings);\n  }\n\n  // realtime_audio_output_node\n\n  getActivityOrThrow(): AgentActivity {\n    if (!this._agentActivity) {\n      throw new Error('Agent activity not found');\n    }\n    return this._agentActivity;\n  }\n\n  async updateChatCtx(chatCtx: ChatContext): Promise<void> {\n    if (!this._agentActivity) {\n      this._chatCtx = chatCtx.copy({ toolCtx: this.toolCtx });\n      return;\n    }\n\n    this._agentActivity.updateChatCtx(chatCtx);\n  }\n\n  static default = {\n    async sttNode(\n      agent: Agent,\n      audio: ReadableStream<AudioFrame>,\n      _modelSettings: ModelSettings,\n    ): Promise<ReadableStream<SpeechEvent | string> | null> {\n      const activity = agent.getActivityOrThrow();\n      if (!activity.stt) {\n        throw new Error('sttNode called but no STT node is available');\n      }\n\n      let wrapped_stt = activity.stt;\n\n      if (!wrapped_stt.capabilities.streaming) {\n        const vad = agent.vad || activity.vad;\n        if (!vad) {\n          throw new Error(\n            'STT does not support streaming, add a VAD to the AgentTask/VoiceAgent to enable streaming',\n          );\n        }\n        wrapped_stt = new STTStreamAdapter(wrapped_stt, vad);\n      }\n\n      const connOptions = activity.agentSession.connOptions.sttConnOptions;\n      const stream = wrapped_stt.stream({ connOptions });\n\n      // Set startTimeOffset to provide linear timestamps across reconnections\n      const audioInputStartedAt =\n        activity.agentSession._recorderIO?.recordingStartedAt ?? // Use recording start time if available\n        activity.agentSession._startedAt ?? // Fallback to session start time\n        Date.now(); // Fallback to current time\n\n      stream.startTimeOffset = (Date.now() - audioInputStartedAt) / 1000;\n\n      stream.updateInputStream(audio);\n\n      let cleaned = false;\n      const cleanup = () => {\n        if (cleaned) return;\n        cleaned = true;\n        stream.detachInputStream();\n        stream.close();\n      };\n\n      return new ReadableStream({\n        async start(controller) {\n          try {\n            for await (const event of stream) {\n              controller.enqueue(event);\n            }\n            controller.close();\n          } finally {\n            // Always clean up the STT stream, whether it ends naturally or is cancelled\n            cleanup();\n          }\n        },\n        cancel() {\n          cleanup();\n        },\n      });\n    },\n\n    async llmNode(\n      agent: Agent,\n      chatCtx: ChatContext,\n      toolCtx: ToolContext,\n      modelSettings: ModelSettings,\n    ): Promise<ReadableStream<ChatChunk | string> | null> {\n      const activity = agent.getActivityOrThrow();\n      if (!activity.llm) {\n        throw new Error('llmNode called but no LLM node is available');\n      }\n\n      if (!(activity.llm instanceof LLM)) {\n        throw new Error(\n          'llmNode should only be used with LLM (non-multimodal/realtime APIs) nodes',\n        );\n      }\n\n      // TODO(brian): make parallelToolCalls configurable\n      const { toolChoice } = modelSettings;\n      const connOptions = activity.agentSession.connOptions.llmConnOptions;\n\n      const stream = activity.llm.chat({\n        chatCtx,\n        toolCtx,\n        toolChoice,\n        connOptions,\n        parallelToolCalls: true,\n      });\n\n      let cleaned = false;\n      const cleanup = () => {\n        if (cleaned) return;\n        cleaned = true;\n        stream.close();\n      };\n\n      return new ReadableStream({\n        async start(controller) {\n          try {\n            for await (const chunk of stream) {\n              controller.enqueue(chunk);\n            }\n            controller.close();\n          } finally {\n            cleanup();\n          }\n        },\n        cancel() {\n          cleanup();\n        },\n      });\n    },\n\n    async ttsNode(\n      agent: Agent,\n      text: ReadableStream<string>,\n      _modelSettings: ModelSettings,\n    ): Promise<ReadableStream<AudioFrame> | null> {\n      const activity = agent.getActivityOrThrow();\n      if (!activity.tts) {\n        throw new Error('ttsNode called but no TTS node is available');\n      }\n\n      let wrapped_tts = activity.tts;\n\n      if (!activity.tts.capabilities.streaming) {\n        wrapped_tts = new TTSStreamAdapter(wrapped_tts, new BasicSentenceTokenizer());\n      }\n\n      const connOptions = activity.agentSession.connOptions.ttsConnOptions;\n      const stream = wrapped_tts.stream({ connOptions });\n      stream.updateInputStream(text);\n\n      let cleaned = false;\n      const cleanup = () => {\n        if (cleaned) return;\n        cleaned = true;\n        stream.close();\n      };\n\n      return new ReadableStream({\n        async start(controller) {\n          try {\n            for await (const chunk of stream) {\n              if (chunk === SynthesizeStream.END_OF_STREAM) {\n                break;\n              }\n              controller.enqueue(chunk.frame);\n            }\n            controller.close();\n          } finally {\n            cleanup();\n          }\n        },\n        cancel() {\n          cleanup();\n        },\n      });\n    },\n\n    async transcriptionNode(\n      agent: Agent,\n      text: ReadableStream<string>,\n      _modelSettings: ModelSettings,\n    ): Promise<ReadableStream<string> | null> {\n      return text;\n    },\n\n    async realtimeAudioOutputNode(\n      _agent: Agent,\n      audio: ReadableStream<AudioFrame>,\n      _modelSettings: ModelSettings,\n    ): Promise<ReadableStream<AudioFrame> | null> {\n      return audio;\n    },\n  };\n}\n"],"mappings":"AAIA,SAAS,yBAAyB;AAClC,SAAS,sBAAsB;AAC/B;AAAA,EACE,OAAO;AAAA,EACP,OAAO;AAAA,EACP,OAAO;AAAA,OAIF;AACP,SAAS,2BAA2B;AAEpC;AAAA,EAEE;AAAA,EACA;AAAA,OAGK;AAEP,SAAS,iBAAiB,wBAAwB;AAClD,SAAS,qBAAqB,8BAA8B;AAE5D,SAAS,kBAAkB,iBAAiB,wBAAwB;AAK7D,MAAM,oBAAoB,IAAI,kBAAmD;AACjF,MAAM,uBAAuB,OAAO,cAAc;AAElD,MAAM,qBAAqB,MAAM;AAAA,EACtC,cAAc;AACZ,UAAM;AACN,SAAK,OAAO;AAEZ,WAAO,eAAe,MAAM,sBAAsB;AAAA,MAChD,OAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AAEO,SAAS,eAAe,OAAuC;AACpE,SACE,UAAU,UACV,UAAU,QACV,OAAO,UAAU,YACjB,wBAAwB;AAE5B;AAqBO,MAAM,MAAsB;AAAA,EACzB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAGR;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA;AAAA,EAEA,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAA2B;AACzB,QAAI,IAAI;AACN,WAAK,MAAM;AAAA,IACb,OAAO;AAEL,YAAM,YAAY,KAAK,YAAY;AACnC,UAAI,cAAc,SAAS;AACzB,aAAK,MAAM;AAAA,MACb,OAAO;AACL,aAAK,MAAM,UACR,QAAQ,YAAY,KAAK,EACzB,YAAY,EACZ,QAAQ,MAAM,EAAE;AAAA,MACrB;AAAA,IACF;AAEA,SAAK,gBAAgB;AACrB,SAAK,SAAS,EAAE,GAAG,MAAM;AACzB,SAAK,WAAW,UACZ,QAAQ,KAAK;AAAA,MACX,SAAS,KAAK;AAAA,IAChB,CAAC,IACD,YAAY,MAAM;AAEtB,SAAK,gBAAgB;AACrB,SAAK,OAAO;AAEZ,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,SAAK,iBAAiB;AAAA,EACxB;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuC;AACzC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAA+B;AACjC,WAAO,IAAI,oBAAoB,KAAK,SAAS,KAAK;AAAA,EACpD;AAAA,EAEA,IAAI,KAAa;AACf,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAiC;AACnC,WAAO,EAAE,GAAG,KAAK,OAAO;AAAA,EAC1B;AAAA,EAEA,IAAI,UAAkC;AACpC,WAAO,KAAK,mBAAmB,EAAE;AAAA,EACnC;AAAA,EAEA,MAAM,UAAyB;AAAA,EAAC;AAAA,EAEhC,MAAM,SAAwB;AAAA,EAAC;AAAA,EAE/B,MAAM,kBACJ,MACA,eACwC;AACxC,WAAO,MAAM,QAAQ,kBAAkB,MAAM,MAAM,aAAa;AAAA,EAClE;AAAA,EAEA,MAAM,oBAAoB,UAAuB,aAAyC;AAAA,EAAC;AAAA,EAE3F,MAAM,QACJ,OACA,eACsD;AACtD,WAAO,MAAM,QAAQ,QAAQ,MAAM,OAAO,aAAa;AAAA,EACzD;AAAA,EAEA,MAAM,QACJ,SACA,SACA,eACoD;AACpD,WAAO,MAAM,QAAQ,QAAQ,MAAM,SAAS,SAAS,aAAa;AAAA,EACpE;AAAA,EAEA,MAAM,QACJ,MACA,eAC4C;AAC5C,WAAO,MAAM,QAAQ,QAAQ,MAAM,MAAM,aAAa;AAAA,EACxD;AAAA,EAEA,MAAM,wBACJ,OACA,eAC4C;AAC5C,WAAO,MAAM,QAAQ,wBAAwB,MAAM,OAAO,aAAa;AAAA,EACzE;AAAA;AAAA,EAIA,qBAAoC;AAClC,QAAI,CAAC,KAAK,gBAAgB;AACxB,YAAM,IAAI,MAAM,0BAA0B;AAAA,IAC5C;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,cAAc,SAAqC;AACvD,QAAI,CAAC,KAAK,gBAAgB;AACxB,WAAK,WAAW,QAAQ,KAAK,EAAE,SAAS,KAAK,QAAQ,CAAC;AACtD;AAAA,IACF;AAEA,SAAK,eAAe,cAAc,OAAO;AAAA,EAC3C;AAAA,EAEA,OAAO,UAAU;AAAA,IACf,MAAM,QACJ,OACA,OACA,gBACsD;AA7P5D;AA8PM,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,cAAc,SAAS;AAE3B,UAAI,CAAC,YAAY,aAAa,WAAW;AACvC,cAAM,MAAM,MAAM,OAAO,SAAS;AAClC,YAAI,CAAC,KAAK;AACR,gBAAM,IAAI;AAAA,YACR;AAAA,UACF;AAAA,QACF;AACA,sBAAc,IAAI,iBAAiB,aAAa,GAAG;AAAA,MACrD;AAEA,YAAM,cAAc,SAAS,aAAa,YAAY;AACtD,YAAM,SAAS,YAAY,OAAO,EAAE,YAAY,CAAC;AAGjD,YAAM,wBACJ,cAAS,aAAa,gBAAtB,mBAAmC;AAAA,MACnC,SAAS,aAAa;AAAA,MACtB,KAAK,IAAI;AAEX,aAAO,mBAAmB,KAAK,IAAI,IAAI,uBAAuB;AAE9D,aAAO,kBAAkB,KAAK;AAE9B,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,kBAAkB;AACzB,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,yBAAW,QAAQ,KAAK;AAAA,YAC1B;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AAEA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,QACJ,OACA,SACA,SACA,eACoD;AACpD,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,EAAE,SAAS,eAAe,MAAM;AAClC,cAAM,IAAI;AAAA,UACR;AAAA,QACF;AAAA,MACF;AAGA,YAAM,EAAE,WAAW,IAAI;AACvB,YAAM,cAAc,SAAS,aAAa,YAAY;AAEtD,YAAM,SAAS,SAAS,IAAI,KAAK;AAAA,QAC/B;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA,mBAAmB;AAAA,MACrB,CAAC;AAED,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,yBAAW,QAAQ,KAAK;AAAA,YAC1B;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AACA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,QACJ,OACA,MACA,gBAC4C;AAC5C,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,cAAc,SAAS;AAE3B,UAAI,CAAC,SAAS,IAAI,aAAa,WAAW;AACxC,sBAAc,IAAI,iBAAiB,aAAa,IAAI,uBAAuB,CAAC;AAAA,MAC9E;AAEA,YAAM,cAAc,SAAS,aAAa,YAAY;AACtD,YAAM,SAAS,YAAY,OAAO,EAAE,YAAY,CAAC;AACjD,aAAO,kBAAkB,IAAI;AAE7B,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,kBAAI,UAAU,iBAAiB,eAAe;AAC5C;AAAA,cACF;AACA,yBAAW,QAAQ,MAAM,KAAK;AAAA,YAChC;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AACA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,kBACJ,OACA,MACA,gBACwC;AACxC,aAAO;AAAA,IACT;AAAA,IAEA,MAAM,wBACJ,QACA,OACA,gBAC4C;AAC5C,aAAO;AAAA,IACT;AAAA,EACF;AACF;","names":[]}
+{"version":3,"sources":["../../src/voice/agent.ts"],"sourcesContent":["// SPDX-FileCopyrightText: 2025 LiveKit, Inc.\n//\n// SPDX-License-Identifier: Apache-2.0\nimport type { AudioFrame } from '@livekit/rtc-node';\nimport { AsyncLocalStorage } from 'node:async_hooks';\nimport { 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 { ReadonlyChatContext } from '../llm/chat_context.js';\nimport type { ChatMessage, FunctionCall, RealtimeModel } from '../llm/index.js';\nimport {\n  type ChatChunk,\n  ChatContext,\n  LLM,\n  type ToolChoice,\n  type ToolContext,\n} from '../llm/index.js';\nimport type { STT, SpeechEvent } from '../stt/index.js';\nimport { StreamAdapter as STTStreamAdapter } from '../stt/index.js';\nimport { SentenceTokenizer as BasicSentenceTokenizer } from '../tokenize/basic/index.js';\nimport type { TTS } from '../tts/index.js';\nimport { SynthesizeStream, StreamAdapter as TTSStreamAdapter } from '../tts/index.js';\nimport type { VAD } from '../vad.js';\nimport type { AgentActivity } from './agent_activity.js';\nimport type { AgentSession, TurnDetectionMode } from './agent_session.js';\n\nexport const asyncLocalStorage = new AsyncLocalStorage<{ functionCall?: FunctionCall }>();\nexport const STOP_RESPONSE_SYMBOL = Symbol('StopResponse');\n\nexport class StopResponse extends Error {\n  constructor() {\n    super();\n    this.name = 'StopResponse';\n\n    Object.defineProperty(this, STOP_RESPONSE_SYMBOL, {\n      value: true,\n    });\n  }\n}\n\nexport function isStopResponse(value: unknown): value is StopResponse {\n  return (\n    value !== undefined &&\n    value !== null &&\n    typeof value === 'object' &&\n    STOP_RESPONSE_SYMBOL in value\n  );\n}\n\nexport interface ModelSettings {\n  /** The tool choice to use when calling the LLM. */\n  toolChoice?: ToolChoice;\n}\n\nexport interface AgentOptions<UserData> {\n  id?: string;\n  instructions: string;\n  chatCtx?: ChatContext;\n  tools?: ToolContext<UserData>;\n  turnDetection?: TurnDetectionMode;\n  stt?: STT | STTModelString;\n  vad?: VAD;\n  llm?: LLM | RealtimeModel | LLMModels;\n  tts?: TTS | TTSModelString;\n  allowInterruptions?: boolean;\n  minConsecutiveSpeechDelay?: number;\n}\n\nexport class Agent<UserData = any> {\n  private _id: string;\n  private turnDetection?: TurnDetectionMode;\n  private _stt?: STT;\n  private _vad?: VAD;\n  private _llm?: LLM | RealtimeModel;\n  private _tts?: TTS;\n\n  /** @internal */\n  _agentActivity?: AgentActivity;\n\n  /** @internal */\n  _chatCtx: ChatContext;\n\n  /** @internal */\n  _instructions: string;\n\n  /** @internal */\n  _tools?: ToolContext<UserData>;\n\n  constructor({\n    id,\n    instructions,\n    chatCtx,\n    tools,\n    turnDetection,\n    stt,\n    vad,\n    llm,\n    tts,\n  }: AgentOptions<UserData>) {\n    if (id) {\n      this._id = id;\n    } else {\n      // Convert class name to snake_case\n      const className = this.constructor.name;\n      if (className === 'Agent') {\n        this._id = 'default_agent';\n      } else {\n        this._id = className\n          .replace(/([A-Z])/g, '_$1')\n          .toLowerCase()\n          .replace(/^_/, '');\n      }\n    }\n\n    this._instructions = instructions;\n    this._tools = { ...tools };\n    this._chatCtx = chatCtx\n      ? chatCtx.copy({\n          toolCtx: this._tools,\n        })\n      : ChatContext.empty();\n\n    this.turnDetection = turnDetection;\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._agentActivity = undefined;\n  }\n\n  get vad(): VAD | undefined {\n    return this._vad;\n  }\n\n  get stt(): STT | undefined {\n    return this._stt;\n  }\n\n  get llm(): LLM | RealtimeModel | undefined {\n    return this._llm;\n  }\n\n  get tts(): TTS | undefined {\n    return this._tts;\n  }\n\n  get chatCtx(): ReadonlyChatContext {\n    return new ReadonlyChatContext(this._chatCtx.items);\n  }\n\n  get id(): string {\n    return this._id;\n  }\n\n  get instructions(): string {\n    return this._instructions;\n  }\n\n  get toolCtx(): ToolContext<UserData> {\n    return { ...this._tools };\n  }\n\n  get session(): AgentSession<UserData> {\n    return this.getActivityOrThrow().agentSession as AgentSession<UserData>;\n  }\n\n  async onEnter(): Promise<void> {}\n\n  async onExit(): Promise<void> {}\n\n  async transcriptionNode(\n    text: ReadableStream<string>,\n    modelSettings: ModelSettings,\n  ): Promise<ReadableStream<string> | null> {\n    return Agent.default.transcriptionNode(this, text, modelSettings);\n  }\n\n  async onUserTurnCompleted(_chatCtx: ChatContext, _newMessage: ChatMessage): Promise<void> {}\n\n  async sttNode(\n    audio: ReadableStream<AudioFrame>,\n    modelSettings: ModelSettings,\n  ): Promise<ReadableStream<SpeechEvent | string> | null> {\n    return Agent.default.sttNode(this, audio, modelSettings);\n  }\n\n  async llmNode(\n    chatCtx: ChatContext,\n    toolCtx: ToolContext,\n    modelSettings: ModelSettings,\n  ): Promise<ReadableStream<ChatChunk | string> | null> {\n    return Agent.default.llmNode(this, chatCtx, toolCtx, modelSettings);\n  }\n\n  async ttsNode(\n    text: ReadableStream<string>,\n    modelSettings: ModelSettings,\n  ): Promise<ReadableStream<AudioFrame> | null> {\n    return Agent.default.ttsNode(this, text, modelSettings);\n  }\n\n  async realtimeAudioOutputNode(\n    audio: ReadableStream<AudioFrame>,\n    modelSettings: ModelSettings,\n  ): Promise<ReadableStream<AudioFrame> | null> {\n    return Agent.default.realtimeAudioOutputNode(this, audio, modelSettings);\n  }\n\n  // realtime_audio_output_node\n\n  getActivityOrThrow(): AgentActivity {\n    if (!this._agentActivity) {\n      throw new Error('Agent activity not found');\n    }\n    return this._agentActivity;\n  }\n\n  async updateChatCtx(chatCtx: ChatContext): Promise<void> {\n    if (!this._agentActivity) {\n      this._chatCtx = chatCtx.copy({ toolCtx: this.toolCtx });\n      return;\n    }\n\n    this._agentActivity.updateChatCtx(chatCtx);\n  }\n\n  static default = {\n    async sttNode(\n      agent: Agent,\n      audio: ReadableStream<AudioFrame>,\n      _modelSettings: ModelSettings,\n    ): Promise<ReadableStream<SpeechEvent | string> | null> {\n      const activity = agent.getActivityOrThrow();\n      if (!activity.stt) {\n        throw new Error('sttNode called but no STT node is available');\n      }\n\n      let wrapped_stt = activity.stt;\n\n      if (!wrapped_stt.capabilities.streaming) {\n        const vad = agent.vad || activity.vad;\n        if (!vad) {\n          throw new Error(\n            'STT does not support streaming, add a VAD to the AgentTask/VoiceAgent to enable streaming',\n          );\n        }\n        wrapped_stt = new STTStreamAdapter(wrapped_stt, vad);\n      }\n\n      const connOptions = activity.agentSession.connOptions.sttConnOptions;\n      const stream = wrapped_stt.stream({ connOptions });\n\n      // Set startTimeOffset to provide linear timestamps across reconnections\n      const audioInputStartedAt =\n        activity.agentSession._recorderIO?.recordingStartedAt ?? // Use recording start time if available\n        activity.agentSession._startedAt ?? // Fallback to session start time\n        Date.now(); // Fallback to current time\n\n      stream.startTimeOffset = (Date.now() - audioInputStartedAt) / 1000;\n\n      stream.updateInputStream(audio);\n\n      let cleaned = false;\n      const cleanup = () => {\n        if (cleaned) return;\n        cleaned = true;\n        stream.detachInputStream();\n        stream.close();\n      };\n\n      return new ReadableStream({\n        async start(controller) {\n          try {\n            for await (const event of stream) {\n              controller.enqueue(event);\n            }\n            controller.close();\n          } finally {\n            // Always clean up the STT stream, whether it ends naturally or is cancelled\n            cleanup();\n          }\n        },\n        cancel() {\n          cleanup();\n        },\n      });\n    },\n\n    async llmNode(\n      agent: Agent,\n      chatCtx: ChatContext,\n      toolCtx: ToolContext,\n      modelSettings: ModelSettings,\n    ): Promise<ReadableStream<ChatChunk | string> | null> {\n      const activity = agent.getActivityOrThrow();\n      if (!activity.llm) {\n        throw new Error('llmNode called but no LLM node is available');\n      }\n\n      if (!(activity.llm instanceof LLM)) {\n        throw new Error(\n          'llmNode should only be used with LLM (non-multimodal/realtime APIs) nodes',\n        );\n      }\n\n      const { toolChoice } = modelSettings;\n      const connOptions = activity.agentSession.connOptions.llmConnOptions;\n\n      // parallelToolCalls is not passed here - it will use the value from LLM's modelOptions\n      // This allows users to configure it via: new inference.LLM({ modelOptions: { parallel_tool_calls: false } })\n      const stream = activity.llm.chat({\n        chatCtx,\n        toolCtx,\n        toolChoice,\n        connOptions,\n      });\n\n      let cleaned = false;\n      const cleanup = () => {\n        if (cleaned) return;\n        cleaned = true;\n        stream.close();\n      };\n\n      return new ReadableStream({\n        async start(controller) {\n          try {\n            for await (const chunk of stream) {\n              controller.enqueue(chunk);\n            }\n            controller.close();\n          } finally {\n            cleanup();\n          }\n        },\n        cancel() {\n          cleanup();\n        },\n      });\n    },\n\n    async ttsNode(\n      agent: Agent,\n      text: ReadableStream<string>,\n      _modelSettings: ModelSettings,\n    ): Promise<ReadableStream<AudioFrame> | null> {\n      const activity = agent.getActivityOrThrow();\n      if (!activity.tts) {\n        throw new Error('ttsNode called but no TTS node is available');\n      }\n\n      let wrapped_tts = activity.tts;\n\n      if (!activity.tts.capabilities.streaming) {\n        wrapped_tts = new TTSStreamAdapter(wrapped_tts, new BasicSentenceTokenizer());\n      }\n\n      const connOptions = activity.agentSession.connOptions.ttsConnOptions;\n      const stream = wrapped_tts.stream({ connOptions });\n      stream.updateInputStream(text);\n\n      let cleaned = false;\n      const cleanup = () => {\n        if (cleaned) return;\n        cleaned = true;\n        stream.close();\n      };\n\n      return new ReadableStream({\n        async start(controller) {\n          try {\n            for await (const chunk of stream) {\n              if (chunk === SynthesizeStream.END_OF_STREAM) {\n                break;\n              }\n              controller.enqueue(chunk.frame);\n            }\n            controller.close();\n          } finally {\n            cleanup();\n          }\n        },\n        cancel() {\n          cleanup();\n        },\n      });\n    },\n\n    async transcriptionNode(\n      agent: Agent,\n      text: ReadableStream<string>,\n      _modelSettings: ModelSettings,\n    ): Promise<ReadableStream<string> | null> {\n      return text;\n    },\n\n    async realtimeAudioOutputNode(\n      _agent: Agent,\n      audio: ReadableStream<AudioFrame>,\n      _modelSettings: ModelSettings,\n    ): Promise<ReadableStream<AudioFrame> | null> {\n      return audio;\n    },\n  };\n}\n"],"mappings":"AAIA,SAAS,yBAAyB;AAClC,SAAS,sBAAsB;AAC/B;AAAA,EACE,OAAO;AAAA,EACP,OAAO;AAAA,EACP,OAAO;AAAA,OAIF;AACP,SAAS,2BAA2B;AAEpC;AAAA,EAEE;AAAA,EACA;AAAA,OAGK;AAEP,SAAS,iBAAiB,wBAAwB;AAClD,SAAS,qBAAqB,8BAA8B;AAE5D,SAAS,kBAAkB,iBAAiB,wBAAwB;AAK7D,MAAM,oBAAoB,IAAI,kBAAmD;AACjF,MAAM,uBAAuB,OAAO,cAAc;AAElD,MAAM,qBAAqB,MAAM;AAAA,EACtC,cAAc;AACZ,UAAM;AACN,SAAK,OAAO;AAEZ,WAAO,eAAe,MAAM,sBAAsB;AAAA,MAChD,OAAO;AAAA,IACT,CAAC;AAAA,EACH;AACF;AAEO,SAAS,eAAe,OAAuC;AACpE,SACE,UAAU,UACV,UAAU,QACV,OAAO,UAAU,YACjB,wBAAwB;AAE5B;AAqBO,MAAM,MAAsB;AAAA,EACzB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA;AAAA,EAGR;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA;AAAA;AAAA,EAGA;AAAA,EAEA,YAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF,GAA2B;AACzB,QAAI,IAAI;AACN,WAAK,MAAM;AAAA,IACb,OAAO;AAEL,YAAM,YAAY,KAAK,YAAY;AACnC,UAAI,cAAc,SAAS;AACzB,aAAK,MAAM;AAAA,MACb,OAAO;AACL,aAAK,MAAM,UACR,QAAQ,YAAY,KAAK,EACzB,YAAY,EACZ,QAAQ,MAAM,EAAE;AAAA,MACrB;AAAA,IACF;AAEA,SAAK,gBAAgB;AACrB,SAAK,SAAS,EAAE,GAAG,MAAM;AACzB,SAAK,WAAW,UACZ,QAAQ,KAAK;AAAA,MACX,SAAS,KAAK;AAAA,IAChB,CAAC,IACD,YAAY,MAAM;AAEtB,SAAK,gBAAgB;AACrB,SAAK,OAAO;AAEZ,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,QAAI,OAAO,QAAQ,UAAU;AAC3B,WAAK,OAAO,aAAa,gBAAgB,GAAG;AAAA,IAC9C,OAAO;AACL,WAAK,OAAO;AAAA,IACd;AAEA,SAAK,iBAAiB;AAAA,EACxB;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuC;AACzC,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,MAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAA+B;AACjC,WAAO,IAAI,oBAAoB,KAAK,SAAS,KAAK;AAAA,EACpD;AAAA,EAEA,IAAI,KAAa;AACf,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,eAAuB;AACzB,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,IAAI,UAAiC;AACnC,WAAO,EAAE,GAAG,KAAK,OAAO;AAAA,EAC1B;AAAA,EAEA,IAAI,UAAkC;AACpC,WAAO,KAAK,mBAAmB,EAAE;AAAA,EACnC;AAAA,EAEA,MAAM,UAAyB;AAAA,EAAC;AAAA,EAEhC,MAAM,SAAwB;AAAA,EAAC;AAAA,EAE/B,MAAM,kBACJ,MACA,eACwC;AACxC,WAAO,MAAM,QAAQ,kBAAkB,MAAM,MAAM,aAAa;AAAA,EAClE;AAAA,EAEA,MAAM,oBAAoB,UAAuB,aAAyC;AAAA,EAAC;AAAA,EAE3F,MAAM,QACJ,OACA,eACsD;AACtD,WAAO,MAAM,QAAQ,QAAQ,MAAM,OAAO,aAAa;AAAA,EACzD;AAAA,EAEA,MAAM,QACJ,SACA,SACA,eACoD;AACpD,WAAO,MAAM,QAAQ,QAAQ,MAAM,SAAS,SAAS,aAAa;AAAA,EACpE;AAAA,EAEA,MAAM,QACJ,MACA,eAC4C;AAC5C,WAAO,MAAM,QAAQ,QAAQ,MAAM,MAAM,aAAa;AAAA,EACxD;AAAA,EAEA,MAAM,wBACJ,OACA,eAC4C;AAC5C,WAAO,MAAM,QAAQ,wBAAwB,MAAM,OAAO,aAAa;AAAA,EACzE;AAAA;AAAA,EAIA,qBAAoC;AAClC,QAAI,CAAC,KAAK,gBAAgB;AACxB,YAAM,IAAI,MAAM,0BAA0B;AAAA,IAC5C;AACA,WAAO,KAAK;AAAA,EACd;AAAA,EAEA,MAAM,cAAc,SAAqC;AACvD,QAAI,CAAC,KAAK,gBAAgB;AACxB,WAAK,WAAW,QAAQ,KAAK,EAAE,SAAS,KAAK,QAAQ,CAAC;AACtD;AAAA,IACF;AAEA,SAAK,eAAe,cAAc,OAAO;AAAA,EAC3C;AAAA,EAEA,OAAO,UAAU;AAAA,IACf,MAAM,QACJ,OACA,OACA,gBACsD;AA7P5D;AA8PM,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,cAAc,SAAS;AAE3B,UAAI,CAAC,YAAY,aAAa,WAAW;AACvC,cAAM,MAAM,MAAM,OAAO,SAAS;AAClC,YAAI,CAAC,KAAK;AACR,gBAAM,IAAI;AAAA,YACR;AAAA,UACF;AAAA,QACF;AACA,sBAAc,IAAI,iBAAiB,aAAa,GAAG;AAAA,MACrD;AAEA,YAAM,cAAc,SAAS,aAAa,YAAY;AACtD,YAAM,SAAS,YAAY,OAAO,EAAE,YAAY,CAAC;AAGjD,YAAM,wBACJ,cAAS,aAAa,gBAAtB,mBAAmC;AAAA,MACnC,SAAS,aAAa;AAAA,MACtB,KAAK,IAAI;AAEX,aAAO,mBAAmB,KAAK,IAAI,IAAI,uBAAuB;AAE9D,aAAO,kBAAkB,KAAK;AAE9B,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,kBAAkB;AACzB,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,yBAAW,QAAQ,KAAK;AAAA,YAC1B;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AAEA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,QACJ,OACA,SACA,SACA,eACoD;AACpD,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,EAAE,SAAS,eAAe,MAAM;AAClC,cAAM,IAAI;AAAA,UACR;AAAA,QACF;AAAA,MACF;AAEA,YAAM,EAAE,WAAW,IAAI;AACvB,YAAM,cAAc,SAAS,aAAa,YAAY;AAItD,YAAM,SAAS,SAAS,IAAI,KAAK;AAAA,QAC/B;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF,CAAC;AAED,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,yBAAW,QAAQ,KAAK;AAAA,YAC1B;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AACA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,QACJ,OACA,MACA,gBAC4C;AAC5C,YAAM,WAAW,MAAM,mBAAmB;AAC1C,UAAI,CAAC,SAAS,KAAK;AACjB,cAAM,IAAI,MAAM,6CAA6C;AAAA,MAC/D;AAEA,UAAI,cAAc,SAAS;AAE3B,UAAI,CAAC,SAAS,IAAI,aAAa,WAAW;AACxC,sBAAc,IAAI,iBAAiB,aAAa,IAAI,uBAAuB,CAAC;AAAA,MAC9E;AAEA,YAAM,cAAc,SAAS,aAAa,YAAY;AACtD,YAAM,SAAS,YAAY,OAAO,EAAE,YAAY,CAAC;AACjD,aAAO,kBAAkB,IAAI;AAE7B,UAAI,UAAU;AACd,YAAM,UAAU,MAAM;AACpB,YAAI,QAAS;AACb,kBAAU;AACV,eAAO,MAAM;AAAA,MACf;AAEA,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,cAAI;AACF,6BAAiB,SAAS,QAAQ;AAChC,kBAAI,UAAU,iBAAiB,eAAe;AAC5C;AAAA,cACF;AACA,yBAAW,QAAQ,MAAM,KAAK;AAAA,YAChC;AACA,uBAAW,MAAM;AAAA,UACnB,UAAE;AACA,oBAAQ;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS;AACP,kBAAQ;AAAA,QACV;AAAA,MACF,CAAC;AAAA,IACH;AAAA,IAEA,MAAM,kBACJ,OACA,MACA,gBACwC;AACxC,aAAO;AAAA,IACT;AAAA,IAEA,MAAM,wBACJ,QACA,OACA,gBAC4C;AAC5C,aAAO;AAAA,IACT;AAAA,EACF;AACF;","names":[]}

package/dist/voice/agent_activity.cjs

@@ -122,9 +122,9 @@ class AgentActivity {
       );
       this.turnDetectionMode = void 0;
     }
-    if (!this.vad && this.stt && this.llm instanceof import_llm.LLM && this.allowInterruptions && this.turnDetectionMode === void 0) {
+    if (!this.vad && this.stt && !this.stt.capabilities.streaming && this.llm instanceof import_llm.LLM && this.allowInterruptions && this.turnDetectionMode === void 0) {
       this.logger.warn(
-        "VAD is not set. Enabling VAD is recommended when using LLM and STT for more responsive interruption handling."
+        "VAD is not set. Enabling VAD is recommended when using LLM and non-streaming STT for more responsive interruption handling."
       );
     }
   }
@@ -473,14 +473,16 @@ class AgentActivity {
     this.agentSession._updateUserState("listening", speechEndTime);
   }
   onVADInferenceDone(ev) {
-    var _a, _b;
     if (this.turnDetection === "manual" || this.turnDetection === "realtime_llm") {
       return;
     }
-    if (this.llm instanceof import_llm.RealtimeModel && this.llm.capabilities.turnDetection) {
-      return;
+    if (ev.speechDuration >= this.agentSession.options.minInterruptionDuration) {
+      this.interruptByAudioActivity();
     }
-    if (ev.speechDuration < this.agentSession.options.minInterruptionDuration) {
+  }
+  interruptByAudioActivity() {
+    var _a, _b;
+    if (this.llm instanceof import_llm.RealtimeModel && this.llm.capabilities.turnDetection) {
       return;
     }
     if (this.stt && this.agentSession.options.minInterruptionWords > 0 && this.audioRecognition) {
@@ -493,7 +495,10 @@ class AgentActivity {
     }
     (_a = this.realtimeSession) == null ? void 0 : _a.startUserActivity();
     if (this._currentSpeech && !this._currentSpeech.interrupted && this._currentSpeech.allowInterruptions) {
-      this.logger.info({ "speech id": this._currentSpeech.id }, "speech interrupted by VAD");
+      this.logger.info(
+        { "speech id": this._currentSpeech.id },
+        "speech interrupted by audio activity"
+      );
       (_b = this.realtimeSession) == null ? void 0 : _b.interrupt();
       this._currentSpeech.interrupt();
     }
@@ -511,6 +516,9 @@ class AgentActivity {
         // TODO(AJS-106): add multi participant support
       })
     );
+    if (ev.alternatives[0].text) {
+      this.interruptByAudioActivity();
+    }
   }
   onFinalTranscript(ev) {
     if (this.llm instanceof import_llm.RealtimeModel && this.llm.capabilities.userTranscription) {
@@ -525,6 +533,9 @@ class AgentActivity {
         // TODO(AJS-106): add multi participant support
       })
     );
+    if (this.audioRecognition && this.turnDetection !== "manual" && this.turnDetection !== "realtime_llm") {
+      this.interruptByAudioActivity();
+    }
   }
   onPreemptiveGeneration(info) {
     if (!this.agentSession.options.preemptiveGeneration || this.draining || this._currentSpeech !== void 0 && !this._currentSpeech.interrupted || !(this.llm instanceof import_llm.LLM)) {
@@ -1457,7 +1468,6 @@ ${instructions}` : instructions,
     await speechHandle.waitIfNotInterrupted(tasks.map((task) => task.result));
     if (audioOutput) {
       await speechHandle.waitIfNotInterrupted([audioOutput.waitForPlayout()]);
-      this.agentSession._updateAgentState("listening");
     }
     if (speechHandle.interrupted) {
       this.logger.debug(
@@ -1525,14 +1535,13 @@ ${instructions}` : instructions,
       this.agentSession._conversationItemAdded(message);
     }
     speechHandle._markGenerationDone();
-    toolOutput.firstToolStartedFuture.await.finally(() => {
-      this.agentSession._updateAgentState("thinking");
-    });
     await executeToolsTask.result;
+    if (toolOutput.output.length > 0) {
+      this.agentSession._updateAgentState("thinking");
+    } else if (this.agentSession.agentState === "speaking") {
+      this.agentSession._updateAgentState("listening");
+    }
     if (toolOutput.output.length === 0) {
-      if (!speechHandle.interrupted) {
-        this.agentSession._updateAgentState("listening");
-      }
       return;
     }
     const { maxToolSteps } = this.agentSession.options;