@livekit/agents 0.4.6 → 0.5.0

package/dist/multimodal/multimodal_agent.js

@@ -1,353 +1,374 @@
-import { AudioSource, AudioStream, LocalAudioTrack, RoomEvent, TrackPublishOptions, TrackSource, } from '@livekit/rtc-node';
-import { EventEmitter } from 'node:events';
-import { AudioByteStream } from '../audio.js';
-import * as llm from '../llm/index.js';
-import { log } from '../log.js';
-import { BasicTranscriptionForwarder } from '../transcription.js';
-import { findMicroTrackId } from '../utils.js';
-import { AgentPlayout } from './agent_playout.js';
-/**
- * @internal
- * @beta
- */
-export class RealtimeSession extends EventEmitter {
+import {
+  AudioSource,
+  AudioStream,
+  LocalAudioTrack,
+  RoomEvent,
+  TrackPublishOptions,
+  TrackSource
+} from "@livekit/rtc-node";
+import { EventEmitter } from "node:events";
+import { AudioByteStream } from "../audio.js";
+import * as llm from "../llm/index.js";
+import { log } from "../log.js";
+import { BasicTranscriptionForwarder } from "../transcription.js";
+import { findMicroTrackId } from "../utils.js";
+import { AgentPlayout } from "./agent_playout.js";
+class RealtimeSession extends EventEmitter {
 }
-/**
- * @internal
- * @beta
- */
-export class RealtimeModel {
+class RealtimeModel {
 }
-export const AGENT_STATE_ATTRIBUTE = 'lk.agent.state';
-/** @beta */
-export class MultimodalAgent extends EventEmitter {
-    model;
-    room = null;
-    linkedParticipant = null;
-    subscribedTrack = null;
-    readMicroTask = null;
-    constructor({ model, chatCtx, fncCtx, }) {
-        super();
-        this.model = model;
-        this.#chatCtx = chatCtx;
-        this.#fncCtx = fncCtx;
+const AGENT_STATE_ATTRIBUTE = "lk.agent.state";
+class MultimodalAgent extends EventEmitter {
+  model;
+  room = null;
+  linkedParticipant = null;
+  subscribedTrack = null;
+  readMicroTask = null;
+  constructor({
+    model,
+    chatCtx,
+    fncCtx
+  }) {
+    super();
+    this.model = model;
+    this.#chatCtx = chatCtx;
+    this.#fncCtx = fncCtx;
+  }
+  #participant = null;
+  #agentPublication = null;
+  #localTrackSid = null;
+  #localSource = null;
+  #agentPlayout = null;
+  #playingHandle = void 0;
+  #logger = log();
+  #session = null;
+  #fncCtx = void 0;
+  #chatCtx = void 0;
+  #_started = false;
+  #_pendingFunctionCalls = /* @__PURE__ */ new Set();
+  #_speaking = false;
+  get fncCtx() {
+    return this.#fncCtx;
+  }
+  set fncCtx(ctx) {
+    this.#fncCtx = ctx;
+    if (this.#session) {
+      this.#session.fncCtx = ctx;
     }
-    #participant = null;
-    #agentPublication = null;
-    #localTrackSid = null;
-    #localSource = null;
-    #agentPlayout = null;
-    #playingHandle = undefined;
-    #logger = log();
-    #session = null;
-    #fncCtx = undefined;
-    #chatCtx = undefined;
-    #_started = false;
-    #_pendingFunctionCalls = new Set();
-    #_speaking = false;
-    get fncCtx() {
-        return this.#fncCtx;
-    }
-    set fncCtx(ctx) {
-        this.#fncCtx = ctx;
-        if (this.#session) {
-            this.#session.fncCtx = ctx;
+  }
+  get #pendingFunctionCalls() {
+    return this.#_pendingFunctionCalls;
+  }
+  set #pendingFunctionCalls(calls) {
+    this.#_pendingFunctionCalls = calls;
+    this.#updateState();
+  }
+  get #speaking() {
+    return this.#_speaking;
+  }
+  set #speaking(isSpeaking) {
+    this.#_speaking = isSpeaking;
+    this.#updateState();
+  }
+  get #started() {
+    return this.#_started;
+  }
+  set #started(started) {
+    this.#_started = started;
+    this.#updateState();
+  }
+  start(room, participant = null) {
+    return new Promise(async (resolve, reject) => {
+      var _a;
+      if (this.#started) {
+        reject(new Error("MultimodalAgent already started"));
+      }
+      this.#updateState();
+      room.on(RoomEvent.ParticipantConnected, (participant2) => {
+        if (this.linkedParticipant) {
+          return;
+        }
+        this.#linkParticipant(participant2.identity);
+      });
+      room.on(
+        RoomEvent.TrackPublished,
+        (trackPublication, participant2) => {
+          if (this.linkedParticipant && participant2.identity === this.linkedParticipant.identity && trackPublication.source === TrackSource.SOURCE_MICROPHONE && !trackPublication.subscribed) {
+            trackPublication.setSubscribed(true);
+          }
+        }
+      );
+      room.on(RoomEvent.TrackSubscribed, this.#handleTrackSubscription.bind(this));
+      this.room = room;
+      this.#participant = participant;
+      this.#localSource = new AudioSource(this.model.sampleRate, this.model.numChannels);
+      this.#agentPlayout = new AgentPlayout(
+        this.#localSource,
+        this.model.sampleRate,
+        this.model.numChannels,
+        this.model.inFrameSize,
+        this.model.outFrameSize
+      );
+      const onPlayoutStarted = () => {
+        this.emit("agent_started_speaking");
+        this.#speaking = true;
+      };
+      const onPlayoutStopped = (interrupted) => {
+        this.emit("agent_stopped_speaking");
+        this.#speaking = false;
+        if (this.#playingHandle) {
+          let text = this.#playingHandle.transcriptionFwd.text;
+          if (interrupted) {
+            text += "\u2026";
+          }
+          const msg = llm.ChatMessage.create({
+            role: llm.ChatRole.ASSISTANT,
+            text
+          });
+          if (interrupted) {
+            this.emit("agent_speech_interrupted", msg);
+          } else {
+            this.emit("agent_speech_committed", msg);
+          }
+          this.#logger.child({ transcription: text, interrupted }).debug("committed agent speech");
+        }
+      };
+      this.#agentPlayout.on("playout_started", onPlayoutStarted);
+      this.#agentPlayout.on("playout_stopped", onPlayoutStopped);
+      const track = LocalAudioTrack.createAudioTrack("assistant_voice", this.#localSource);
+      const options = new TrackPublishOptions();
+      options.source = TrackSource.SOURCE_MICROPHONE;
+      this.#agentPublication = await ((_a = room.localParticipant) == null ? void 0 : _a.publishTrack(track, options)) || null;
+      if (!this.#agentPublication) {
+        this.#logger.error("Failed to publish track");
+        reject(new Error("Failed to publish track"));
+        return;
+      }
+      await this.#agentPublication.waitForSubscription();
+      if (participant) {
+        if (typeof participant === "string") {
+          this.#linkParticipant(participant);
+        } else {
+          this.#linkParticipant(participant.identity);
+        }
+      } else {
+        for (const participant2 of room.remoteParticipants.values()) {
+          this.#linkParticipant(participant2.identity);
+          break;
+        }
+      }
+      this.#session = this.model.session({ fncCtx: this.#fncCtx, chatCtx: this.#chatCtx });
+      this.#started = true;
+      this.#session.on("response_content_added", (message) => {
+        var _a2;
+        const trFwd = new BasicTranscriptionForwarder(
+          this.room,
+          this.room.localParticipant.identity,
+          this.#getLocalTrackSid(),
+          message.responseId
+        );
+        const handle = (_a2 = this.#agentPlayout) == null ? void 0 : _a2.play(
+          message.itemId,
+          message.contentIndex,
+          trFwd,
+          message.textStream,
+          message.audioStream
+        );
+        this.#playingHandle = handle;
+      });
+      this.#session.on("input_speech_committed", (ev) => {
+        var _a2, _b;
+        const participantIdentity = (_a2 = this.linkedParticipant) == null ? void 0 : _a2.identity;
+        const trackSid = (_b = this.subscribedTrack) == null ? void 0 : _b.sid;
+        if (participantIdentity && trackSid) {
+          this.#publishTranscription(participantIdentity, trackSid, "\u2026", false, ev.itemId);
+        } else {
+          this.#logger.error("Participant or track not set");
+        }
+      });
+      this.#session.on("input_speech_transcription_completed", (ev) => {
+        var _a2, _b;
+        const transcription = ev.transcript;
+        const participantIdentity = (_a2 = this.linkedParticipant) == null ? void 0 : _a2.identity;
+        const trackSid = (_b = this.subscribedTrack) == null ? void 0 : _b.sid;
+        if (participantIdentity && trackSid) {
+          this.#publishTranscription(participantIdentity, trackSid, transcription, true, ev.itemId);
+        } else {
+          this.#logger.error("Participant or track not set");
+        }
+        const userMsg = llm.ChatMessage.create({
+          role: llm.ChatRole.USER,
+          text: transcription
+        });
+        this.emit("user_speech_committed", userMsg);
+        this.#logger.child({ transcription }).debug("committed user speech");
+      });
+      this.#session.on("input_speech_started", (ev) => {
+        var _a2, _b;
+        if (this.#playingHandle && !this.#playingHandle.done) {
+          this.#playingHandle.interrupt();
+          this.#session.conversation.item.truncate(
+            this.#playingHandle.itemId,
+            this.#playingHandle.contentIndex,
+            Math.floor(this.#playingHandle.audioSamples / 24e3 * 1e3)
+          );
+          this.#playingHandle = void 0;
         }
+        const participantIdentity = (_a2 = this.linkedParticipant) == null ? void 0 : _a2.identity;
+        const trackSid = (_b = this.subscribedTrack) == null ? void 0 : _b.sid;
+        if (participantIdentity && trackSid) {
+          this.#publishTranscription(participantIdentity, trackSid, "\u2026", false, ev.itemId);
+        }
+      });
+      this.#session.on("input_speech_stopped", (ev) => {
+        this.emit("user_stopped_speaking");
+      });
+      this.#session.on("function_call_started", (ev) => {
+        this.#pendingFunctionCalls.add(ev.callId);
+        this.#updateState();
+      });
+      this.#session.on("function_call_completed", (ev) => {
+        this.#pendingFunctionCalls.delete(ev.callId);
+        this.#updateState();
+      });
+      this.#session.on("function_call_failed", (ev) => {
+        this.#pendingFunctionCalls.delete(ev.callId);
+        this.#updateState();
+      });
+      resolve(this.#session);
+    });
+  }
+  #linkParticipant(participantIdentity) {
+    if (!this.room) {
+      this.#logger.error("Room is not set");
+      return;
     }
-    get #pendingFunctionCalls() {
-        return this.#_pendingFunctionCalls;
+    this.linkedParticipant = this.room.remoteParticipants.get(participantIdentity) || null;
+    if (!this.linkedParticipant) {
+      this.#logger.error(`Participant with identity ${participantIdentity} not found`);
+      return;
     }
-    set #pendingFunctionCalls(calls) {
-        this.#_pendingFunctionCalls = calls;
-        this.#updateState();
+    if (this.linkedParticipant.trackPublications.size > 0) {
+      this.#subscribeToMicrophone();
     }
-    get #speaking() {
-        return this.#_speaking;
+    for (const publication of this.linkedParticipant.trackPublications.values()) {
+      if (publication.source === TrackSource.SOURCE_MICROPHONE && publication.track) {
+        this.#handleTrackSubscription(publication.track, publication, this.linkedParticipant);
+        break;
+      }
     }
-    set #speaking(isSpeaking) {
-        this.#_speaking = isSpeaking;
-        this.#updateState();
+  }
+  #subscribeToMicrophone() {
+    if (!this.linkedParticipant) {
+      this.#logger.error("Participant is not set");
+      return;
     }
-    get #started() {
-        return this.#_started;
+    let microphonePublication = void 0;
+    for (const publication of this.linkedParticipant.trackPublications.values()) {
+      if (publication.source === TrackSource.SOURCE_MICROPHONE) {
+        microphonePublication = publication;
+        break;
+      }
     }
-    set #started(started) {
-        this.#_started = started;
-        this.#updateState();
+    if (!microphonePublication) {
+      return;
     }
-    start(room, participant = null) {
-        return new Promise(async (resolve, reject) => {
-            if (this.#started) {
-                reject(new Error('MultimodalAgent already started'));
-            }
-            this.#updateState();
-            room.on(RoomEvent.ParticipantConnected, (participant) => {
-                // automatically link to the first participant that connects, if not already linked
-                if (this.linkedParticipant) {
-                    return;
-                }
-                this.#linkParticipant(participant.identity);
-            });
-            room.on(RoomEvent.TrackPublished, (trackPublication, participant) => {
-                if (this.linkedParticipant &&
-                    participant.identity === this.linkedParticipant.identity &&
-                    trackPublication.source === TrackSource.SOURCE_MICROPHONE &&
-                    !trackPublication.subscribed) {
-                    trackPublication.setSubscribed(true);
-                }
-            });
-            room.on(RoomEvent.TrackSubscribed, this.#handleTrackSubscription.bind(this));
-            this.room = room;
-            this.#participant = participant;
-            this.#localSource = new AudioSource(this.model.sampleRate, this.model.numChannels);
-            this.#agentPlayout = new AgentPlayout(this.#localSource, this.model.sampleRate, this.model.numChannels, this.model.inFrameSize, this.model.outFrameSize);
-            const onPlayoutStarted = () => {
-                this.emit('agent_started_speaking');
-                this.#speaking = true;
-            };
-            const onPlayoutStopped = (interrupted) => {
-                this.emit('agent_stopped_speaking');
-                this.#speaking = false;
-                if (this.#playingHandle) {
-                    let text = this.#playingHandle.transcriptionFwd.text;
-                    if (interrupted) {
-                        text += '…';
-                    }
-                    const msg = llm.ChatMessage.create({
-                        role: llm.ChatRole.ASSISTANT,
-                        text,
-                    });
-                    if (interrupted) {
-                        this.emit('agent_speech_interrupted', msg);
-                    }
-                    else {
-                        this.emit('agent_speech_committed', msg);
-                    }
-                    this.#logger.child({ transcription: text, interrupted }).debug('committed agent speech');
-                }
-            };
-            this.#agentPlayout.on('playout_started', onPlayoutStarted);
-            this.#agentPlayout.on('playout_stopped', onPlayoutStopped);
-            const track = LocalAudioTrack.createAudioTrack('assistant_voice', this.#localSource);
-            const options = new TrackPublishOptions();
-            options.source = TrackSource.SOURCE_MICROPHONE;
-            this.#agentPublication = (await room.localParticipant?.publishTrack(track, options)) || null;
-            if (!this.#agentPublication) {
-                this.#logger.error('Failed to publish track');
-                reject(new Error('Failed to publish track'));
-                return;
-            }
-            await this.#agentPublication.waitForSubscription();
-            if (participant) {
-                if (typeof participant === 'string') {
-                    this.#linkParticipant(participant);
-                }
-                else {
-                    this.#linkParticipant(participant.identity);
-                }
-            }
-            else {
-                // No participant specified, try to find the first participant in the room
-                for (const participant of room.remoteParticipants.values()) {
-                    this.#linkParticipant(participant.identity);
-                    break;
-                }
-            }
-            this.#session = this.model.session({ fncCtx: this.#fncCtx, chatCtx: this.#chatCtx });
-            this.#started = true;
-            // eslint-disable-next-line @typescript-eslint/no-explicit-any
-            this.#session.on('response_content_added', (message) => {
-                // openai.realtime.RealtimeContent
-                const trFwd = new BasicTranscriptionForwarder(this.room, this.room.localParticipant.identity, this.#getLocalTrackSid(), message.responseId);
-                const handle = this.#agentPlayout?.play(message.itemId, message.contentIndex, trFwd, message.textStream, message.audioStream);
-                this.#playingHandle = handle;
-            });
-            // eslint-disable-next-line @typescript-eslint/no-explicit-any
-            this.#session.on('input_speech_committed', (ev) => {
-                // openai.realtime.InputSpeechCommittedEvent
-                const participantIdentity = this.linkedParticipant?.identity;
-                const trackSid = this.subscribedTrack?.sid;
-                if (participantIdentity && trackSid) {
-                    this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);
-                }
-                else {
-                    this.#logger.error('Participant or track not set');
-                }
-            });
-            // eslint-disable-next-line @typescript-eslint/no-explicit-any
-            this.#session.on('input_speech_transcription_completed', (ev) => {
-                // openai.realtime.InputSpeechTranscriptionCompletedEvent
-                const transcription = ev.transcript;
-                const participantIdentity = this.linkedParticipant?.identity;
-                const trackSid = this.subscribedTrack?.sid;
-                if (participantIdentity && trackSid) {
-                    this.#publishTranscription(participantIdentity, trackSid, transcription, true, ev.itemId);
-                }
-                else {
-                    this.#logger.error('Participant or track not set');
-                }
-                const userMsg = llm.ChatMessage.create({
-                    role: llm.ChatRole.USER,
-                    text: transcription,
-                });
-                this.emit('user_speech_committed', userMsg);
-                this.#logger.child({ transcription }).debug('committed user speech');
-            });
-            this.#session.on('input_speech_started', (ev) => {
-                if (this.#playingHandle && !this.#playingHandle.done) {
-                    this.#playingHandle.interrupt();
-                    this.#session.conversation.item.truncate(this.#playingHandle.itemId, this.#playingHandle.contentIndex, Math.floor((this.#playingHandle.audioSamples / 24000) * 1000));
-                    this.#playingHandle = undefined;
-                }
-                const participantIdentity = this.linkedParticipant?.identity;
-                const trackSid = this.subscribedTrack?.sid;
-                if (participantIdentity && trackSid) {
-                    this.#publishTranscription(participantIdentity, trackSid, '…', false, ev.itemId);
-                }
-            });
-            // eslint-disable-next-line @typescript-eslint/no-unused-vars
-            this.#session.on('input_speech_stopped', (ev) => {
-                this.emit('user_stopped_speaking');
-            });
-            // eslint-disable-next-line @typescript-eslint/no-explicit-any
-            this.#session.on('function_call_started', (ev) => {
-                this.#pendingFunctionCalls.add(ev.callId);
-                this.#updateState();
-            });
-            // eslint-disable-next-line @typescript-eslint/no-explicit-any
-            this.#session.on('function_call_completed', (ev) => {
-                this.#pendingFunctionCalls.delete(ev.callId);
-                this.#updateState();
-            });
-            // eslint-disable-next-line @typescript-eslint/no-explicit-any
-            this.#session.on('function_call_failed', (ev) => {
-                this.#pendingFunctionCalls.delete(ev.callId);
-                this.#updateState();
-            });
-            resolve(this.#session);
-        });
+    if (!microphonePublication.subscribed) {
+      microphonePublication.setSubscribed(true);
     }
-    #linkParticipant(participantIdentity) {
-        if (!this.room) {
-            this.#logger.error('Room is not set');
-            return;
-        }
-        this.linkedParticipant = this.room.remoteParticipants.get(participantIdentity) || null;
-        if (!this.linkedParticipant) {
-            this.#logger.error(`Participant with identity ${participantIdentity} not found`);
-            return;
-        }
-        if (this.linkedParticipant.trackPublications.size > 0) {
-            this.#subscribeToMicrophone();
-        }
-        // also check if already subscribed
-        for (const publication of this.linkedParticipant.trackPublications.values()) {
-            if (publication.source === TrackSource.SOURCE_MICROPHONE && publication.track) {
-                this.#handleTrackSubscription(publication.track, publication, this.linkedParticipant);
-                break;
-            }
-        }
+  }
+  #handleTrackSubscription(track, publication, participant) {
+    var _a;
+    if (publication.source !== TrackSource.SOURCE_MICROPHONE || participant.identity !== ((_a = this.linkedParticipant) == null ? void 0 : _a.identity)) {
+      return;
     }
-    #subscribeToMicrophone() {
-        if (!this.linkedParticipant) {
-            this.#logger.error('Participant is not set');
-            return;
-        }
-        let microphonePublication = undefined;
-        for (const publication of this.linkedParticipant.trackPublications.values()) {
-            if (publication.source === TrackSource.SOURCE_MICROPHONE) {
-                microphonePublication = publication;
-                break;
-            }
-        }
-        if (!microphonePublication) {
-            return;
-        }
-        if (!microphonePublication.subscribed) {
-            microphonePublication.setSubscribed(true);
+    const readAudioStreamTask = async (audioStream) => {
+      const bstream = new AudioByteStream(
+        this.model.sampleRate,
+        this.model.numChannels,
+        this.model.inFrameSize
+      );
+      for await (const frame of audioStream) {
+        const audioData = frame.data;
+        for (const frame2 of bstream.write(audioData.buffer)) {
+          this.#session.inputAudioBuffer.append(frame2);
         }
+      }
+    };
+    this.subscribedTrack = track;
+    if (this.readMicroTask) {
+      this.readMicroTask.cancel();
     }
-    #handleTrackSubscription(track, publication, participant) {
-        if (publication.source !== TrackSource.SOURCE_MICROPHONE ||
-            participant.identity !== this.linkedParticipant?.identity) {
-            return;
-        }
-        const readAudioStreamTask = async (audioStream) => {
-            const bstream = new AudioByteStream(this.model.sampleRate, this.model.numChannels, this.model.inFrameSize);
-            for await (const frame of audioStream) {
-                const audioData = frame.data;
-                for (const frame of bstream.write(audioData.buffer)) {
-                    this.#session.inputAudioBuffer.append(frame);
-                }
-            }
-        };
-        this.subscribedTrack = track;
-        if (this.readMicroTask) {
-            this.readMicroTask.cancel();
-        }
-        let cancel;
-        this.readMicroTask = {
-            promise: new Promise((resolve, reject) => {
-                cancel = () => {
-                    reject(new Error('Task cancelled'));
-                };
-                readAudioStreamTask(new AudioStream(track, this.model.sampleRate, this.model.numChannels))
-                    .then(resolve)
-                    .catch(reject);
-            }),
-            cancel: () => cancel(),
+    let cancel;
+    this.readMicroTask = {
+      promise: new Promise((resolve, reject) => {
+        cancel = () => {
+          reject(new Error("Task cancelled"));
         };
+        readAudioStreamTask(new AudioStream(track, this.model.sampleRate, this.model.numChannels)).then(resolve).catch(reject);
+      }),
+      cancel: () => cancel()
+    };
+  }
+  #getLocalTrackSid() {
+    var _a;
+    if (!this.#localTrackSid && this.room && this.room.localParticipant) {
+      this.#localTrackSid = findMicroTrackId(this.room, (_a = this.room.localParticipant) == null ? void 0 : _a.identity);
     }
-    #getLocalTrackSid() {
-        if (!this.#localTrackSid && this.room && this.room.localParticipant) {
-            this.#localTrackSid = findMicroTrackId(this.room, this.room.localParticipant?.identity);
-        }
-        return this.#localTrackSid;
-    }
-    #publishTranscription(participantIdentity, trackSid, text, isFinal, id) {
-        this.#logger.debug(`Publishing transcription ${participantIdentity} ${trackSid} ${text} ${isFinal} ${id}`);
-        if (!this.room?.localParticipant) {
-            this.#logger.error('Room or local participant not set');
-            return;
-        }
-        this.room.localParticipant.publishTranscription({
-            participantIdentity,
-            trackSid,
-            segments: [
-                {
-                    text,
-                    final: isFinal,
-                    id,
-                    startTime: BigInt(0),
-                    endTime: BigInt(0),
-                    language: '',
-                },
-            ],
-        });
+    return this.#localTrackSid;
+  }
+  #publishTranscription(participantIdentity, trackSid, text, isFinal, id) {
+    var _a;
+    this.#logger.debug(
+      `Publishing transcription ${participantIdentity} ${trackSid} ${text} ${isFinal} ${id}`
+    );
+    if (!((_a = this.room) == null ? void 0 : _a.localParticipant)) {
+      this.#logger.error("Room or local participant not set");
+      return;
     }
-    #updateState() {
-        let newState = 'initializing';
-        if (this.#pendingFunctionCalls.size > 0) {
-            newState = 'thinking';
+    this.room.localParticipant.publishTranscription({
+      participantIdentity,
+      trackSid,
+      segments: [
+        {
+          text,
+          final: isFinal,
+          id,
+          startTime: BigInt(0),
+          endTime: BigInt(0),
+          language: ""
         }
-        else if (this.#speaking) {
-            newState = 'speaking';
-        }
-        else if (this.#started) {
-            newState = 'listening';
-        }
-        this.#setState(newState);
+      ]
+    });
+  }
+  #updateState() {
+    let newState = "initializing";
+    if (this.#pendingFunctionCalls.size > 0) {
+      newState = "thinking";
+    } else if (this.#speaking) {
+      newState = "speaking";
+    } else if (this.#started) {
+      newState = "listening";
     }
-    #setState(state) {
-        if (this.room?.isConnected && this.room.localParticipant) {
-            const currentState = this.room.localParticipant.attributes[AGENT_STATE_ATTRIBUTE];
-            if (currentState !== state) {
-                this.room.localParticipant.setAttributes({
-                    [AGENT_STATE_ATTRIBUTE]: state,
-                });
-                this.#logger.debug(`${AGENT_STATE_ATTRIBUTE}: ${currentState} ->${state}`);
-            }
-        }
+    this.#setState(newState);
+  }
+  #setState(state) {
+    var _a;
+    if (((_a = this.room) == null ? void 0 : _a.isConnected) && this.room.localParticipant) {
+      const currentState = this.room.localParticipant.attributes[AGENT_STATE_ATTRIBUTE];
+      if (currentState !== state) {
+        this.room.localParticipant.setAttributes({
+          [AGENT_STATE_ATTRIBUTE]: state
+        });
+        this.#logger.debug(`${AGENT_STATE_ATTRIBUTE}: ${currentState} ->${state}`);
+      }
     }
+  }
 }
+export {
+  AGENT_STATE_ATTRIBUTE,
+  MultimodalAgent,
+  RealtimeModel,
+  RealtimeSession
+};
 //# sourceMappingURL=multimodal_agent.js.map