@livekit/agents 0.7.9 → 1.0.0-next.0

package/dist/voice/agent_activity.cjs

@@ -0,0 +1,1453 @@
+"use strict";
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+var agent_activity_exports = {};
+__export(agent_activity_exports, {
+  AgentActivity: () => AgentActivity
+});
+module.exports = __toCommonJS(agent_activity_exports);
+var import_mutex = require("@livekit/mutex");
+var import_heap_js = require("heap-js");
+var import_node_async_hooks = require("node:async_hooks");
+var import_web = require("node:stream/web");
+var import_chat_context = require("../llm/chat_context.cjs");
+var import_llm = require("../llm/index.cjs");
+var import_log = require("../log.cjs");
+var import_deferred_stream = require("../stream/deferred_stream.cjs");
+var import_stt = require("../stt/stt.cjs");
+var import_word = require("../tokenize/basic/word.cjs");
+var import_tts = require("../tts/tts.cjs");
+var import_utils = require("../utils.cjs");
+var import_vad = require("../vad.cjs");
+var import_agent = require("./agent.cjs");
+var import_agent_session = require("./agent_session.cjs");
+var import_audio_recognition = require("./audio_recognition.cjs");
+var import_events = require("./events.cjs");
+var import_generation = require("./generation.cjs");
+var import_speech_handle = require("./speech_handle.cjs");
+const speechHandleStorage = new import_node_async_hooks.AsyncLocalStorage();
+class AgentActivity {
+  static REPLY_TASK_CANCEL_TIMEOUT = 5e3;
+  started = false;
+  audioRecognition;
+  realtimeSession;
+  turnDetectionMode;
+  logger = (0, import_log.log)();
+  _draining = false;
+  _currentSpeech;
+  speechQueue;
+  // [priority, timestamp, speechHandle]
+  q_updated;
+  speechTasks = /* @__PURE__ */ new Set();
+  lock = new import_mutex.Mutex();
+  audioStream = new import_deferred_stream.DeferredReadableStream();
+  // default to null as None, which maps to the default provider tool choice value
+  toolChoice = null;
+  agent;
+  agentSession;
+  /** @internal */
+  _mainTask;
+  _userTurnCompletedTask;
+  constructor(agent, agentSession) {
+    this.agent = agent;
+    this.agentSession = agentSession;
+    this.speechQueue = new import_heap_js.Heap(([p1, t1, _], [p2, t2, __]) => {
+      return p1 === p2 ? t1 - t2 : p2 - p1;
+    });
+    this.q_updated = new import_utils.Future();
+    this.turnDetectionMode = typeof this.turnDetection === "string" ? this.turnDetection : void 0;
+    if (this.turnDetectionMode === "vad" && this.vad === void 0) {
+      this.logger.warn(
+        'turnDetection is set to "vad", but no VAD model is provided, ignoring the turnDdetection setting'
+      );
+      this.turnDetectionMode = void 0;
+    }
+    if (this.turnDetectionMode === "stt" && this.stt === void 0) {
+      this.logger.warn(
+        'turnDetection is set to "stt", but no STT model is provided, ignoring the turnDetection setting'
+      );
+      this.turnDetectionMode = void 0;
+    }
+    if (this.llm instanceof import_llm.RealtimeModel) {
+      if (this.llm.capabilities.turnDetection && !this.allowInterruptions) {
+        this.logger.warn(
+          "the RealtimeModel uses a server-side turn detection, allowInterruptions cannot be false, disable turnDetection in the RealtimeModel and use VAD on the AgentSession instead"
+        );
+      }
+      if (this.turnDetectionMode === "realtime_llm" && !this.llm.capabilities.turnDetection) {
+        this.logger.warn(
+          'turnDetection is set to "realtime_llm", but the LLM is not a RealtimeModel or the server-side turn detection is not supported/enabled, ignoring the turnDetection setting'
+        );
+        this.turnDetectionMode = void 0;
+      }
+      if (this.turnDetectionMode === "stt") {
+        this.logger.warn(
+          'turnDetection is set to "stt", but the LLM is a RealtimeModel, ignoring the turnDetection setting'
+        );
+        this.turnDetectionMode = void 0;
+      }
+      if (this.turnDetectionMode && this.turnDetectionMode !== "realtime_llm" && this.llm.capabilities.turnDetection) {
+        this.logger.warn(
+          `turnDetection is set to "${this.turnDetectionMode}", but the LLM is a RealtimeModel and server-side turn detection enabled, ignoring the turnDetection setting`
+        );
+        this.turnDetectionMode = void 0;
+      }
+      if (!this.llm.capabilities.turnDetection && this.vad && this.turnDetectionMode === void 0) {
+        this.turnDetectionMode = "vad";
+      }
+    } else if (this.turnDetectionMode === "realtime_llm") {
+      this.logger.warn(
+        'turnDetection is set to "realtime_llm", but the LLM is not a RealtimeModel'
+      );
+      this.turnDetectionMode = void 0;
+    }
+    if (!this.vad && this.stt && 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."
+      );
+    }
+  }
+  async start() {
+    const unlock = await this.lock.lock();
+    try {
+      this.agent._agentActivity = this;
+      if (this.llm instanceof import_llm.RealtimeModel) {
+        this.realtimeSession = this.llm.session();
+        this.realtimeSession.on("generation_created", (ev) => this.onGenerationCreated(ev));
+        this.realtimeSession.on("input_speech_started", (ev) => this.onInputSpeechStarted(ev));
+        this.realtimeSession.on("input_speech_stopped", (ev) => this.onInputSpeechStopped(ev));
+        this.realtimeSession.on(
+          "input_audio_transcription_completed",
+          (ev) => this.onInputAudioTranscriptionCompleted(ev)
+        );
+        this.realtimeSession.on("metrics_collected", (ev) => this.onMetricsCollected(ev));
+        this.realtimeSession.on("error", (ev) => this.onError(ev));
+        (0, import_generation.removeInstructions)(this.agent._chatCtx);
+        try {
+          await this.realtimeSession.updateInstructions(this.agent.instructions);
+        } catch (error) {
+          this.logger.error(error, "failed to update the instructions");
+        }
+        try {
+          await this.realtimeSession.updateChatCtx(this.agent.chatCtx);
+        } catch (error) {
+          this.logger.error(error, "failed to update the chat context");
+        }
+        try {
+          await this.realtimeSession.updateTools(this.tools);
+        } catch (error) {
+          this.logger.error(error, "failed to update the tools");
+        }
+      } else if (this.llm instanceof import_llm.LLM) {
+        try {
+          (0, import_generation.updateInstructions)({
+            chatCtx: this.agent._chatCtx,
+            instructions: this.agent.instructions,
+            addIfMissing: true
+          });
+        } catch (error) {
+          this.logger.error("failed to update the instructions", error);
+        }
+      }
+      if (this.llm instanceof import_llm.LLM) {
+        this.llm.on("metrics_collected", (ev) => this.onMetricsCollected(ev));
+        this.llm.on("error", (ev) => this.onError(ev));
+      }
+      if (this.stt instanceof import_stt.STT) {
+        this.stt.on("metrics_collected", (ev) => this.onMetricsCollected(ev));
+        this.stt.on("error", (ev) => this.onError(ev));
+      }
+      if (this.tts instanceof import_tts.TTS) {
+        this.tts.on("metrics_collected", (ev) => this.onMetricsCollected(ev));
+        this.tts.on("error", (ev) => this.onError(ev));
+      }
+      if (this.vad instanceof import_vad.VAD) {
+        this.vad.on("metrics_collected", (ev) => this.onMetricsCollected(ev));
+      }
+      this.audioRecognition = new import_audio_recognition.AudioRecognition({
+        recognitionHooks: this,
+        // Disable stt node if stt is not provided
+        stt: this.stt ? (...args) => this.agent.sttNode(...args) : void 0,
+        vad: this.vad,
+        turnDetector: typeof this.turnDetection === "string" ? void 0 : this.turnDetection,
+        turnDetectionMode: this.turnDetectionMode,
+        minEndpointingDelay: this.agentSession.options.minEndpointingDelay,
+        maxEndpointingDelay: this.agentSession.options.maxEndpointingDelay
+      });
+      this.audioRecognition.start();
+      this.started = true;
+      this._mainTask = import_utils.Task.from(({ signal }) => this.mainTask(signal));
+      this.createSpeechTask({
+        promise: this.agent.onEnter(),
+        name: "AgentActivity_onEnter"
+      });
+    } finally {
+      unlock();
+    }
+  }
+  get currentSpeech() {
+    return this._currentSpeech;
+  }
+  get vad() {
+    return this.agent.vad || this.agentSession.vad;
+  }
+  get stt() {
+    return this.agent.stt || this.agentSession.stt;
+  }
+  get llm() {
+    return this.agent.llm || this.agentSession.llm;
+  }
+  get tts() {
+    return this.agent.tts || this.agentSession.tts;
+  }
+  get tools() {
+    return this.agent.toolCtx;
+  }
+  get draining() {
+    return this._draining;
+  }
+  get realtimeLLMSession() {
+    return this.realtimeSession;
+  }
+  get allowInterruptions() {
+    return this.agentSession.options.allowInterruptions;
+  }
+  get turnDetection() {
+    return this.agentSession.turnDetection;
+  }
+  get toolCtx() {
+    return this.agent.toolCtx;
+  }
+  async updateChatCtx(chatCtx) {
+    chatCtx = chatCtx.copy({ toolCtx: this.toolCtx });
+    this.agent._chatCtx = chatCtx;
+    if (this.realtimeSession) {
+      (0, import_generation.removeInstructions)(chatCtx);
+      this.realtimeSession.updateChatCtx(chatCtx);
+    } else {
+      (0, import_generation.updateInstructions)({
+        chatCtx,
+        instructions: this.agent.instructions,
+        addIfMissing: true
+      });
+    }
+  }
+  updateOptions({ toolChoice }) {
+    if (toolChoice !== void 0) {
+      this.toolChoice = toolChoice;
+    }
+    if (this.realtimeSession) {
+      this.realtimeSession.updateOptions({ toolChoice: this.toolChoice });
+    }
+  }
+  attachAudioInput(audioStream) {
+    if (this.audioStream.isSourceSet) {
+      this.logger.debug("detaching existing audio input in agent activity");
+      this.audioStream.detachSource();
+    }
+    this.audioStream.setSource(audioStream);
+    const [realtimeAudioStream, recognitionAudioStream] = this.audioStream.stream.tee();
+    if (this.realtimeSession) {
+      this.realtimeSession.setInputAudioStream(realtimeAudioStream);
+    }
+    if (this.audioRecognition) {
+      this.audioRecognition.setInputAudioStream(recognitionAudioStream);
+    }
+  }
+  detachAudioInput() {
+    this.audioStream.detachSource();
+  }
+  commitUserTurn() {
+    if (!this.audioRecognition) {
+      throw new Error("AudioRecognition is not initialized");
+    }
+    const audioDetached = false;
+    this.audioRecognition.commitUserTurn(audioDetached);
+  }
+  clearUserTurn() {
+    var _a, _b;
+    (_a = this.audioRecognition) == null ? void 0 : _a.clearUserTurn();
+    (_b = this.realtimeSession) == null ? void 0 : _b.clearAudio();
+  }
+  say(text, options) {
+    const {
+      audio,
+      allowInterruptions: defaultAllowInterruptions,
+      addToChatCtx = true
+    } = options ?? {};
+    let allowInterruptions = defaultAllowInterruptions;
+    if (!audio && !this.tts && this.agentSession.output.audio && this.agentSession.output.audioEnabled) {
+      throw new Error("trying to generate speech from text without a TTS model");
+    }
+    if (this.llm instanceof import_llm.RealtimeModel && this.llm.capabilities.turnDetection && allowInterruptions === false) {
+      this.logger.warn(
+        "the RealtimeModel uses a server-side turn detection, allowInterruptions cannot be false when using VoiceAgent.say(), disable turnDetection in the RealtimeModel and use VAD on the AgentTask/VoiceAgent instead"
+      );
+      allowInterruptions = true;
+    }
+    const handle = import_speech_handle.SpeechHandle.create({
+      allowInterruptions: allowInterruptions ?? this.allowInterruptions
+    });
+    this.agentSession.emit(
+      import_events.AgentSessionEventTypes.SpeechCreated,
+      (0, import_events.createSpeechCreatedEvent)({
+        userInitiated: true,
+        source: "say",
+        speechHandle: handle
+      })
+    );
+    const task = this.createSpeechTask({
+      promise: this.ttsTask(handle, text, addToChatCtx, {}, audio),
+      ownedSpeechHandle: handle,
+      name: "AgentActivity.say_tts"
+    });
+    task.finally(() => this.onPipelineReplyDone());
+    this.scheduleSpeech(handle, import_speech_handle.SpeechHandle.SPEECH_PRIORITY_NORMAL);
+    return handle;
+  }
+  // -- Metrics and errors --
+  onMetricsCollected = (ev) => {
+    const speechHandle = speechHandleStorage.getStore();
+    if (speechHandle && (ev.type === "llm_metrics" || ev.type === "tts_metrics")) {
+      ev.speechId = speechHandle.id;
+    }
+    this.agentSession.emit(
+      import_events.AgentSessionEventTypes.MetricsCollected,
+      (0, import_events.createMetricsCollectedEvent)({ metrics: ev })
+    );
+  };
+  onError(ev) {
+    if (ev.type === "realtime_model_error") {
+      const errorEvent = (0, import_events.createErrorEvent)(ev.error, this.llm);
+      this.agentSession.emit(import_events.AgentSessionEventTypes.Error, errorEvent);
+    } else if (ev.type === "stt_error") {
+      const errorEvent = (0, import_events.createErrorEvent)(ev.error, this.stt);
+      this.agentSession.emit(import_events.AgentSessionEventTypes.Error, errorEvent);
+    } else if (ev.type === "tts_error") {
+      const errorEvent = (0, import_events.createErrorEvent)(ev.error, this.tts);
+      this.agentSession.emit(import_events.AgentSessionEventTypes.Error, errorEvent);
+    } else if (ev.type === "llm_error") {
+      const errorEvent = (0, import_events.createErrorEvent)(ev.error, this.llm);
+      this.agentSession.emit(import_events.AgentSessionEventTypes.Error, errorEvent);
+    }
+    this.agentSession._onError(ev);
+  }
+  // -- Realtime Session events --
+  onInputSpeechStarted(_ev) {
+    this.logger.info("onInputSpeechStarted");
+    if (!this.vad) {
+      this.agentSession._updateUserState("speaking");
+    }
+    try {
+      this.interrupt();
+    } catch (error) {
+      this.logger.error(
+        "RealtimeAPI input_speech_started, but current speech is not interruptable, this should never happen!",
+        error
+      );
+    }
+  }
+  onInputSpeechStopped(ev) {
+    this.logger.info(ev, "onInputSpeechStopped");
+    if (!this.vad) {
+      this.agentSession._updateUserState("listening");
+    }
+    if (ev.userTranscriptionEnabled) {
+      this.agentSession.emit(
+        import_events.AgentSessionEventTypes.UserInputTranscribed,
+        (0, import_events.createUserInputTranscribedEvent)({
+          isFinal: false,
+          transcript: ""
+        })
+      );
+    }
+  }
+  onInputAudioTranscriptionCompleted(ev) {
+    this.agentSession.emit(
+      import_events.AgentSessionEventTypes.UserInputTranscribed,
+      (0, import_events.createUserInputTranscribedEvent)({
+        transcript: ev.transcript,
+        isFinal: ev.isFinal
+      })
+    );
+    if (ev.isFinal) {
+      const message = import_chat_context.ChatMessage.create({
+        role: "user",
+        content: ev.transcript,
+        id: ev.itemId
+      });
+      this.agent._chatCtx.items.push(message);
+      this.agentSession._conversationItemAdded(message);
+    }
+  }
+  onGenerationCreated(ev) {
+    if (ev.userInitiated) {
+      return;
+    }
+    if (this.draining) {
+      this.logger.warn("skipping new realtime generation, the agent is draining");
+      return;
+    }
+    const handle = import_speech_handle.SpeechHandle.create({
+      allowInterruptions: this.allowInterruptions
+    });
+    this.agentSession.emit(
+      import_events.AgentSessionEventTypes.SpeechCreated,
+      (0, import_events.createSpeechCreatedEvent)({
+        userInitiated: false,
+        source: "generate_reply",
+        speechHandle: handle
+      })
+    );
+    this.logger.info({ speech_id: handle.id }, "Creating speech handle");
+    this.createSpeechTask({
+      promise: this.realtimeGenerationTask(handle, ev, {}),
+      ownedSpeechHandle: handle,
+      name: "AgentActivity.realtimeGeneration"
+    });
+    this.scheduleSpeech(handle, import_speech_handle.SpeechHandle.SPEECH_PRIORITY_NORMAL);
+  }
+  // recognition hooks
+  onStartOfSpeech(_ev) {
+    this.agentSession._updateUserState("speaking");
+  }
+  onEndOfSpeech(_ev) {
+    this.agentSession._updateUserState("listening");
+  }
+  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) {
+      return;
+    }
+    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) {
+        return;
+      }
+    }
+    (_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");
+      (_b = this.realtimeSession) == null ? void 0 : _b.interrupt();
+      this._currentSpeech.interrupt();
+    }
+  }
+  onInterimTranscript(ev) {
+    if (this.llm instanceof import_llm.RealtimeModel && this.llm.capabilities.userTranscription) {
+      return;
+    }
+    this.agentSession.emit(
+      import_events.AgentSessionEventTypes.UserInputTranscribed,
+      (0, import_events.createUserInputTranscribedEvent)({
+        transcript: ev.alternatives[0].text,
+        isFinal: false
+        // TODO(AJS-106): add multi participant support
+      })
+    );
+  }
+  onFinalTranscript(ev) {
+    if (this.llm instanceof import_llm.RealtimeModel && this.llm.capabilities.userTranscription) {
+      return;
+    }
+    this.agentSession.emit(
+      import_events.AgentSessionEventTypes.UserInputTranscribed,
+      (0, import_events.createUserInputTranscribedEvent)({
+        transcript: ev.alternatives[0].text,
+        isFinal: true
+        // TODO(AJS-106): add multi participant support
+      })
+    );
+  }
+  createSpeechTask(options) {
+    const { promise, ownedSpeechHandle } = options;
+    this.speechTasks.add(promise);
+    promise.finally(() => {
+      this.speechTasks.delete(promise);
+      if (ownedSpeechHandle) {
+        ownedSpeechHandle._markPlayoutDone();
+      }
+      this.wakeupMainTask();
+    });
+    return promise;
+  }
+  async onEndOfTurn(info) {
+    if (this.draining) {
+      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.logger.info("skipping user input, new_transcript is too short");
+      return false;
+    }
+    const oldTask = this._userTurnCompletedTask;
+    this._userTurnCompletedTask = this.createSpeechTask({
+      promise: this.userTurnCompleted(info, oldTask),
+      name: "AgentActivity.userTurnCompleted"
+    });
+    return true;
+  }
+  retrieveChatCtx() {
+    return this.agentSession.chatCtx;
+  }
+  async mainTask(signal) {
+    const abortFuture = new import_utils.Future();
+    const abortHandler = () => {
+      abortFuture.resolve();
+      signal.removeEventListener("abort", abortHandler);
+    };
+    signal.addEventListener("abort", abortHandler);
+    while (true) {
+      await Promise.race([this.q_updated.await, abortFuture.await]);
+      if (signal.aborted) break;
+      while (this.speechQueue.size() > 0) {
+        if (signal.aborted) break;
+        const heapItem = this.speechQueue.pop();
+        if (!heapItem) {
+          throw new Error("Speech queue is empty");
+        }
+        const speechHandle = heapItem[2];
+        this._currentSpeech = speechHandle;
+        speechHandle._authorizePlayout();
+        await speechHandle.waitForPlayout();
+        this._currentSpeech = void 0;
+      }
+      if (this.draining && this.speechTasks.size === 0) {
+        this.logger.info("mainTask: draining and no more speech tasks");
+        break;
+      }
+      this.q_updated = new import_utils.Future();
+    }
+    this.logger.info("AgentActivity mainTask: exiting");
+  }
+  wakeupMainTask() {
+    this.q_updated.resolve();
+  }
+  generateReply(options) {
+    var _a;
+    const {
+      userMessage,
+      chatCtx,
+      instructions: defaultInstructions,
+      toolChoice: defaultToolChoice,
+      allowInterruptions: defaultAllowInterruptions
+    } = options;
+    let instructions = defaultInstructions;
+    let toolChoice = defaultToolChoice;
+    let allowInterruptions = defaultAllowInterruptions;
+    if (this.llm instanceof import_llm.RealtimeModel && this.llm.capabilities.turnDetection && allowInterruptions === false) {
+      this.logger.warn(
+        "the RealtimeModel uses a server-side turn detection, allowInterruptions cannot be false when using VoiceAgent.generateReply(), disable turnDetection in the RealtimeModel and use VAD on the AgentTask/VoiceAgent instead"
+      );
+      allowInterruptions = true;
+    }
+    if (this.llm === void 0) {
+      throw new Error("trying to generate reply without an LLM model");
+    }
+    const functionCall = (_a = import_agent.asyncLocalStorage.getStore()) == null ? void 0 : _a.functionCall;
+    if (toolChoice === void 0 && functionCall !== void 0) {
+      toolChoice = "none";
+    }
+    const handle = import_speech_handle.SpeechHandle.create({
+      allowInterruptions: allowInterruptions ?? this.allowInterruptions
+    });
+    this.agentSession.emit(
+      import_events.AgentSessionEventTypes.SpeechCreated,
+      (0, import_events.createSpeechCreatedEvent)({
+        userInitiated: true,
+        source: "generate_reply",
+        speechHandle: handle
+      })
+    );
+    this.logger.info({ speech_id: handle.id }, "Creating speech handle");
+    if (this.llm instanceof import_llm.RealtimeModel) {
+      this.createSpeechTask({
+        promise: this.realtimeReplyTask({
+          speechHandle: handle,
+          // TODO(brian): support llm.ChatMessage for the realtime model
+          userInput: userMessage == null ? void 0 : userMessage.textContent,
+          instructions,
+          modelSettings: {
+            // isGiven(toolChoice) = toolChoice !== undefined
+            toolChoice: toOaiToolChoice(toolChoice !== void 0 ? toolChoice : this.toolChoice)
+          }
+        }),
+        ownedSpeechHandle: handle,
+        name: "AgentActivity.realtimeReply"
+      });
+    } else if (this.llm instanceof import_llm.LLM) {
+      if (instructions) {
+        instructions = `${this.agent.instructions}
+${instructions}`;
+      }
+      const task = this.createSpeechTask({
+        promise: this.pipelineReplyTask(
+          handle,
+          chatCtx ?? this.agent.chatCtx,
+          this.agent.toolCtx,
+          { toolChoice: toOaiToolChoice(toolChoice !== void 0 ? toolChoice : this.toolChoice) },
+          instructions ? `${this.agent.instructions}
+${instructions}` : instructions,
+          userMessage
+        ),
+        ownedSpeechHandle: handle,
+        name: "AgentActivity.pipelineReply"
+      });
+      task.finally(() => this.onPipelineReplyDone());
+    }
+    this.scheduleSpeech(handle, import_speech_handle.SpeechHandle.SPEECH_PRIORITY_NORMAL);
+    return handle;
+  }
+  interrupt() {
+    var _a;
+    const future = new import_utils.Future();
+    const currentSpeech = this._currentSpeech;
+    currentSpeech == null ? void 0 : currentSpeech.interrupt();
+    for (const [_, __, speech] of this.speechQueue) {
+      speech.interrupt();
+    }
+    (_a = this.realtimeSession) == null ? void 0 : _a.interrupt();
+    if (currentSpeech === void 0) {
+      future.resolve();
+    } else {
+      currentSpeech.then(() => {
+        if (future.done) return;
+        future.resolve();
+      });
+    }
+    return future;
+  }
+  onPipelineReplyDone() {
+    if (!this.speechQueue.peek() && (!this._currentSpeech || this._currentSpeech.done)) {
+      this.agentSession._updateAgentState("listening");
+    }
+  }
+  async userTurnCompleted(info, oldTask) {
+    var _a, _b;
+    if (oldTask) {
+      await oldTask;
+    }
+    if (this.llm instanceof import_llm.RealtimeModel) {
+      if (this.llm.capabilities.turnDetection) {
+        return;
+      }
+      (_a = this.realtimeSession) == null ? void 0 : _a.commitAudio();
+    }
+    if (this._currentSpeech) {
+      if (!this._currentSpeech.allowInterruptions) {
+        this.logger.warn(
+          { user_input: info.newTranscript },
+          "skipping user input, current speech generation cannot be interrupted"
+        );
+        return;
+      }
+      this.logger.info(
+        { "speech id": this._currentSpeech.id },
+        "speech interrupted, new user turn detected"
+      );
+      this._currentSpeech.interrupt();
+      (_b = this.realtimeSession) == null ? void 0 : _b.interrupt();
+    }
+    let userMessage = import_chat_context.ChatMessage.create({
+      role: "user",
+      content: info.newTranscript
+    });
+    const chatCtx = this.agent.chatCtx.copy();
+    const startTime = Date.now();
+    try {
+      await this.agent.onUserTurnCompleted(chatCtx, userMessage);
+    } catch (e) {
+      if (e instanceof import_agent.StopResponse) {
+        return;
+      }
+      this.logger.error({ error: e }, "error occurred during onUserTurnCompleted");
+    }
+    const callbackDuration = Date.now() - startTime;
+    if (this.llm instanceof import_llm.RealtimeModel) {
+      userMessage = void 0;
+    } else if (this.llm === void 0) {
+      return;
+    }
+    const speechHandle = this.generateReply({ userMessage, chatCtx });
+    const eouMetrics = {
+      type: "eou_metrics",
+      timestamp: Date.now(),
+      endOfUtteranceDelay: info.endOfUtteranceDelay,
+      transcriptionDelay: info.transcriptionDelay,
+      onUserTurnCompletedDelay: callbackDuration,
+      speechId: speechHandle.id
+    };
+    this.agentSession.emit(
+      import_events.AgentSessionEventTypes.MetricsCollected,
+      (0, import_events.createMetricsCollectedEvent)({ metrics: eouMetrics })
+    );
+  }
+  async ttsTask(speechHandle, text, addToChatCtx, modelSettings, audio) {
+    speechHandleStorage.enterWith(speechHandle);
+    const transcriptionOutput = this.agentSession.output.transcriptionEnabled ? this.agentSession.output.transcription : null;
+    const audioOutput = this.agentSession.output.audioEnabled ? this.agentSession.output.audio : null;
+    const replyAbortController = new AbortController();
+    await speechHandle.waitIfNotInterrupted([speechHandle._waitForAuthorization()]);
+    if (speechHandle.interrupted) {
+      return;
+    }
+    let baseStream;
+    if (text instanceof import_web.ReadableStream) {
+      baseStream = text;
+    } else {
+      baseStream = new import_web.ReadableStream({
+        start(controller) {
+          controller.enqueue(text);
+          controller.close();
+        }
+      });
+    }
+    const [textSource, audioSource] = baseStream.tee();
+    const tasks = [];
+    const trNode = await this.agent.transcriptionNode(textSource, {});
+    let textOut = null;
+    if (trNode) {
+      const [textForwardTask, _textOut] = (0, import_generation.performTextForwarding)(
+        trNode,
+        replyAbortController,
+        transcriptionOutput
+      );
+      textOut = _textOut;
+      tasks.push(textForwardTask);
+    }
+    const onFirstFrame = () => {
+      this.agentSession._updateAgentState("speaking");
+    };
+    if (!audioOutput) {
+      if (textOut) {
+        textOut.firstTextFut.await.finally(onFirstFrame);
+      }
+    } else {
+      let audioOut = null;
+      if (!audio) {
+        const [ttsTask, ttsStream] = (0, import_generation.performTTSInference)(
+          (...args) => this.agent.ttsNode(...args),
+          audioSource,
+          modelSettings,
+          replyAbortController
+        );
+        tasks.push(ttsTask);
+        const [forwardTask, _audioOut] = (0, import_generation.performAudioForwarding)(
+          ttsStream,
+          audioOutput,
+          replyAbortController
+        );
+        tasks.push(forwardTask);
+        audioOut = _audioOut;
+      } else {
+        const [forwardTask, _audioOut] = (0, import_generation.performAudioForwarding)(
+          audio,
+          audioOutput,
+          replyAbortController
+        );
+        tasks.push(forwardTask);
+        audioOut = _audioOut;
+      }
+      audioOut.firstFrameFut.await.finally(onFirstFrame);
+    }
+    await speechHandle.waitIfNotInterrupted(tasks.map((task) => task.result));
+    if (audioOutput) {
+      await speechHandle.waitIfNotInterrupted([audioOutput.waitForPlayout()]);
+    }
+    if (speechHandle.interrupted) {
+      replyAbortController.abort();
+      await (0, import_utils.cancelAndWait)(tasks, AgentActivity.REPLY_TASK_CANCEL_TIMEOUT);
+      if (audioOutput) {
+        audioOutput.clearBuffer();
+        await audioOutput.waitForPlayout();
+      }
+    }
+    if (addToChatCtx) {
+      const message = import_chat_context.ChatMessage.create({
+        role: "assistant",
+        content: (textOut == null ? void 0 : textOut.text) || "",
+        interrupted: speechHandle.interrupted
+      });
+      this.agent._chatCtx.insert(message);
+      this.agentSession._conversationItemAdded(message);
+    }
+    if (this.agentSession.agentState === "speaking") {
+      this.agentSession._updateAgentState("listening");
+    }
+  }
+  async pipelineReplyTask(speechHandle, chatCtx, toolCtx, modelSettings, instructions, newMessage, toolsMessages) {
+    var _a, _b, _c;
+    speechHandleStorage.enterWith(speechHandle);
+    const replyAbortController = new AbortController();
+    const audioOutput = this.agentSession.output.audioEnabled ? this.agentSession.output.audio : null;
+    const transcriptionOutput = this.agentSession.output.transcriptionEnabled ? this.agentSession.output.transcription : null;
+    chatCtx = chatCtx.copy();
+    if (newMessage) {
+      chatCtx.insert(newMessage);
+      this.agent._chatCtx.insert(newMessage);
+      this.agentSession._conversationItemAdded(newMessage);
+    }
+    if (instructions) {
+      try {
+        (0, import_generation.updateInstructions)({
+          chatCtx,
+          instructions,
+          addIfMissing: true
+        });
+      } catch (e) {
+        this.logger.error({ error: e }, "error occurred during updateInstructions");
+      }
+    }
+    this.agentSession._updateAgentState("thinking");
+    const tasks = [];
+    const [llmTask, llmGenData] = (0, import_generation.performLLMInference)(
+      // preserve  `this` context in llmNode
+      (...args) => this.agent.llmNode(...args),
+      chatCtx,
+      toolCtx,
+      modelSettings,
+      replyAbortController
+    );
+    tasks.push(llmTask);
+    const [ttsTextInput, llmOutput] = llmGenData.textStream.tee();
+    let ttsTask = null;
+    let ttsStream = null;
+    if (audioOutput) {
+      [ttsTask, ttsStream] = (0, import_generation.performTTSInference)(
+        (...args) => this.agent.ttsNode(...args),
+        ttsTextInput,
+        modelSettings,
+        replyAbortController
+      );
+      tasks.push(ttsTask);
+    }
+    await speechHandle.waitIfNotInterrupted([speechHandle._waitForAuthorization()]);
+    if (speechHandle.interrupted) {
+      replyAbortController.abort();
+      await (0, import_utils.cancelAndWait)(tasks, AgentActivity.REPLY_TASK_CANCEL_TIMEOUT);
+      return;
+    }
+    const replyStartedAt = Date.now();
+    const trNodeResult = await this.agent.transcriptionNode(llmOutput, modelSettings);
+    let textOut = null;
+    if (trNodeResult) {
+      const [textForwardTask, _textOut] = (0, import_generation.performTextForwarding)(
+        trNodeResult,
+        replyAbortController,
+        transcriptionOutput
+      );
+      tasks.push(textForwardTask);
+      textOut = _textOut;
+    }
+    const onFirstFrame = () => {
+      this.agentSession._updateAgentState("speaking");
+    };
+    let audioOut = null;
+    if (audioOutput) {
+      if (ttsStream) {
+        const [forwardTask, _audioOut] = (0, import_generation.performAudioForwarding)(
+          ttsStream,
+          audioOutput,
+          replyAbortController
+        );
+        audioOut = _audioOut;
+        tasks.push(forwardTask);
+        audioOut.firstFrameFut.await.finally(onFirstFrame);
+      } else {
+        throw Error("ttsStream is null when audioOutput is enabled");
+      }
+    } else {
+      textOut == null ? void 0 : textOut.firstTextFut.await.finally(onFirstFrame);
+    }
+    const onToolExecutionStarted = (_) => {
+    };
+    const onToolExecutionCompleted = (_) => {
+    };
+    const [executeToolsTask, toolOutput] = (0, import_generation.performToolExecutions)({
+      session: this.agentSession,
+      speechHandle,
+      toolCtx,
+      toolChoice: modelSettings.toolChoice,
+      toolCallStream: llmGenData.toolCallStream,
+      controller: replyAbortController,
+      onToolExecutionStarted,
+      onToolExecutionCompleted
+    });
+    tasks.push(executeToolsTask);
+    await speechHandle.waitIfNotInterrupted(tasks.map((task) => task.result));
+    if (audioOutput) {
+      await speechHandle.waitIfNotInterrupted([audioOutput.waitForPlayout()]);
+    }
+    if (toolsMessages) {
+      for (const msg of toolsMessages) {
+        msg.createdAt = replyStartedAt;
+      }
+      this.agent._chatCtx.insert(toolsMessages);
+    }
+    if (speechHandle.interrupted) {
+      this.logger.debug(
+        { speech_id: speechHandle.id },
+        "Aborting all pipeline reply tasks due to interruption"
+      );
+      replyAbortController.abort();
+      await Promise.allSettled(
+        tasks.map((task) => task.cancelAndWait(AgentActivity.REPLY_TASK_CANCEL_TIMEOUT))
+      );
+      let forwardedText = (textOut == null ? void 0 : textOut.text) || "";
+      if (audioOutput) {
+        audioOutput.clearBuffer();
+        const playbackEv = await audioOutput.waitForPlayout();
+        if (audioOut == null ? void 0 : audioOut.firstFrameFut.done) {
+          this.logger.info(
+            { speech_id: speechHandle.id, playbackPosition: playbackEv.playbackPosition },
+            "playout interrupted"
+          );
+          if (playbackEv.synchronizedTranscript) {
+            forwardedText = playbackEv.synchronizedTranscript;
+          }
+        } else {
+          forwardedText = "";
+        }
+      }
+      if (forwardedText) {
+        const message = import_chat_context.ChatMessage.create({
+          role: "assistant",
+          content: forwardedText,
+          id: llmGenData.id,
+          interrupted: true,
+          createdAt: replyStartedAt
+        });
+        chatCtx.insert(message);
+        this.agent._chatCtx.insert(message);
+        this.agentSession._conversationItemAdded(message);
+      }
+      if (this.agentSession.agentState === "speaking") {
+        this.agentSession._updateAgentState("listening");
+      }
+      this.logger.info(
+        { speech_id: speechHandle.id, message: forwardedText },
+        "playout completed with interrupt"
+      );
+      speechHandle._markPlayoutDone();
+      await executeToolsTask.cancelAndWait(AgentActivity.REPLY_TASK_CANCEL_TIMEOUT);
+      return;
+    }
+    if (textOut && textOut.text) {
+      const message = import_chat_context.ChatMessage.create({
+        role: "assistant",
+        id: llmGenData.id,
+        interrupted: false,
+        createdAt: replyStartedAt,
+        content: textOut.text
+      });
+      chatCtx.insert(message);
+      this.agent._chatCtx.insert(message);
+      this.agentSession._conversationItemAdded(message);
+      this.logger.info(
+        { speech_id: speechHandle.id, message: textOut.text },
+        "playout completed without interruption"
+      );
+    }
+    if (toolOutput.output.length > 0) {
+      this.agentSession._updateAgentState("thinking");
+    } else if (this.agentSession.agentState === "speaking") {
+      this.agentSession._updateAgentState("listening");
+    }
+    speechHandle._markPlayoutDone();
+    await executeToolsTask.result;
+    if (toolOutput.output.length === 0) return;
+    const { maxToolSteps } = this.agentSession.options;
+    if (speechHandle.stepIndex >= maxToolSteps) {
+      this.logger.warn(
+        { speech_id: speechHandle.id, max_tool_steps: maxToolSteps },
+        "maximum number of function calls steps reached"
+      );
+      return;
+    }
+    const functionToolsExecutedEvent = (0, import_events.createFunctionToolsExecutedEvent)({
+      functionCalls: [],
+      functionCallOutputs: []
+    });
+    let shouldGenerateToolReply = false;
+    let newAgentTask = null;
+    let ignoreTaskSwitch = false;
+    for (const sanitizedOut of toolOutput.output) {
+      if (sanitizedOut.toolCallOutput !== void 0) {
+        functionToolsExecutedEvent.functionCalls.push(sanitizedOut.toolCall);
+        functionToolsExecutedEvent.functionCallOutputs.push(sanitizedOut.toolCallOutput);
+        if (sanitizedOut.replyRequired) {
+          shouldGenerateToolReply = true;
+        }
+      }
+      if (newAgentTask !== null && sanitizedOut.agentTask !== void 0) {
+        this.logger.error("expected to receive only one agent task from the tool executions");
+        ignoreTaskSwitch = true;
+      }
+      newAgentTask = sanitizedOut.agentTask ?? null;
+      this.logger.debug(
+        {
+          speechId: speechHandle.id,
+          name: (_a = sanitizedOut.toolCall) == null ? void 0 : _a.name,
+          args: sanitizedOut.toolCall.args,
+          output: (_b = sanitizedOut.toolCallOutput) == null ? void 0 : _b.output,
+          isError: (_c = sanitizedOut.toolCallOutput) == null ? void 0 : _c.isError
+        },
+        "Tool call execution finished"
+      );
+    }
+    this.agentSession.emit(
+      import_events.AgentSessionEventTypes.FunctionToolsExecuted,
+      functionToolsExecutedEvent
+    );
+    let draining = this.draining;
+    if (!ignoreTaskSwitch && newAgentTask !== null) {
+      this.agentSession.updateAgent(newAgentTask);
+      draining = true;
+    }
+    const toolMessages = [
+      ...functionToolsExecutedEvent.functionCalls,
+      ...functionToolsExecutedEvent.functionCallOutputs
+    ];
+    if (shouldGenerateToolReply) {
+      chatCtx.insert(toolMessages);
+      const handle = import_speech_handle.SpeechHandle.create({
+        allowInterruptions: speechHandle.allowInterruptions,
+        stepIndex: speechHandle.stepIndex + 1,
+        parent: speechHandle
+      });
+      this.agentSession.emit(
+        import_events.AgentSessionEventTypes.SpeechCreated,
+        (0, import_events.createSpeechCreatedEvent)({
+          userInitiated: false,
+          source: "tool_response",
+          speechHandle: handle
+        })
+      );
+      const respondToolChoice = draining || modelSettings.toolChoice === "none" ? "none" : "auto";
+      const toolResponseTask = this.createSpeechTask({
+        promise: this.pipelineReplyTask(
+          handle,
+          chatCtx,
+          toolCtx,
+          { toolChoice: respondToolChoice },
+          instructions,
+          void 0,
+          toolMessages
+        ),
+        ownedSpeechHandle: handle,
+        name: "AgentActivity.pipelineReply"
+      });
+      toolResponseTask.finally(() => this.onPipelineReplyDone());
+      this.scheduleSpeech(handle, import_speech_handle.SpeechHandle.SPEECH_PRIORITY_NORMAL, true);
+    } else if (functionToolsExecutedEvent.functionCallOutputs.length > 0) {
+      for (const msg of toolMessages) {
+        msg.createdAt = replyStartedAt;
+      }
+      this.agent._chatCtx.insert(toolMessages);
+    }
+  }
+  async realtimeGenerationTask(speechHandle, ev, modelSettings) {
+    var _a, _b, _c;
+    speechHandleStorage.enterWith(speechHandle);
+    if (!this.realtimeSession) {
+      throw new Error("realtime session is not initialized");
+    }
+    if (!(this.llm instanceof import_llm.RealtimeModel)) {
+      throw new Error("llm is not a realtime model");
+    }
+    this.logger.debug(
+      { speech_id: speechHandle.id, stepIndex: speechHandle.stepIndex },
+      "realtime generation started"
+    );
+    const audioOutput = this.agentSession.output.audioEnabled ? this.agentSession.output.audio : null;
+    const textOutput = this.agentSession.output.transcriptionEnabled ? this.agentSession.output.transcription : null;
+    const toolCtx = this.realtimeSession.tools;
+    await speechHandle.waitIfNotInterrupted([speechHandle._waitForAuthorization()]);
+    if (speechHandle.interrupted) {
+      return;
+    }
+    const onFirstFrame = () => {
+      this.agentSession._updateAgentState("speaking");
+    };
+    const replyAbortController = new AbortController();
+    const readMessages = async (abortController, outputs) => {
+      const forwardTasks = [];
+      try {
+        for await (const msg of ev.messageStream) {
+          if (forwardTasks.length > 0) {
+            this.logger.warn(
+              "expected to receive only one message generation from the realtime API"
+            );
+            break;
+          }
+          const trNodeResult = await this.agent.transcriptionNode(msg.textStream, modelSettings);
+          let textOut = null;
+          if (trNodeResult) {
+            const [textForwardTask, _textOut] = (0, import_generation.performTextForwarding)(
+              trNodeResult,
+              abortController,
+              textOutput
+            );
+            forwardTasks.push(textForwardTask);
+            textOut = _textOut;
+          }
+          let audioOut = null;
+          if (audioOutput) {
+            const realtimeAudio = await this.agent.realtimeAudioOutputNode(
+              msg.audioStream,
+              modelSettings
+            );
+            if (realtimeAudio) {
+              const [forwardTask, _audioOut] = (0, import_generation.performAudioForwarding)(
+                realtimeAudio,
+                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]);
+        }
+        await (0, import_utils.waitFor)(forwardTasks);
+      } catch (error) {
+        this.logger.error(error, "error reading messages from the realtime API");
+      } finally {
+        await (0, import_utils.cancelAndWait)(forwardTasks, AgentActivity.REPLY_TASK_CANCEL_TIMEOUT);
+      }
+    };
+    const messageOutputs = [];
+    const tasks = [
+      import_utils.Task.from(
+        (controller) => readMessages(controller, messageOutputs),
+        replyAbortController,
+        "AgentActivity.realtime_generation.read_messages"
+      )
+    ];
+    const [toolCallStream, toolCallStreamForTracing] = ev.functionStream.tee();
+    const toolCalls = [];
+    const readToolStreamTask = async (controller, stream) => {
+      const reader = stream.getReader();
+      try {
+        while (!controller.signal.aborted) {
+          const { done, value } = await reader.read();
+          if (done) break;
+          this.logger.debug({ tool_call: value }, "received tool call from the realtime API");
+          toolCalls.push(value);
+        }
+      } finally {
+        reader.releaseLock();
+      }
+    };
+    tasks.push(
+      import_utils.Task.from(
+        (controller) => readToolStreamTask(controller, toolCallStreamForTracing),
+        replyAbortController,
+        "AgentActivity.realtime_generation.read_tool_stream"
+      )
+    );
+    const onToolExecutionStarted = (_) => {
+    };
+    const onToolExecutionCompleted = (_) => {
+    };
+    const [executeToolsTask, toolOutput] = (0, import_generation.performToolExecutions)({
+      session: this.agentSession,
+      speechHandle,
+      toolCtx,
+      toolCallStream,
+      toolChoice: modelSettings.toolChoice,
+      controller: replyAbortController,
+      onToolExecutionStarted,
+      onToolExecutionCompleted
+    });
+    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(
+        { speech_id: speechHandle.id },
+        "Aborting all realtime generation tasks due to interruption"
+      );
+      replyAbortController.abort();
+      await (0, import_utils.cancelAndWait)(tasks, AgentActivity.REPLY_TASK_CANCEL_TIMEOUT);
+      if (messageOutputs.length > 0) {
+        const [msgId, textOut, audioOut] = messageOutputs[0];
+        let forwardedText = (textOut == null ? void 0 : textOut.text) || "";
+        if (audioOutput) {
+          audioOutput.clearBuffer();
+          const playbackEv = await audioOutput.waitForPlayout();
+          let playbackPosition = playbackEv.playbackPosition;
+          if (audioOut == null ? void 0 : audioOut.firstFrameFut.done) {
+            this.logger.info(
+              { speech_id: speechHandle.id, playbackPosition: playbackEv.playbackPosition },
+              "playout interrupted"
+            );
+            if (playbackEv.synchronizedTranscript) {
+              forwardedText = playbackEv.synchronizedTranscript;
+            }
+          } else {
+            forwardedText = "";
+            playbackPosition = 0;
+          }
+          this.realtimeSession.truncate({
+            messageId: msgId,
+            audioEndMs: Math.floor(playbackPosition)
+          });
+        }
+        if (forwardedText) {
+          const message = import_chat_context.ChatMessage.create({
+            role: "assistant",
+            content: forwardedText,
+            id: msgId,
+            interrupted: true
+          });
+          this.agent._chatCtx.insert(message);
+          speechHandle._setChatMessage(message);
+          this.agentSession._conversationItemAdded(message);
+        }
+        this.logger.info(
+          { speech_id: speechHandle.id, message: forwardedText },
+          "playout completed with interrupt"
+        );
+      }
+      speechHandle._markPlayoutDone();
+      await executeToolsTask.cancelAndWait(AgentActivity.REPLY_TASK_CANCEL_TIMEOUT);
+      return;
+    }
+    if (messageOutputs.length > 0) {
+      const [msgId, textOut, _] = messageOutputs[0];
+      const message = import_chat_context.ChatMessage.create({
+        role: "assistant",
+        content: (textOut == null ? void 0 : textOut.text) || "",
+        id: msgId,
+        interrupted: false
+      });
+      this.agent._chatCtx.insert(message);
+      speechHandle._setChatMessage(message);
+      this.agentSession._conversationItemAdded(message);
+    }
+    speechHandle._markPlayoutDone();
+    toolOutput.firstToolStartedFuture.await.finally(() => {
+      this.agentSession._updateAgentState("thinking");
+    });
+    await executeToolsTask.result;
+    if (toolOutput.output.length === 0) return;
+    const { maxToolSteps } = this.agentSession.options;
+    if (speechHandle.stepIndex >= maxToolSteps) {
+      this.logger.warn(
+        { speech_id: speechHandle.id, max_tool_steps: maxToolSteps },
+        "maximum number of function calls steps reached"
+      );
+      return;
+    }
+    const functionToolsExecutedEvent = (0, import_events.createFunctionToolsExecutedEvent)({
+      functionCalls: [],
+      functionCallOutputs: []
+    });
+    let shouldGenerateToolReply = false;
+    let newAgentTask = null;
+    let ignoreTaskSwitch = false;
+    for (const sanitizedOut of toolOutput.output) {
+      if (sanitizedOut.toolCallOutput !== void 0) {
+        functionToolsExecutedEvent.functionCallOutputs.push(sanitizedOut.toolCallOutput);
+        if (sanitizedOut.replyRequired) {
+          shouldGenerateToolReply = true;
+        }
+      }
+      if (newAgentTask !== null && sanitizedOut.agentTask !== void 0) {
+        this.logger.error("expected to receive only one agent task from the tool executions");
+        ignoreTaskSwitch = true;
+      }
+      newAgentTask = sanitizedOut.agentTask ?? null;
+      this.logger.debug(
+        {
+          speechId: speechHandle.id,
+          name: (_a = sanitizedOut.toolCall) == null ? void 0 : _a.name,
+          args: sanitizedOut.toolCall.args,
+          output: (_b = sanitizedOut.toolCallOutput) == null ? void 0 : _b.output,
+          isError: (_c = sanitizedOut.toolCallOutput) == null ? void 0 : _c.isError
+        },
+        "Tool call execution finished"
+      );
+    }
+    this.agentSession.emit(
+      import_events.AgentSessionEventTypes.FunctionToolsExecuted,
+      functionToolsExecutedEvent
+    );
+    let draining = this.draining;
+    if (!ignoreTaskSwitch && newAgentTask !== null) {
+      this.agentSession.updateAgent(newAgentTask);
+      draining = true;
+    }
+    if (functionToolsExecutedEvent.functionCallOutputs.length > 0) {
+      const chatCtx = this.realtimeSession.chatCtx.copy();
+      chatCtx.items.push(...functionToolsExecutedEvent.functionCallOutputs);
+      try {
+        await this.realtimeSession.updateChatCtx(chatCtx);
+      } catch (error) {
+        this.logger.warn(
+          { error },
+          "failed to update chat context before generating the function calls results"
+        );
+      }
+    }
+    if (!shouldGenerateToolReply || this.llm.capabilities.autoToolReplyGeneration) {
+      return;
+    }
+    this.realtimeSession.interrupt();
+    const replySpeechHandle = import_speech_handle.SpeechHandle.create({
+      allowInterruptions: speechHandle.allowInterruptions,
+      stepIndex: speechHandle.stepIndex + 1,
+      parent: speechHandle
+    });
+    this.agentSession.emit(
+      import_events.AgentSessionEventTypes.SpeechCreated,
+      (0, import_events.createSpeechCreatedEvent)({
+        userInitiated: false,
+        source: "tool_response",
+        speechHandle: replySpeechHandle
+      })
+    );
+    const toolChoice = draining || modelSettings.toolChoice === "none" ? "none" : "auto";
+    this.createSpeechTask({
+      promise: this.realtimeReplyTask({
+        speechHandle: replySpeechHandle,
+        modelSettings: { toolChoice }
+      }),
+      ownedSpeechHandle: replySpeechHandle,
+      name: "AgentActivity.realtime_reply"
+    });
+    this.scheduleSpeech(replySpeechHandle, import_speech_handle.SpeechHandle.SPEECH_PRIORITY_NORMAL, true);
+  }
+  async realtimeReplyTask({
+    speechHandle,
+    modelSettings: { toolChoice },
+    userInput,
+    instructions
+  }) {
+    speechHandleStorage.enterWith(speechHandle);
+    if (!this.realtimeSession) {
+      throw new Error("realtime session is not available");
+    }
+    await speechHandle.waitIfNotInterrupted([speechHandle._waitForAuthorization()]);
+    if (userInput) {
+      const chatCtx = this.realtimeSession.chatCtx.copy();
+      const message = chatCtx.addMessage({
+        role: "user",
+        content: userInput
+      });
+      await this.realtimeSession.updateChatCtx(chatCtx);
+      this.agent._chatCtx.insert(message);
+      this.agentSession._conversationItemAdded(message);
+    }
+    const originalToolChoice = this.toolChoice;
+    if (toolChoice !== void 0) {
+      this.realtimeSession.updateOptions({ toolChoice });
+    }
+    try {
+      const generationEvent = await this.realtimeSession.generateReply(instructions);
+      await this.realtimeGenerationTask(speechHandle, generationEvent, { toolChoice });
+    } finally {
+      if (toolChoice !== void 0 && toolChoice !== originalToolChoice) {
+        this.realtimeSession.updateOptions({ toolChoice: originalToolChoice });
+      }
+    }
+  }
+  scheduleSpeech(speechHandle, priority, bypassDraining = false) {
+    if (this.draining && !bypassDraining) {
+      throw new Error("cannot schedule new speech, the agent is draining");
+    }
+    this.speechQueue.push([priority, Number(process.hrtime.bigint()), speechHandle]);
+    this.wakeupMainTask();
+  }
+  async drain() {
+    var _a;
+    const unlock = await this.lock.lock();
+    try {
+      if (this._draining) return;
+      this.createSpeechTask({
+        promise: this.agent.onExit(),
+        name: "AgentActivity_onExit"
+      });
+      this.wakeupMainTask();
+      this._draining = true;
+      await ((_a = this._mainTask) == null ? void 0 : _a.result);
+    } finally {
+      unlock();
+    }
+  }
+  async close() {
+    var _a, _b, _c;
+    const unlock = await this.lock.lock();
+    try {
+      if (!this._draining) {
+        this.logger.warn("task closing without draining");
+      }
+      if (this.llm instanceof import_llm.LLM) {
+        this.llm.off("metrics_collected", this.onMetricsCollected);
+      }
+      if (this.realtimeSession) {
+        this.realtimeSession.off("generation_created", this.onGenerationCreated);
+        this.realtimeSession.off("input_speech_started", this.onInputSpeechStarted);
+        this.realtimeSession.off("input_speech_stopped", this.onInputSpeechStopped);
+        this.realtimeSession.off(
+          "input_audio_transcription_completed",
+          this.onInputAudioTranscriptionCompleted
+        );
+        this.realtimeSession.off("metrics_collected", this.onMetricsCollected);
+      }
+      if (this.stt instanceof import_stt.STT) {
+        this.stt.off("metrics_collected", this.onMetricsCollected);
+      }
+      if (this.tts instanceof import_tts.TTS) {
+        this.tts.off("metrics_collected", this.onMetricsCollected);
+      }
+      if (this.vad instanceof import_vad.VAD) {
+        this.vad.off("metrics_collected", this.onMetricsCollected);
+      }
+      this.detachAudioInput();
+      await ((_a = this.realtimeSession) == null ? void 0 : _a.close());
+      await ((_b = this.audioRecognition) == null ? void 0 : _b.close());
+      await ((_c = this._mainTask) == null ? void 0 : _c.cancelAndWait());
+    } finally {
+      unlock();
+    }
+  }
+}
+function toOaiToolChoice(toolChoice) {
+  return toolChoice !== null ? toolChoice : void 0;
+}
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  AgentActivity
+});
+//# sourceMappingURL=agent_activity.cjs.map