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);

mobile-uni/src/services/tts-voice-preference.ts

@@ -0,0 +1,96 @@
+import type { SpeechVoiceProfile } from '@/types/api';
+
+export const qwenTtsVoiceOptions = [
+  { value: '', label: '跟随场景' },
+  { value: 'longanhuan_v3.6', label: '龙安欢 · 中文女声' },
+  { value: 'longjielidou_v3.6', label: '龙杰力豆 · 童声' },
+  { value: 'loongeva_v3.6', label: 'Loongeva · 英文女声' },
+  { value: 'loongjohn', label: 'loongJohn · 英文男声' }
+] as const;
+
+export const qwenTtsDialectOptions = [
+  { value: '', label: '跟随场景' },
+  { value: 'mandarin', label: '普通话' },
+  { value: 'cantonese', label: '粤语' },
+  { value: 'chongqing', label: '重庆话' },
+  { value: 'northeastern', label: '东北话' },
+  { value: 'gansu', label: '甘肃话' },
+  { value: 'guizhou', label: '贵州话' },
+  { value: 'zhejiang', label: '浙江话' },
+  { value: 'hebei', label: '河北话' },
+  { value: 'henan', label: '河南话' },
+  { value: 'hubei', label: '湖北话' },
+  { value: 'hunan', label: '湖南话' },
+  { value: 'jiangxi', label: '江西话' },
+  { value: 'ningbo', label: '宁波话' },
+  { value: 'ningxia', label: '宁夏话' },
+  { value: 'qingdao', label: '青岛话' },
+  { value: 'shaanxi', label: '陕西话' },
+  { value: 'shanxi', label: '山西话' },
+  { value: 'shandong', label: '山东话' },
+  { value: 'shanghai', label: '上海话' },
+  { value: 'sichuanese', label: '四川话' },
+  { value: 'yunnan', label: '云南话' }
+] as const;
+
+export type TtsVoicePreference = typeof qwenTtsVoiceOptions[number]['value'];
+export type TtsDialectPreference = typeof qwenTtsDialectOptions[number]['value'];
+
+const preferenceKeyPrefix = 'aihr_tts_voice';
+const dialectPreferenceKeyPrefix = 'aihr_tts_dialect';
+
+export const normalizeTtsVoicePreference = (value: unknown): TtsVoicePreference =>
+  qwenTtsVoiceOptions.some((option) => option.value === value) ? value as TtsVoicePreference : '';
+
+export const normalizeTtsDialectPreference = (value: unknown): TtsDialectPreference =>
+  qwenTtsDialectOptions.some((option) => option.value === value) ? value as TtsDialectPreference : '';
+
+export const ttsVoicePreferenceStorageKey = (tenantId: unknown, phone: unknown) => {
+  const tenant = String(tenantId || '').trim();
+  const account = String(phone || '').trim();
+  return tenant && account ? `${preferenceKeyPrefix}:${tenant}:${account}` : '';
+};
+
+export const getTtsVoicePreference = (tenantId: unknown, phone: unknown): TtsVoicePreference => {
+  const key = ttsVoicePreferenceStorageKey(tenantId, phone);
+  return key ? normalizeTtsVoicePreference(uni.getStorageSync(key)) : '';
+};
+
+const ttsDialectPreferenceStorageKey = (tenantId: unknown, phone: unknown) => {
+  const tenant = String(tenantId || '').trim();
+  const account = String(phone || '').trim();
+  return tenant && account ? `${dialectPreferenceKeyPrefix}:${tenant}:${account}` : '';
+};
+
+export const getTtsDialectPreference = (tenantId: unknown, phone: unknown): TtsDialectPreference => {
+  const key = ttsDialectPreferenceStorageKey(tenantId, phone);
+  return key ? normalizeTtsDialectPreference(uni.getStorageSync(key)) : '';
+};
+
+export const setTtsVoicePreference = (tenantId: unknown, phone: unknown, value: unknown): TtsVoicePreference => {
+  const key = ttsVoicePreferenceStorageKey(tenantId, phone);
+  const normalized = normalizeTtsVoicePreference(value);
+  if (key) uni.setStorageSync(key, normalized);
+  return normalized;
+};
+
+export const setTtsDialectPreference = (tenantId: unknown, phone: unknown, value: unknown): TtsDialectPreference => {
+  const key = ttsDialectPreferenceStorageKey(tenantId, phone);
+  const normalized = normalizeTtsDialectPreference(value);
+  if (key) uni.setStorageSync(key, normalized);
+  return normalized;
+};
+
+export const applyTtsVoicePreference = (
+  profile: SpeechVoiceProfile | undefined,
+  preference: unknown,
+  dialectPreference?: unknown
+): SpeechVoiceProfile | undefined => {
+  const voice = normalizeTtsVoicePreference(preference);
+  const dialect = normalizeTtsDialectPreference(dialectPreference);
+  if (!voice && !dialect) return profile;
+  const next: SpeechVoiceProfile = { ...(profile || { role: 'neutral' }) };
+  if (voice) next.voice = voice;
+  if (dialect) next.dialect = dialect;
+  return next;
+};