feat(mobile-app): prepare native app packaging

2026-07-23 14:40:23 +08:00
parent f6609808df
commit 0f5c2e8667
62 changed files with 979 additions and 833 deletions
@@ -159,9 +159,11 @@ import {
  createSpeechPlaybackController,
  INTERVIEWER_VOICE_PROFILE,
  normalizeOssId,
-  transcribeSpeechBlob,
+  transcribeSpeechCapture,
  transcribeSpeechFile
 } from '@/services/speech';
+import { startSpeechCapture } from '@/services/speech-capture';
+import type { SpeechCapture, SpeechCaptureSession } from '@/services/speech-capture';

 const currentPath = '/pages/candidate/interview/index';
 const records = ref<InterviewRecord[]>([]);
@@ -177,8 +179,7 @@ const message = ref('');
 const transcribingQuestionId = ref('');
 const recordingQuestionId = ref('');
 const recordingQuestionStarting = ref(false);
-const mediaRecorder = ref<MediaRecorder | null>(null);
-const mediaStream = ref<MediaStream | null>(null);
+let questionCapture: SpeechCaptureSession | null = null;
 let questionCaptureGeneration = 0;
 let questionTranscriptionGeneration = 0;
 let abortQuestionTranscription: (() => void) | null = null;
@@ -298,11 +299,6 @@ const interviewResultSpeech = () => {
  ].filter(Boolean).join('。');
 };

-const stopQuestionMediaTracks = () => {
-  mediaStream.value?.getTracks().forEach((track) => track.stop());
-  mediaStream.value = null;
-};
-
 const cancelQuestionTranscription = () => {
  questionTranscriptionGeneration += 1;
  abortQuestionTranscription?.();
@@ -314,14 +310,8 @@ const cancelQuestionRecording = () => {
  questionCaptureGeneration += 1;
  cancelQuestionTranscription();
  recordingQuestionStarting.value = false;
-  const recorder = mediaRecorder.value;
-  if (recorder && recorder.state !== 'inactive') {
-    recorder.ondataavailable = null;
-    recorder.onstop = null;
-    recorder.stop();
-  }
-  mediaRecorder.value = null;
-  stopQuestionMediaTracks();
+  questionCapture?.cancel();
+  questionCapture = null;
  recordingQuestionId.value = '';
 };

@@ -361,13 +351,13 @@ const transcribeQuestionFile = async (questionId: string) => {
  }
 };

-const transcribeQuestionBlob = async (questionId: string, blob: Blob) => {
+const transcribeQuestionCapture = async (questionId: string, capture: SpeechCapture) => {
  if (transcribingQuestionId.value) return;
  const generation = ++questionTranscriptionGeneration;
  transcribingQuestionId.value = questionId;
  message.value = '语音转写中';
  try {
-    const result = await transcribeSpeechBlob(blob, 'candidate-answer.webm', (abort) => {
+    const result = await transcribeSpeechCapture(capture, 'candidate-answer.webm', (abort) => {
      if (generation === questionTranscriptionGeneration) abortQuestionTranscription = abort;
    });
    if (generation !== questionTranscriptionGeneration) return;
@@ -385,42 +375,35 @@ const transcribeQuestionBlob = async (questionId: string, blob: Blob) => {

 const startQuestionRecording = async (questionId: string) => {
  if (interviewBusy.value || recordingQuestionStarting.value || recordingQuestionId.value || transcribingQuestionId.value) return;
-  if (typeof navigator === 'undefined' || !navigator.mediaDevices?.getUserMedia || typeof MediaRecorder === 'undefined') {
-    await transcribeQuestionFile(questionId);
-    return;
-  }
  const generation = ++questionCaptureGeneration;
  recordingQuestionStarting.value = true;
+  let fallbackRequested = false;
  try {
-    const stream = await navigator.mediaDevices.getUserMedia({ audio: true });
-    if (generation !== questionCaptureGeneration) {
-      stream.getTracks().forEach((track) => track.stop());
+    const session = await startSpeechCapture({
+      onCapture: (capture) => {
+        if (generation !== questionCaptureGeneration) return;
+        questionCapture = null;
+        recordingQuestionId.value = '';
+        void transcribeQuestionCapture(questionId, capture);
+      },
+      onError: () => {
+        if (generation !== questionCaptureGeneration) return;
+        fallbackRequested = true;
+        questionCapture = null;
+        recordingQuestionId.value = '';
+        void transcribeQuestionFile(questionId);
+      }
+    });
+    if (!session || generation !== questionCaptureGeneration) {
+      session?.cancel();
+      if (generation === questionCaptureGeneration && !fallbackRequested) await transcribeQuestionFile(questionId);
      return;
    }
-    mediaStream.value = stream;
-    const recorder = new MediaRecorder(stream);
-    const chunks: Blob[] = [];
-    mediaRecorder.value = recorder;
+    questionCapture = session;
    recordingQuestionId.value = questionId;
-    recorder.ondataavailable = (event) => {
-      if (event.data?.size) chunks.push(event.data);
-    };
-    recorder.onstop = () => {
-      mediaRecorder.value = null;
-      recordingQuestionId.value = '';
-      stopQuestionMediaTracks();
-      const blob = new Blob(chunks, { type: recorder.mimeType || 'audio/webm' });
-      if (!blob.size) {
-        message.value = '未录到音频，请改用文字输入';
-        return;
-      }
-      void transcribeQuestionBlob(questionId, blob);
-    };
-    recorder.start();
    message.value = '正在录音';
  } catch {
    recordingQuestionId.value = '';
-    stopQuestionMediaTracks();
    if (generation === questionCaptureGeneration) await transcribeQuestionFile(questionId);
  } finally {
    if (generation === questionCaptureGeneration) recordingQuestionStarting.value = false;
@@ -428,10 +411,9 @@ const startQuestionRecording = async (questionId: string) => {
 };

 const stopQuestionRecording = () => {
-  const recorder = mediaRecorder.value;
-  if (!recordingQuestionId.value || !recorder || recorder.state === 'inactive') return;
+  if (!recordingQuestionId.value || !questionCapture) return;
  message.value = '录音结束，正在转写';
-  recorder.stop();
+  questionCapture.stop();
 };

 const toggleQuestionRecording = (questionId: string) => {
@@ -306,7 +306,7 @@
        :maxlength="500"
        :panel-items="practicePanelItems"
        @send="handleComposerSend"
-        @voice-blob="handleSessionVoiceBlob"
+        @voice-capture="handleSessionVoiceCapture"
        @pick-audio="composerPickAudio"
        @panel="handlePanelAction"
      />
@@ -397,7 +397,9 @@ import { getSelectedPosition, isLoggedIn, rememberLoginRedirect } from '@/servic
 import { searchKnowledge } from '@/services/knowledge';
 import { resetPageScroll } from '@/services/navigation';
 import { ensureEmployeePosition } from '@/services/position';
-import { chooseSpeechAudio, createSpeechPlaybackController, customerVoiceProfile, measureAudioDuration, MENTOR_VOICE_PROFILE, normalizeOssId, transcribeSpeechBlob, transcribeSpeechFile } from '@/services/speech';
+import { chooseSpeechAudio, createSpeechPlaybackController, customerVoiceProfile, measureAudioDuration, MENTOR_VOICE_PROFILE, normalizeOssId, transcribeSpeechCapture, transcribeSpeechFile } from '@/services/speech';
+import { startSpeechCapture } from '@/services/speech-capture';
+import type { SpeechCapture, SpeechCaptureSession } from '@/services/speech-capture';
 import { createPageRequestScope, currentAccountKey, type RequestScopeTicket } from '@/services/request-scope';
 import { createRealtimePracticeController, DEFAULT_REALTIME_PRACTICE_VOICE, isRealtimePracticeVoice, REALTIME_PRACTICE_VOICE_OPTIONS, type RealtimePracticeVoice } from '@/services/realtime-practice';
 import { stripInternalCodes } from '@/services/text';
@@ -473,8 +475,7 @@ const transcribing = ref(false);
 const draftAudioUrl = ref('');
 const draftAudioOssId = ref<string | undefined>();
 const speechState = ref<{ key: string; status: SpeechPlaybackStatus }>({ key: '', status: 'idle' });
-const mediaRecorder = ref<MediaRecorder | null>(null);
-const mediaStream = ref<MediaStream | null>(null);
+let recordingCapture: SpeechCaptureSession | null = null;
 let recordingCaptureGeneration = 0;
 let transcriptionGeneration = 0;
 let abortTranscription: (() => void) | null = null;
@@ -665,12 +666,12 @@ const submitTranscribed = async (text: string, audioUrl: string, ossId?: string)
  await submit();
 };

-const handleSessionVoiceBlob = async (blob: Blob) => {
+const handleSessionVoiceCapture = async (capture: SpeechCapture) => {
  if (transcribing.value || busy.value) return;
  const generation = ++transcriptionGeneration;
  transcribing.value = true;
  try {
-    const asr = await transcribeSpeechBlob(blob, 'practice-turn.webm', (abort) => {
+    const asr = await transcribeSpeechCapture(capture, 'practice-turn.webm', (abort) => {
      if (generation === transcriptionGeneration) abortTranscription = abort;
    });
    if (generation !== transcriptionGeneration) return;
@@ -907,11 +908,6 @@ const clearDraftAudio = () => {
  draftAudioOssId.value = undefined;
 };

-const stopMediaTracks = () => {
-  mediaStream.value?.getTracks().forEach((track) => track.stop());
-  mediaStream.value = null;
-};
-
 const cancelTranscription = () => {
  transcriptionGeneration += 1;
  abortTranscription?.();
@@ -923,14 +919,8 @@ const cancelRecording = () => {
  recordingCaptureGeneration += 1;
  cancelTranscription();
  recordingStarting.value = false;
-  const recorder = mediaRecorder.value;
-  if (recorder && recorder.state !== 'inactive') {
-    recorder.ondataavailable = null;
-    recorder.onstop = null;
-    recorder.stop();
-  }
-  mediaRecorder.value = null;
-  stopMediaTracks();
+  recordingCapture?.cancel();
+  recordingCapture = null;
  recording.value = false;
  recordingAssignmentId.value = null;
 };
@@ -965,13 +955,13 @@ const applyDailyAsrResult = (assignmentId: number, result: { text?: string; audi
  message.value = '每日一练语音已转成文字，请确认后提交';
 };

-const transcribeBlob = async (blob: Blob, assignmentId?: number) => {
+const transcribeCapture = async (capture: SpeechCapture, assignmentId?: number) => {
  if (transcribing.value) return;
  const generation = ++transcriptionGeneration;
  transcribing.value = true;
  message.value = '语音转写中';
  try {
-    const result = await transcribeSpeechBlob(blob, 'practice-audio.webm', (abort) => {
+    const result = await transcribeSpeechCapture(capture, 'practice-audio.webm', (abort) => {
      if (generation === transcriptionGeneration) abortTranscription = abort;
    });
    if (generation !== transcriptionGeneration) return;
@@ -1023,45 +1013,40 @@ const pickAudio = async (assignmentId?: number) => {
 const startRecording = async (assignmentId?: number) => {
  if (busy.value || recordingStarting.value || recording.value || transcribing.value) return;
  if (!requireLogin()) return;
-  if (typeof navigator === 'undefined' || !navigator.mediaDevices?.getUserMedia || typeof MediaRecorder === 'undefined') {
-    await pickAudio(assignmentId);
-    return;
-  }
  const generation = ++recordingCaptureGeneration;
  recordingStarting.value = true;
+  let fallbackRequested = false;
  try {
-    const stream = await navigator.mediaDevices.getUserMedia({ audio: true });
-    if (generation !== recordingCaptureGeneration) {
-      stream.getTracks().forEach((track) => track.stop());
+    const session = await startSpeechCapture({
+      onCapture: (capture) => {
+        if (generation !== recordingCaptureGeneration) return;
+        recordingCapture = null;
+        recording.value = false;
+        const targetAssignmentId = assignmentId;
+        recordingAssignmentId.value = null;
+        void transcribeCapture(capture, targetAssignmentId);
+      },
+      onError: () => {
+        if (generation !== recordingCaptureGeneration) return;
+        fallbackRequested = true;
+        recordingCapture = null;
+        recording.value = false;
+        recordingAssignmentId.value = null;
+        void pickAudio(assignmentId);
+      }
+    });
+    if (!session || generation !== recordingCaptureGeneration) {
+      session?.cancel();
+      if (generation === recordingCaptureGeneration && !fallbackRequested) await pickAudio(assignmentId);
      return;
    }
-    mediaStream.value = stream;
-    const recorder = new MediaRecorder(stream);
-    const chunks: Blob[] = [];
-    mediaRecorder.value = recorder;
+    recordingCapture = session;
    recordingAssignmentId.value = assignmentId ?? null;
-    recorder.ondataavailable = (event) => {
-      if (event.data?.size) chunks.push(event.data);
-    };
-    recorder.onstop = () => {
-      recording.value = false;
-      mediaRecorder.value = null;
-      stopMediaTracks();
-      const blob = new Blob(chunks, { type: recorder.mimeType || 'audio/webm' });
-      if (!blob.size) {
-        message.value = '未录到音频，请改用文字输入';
-        return;
-      }
-      const targetAssignmentId = assignmentId;
-      recordingAssignmentId.value = null;
-      void transcribeBlob(blob, targetAssignmentId);
-    };
-    recorder.start();
    recording.value = true;
    message.value = '正在录音';
  } catch (error) {
    recordingAssignmentId.value = null;
-    stopMediaTracks();
+    recordingCapture = null;
    if (generation === recordingCaptureGeneration) await pickAudio(assignmentId);
  } finally {
    if (generation === recordingCaptureGeneration) recordingStarting.value = false;
@@ -1069,10 +1054,9 @@ const startRecording = async (assignmentId?: number) => {
 };

 const stopRecording = () => {
-  const recorder = mediaRecorder.value;
-  if (!recording.value || !recorder || recorder.state === 'inactive') return;
+  if (!recording.value || !recordingCapture) return;
  message.value = '录音结束，正在转写';
-  recorder.stop();
+  recordingCapture.stop();
 };

 const toggleRecording = (assignmentId?: number) => {
@@ -219,7 +219,7 @@
        :maxlength="500"
        :panel-items="composerPanelItems"
        @send="handleComposerSend"
-        @voice-blob="handleVoiceBlob"
+        @voice-capture="handleVoiceCapture"
        @pick-audio="pickVoiceAudio"
        @panel="handlePanelAction"
      />
@@ -239,9 +239,10 @@ import {
  chooseSpeechAudio,
  createSpeechPlaybackController,
  MENTOR_VOICE_PROFILE,
-  transcribeSpeechBlob,
+  transcribeSpeechCapture,
  transcribeSpeechFile
 } from '@/services/speech';
+import type { SpeechCapture } from '@/services/speech-capture';
 import type { SpeechPlaybackStatus } from '@/types/api';
 import {
  listMyWorkReports,
@@ -473,13 +474,13 @@ const cancelVoiceTranscription = () => {
  voiceTranscribing.value = false;
 };

-const handleVoiceBlob = async (blob: Blob) => {
+const handleVoiceCapture = async (capture: SpeechCapture) => {
  if (voiceTranscribing.value || composerDisabled.value) return;
  const generation = ++voiceTranscriptionGeneration;
  const accountKey = currentAccountKey();
  voiceTranscribing.value = true;
  try {
-    const result = await transcribeSpeechBlob(blob, 'work-report.webm', (abort) => {
+    const result = await transcribeSpeechCapture(capture, 'work-report.webm', (abort) => {
      if (generation === voiceTranscriptionGeneration) abortVoiceTranscription = abort;
    });
    if (!requestIsCurrent(generation, voiceTranscriptionGeneration, accountKey)) return;
@@ -282,7 +282,7 @@
      :tool-label="broadcastContext ? '' : defaultToolCode === 'TEAM_PRACTICE_SUMMARY' ? '团队训练概况' : '我的训练概况'"
      :panel-items="composerPanelItems"
      @send="handleComposerSend"
-      @voice-blob="handleVoiceBlob"
+      @voice-capture="handleVoiceCapture"
      @pick-audio="pickVoiceAudio"
      @pick-media="pickMedia"
      @clear-media="clearMedia"
@@ -440,7 +440,8 @@ import {
 import { resetPageScroll } from '@/services/navigation';
 import { ApiError, authenticatedFileUrl, releaseAuthenticatedFileUrl } from '@/services/api';
 import { ensureEmployeePosition, fetchMobileIdentity } from '@/services/position';
-import { chooseSpeechAudio, createSpeechPlaybackController, measureAudioDuration, MENTOR_VOICE_PROFILE, transcribeSpeechBlob, transcribeSpeechFile } from '@/services/speech';
+import { chooseSpeechAudio, createSpeechPlaybackController, measureAudioDuration, MENTOR_VOICE_PROFILE, transcribeSpeechCapture, transcribeSpeechFile } from '@/services/speech';
+import type { SpeechCapture } from '@/services/speech-capture';
 import { downloadSummaryCardImage } from '@/services/summary-card-image';
 import { getMemoryCandidates } from '@/services/personal-assistant';
 import { currentAccountKey } from '@/services/request-scope';
@@ -1089,12 +1090,12 @@ const cancelVoiceTranscription = () => {
  voiceTranscribing.value = false;
 };

-const handleVoiceBlob = async (blob: Blob) => {
+const handleVoiceCapture = async (capture: SpeechCapture) => {
  if (voiceTranscribing.value) return;
  const generation = ++voiceTranscriptionGeneration;
  voiceTranscribing.value = true;
  try {
-    const result = await transcribeSpeechBlob(blob, 'sop-question.webm', (abort) => {
+    const result = await transcribeSpeechCapture(capture, 'sop-question.webm', (abort) => {
      if (generation === voiceTranscriptionGeneration) abortVoiceTranscription = abort;
    });
    if (generation !== voiceTranscriptionGeneration) return;
@@ -1552,14 +1553,19 @@ onUnload(() => {
 }

 .broadcast-context-action {
+  display: flex;
  min-height: 38px;
+  align-items: center;
+  justify-content: center;
  margin: 0;
+  padding: 0;
  border: 1px solid #e7aeb7;
  border-radius: 10px;
  background: #ffffff;
  color: #b51c2b;
  font-size: 12px;
  font-weight: 700;
+  line-height: 1;
 }

 .broadcast-context-action.end {
@@ -211,7 +211,9 @@ import { getMobileHome } from '@/services/home';
 import { getPracticeAssignments } from '@/services/practice';
 import { getSelectedPositionProfile, isLoggedIn, rememberLoginRedirect } from '@/services/auth';
 import { resetPageScroll } from '@/services/navigation';
-import { chooseSpeechAudio, transcribeSpeechBlob, transcribeSpeechFile } from '@/services/speech';
+import { chooseSpeechAudio, transcribeSpeechCapture, transcribeSpeechFile } from '@/services/speech';
+import { startSpeechCapture } from '@/services/speech-capture';
+import type { SpeechCapture, SpeechCaptureSession } from '@/services/speech-capture';
 import { stripInternalCodes } from '@/services/text';
 import { ensureEmployeePosition } from '@/services/position';
 import { createPageRequestScope, currentAccountKey } from '@/services/request-scope';
@@ -393,8 +395,7 @@ const loadBroadcastUnreadCount = async () => {
 const recording = ref(false);
 const cancelling = ref(false);
 const transcribing = ref(false);
-let recorder: MediaRecorder | null = null;
-let recordStream: MediaStream | null = null;
+let holdCapture: SpeechCaptureSession | null = null;
 let holdGeneration = 0;
 let discardRecording = false;
 let holdStartY = 0;
@@ -411,57 +412,45 @@ const voiceSub = computed(() => {
  return '松开发送，上滑取消';
 });

-const supportsRecording = () =>
-  typeof navigator !== 'undefined'
-  && Boolean(navigator.mediaDevices?.getUserMedia)
-  && typeof MediaRecorder !== 'undefined';
-
-const stopRecordTracks = () => {
-  recordStream?.getTracks().forEach((track) => track.stop());
-  recordStream = null;
-};
-
 const startHold = async (event?: TouchEvent | MouseEvent) => {
  if (recording.value || transcribing.value) return;
  if (!loggedIn.value) {
    goLogin();
    return;
  }
-  if (!supportsRecording()) {
-    void pickAudioFallback();
-    return;
-  }
  const generation = ++holdGeneration;
  discardRecording = false;
  cancelling.value = false;
+  let fallbackRequested = false;
  if (event && 'touches' in event && event.touches[0]) {
    holdStartY = event.touches[0].clientY;
  }
  try {
-    const mediaStream = await navigator.mediaDevices.getUserMedia({ audio: true });
-    if (generation !== holdGeneration) {
-      mediaStream.getTracks().forEach((track) => track.stop());
+    const session = await startSpeechCapture({
+      onCapture: (capture) => {
+        if (generation !== holdGeneration) return;
+        holdCapture = null;
+        recording.value = false;
+        if (!discardRecording) void handleVoiceCapture(capture);
+      },
+      onError: (error) => {
+        if (generation !== holdGeneration) return;
+        fallbackRequested = true;
+        holdCapture = null;
+        recording.value = false;
+        message.value = error.message;
+        void pickAudioFallback();
+      }
+    });
+    if (!session || generation !== holdGeneration) {
+      session?.cancel();
+      if (generation === holdGeneration && !fallbackRequested) void pickAudioFallback();
      return;
    }
-    recordStream = mediaStream;
-    const mediaRecorder = new MediaRecorder(mediaStream);
-    const chunks: Blob[] = [];
-    recorder = mediaRecorder;
-    mediaRecorder.ondataavailable = (e) => {
-      if (e.data?.size) chunks.push(e.data);
-    };
-    mediaRecorder.onstop = () => {
-      recording.value = false;
-      recorder = null;
-      stopRecordTracks();
-      if (discardRecording) return;
-      const blob = new Blob(chunks, { type: mediaRecorder.mimeType || 'audio/webm' });
-      if (blob.size) void handleVoiceBlob(blob);
-    };
-    mediaRecorder.start();
+    holdCapture = session;
    recording.value = true;
  } catch {
-    stopRecordTracks();
+    holdCapture = null;
    if (generation === holdGeneration) void pickAudioFallback();
  }
 };
@@ -475,20 +464,22 @@ const moveHold = (event: TouchEvent) => {
 const endHold = () => {
  if (!recording.value) return;
  discardRecording = cancelling.value;
-  if (recorder && recorder.state !== 'inactive') recorder.stop();
+  holdCapture?.stop();
 };

 const cancelHold = () => {
  if (!recording.value) return;
  discardRecording = true;
-  if (recorder && recorder.state !== 'inactive') recorder.stop();
+  holdCapture?.cancel();
+  holdCapture = null;
+  recording.value = false;
 };

-const handleVoiceBlob = async (blob: Blob) => {
+const handleVoiceCapture = async (capture: SpeechCapture) => {
  transcribing.value = true;
  message.value = '';
  try {
-    const result = await transcribeSpeechBlob(blob, 'today-ask.webm');
+    const result = await transcribeSpeechCapture(capture, 'today-ask.webm');
    const text = result.text?.trim();
    if (text) ask(text);
    else message.value = '没听清，请再说一次';
@@ -105,7 +105,7 @@
        { key: 'audio', icon: 'sound', label: '选音频转文字' }
      ]"
      @send="handleComposerSend"
-      @voice-blob="handleVoiceBlob"
+      @voice-capture="handleVoiceCapture"
      @panel="handlePanelAction"
      @clear-media="clearImage"
    />
@@ -120,7 +120,8 @@ import AskSourceHeader from '@/components/chat/AskSourceHeader.vue';
 import ChatComposer from '@/components/chat/ChatComposer.vue';
 import { isLoggedIn, rememberLoginRedirect } from '@/services/auth';
 import { chooseKnowledgeMedia } from '@/services/knowledge';
-import { chooseSpeechAudio, createSpeechPlaybackController, MENTOR_VOICE_PROFILE, transcribeSpeechBlob, transcribeSpeechFile } from '@/services/speech';
+import { chooseSpeechAudio, createSpeechPlaybackController, MENTOR_VOICE_PROFILE, transcribeSpeechCapture, transcribeSpeechFile } from '@/services/speech';
+import type { SpeechCapture } from '@/services/speech-capture';
 import { getWebAiCapabilities, queryWebAi, queryWebAiWithImage } from '@/services/web-ai';
 import { renderMarkdown } from '@/utils/markdown';

@@ -290,12 +291,12 @@ const handleComposerSend = (text: string) => {
  void sendQuestion(text);
 };

-const handleVoiceBlob = async (blob: Blob) => {
+const handleVoiceCapture = async (capture: SpeechCapture) => {
  if (voiceTranscribing.value) return;
  const generation = ++voiceTranscriptionGeneration;
  voiceTranscribing.value = true;
  try {
-    const asr = await transcribeSpeechBlob(blob, 'web-ai-question.webm', (abort) => {
+    const asr = await transcribeSpeechCapture(capture, 'web-ai-question.webm', (abort) => {
      if (generation === voiceTranscriptionGeneration) abortVoiceTranscription = abort;
    });
    if (generation !== voiceTranscriptionGeneration) return;