@livekit/agents 1.0.17 → 1.0.19

package/dist/voice/agent.js

@@ -28,6 +28,7 @@ function isStopResponse(value) {
   return value !== void 0 && value !== null && typeof value === "object" && STOP_RESPONSE_SYMBOL in value;
 }
 class Agent {
+  _id;
   turnDetection;
   _stt;
   _vad;
@@ -42,6 +43,7 @@ class Agent {
   /** @internal */
   _tools;
   constructor({
+    id,
     instructions,
     chatCtx,
     tools,
@@ -51,6 +53,16 @@ class Agent {
     llm,
     tts
   }) {
+    if (id) {
+      this._id = id;
+    } else {
+      const className = this.constructor.name;
+      if (className === "Agent") {
+        this._id = "default_agent";
+      } else {
+        this._id = className.replace(/([A-Z])/g, "_$1").toLowerCase().replace(/^_/, "");
+      }
+    }
     this._instructions = instructions;
     this._tools = { ...tools };
     this._chatCtx = chatCtx ? chatCtx.copy({
@@ -90,6 +102,9 @@ class Agent {
   get chatCtx() {
     return new ReadonlyChatContext(this._chatCtx.items);
   }
+  get id() {
+    return this._id;
+  }
   get instructions() {
     return this._instructions;
   }

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  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 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    instructions,\n    chatCtx,\n    tools,\n    turnDetection,\n    stt,\n    vad,\n    llm,\n    tts,\n  }: AgentOptions<UserData>) {\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 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        if (!agent.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, agent.vad);\n      }\n\n      const stream = wrapped_stt.stream();\n      stream.updateInputStream(audio);\n\n      return new ReadableStream({\n        async start(controller) {\n          for await (const event of stream) {\n            controller.enqueue(event);\n          }\n          controller.close();\n        },\n        cancel() {\n          stream.detachInputStream();\n          stream.close();\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\n      const stream = activity.llm.chat({\n        chatCtx,\n        toolCtx,\n        toolChoice,\n        parallelToolCalls: true,\n      });\n      return new ReadableStream({\n        async start(controller) {\n          for await (const chunk of stream) {\n            controller.enqueue(chunk);\n          }\n          controller.close();\n        },\n        cancel() {\n          stream.close();\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 stream = wrapped_tts.stream();\n      stream.updateInputStream(text);\n\n      return new ReadableStream({\n        async start(controller) {\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        },\n        cancel() {\n          stream.close();\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;AAoBO,MAAM,MAAsB;AAAA,EACzB;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,EACF,GAA2B;AACzB,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,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;AACtD,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,YAAI,CAAC,MAAM,KAAK;AACd,gBAAM,IAAI;AAAA,YACR;AAAA,UACF;AAAA,QACF;AACA,sBAAc,IAAI,iBAAiB,aAAa,MAAM,GAAG;AAAA,MAC3D;AAEA,YAAM,SAAS,YAAY,OAAO;AAClC,aAAO,kBAAkB,KAAK;AAE9B,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,uBAAW,QAAQ,KAAK;AAAA,UAC1B;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,kBAAkB;AACzB,iBAAO,MAAM;AAAA,QACf;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;AAEvB,YAAM,SAAS,SAAS,IAAI,KAAK;AAAA,QAC/B;AAAA,QACA;AAAA,QACA;AAAA,QACA,mBAAmB;AAAA,MACrB,CAAC;AACD,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,uBAAW,QAAQ,KAAK;AAAA,UAC1B;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,MAAM;AAAA,QACf;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,SAAS,YAAY,OAAO;AAClC,aAAO,kBAAkB,IAAI;AAE7B,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,gBAAI,UAAU,iBAAiB,eAAe;AAC5C;AAAA,YACF;AACA,uBAAW,QAAQ,MAAM,KAAK;AAAA,UAChC;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,MAAM;AAAA,QACf;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        if (!agent.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, agent.vad);\n      }\n\n      const stream = wrapped_stt.stream();\n      stream.updateInputStream(audio);\n\n      return new ReadableStream({\n        async start(controller) {\n          for await (const event of stream) {\n            controller.enqueue(event);\n          }\n          controller.close();\n        },\n        cancel() {\n          stream.detachInputStream();\n          stream.close();\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\n      const stream = activity.llm.chat({\n        chatCtx,\n        toolCtx,\n        toolChoice,\n        parallelToolCalls: true,\n      });\n      return new ReadableStream({\n        async start(controller) {\n          for await (const chunk of stream) {\n            controller.enqueue(chunk);\n          }\n          controller.close();\n        },\n        cancel() {\n          stream.close();\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 stream = wrapped_tts.stream();\n      stream.updateInputStream(text);\n\n      return new ReadableStream({\n        async start(controller) {\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        },\n        cancel() {\n          stream.close();\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;AACtD,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,YAAI,CAAC,MAAM,KAAK;AACd,gBAAM,IAAI;AAAA,YACR;AAAA,UACF;AAAA,QACF;AACA,sBAAc,IAAI,iBAAiB,aAAa,MAAM,GAAG;AAAA,MAC3D;AAEA,YAAM,SAAS,YAAY,OAAO;AAClC,aAAO,kBAAkB,KAAK;AAE9B,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,uBAAW,QAAQ,KAAK;AAAA,UAC1B;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,kBAAkB;AACzB,iBAAO,MAAM;AAAA,QACf;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;AAEvB,YAAM,SAAS,SAAS,IAAI,KAAK;AAAA,QAC/B;AAAA,QACA;AAAA,QACA;AAAA,QACA,mBAAmB;AAAA,MACrB,CAAC;AACD,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,uBAAW,QAAQ,KAAK;AAAA,UAC1B;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,MAAM;AAAA,QACf;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,SAAS,YAAY,OAAO;AAClC,aAAO,kBAAkB,IAAI;AAE7B,aAAO,IAAI,eAAe;AAAA,QACxB,MAAM,MAAM,YAAY;AACtB,2BAAiB,SAAS,QAAQ;AAChC,gBAAI,UAAU,iBAAiB,eAAe;AAC5C;AAAA,YACF;AACA,uBAAW,QAAQ,MAAM,KAAK;AAAA,UAChC;AACA,qBAAW,MAAM;AAAA,QACnB;AAAA,QACA,SAAS;AACP,iBAAO,MAAM;AAAA,QACf;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

@@ -155,6 +155,11 @@ class AgentActivity {
         } catch (error) {
           this.logger.error(error, "failed to update the tools");
         }
+        if (!this.llm.capabilities.audioOutput && !this.tts && this.agentSession.output.audio) {
+          this.logger.error(
+            "audio output is enabled but RealtimeModel has no audio modality and no TTS is set. Either enable audio modality in the RealtimeModel or set a TTS model."
+          );
+        }
       } else if (this.llm instanceof import_llm.LLM) {
         try {
           (0, import_generation.updateInstructions)({
@@ -452,7 +457,9 @@ class AgentActivity {
     }
     if (this.stt && this.agentSession.options.minInterruptionWords > 0 && this.audioRecognition) {
       const text = this.audioRecognition.currentTranscript;
-      if (text && (0, import_word.splitWords)(text, true).length < this.agentSession.options.minInterruptionWords) {
+      const normalizedText = text ?? "";
+      const wordCount = (0, import_word.splitWords)(normalizedText, true).length;
+      if (wordCount < this.agentSession.options.minInterruptionWords) {
         return;
       }
     }
@@ -554,10 +561,19 @@ class AgentActivity {
       this.logger.warn({ user_input: info.newTranscript }, "skipping user input, task is draining");
       return true;
     }
-    if (this.stt && this.turnDetection !== "manual" && this._currentSpeech && this._currentSpeech.allowInterruptions && !this._currentSpeech.interrupted && this.agentSession.options.minInterruptionWords > 0 && info.newTranscript.split(" ").length < this.agentSession.options.minInterruptionWords) {
-      this.cancelPreemptiveGeneration();
-      this.logger.info("skipping user input, new_transcript is too short");
-      return false;
+    if (this.stt && this.turnDetection !== "manual" && this._currentSpeech && this._currentSpeech.allowInterruptions && !this._currentSpeech.interrupted && this.agentSession.options.minInterruptionWords > 0) {
+      const wordCount = (0, import_word.splitWords)(info.newTranscript, true).length;
+      if (wordCount < this.agentSession.options.minInterruptionWords) {
+        this.cancelPreemptiveGeneration();
+        this.logger.info(
+          {
+            wordCount,
+            minInterruptionWords: this.agentSession.options.minInterruptionWords
+          },
+          "skipping user input, word count below minimum interruption threshold"
+        );
+        return false;
+      }
     }
     const oldTask = this._userTurnCompletedTask;
     this._userTurnCompletedTask = this.createSpeechTask({
@@ -887,6 +903,7 @@ ${instructions}` : instructions,
       this.agentSession._updateAgentState("listening");
     }
   }
+  // TODO(brian): PR3 - Wrap entire pipelineReplyTask() method with tracer.startActiveSpan('agent_turn')
   async pipelineReplyTask(speechHandle, chatCtx, toolCtx, modelSettings, replyAbortController, instructions, newMessage, toolsMessages) {
     var _a, _b, _c;
     speechHandleStorage.enterWith(speechHandle);
@@ -1200,7 +1217,22 @@ ${instructions}` : instructions,
             );
             break;
           }
-          const trNodeResult = await this.agent.transcriptionNode(msg.textStream, modelSettings);
+          const msgModalities = msg.modalities ? await msg.modalities : void 0;
+          let ttsTextInput = null;
+          let trTextInput;
+          if (msgModalities && !msgModalities.includes("audio") && this.tts) {
+            if (this.llm instanceof import_llm.RealtimeModel && this.llm.capabilities.audioOutput) {
+              this.logger.warn(
+                "text response received from realtime API, falling back to use a TTS model."
+              );
+            }
+            const [_ttsTextInput, _trTextInput] = msg.textStream.tee();
+            ttsTextInput = _ttsTextInput;
+            trTextInput = _trTextInput;
+          } else {
+            trTextInput = msg.textStream;
+          }
+          const trNodeResult = await this.agent.transcriptionNode(trTextInput, modelSettings);
           let textOut = null;
           if (trNodeResult) {
             const [textForwardTask, _textOut] = (0, import_generation.performTextForwarding)(
@@ -1213,28 +1245,44 @@ ${instructions}` : instructions,
           }
           let audioOut = null;
           if (audioOutput) {
-            const realtimeAudio = await this.agent.realtimeAudioOutputNode(
-              msg.audioStream,
-              modelSettings
-            );
-            if (realtimeAudio) {
+            let realtimeAudioResult = null;
+            if (ttsTextInput) {
+              const [ttsTask, ttsStream] = (0, import_generation.performTTSInference)(
+                (...args) => this.agent.ttsNode(...args),
+                ttsTextInput,
+                modelSettings,
+                abortController
+              );
+              tasks.push(ttsTask);
+              realtimeAudioResult = ttsStream;
+            } else if (msgModalities && msgModalities.includes("audio")) {
+              realtimeAudioResult = await this.agent.realtimeAudioOutputNode(
+                msg.audioStream,
+                modelSettings
+              );
+            } else if (this.llm instanceof import_llm.RealtimeModel && this.llm.capabilities.audioOutput) {
+              this.logger.error(
+                "Text message received from Realtime API with audio modality. This usually happens when text chat context is synced to the API. Try to add a TTS model as fallback or use text modality with TTS instead."
+              );
+            } else {
+              this.logger.warn(
+                "audio output is enabled but neither tts nor realtime audio is available"
+              );
+            }
+            if (realtimeAudioResult) {
               const [forwardTask, _audioOut] = (0, import_generation.performAudioForwarding)(
-                realtimeAudio,
+                realtimeAudioResult,
                 audioOutput,
                 abortController
               );
               forwardTasks.push(forwardTask);
               audioOut = _audioOut;
               audioOut.firstFrameFut.await.finally(onFirstFrame);
-            } else {
-              this.logger.warn(
-                "audio output is enabled but neither tts nor realtime audio is available"
-              );
             }
           } else if (textOut) {
             textOut.firstTextFut.await.finally(onFirstFrame);
           }
-          outputs.push([msg.messageId, textOut, audioOut]);
+          outputs.push([msg.messageId, textOut, audioOut, msgModalities]);
         }
         await (0, import_utils.waitFor)(forwardTasks);
       } catch (error) {
@@ -1304,7 +1352,7 @@ ${instructions}` : instructions,
       replyAbortController.abort();
       await (0, import_utils.cancelAndWait)(tasks, AgentActivity.REPLY_TASK_CANCEL_TIMEOUT);
       if (messageOutputs.length > 0) {
-        const [msgId, textOut, audioOut] = messageOutputs[0];
+        const [msgId, textOut, audioOut, msgModalities] = messageOutputs[0];
         let forwardedText = (textOut == null ? void 0 : textOut.text) || "";
         if (audioOutput) {
           audioOutput.clearBuffer();
@@ -1324,7 +1372,9 @@ ${instructions}` : instructions,
           }
           this.realtimeSession.truncate({
             messageId: msgId,
-            audioEndMs: Math.floor(playbackPosition)
+            audioEndMs: Math.floor(playbackPosition),
+            modalities: msgModalities,
+            audioTranscript: forwardedText
           });
         }
         if (forwardedText) {
@@ -1348,7 +1398,7 @@ ${instructions}` : instructions,
       return;
     }
     if (messageOutputs.length > 0) {
-      const [msgId, textOut, _] = messageOutputs[0];
+      const [msgId, textOut, _, __] = messageOutputs[0];
       const message = import_chat_context.ChatMessage.create({
         role: "assistant",
         content: (textOut == null ? void 0 : textOut.text) || "",
@@ -1510,6 +1560,7 @@ ${instructions}` : instructions,
     speechHandle._markScheduled();
     this.wakeupMainTask();
   }
+  // TODO(brian): PR3 - Wrap entire drain() method with tracer.startActiveSpan('drain_agent_activity', { attributes: { 'lk.agent_label': this.agent.label } })
   async drain() {
     var _a;
     const unlock = await this.lock.lock();