feat(speech): add qwen tts voice preferences

mobile-uni/src/services/speech.ts

@@ -1,6 +1,8 @@
 import type { AsrResponse, PracticeTtsContext, SpeechPlaybackStatus, SpeechSelectedFile, SpeechVoiceProfile, TtsResponse } from '@/types/api';
 import { apiRequest, apiUrl, authHeaders, readTextPayload } from './api';
+import { getAuth } from './auth';
 import type { SpeechCapture } from './speech-capture';
+import { applyTtsVoicePreference, getTtsDialectPreference, getTtsVoicePreference } from './tts-voice-preference';
 
 const audioExtensions = ['.mp3', '.wav', '.m4a', '.webm', '.aac', '.ogg'];
 
@@ -103,11 +105,20 @@ export const transcribeSpeechCapture = (
   ? transcribeSpeechBlob(capture.blob, filename, registerAbort)
   : transcribeSpeechFile(capture.file, registerAbort);
 
+export const resolveTtsVoiceProfile = (voiceProfile?: SpeechVoiceProfile) => {
+  const auth = getAuth();
+  return applyTtsVoicePreference(
+    voiceProfile,
+    getTtsVoicePreference(auth.tenantId, auth.phone),
+    getTtsDialectPreference(auth.tenantId, auth.phone)
+  );
+};
+
 export const synthesizeSpeech = (text: string, voiceProfile?: SpeechVoiceProfile, practiceContext?: PracticeTtsContext) =>
   apiRequest<TtsResponse>({
     url: '/api/ai/tts',
     method: 'POST',
-    data: { text: text.trim().slice(0, 300), voiceProfile, practiceContext },
+    data: { text: text.trim().slice(0, 300), voiceProfile: resolveTtsVoiceProfile(voiceProfile), practiceContext },
     timeout: 60000
   });
 
@@ -118,7 +129,7 @@ export interface SpeechPlaybackSnapshot {
 
 const browserDialectLanguage = (dialect: SpeechVoiceProfile['dialect']) => {
   if (dialect === 'cantonese') return 'zh-HK';
-  if (dialect === 'sichuanese') return '';
+  if (dialect && dialect !== 'mandarin') return '';
   return 'zh-CN';
 };
 
@@ -221,8 +232,8 @@ export const createSpeechPlaybackController = (
     stop();
     const value = text.trim();
     if (!value) return;
-    const profile = voiceProfile || defaultVoiceProfile;
-    const sourceKey = cacheKey(value, profile, practiceContext);
+    const effectiveProfile = resolveTtsVoiceProfile(voiceProfile || defaultVoiceProfile);
+    const sourceKey = cacheKey(value, effectiveProfile, practiceContext);
     const currentGeneration = generation;
     publish(key, 'loading');
     try {
@@ -233,7 +244,7 @@ export const createSpeechPlaybackController = (
         playAudioSource(key, cachedSource, sourceKey, currentGeneration);
         return;
       }
-      const result = await synthesizeSpeech(value, profile, practiceContext);
+      const result = await synthesizeSpeech(value, effectiveProfile, practiceContext);
       if (currentGeneration !== generation || activeKey !== key) return;
       const source = result.inlineAudioUrl || result.audioUrl;
       if (!source) throw new Error('语音合成未返回音频');
@@ -246,7 +257,7 @@ export const createSpeechPlaybackController = (
       playAudioSource(key, source, sourceKey, currentGeneration);
     } catch (error) {
       if (currentGeneration !== generation) return;
-      const browserFallback = playWithBrowserSpeech(key, value, currentGeneration, profile);
+      const browserFallback = playWithBrowserSpeech(key, value, currentGeneration, effectiveProfile);
       if (browserFallback.played) return;
       stop();
       onError(browserFallback.message || (error instanceof Error ? error.message : '语音生成失败'));
@@ -261,12 +272,12 @@ export const createSpeechPlaybackController = (
   const preload = async (text: string, voiceProfile?: SpeechVoiceProfile, practiceContext?: PracticeTtsContext): Promise<string | null> => {
     const value = text.trim();
     if (!value) return null;
-    const profile = voiceProfile || defaultVoiceProfile;
-    const sourceKey = cacheKey(value, profile, practiceContext);
+    const effectiveProfile = resolveTtsVoiceProfile(voiceProfile || defaultVoiceProfile);
+    const sourceKey = cacheKey(value, effectiveProfile, practiceContext);
     const cached = sourceCache.get(sourceKey);
     if (cached) return cached;
     try {
-      const result = await synthesizeSpeech(value, profile, practiceContext);
+      const result = await synthesizeSpeech(value, effectiveProfile, practiceContext);
       const source = result.inlineAudioUrl || result.audioUrl;
       if (!source) return null;
       sourceCache.set(sourceKey, source);