@webex/internal-plugin-voicea 3.12.0-llmrefactor.1 → 3.12.0-llmrefactor.11

package/dist/voicea.types.js.map

@@ -1 +1 @@
-{"version":3,"names":[],"sources":["voicea.types.ts"],"sourcesContent":["/* eslint-disable camelcase */\n/* eslint-disable no-undef */\n\n/**\n * Type for payload\n */\ninterface AnnouncementPayload {\n  translation: {\n    max_languages: number;\n    allowed_languages: string[];\n  };\n\n  ASR: {\n    spoken_languages: string[];\n  };\n}\n\n// eslint-disable-next-line no-undef\ntype TRANSCRIPTION_TYPES =\n  | 'unknown'\n  | 'eva_wake'\n  | 'eva_thanks'\n  | 'eva_cancel'\n  | 'highlight_created'\n  | 'transcript_interim_results'\n  | 'transcript_final_result'\n  | 'manual_caption_interim_results'\n  | 'manual_caption_interim_result'\n  | 'manual_caption_final_result';\n\n/**\n * Class for an Transcription Object\n */\ninterface Transcription {\n  start_millis: number;\n  end_millis: number;\n  text: string;\n  transcript_language_code: string;\n  translations: {[x: string]: string};\n  csis: number[];\n  last_packet_timestamp_ms: number;\n  timestamp: string;\n}\n\n/**\n * Highlights\n */\ninterface Highlight {\n  highlight_id: string;\n  transcript: string;\n  highlight_label: string;\n  highlight_source: string;\n  start_millis: number;\n  end_millis: number;\n  csis: number[];\n}\n/**\n * Type for Transcription message\n */\ninterface TranscriptionResponse {\n  type: TRANSCRIPTION_TYPES;\n  id: string;\n  sender: string;\n  data_source: string;\n  transcript_id: string;\n  translations?: {[x: string]: string};\n  transcripts?: Transcription[];\n  transcript?: Transcription;\n  highlight?: Highlight;\n  csis: number[];\n  data: string;\n  command_response: string;\n  language: string;\n}\n/**\n * Type for CaptionLanguageResponse\n */\ninterface CaptionLanguageResponse {\n  requestId: string;\n  statusCode: number;\n  errorCode: number;\n  message: string;\n}\n\ninterface IVoiceaChannel {\n  setSpokenLanguage: (languageCode: string) => Promise<void>;\n  onSpokenLanguageUpdate: (languageCode: string, meetingId: string) => void;\n  requestLanguage: (languageCode: string) => void;\n  turnOnCaptions: () => undefined | Promise<void>;\n  toggleTranscribing: (activate: boolean, spokenLanguage: string) => undefined | Promise<void>;\n  deregisterEvents: () => void;\n  toggleManualCaption: (enable: boolean) => undefined | Promise<void>;\n  sendManualClosedCaption: (\n    text: string,\n    timeStamp: number,\n    csis: number[],\n    isFinal: boolean\n  ) => void;\n  getKeepTranscriptionSubscribed: () => boolean;\n}\n\ntype MeetingTranscripts = {\n  start_mills?: number;\n  end_mills?: number;\n  text: string;\n  last_packet_timestamp_ms?: number;\n  csis: Array<number>;\n  transcript_language_code: string;\n  translations?: {\n    [key: string]: string;\n  };\n  timestamp?: string;\n};\n\ntype MeetingTranscriptPayload = {\n  isFinal: boolean;\n  transcriptId: string;\n  transcripts: Array<MeetingTranscripts>;\n};\n\nexport type {\n  AnnouncementPayload,\n  CaptionLanguageResponse,\n  TranscriptionResponse,\n  Transcription,\n  Highlight,\n  IVoiceaChannel,\n  MeetingTranscriptPayload,\n};\n"],"mappings":"","ignoreList":[]}
+{"version":3,"names":[],"sources":["voicea.types.ts"],"sourcesContent":["/* eslint-disable camelcase */\n/* eslint-disable no-undef */\n\n/**\n * Type for payload\n */\ninterface AnnouncementPayload {\n  translation: {\n    max_languages: number;\n    allowed_languages: string[];\n  };\n\n  ASR: {\n    spoken_languages: string[];\n  };\n}\n\n// eslint-disable-next-line no-undef\ntype TRANSCRIPTION_TYPES =\n  | 'unknown'\n  | 'eva_wake'\n  | 'eva_thanks'\n  | 'eva_cancel'\n  | 'highlight_created'\n  | 'transcript_interim_results'\n  | 'transcript_final_result'\n  | 'manual_caption_interim_results'\n  | 'manual_caption_interim_result'\n  | 'manual_caption_final_result';\n\n/**\n * Class for an Transcription Object\n */\ninterface Transcription {\n  start_millis: number;\n  end_millis: number;\n  text: string;\n  transcript_language_code: string;\n  translations: {[x: string]: string};\n  csis: number[];\n  last_packet_timestamp_ms: number;\n  timestamp: string;\n  taggedSpeaker?: {\n    speakerId?: string[];\n    newName?: string;\n  };\n}\n\ninterface TaggedSpeaker {\n  csiId: number;\n  speakerId: string;\n  newName: string;\n}\n\ninterface SpeakerNameUpdatePayload {\n  id?: string;\n  taggedSpeakers: TaggedSpeaker[];\n}\n\n/**\n * Highlights\n */\ninterface Highlight {\n  highlight_id: string;\n  transcript: string;\n  highlight_label: string;\n  highlight_source: string;\n  start_millis: number;\n  end_millis: number;\n  csis: number[];\n}\n/**\n * Type for Transcription message\n */\ninterface TranscriptionResponse {\n  type: TRANSCRIPTION_TYPES;\n  id: string;\n  sender: string;\n  data_source: string;\n  transcript_id: string;\n  translations?: {[x: string]: string};\n  transcripts?: Transcription[];\n  transcript?: Transcription;\n  highlight?: Highlight;\n  csis: number[];\n  data: string;\n  command_response: string;\n  language: string;\n}\n/**\n * Type for CaptionLanguageResponse\n */\ninterface CaptionLanguageResponse {\n  requestId: string;\n  statusCode: number;\n  errorCode: number;\n  message: string;\n}\n\ninterface IVoiceaChannel {\n  setSpokenLanguage: (languageCode: string) => Promise<void>;\n  onSpokenLanguageUpdate: (languageCode: string, meetingId: string) => void;\n  requestLanguage: (languageCode: string) => void;\n  turnOnCaptions: () => undefined | Promise<void>;\n  toggleTranscribing: (activate: boolean, spokenLanguage: string) => undefined | Promise<void>;\n  deregisterEvents: () => void;\n  toggleManualCaption: (enable: boolean) => undefined | Promise<void>;\n  sendManualClosedCaption: (\n    text: string,\n    timeStamp: number,\n    csis: number[],\n    isFinal: boolean\n  ) => void;\n  getKeepTranscriptionSubscribed: () => boolean;\n}\n\ntype MeetingTranscripts = {\n  start_mills?: number;\n  end_mills?: number;\n  text: string;\n  last_packet_timestamp_ms?: number;\n  csis: Array<number>;\n  transcript_language_code: string;\n  translations?: {\n    [key: string]: string;\n  };\n  timestamp?: string;\n  taggedSpeaker?: {\n    speakerId?: string[];\n    newName?: string;\n  };\n};\n\ntype MeetingTranscriptPayload = {\n  isFinal: boolean;\n  transcriptId: string;\n  transcripts: Array<MeetingTranscripts>;\n};\n\nexport type {\n  AnnouncementPayload,\n  CaptionLanguageResponse,\n  TranscriptionResponse,\n  Transcription,\n  Highlight,\n  TaggedSpeaker,\n  SpeakerNameUpdatePayload,\n  IVoiceaChannel,\n  MeetingTranscriptPayload,\n};\n"],"mappings":"","ignoreList":[]}

package/package.json

@@ -14,9 +14,9 @@
     "node": ">=18"
   },
   "dependencies": {
-    "@webex/internal-plugin-llm": "3.12.0-llmrefactor.1",
-    "@webex/internal-plugin-mercury": "3.12.0-llmrefactor.1",
-    "@webex/webex-core": "3.12.0-llmrefactor.1",
+    "@webex/internal-plugin-llm": "3.12.0-llmrefactor.9",
+    "@webex/internal-plugin-mercury": "3.12.0-llmrefactor.9",
+    "@webex/webex-core": "3.12.0-llmrefactor.7",
     "uuid": "^3.3.2"
   },
   "browserify": {
@@ -31,11 +31,11 @@
     "@webex/eslint-config-legacy": "0.0.0",
     "@webex/jest-config-legacy": "0.0.0",
     "@webex/legacy-tools": "0.0.0",
-    "@webex/test-helper-chai": "3.12.0-llmrefactor.1",
-    "@webex/test-helper-mocha": "3.12.0-llmrefactor.1",
-    "@webex/test-helper-mock-web-socket": "3.12.0-llmrefactor.1",
+    "@webex/test-helper-chai": "3.12.0-llmrefactor.2",
+    "@webex/test-helper-mocha": "3.12.0-llmrefactor.2",
+    "@webex/test-helper-mock-web-socket": "3.12.0-llmrefactor.2",
     "@webex/test-helper-mock-webex": "3.12.0-llmrefactor.1",
-    "@webex/test-helper-test-users": "3.12.0-llmrefactor.1",
+    "@webex/test-helper-test-users": "3.12.0-llmrefactor.0",
     "eslint": "^8.24.0",
     "prettier": "^2.7.1",
     "sinon": "^9.2.4"
@@ -49,5 +49,5 @@
     "test:style": "eslint ./src/**/*.*",
     "test:unit": "webex-legacy-tools test --unit --runner jest"
   },
-  "version": "3.12.0-llmrefactor.1"
+  "version": "3.12.0-llmrefactor.11"
 }

package/src/constants.ts

@@ -7,6 +7,7 @@ export const EVENT_TRIGGERS = {
   TRANSCRIBING_OFF: 'voicea:transcribingOff',
 
   NEW_CAPTION: 'voicea:newCaption',
+  SPEAKER_NAME_UPDATED: 'voicea:speakerNameUpdated',
   EVA_COMMAND: 'voicea:wxa',
   HIGHLIGHT_CREATED: 'voicea:highlightCreated',
   NEW_MANUAL_CAPTION: 'aibridge:newManualCaption',
@@ -21,6 +22,7 @@ export const AIBRIDGE_RELAY_TYPES = {
     TRANSLATION_REQUEST: 'voicea.transl.req',
     TRANSLATION_RESPONSE: 'voicea.transl.rsp',
     TRANSCRIPTION: 'voicea.transcription',
+    SPEAKER_NAME_UPDATE: 'voicea.update_speakername',
   },
   MANUAL: {
     TRANSCRIPTION: 'aibridge.manual_transcription',

package/src/index.ts

@@ -2,11 +2,11 @@ import * as WebexCore from '@webex/webex-core';
 
 import Voicea from './voicea-plugin';
 import {VoiceaChannel} from './voicea';
-import type {MeetingTranscriptPayload} from './voicea.types';
+import type {MeetingTranscriptPayload, SpeakerNameUpdatePayload} from './voicea.types';
 
 WebexCore.registerInternalPlugin('voicea', Voicea, {});
 
 export {Voicea as VoiceaPlugin, VoiceaChannel};
 export default Voicea;
-export {type MeetingTranscriptPayload};
+export {type MeetingTranscriptPayload, type SpeakerNameUpdatePayload};
 export {EVENT_TRIGGERS, TURN_ON_CAPTION_STATUS} from './constants';

package/src/voicea-plugin.ts

@@ -13,16 +13,16 @@ export class VoiceaPlugin extends WebexPlugin {
   namespace = VOICEA;
 
   /**
-   * Creates a VoiceaChannel bound to the given LLMChannel.
+   * Creates a VoiceaChannel, optionally bound to an LLMChannel.
    *
    * Note: VoiceaChannel does not take ownership of the LLMChannel.
    * The caller retains ownership and is responsible for the LLMChannel's
    * lifecycle (connection, disconnection, cleanup).
    *
-   * @param {LLMChannel} llmChannel - The LLM channel to use for voicea
+   * @param {LLMChannel} [llmChannel] - The LLM channel to use for voicea (optional)
    * @returns {VoiceaChannel} A new VoiceaChannel instance
    */
-  public createChannel(llmChannel: LLMChannel): VoiceaChannel {
+  public createChannel(llmChannel?: LLMChannel): VoiceaChannel {
     // @ts-ignore - webex is available on WebexPlugin
     const channel = new VoiceaChannel(llmChannel, this.webex.request.bind(this.webex));
 

package/src/voicea.ts

@@ -17,6 +17,7 @@ import {
 import {
   AnnouncementPayload,
   CaptionLanguageResponse,
+  SpeakerNameUpdatePayload,
   TranscriptionResponse,
   IVoiceaChannel,
 } from './voicea.types';
@@ -31,7 +32,7 @@ import {millisToMinutesAndSeconds} from './utils';
  */
 export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
   private request: (options: {method: string; url: string; body?: object}) => Promise<any>;
-  private llmChannel: LLMChannel;
+  private llmChannel?: LLMChannel;
 
   private seqNum: number;
   private areCaptionsEnabled: boolean;
@@ -47,13 +48,21 @@ export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
   private spokenLanguages: string[] = [];
   private currentCaptionLanguage?: string;
 
+  // Target channel for reconciler pattern - the channel voicea WANTS to be bound to
+  private targetLLMChannel?: LLMChannel;
+  // Single pending 'online' listener for deferred reconciliation
+  private _pendingOnlineListener?: () => void;
+  // Tracks whether caption restoration is pending (set on switch, cleared on restore)
+  private _pendingCaptionRestore = false;
+
   /**
-   * Creates a VoiceaChannel bound to the given LLMChannel
-   * @param {LLMChannel} llmChannel - The LLM channel to use
+   * Creates a VoiceaChannel, optionally bound to an LLMChannel.
+   * If no llmChannel is provided, call switchLLMChannel() later to attach one.
+   * @param {LLMChannel} [llmChannel] - The LLM channel to use (optional)
    * @param {Function} request - The request function for making API calls (typically webex.request bound to webex)
    */
   constructor(
-    llmChannel: LLMChannel,
+    llmChannel: LLMChannel | undefined,
     request: (options: {method: string; url: string; body?: object}) => Promise<any>
   ) {
     super();
@@ -70,9 +79,25 @@ export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
     this.currentCaptionLanguage = undefined;
     this.keepTranscriptionSubscribed = false;
 
-    // Subscribe to relay events from the LLM channel
-    this.llmChannel.on('event:relay.event', this.eventProcessor);
-    this.hasSubscribedToEvents = true;
+    // Subscribe to relay events from the LLM channel if provided
+    if (this.llmChannel) {
+      this.llmChannel.on('event:relay.event', this.eventProcessor);
+      this.hasSubscribedToEvents = true;
+    }
+  }
+
+  /**
+   * Returns the LLM channel, throwing if not available.
+   * @private
+   * @returns {LLMChannel}
+   * @throws {Error} If LLM channel is not available
+   */
+  private requireLLMChannel(): LLMChannel {
+    if (!this.llmChannel) {
+      throw new Error('VoiceaChannel: LLM channel not available');
+    }
+
+    return this.llmChannel;
   }
 
   /**
@@ -94,6 +119,9 @@ export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
       case AIBRIDGE_RELAY_TYPES.VOICEA.TRANSCRIPTION:
         this.processTranscription(e.data.voiceaPayload);
         break;
+      case AIBRIDGE_RELAY_TYPES.VOICEA.SPEAKER_NAME_UPDATE:
+        this.processSpeakerNameUpdate(e.data.voiceaPayload);
+        break;
       case AIBRIDGE_RELAY_TYPES.MANUAL.TRANSCRIPTION:
       case AIBRIDGE_RELAY_TYPES.MANUAL.CAPTIONER:
         this.processManualTranscription({
@@ -115,9 +143,14 @@ export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
     this.areCaptionsEnabled = false;
     this.keepTranscriptionSubscribed = false;
     this.captionServiceId = undefined;
+    this.targetLLMChannel = undefined;
+    this._pendingCaptionRestore = false;
+
+    // Remove any pending online listener
+    this.detachPendingOnline();
 
     if (this.hasSubscribedToEvents) {
-      this.llmChannel.off('event:relay.event', this.eventProcessor);
+      this.llmChannel?.off('event:relay.event', this.eventProcessor);
       this.hasSubscribedToEvents = false;
     }
 
@@ -129,42 +162,101 @@ export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
   }
 
   /**
-   * Switch to a different LLM channel while preserving caption state.
-   * Used when transitioning between main meeting and practice session.
-   * - Preserves keepTranscriptionSubscribed and spokenLanguage state
-   * - Unsubscribes from old channel, subscribes to new channel
-   * - Re-announces and re-enables captions if they were on
-   * @param {LLMChannel} newLLMChannel - The new LLM channel to switch to
-   * @returns {Promise<void>}
+   * Attach a pending 'online' listener to a channel for deferred reconciliation.
+   * Detaches any existing listener first.
+   * @param {LLMChannel} channel - The channel to attach to
+   * @param {Function} callback - The callback to run when 'online' fires
+   * @private
+   * @returns {void}
    */
-  public async switchLLMChannel(newLLMChannel: LLMChannel): Promise<void> {
-    // Save current state
-    const captionsWereOn = this.keepTranscriptionSubscribed;
-    const spokenLanguage = this.currentSpokenLanguage;
+  private attachPendingOnline(channel: LLMChannel, callback: () => void): void {
+    this.detachPendingOnline();
+    this._pendingOnlineListener = callback;
+    channel.once('online', this._pendingOnlineListener);
+  }
 
-    // Unsubscribe from old channel
-    if (this.hasSubscribedToEvents && this.llmChannel) {
-      this.llmChannel.off('event:relay.event', this.eventProcessor);
-      this.hasSubscribedToEvents = false;
+  /**
+   * Detach any pending 'online' listener.
+   * @private
+   * @returns {void}
+   */
+  private detachPendingOnline(): void {
+    if (this._pendingOnlineListener && this.llmChannel) {
+      this.llmChannel.off('online', this._pendingOnlineListener);
     }
+    this._pendingOnlineListener = undefined;
+  }
 
-    // Switch to new channel
-    this.llmChannel = newLLMChannel;
-
-    // Subscribe to new channel
-    this.llmChannel.on('event:relay.event', this.eventProcessor);
-    this.hasSubscribedToEvents = true;
-
-    // Reset announcement state for new connection
+  /**
+   * Reset announcement state for a new connection.
+   * @private
+   * @returns {void}
+   */
+  private resetAnnounceState(): void {
     this.announceStatus = ANNOUNCE_STATUS.IDLE;
     this.captionStatus = TURN_ON_CAPTION_STATUS.IDLE;
     this.captionServiceId = undefined;
     this.areCaptionsEnabled = false;
+  }
 
-    // Re-announce and re-enable captions if they were on
-    if (captionsWereOn) {
-      await this.turnOnCaptions(spokenLanguage);
+  /**
+   * Reconcile voicea binding and caption state to match the desired target.
+   * This is idempotent - always re-reads targetLLMChannel and current state.
+   * A deferred 'online' callback that fires late self-corrects automatically.
+   * @private
+   * @returns {Promise<void>}
+   */
+  private reconcile(): Promise<void> {
+    const target = this.targetLLMChannel;
+    if (!target) return Promise.resolve();
+
+    // 1. Rebind relay-event subscription if actual != desired
+    if (this.llmChannel !== target) {
+      if (this.hasSubscribedToEvents && this.llmChannel) {
+        this.llmChannel.off('event:relay.event', this.eventProcessor);
+      }
+      this.detachPendingOnline();
+      this.llmChannel = target;
+      this.llmChannel.on('event:relay.event', this.eventProcessor);
+      this.hasSubscribedToEvents = true;
+      this.resetAnnounceState();
+      // Mark caption restoration as pending if captions were subscribed
+      this._pendingCaptionRestore = this.keepTranscriptionSubscribed;
     }
+
+    // 2. Restore captions if pending
+    if (!this._pendingCaptionRestore) return Promise.resolve();
+
+    if (this.isLLMConnected()) {
+      // Channel is connected, restore captions immediately
+      this._pendingCaptionRestore = false;
+
+      return this.turnOnCaptions(this.currentSpokenLanguage)
+        .then(() => undefined)
+        .catch(() => {
+          // Best-effort restoration
+        });
+    }
+    // Channel not yet connected - defer until 'online' then re-reconcile.
+    // If target changes meanwhile, reconcile() self-corrects.
+    this.attachPendingOnline(target, () => {
+      this.reconcile();
+    });
+
+    return Promise.resolve();
+  }
+
+  /**
+   * Switch to a different LLM channel while preserving caption state.
+   * Used when transitioning between main meeting and practice session.
+   * This is a thin intent-setter - actual binding happens in reconcile().
+   * @param {LLMChannel} newLLMChannel - The new LLM channel to switch to
+   * @returns {Promise<void>}
+   */
+  public switchLLMChannel(newLLMChannel: LLMChannel): Promise<void> {
+    this.targetLLMChannel = newLLMChannel;
+
+    return this.reconcile();
   }
 
   /**
@@ -187,6 +279,16 @@ export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
     }
   };
 
+  /**
+   * Process speaker name update and send alert
+   * @param {SpeakerNameUpdatePayload} voiceaPayload
+   * @returns {void}
+   */
+  private processSpeakerNameUpdate = (voiceaPayload: SpeakerNameUpdatePayload): void => {
+    // @ts-ignore
+    this.emit(EVENT_TRIGGERS.SPEAKER_NAME_UPDATED, voiceaPayload);
+  };
+
   /**
    * Process Transcript and send alert
    * @param {TranscriptionResponse} voiceaPayload
@@ -290,7 +392,7 @@ export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
    * Indicates whether the LLM channel is connected.
    * @returns {boolean}
    */
-  public isLLMConnected = (): boolean => this.llmChannel.isConnected();
+  public isLLMConnected = (): boolean => this.llmChannel?.isConnected() ?? false;
 
   public getKeepTranscriptionSubscribed = (): boolean => this.keepTranscriptionSubscribed;
 
@@ -299,9 +401,10 @@ export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
    * @returns {void}
    */
   public sendAnnouncement = (): void => {
+    const llm = this.requireLLMChannel();
     this.announceStatus = ANNOUNCE_STATUS.JOINING;
-    const socket = this.llmChannel.getSocket();
-    const binding = this.llmChannel.getBinding();
+    const socket = llm.getSocket();
+    const binding = llm.getBinding();
 
     const payload = {
       id: `${this.seqNum}`,
@@ -338,7 +441,7 @@ export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
   ): Promise<void> =>
     this.request({
       method: 'PUT',
-      url: `${this.llmChannel.getLocusUrl()}/controls/`,
+      url: `${this.requireLLMChannel().getLocusUrl()}/controls/`,
       body: {
         transcribe: {
           spokenLanguage: languageCode,
@@ -359,8 +462,9 @@ export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
       return;
     }
 
-    const socket = this.llmChannel.getSocket();
-    const binding = this.llmChannel.getBinding();
+    const llm = this.requireLLMChannel();
+    const socket = llm.getSocket();
+    const binding = llm.getBinding();
 
     socket.send({
       id: `${this.seqNum}`,
@@ -405,8 +509,9 @@ export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
       return;
     }
 
-    const socket = this.llmChannel.getSocket();
-    const binding = this.llmChannel.getBinding();
+    const llm = this.requireLLMChannel();
+    const socket = llm.getSocket();
+    const binding = llm.getBinding();
 
     socket?.send({
       id: `${this.seqNum}`,
@@ -449,7 +554,7 @@ export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
   private requestTurnOnCaptions = (languageCode?: string): undefined | Promise<void> => {
     this.captionStatus = TURN_ON_CAPTION_STATUS.SENDING;
 
-    const locusUrl = this.llmChannel.getLocusUrl();
+    const locusUrl = this.requireLLMChannel().getLocusUrl();
 
     const body = {
       transcribe: {caption: true},
@@ -471,7 +576,7 @@ export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
       })
       .catch((error) => {
         this.captionStatus = TURN_ON_CAPTION_STATUS.IDLE;
-        throw new Error('turn on captions fail');
+        throw new Error('turn on captions fail', {cause: error});
       });
   };
 
@@ -515,7 +620,7 @@ export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
    * @returns {Promise}
    */
   public turnOnCaptions = async (spokenLanguage?: string): Promise<void | undefined> => {
-    if (this.captionStatus === TURN_ON_CAPTION_STATUS.SENDING) return undefined;
+    if (this.isCaptionProcessing()) return undefined;
 
     if (!this.isLLMConnected()) {
       throw new Error('can not turn on captions before llm connected');
@@ -536,7 +641,7 @@ export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
   ): undefined | Promise<void> => {
     return this.request({
       method: 'PUT',
-      url: `${this.llmChannel.getLocusUrl()}/controls/`,
+      url: `${this.requireLLMChannel().getLocusUrl()}/controls/`,
       body: {
         transcribe: {
           transcribing: activate,
@@ -564,7 +669,7 @@ export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
 
     return this.request({
       method: 'PUT',
-      url: `${this.llmChannel.getLocusUrl()}/controls/`,
+      url: `${this.requireLLMChannel().getLocusUrl()}/controls/`,
       body: {
         manualCaption: {
           enable,
@@ -638,11 +743,12 @@ export class VoiceaChannel extends EventEmitter implements IVoiceaChannel {
   } = {}): Promise<void> => {
     if (!this.isLLMConnected()) return;
 
-    const isDataChannelTokenEnabled = await this.llmChannel.isDataChannelTokenEnabled();
+    const llm = this.requireLLMChannel();
+    const isDataChannelTokenEnabled = await llm.isDataChannelTokenEnabled();
     if (!isDataChannelTokenEnabled) return;
 
-    const socket = this.llmChannel.getSocket();
-    const datachannelUrl = this.llmChannel.getDatachannelUrl();
+    const socket = llm.getSocket();
+    const datachannelUrl = llm.getDatachannelUrl();
 
     socket.send({
       id: `${this.seqNum}`,

package/src/voicea.types.ts

@@ -40,6 +40,21 @@ interface Transcription {
   csis: number[];
   last_packet_timestamp_ms: number;
   timestamp: string;
+  taggedSpeaker?: {
+    speakerId?: string[];
+    newName?: string;
+  };
+}
+
+interface TaggedSpeaker {
+  csiId: number;
+  speakerId: string;
+  newName: string;
+}
+
+interface SpeakerNameUpdatePayload {
+  id?: string;
+  taggedSpeakers: TaggedSpeaker[];
 }
 
 /**
@@ -110,6 +125,10 @@ type MeetingTranscripts = {
     [key: string]: string;
   };
   timestamp?: string;
+  taggedSpeaker?: {
+    speakerId?: string[];
+    newName?: string;
+  };
 };
 
 type MeetingTranscriptPayload = {
@@ -124,6 +143,8 @@ export type {
   TranscriptionResponse,
   Transcription,
   Highlight,
+  TaggedSpeaker,
+  SpeakerNameUpdatePayload,
   IVoiceaChannel,
   MeetingTranscriptPayload,
 };

package/test/unit/spec/voicea-plugin.js

@@ -51,6 +51,18 @@ describe('plugin-voicea', () => {
         assert.instanceOf(channel, VoiceaChannel);
       });
 
+      it('creates a new VoiceaChannel instance without llmChannel', () => {
+        const channel = plugin.createChannel();
+
+        assert.instanceOf(channel, VoiceaChannel);
+      });
+
+      it('creates a new VoiceaChannel instance with undefined llmChannel', () => {
+        const channel = plugin.createChannel(undefined);
+
+        assert.instanceOf(channel, VoiceaChannel);
+      });
+
       it('creates independent channels for multiple calls', () => {
         const mockLLMChannel1 = createMockLLMChannel();
         const mockLLMChannel2 = createMockLLMChannel();