From bf82ff969118538b5cf7675bcd001a6d4ba8d0e6 Mon Sep 17 00:00:00 2001
From: let5sne <let5sne@gmail.com>
Date: Thu, 23 Jul 2026 22:12:42 +0800
Subject: [PATCH] chore: prepare local speech context for isolated tts work

---
 mobile-uni/src/services/speech.ts | 21 +++++++++++----------
 mobile-uni/src/types/api.ts       |  7 +++++++
 2 files changed, 18 insertions(+), 10 deletions(-)

diff --git a/mobile-uni/src/services/speech.ts b/mobile-uni/src/services/speech.ts
index bb8b533b..c2c02d81 100644
--- a/mobile-uni/src/services/speech.ts
+++ b/mobile-uni/src/services/speech.ts
@@ -1,4 +1,4 @@
-import type { AsrResponse, SpeechPlaybackStatus, SpeechSelectedFile, SpeechVoiceProfile, TtsResponse } from '@/types/api';
+import type { AsrResponse, PracticeTtsContext, SpeechPlaybackStatus, SpeechSelectedFile, SpeechVoiceProfile, TtsResponse } from '@/types/api';
 import { apiRequest, apiUrl, authHeaders, readTextPayload } from './api';
 import type { SpeechCapture } from './speech-capture';
 
@@ -103,11 +103,11 @@ export const transcribeSpeechCapture = (
   ? transcribeSpeechBlob(capture.blob, filename, registerAbort)
   : transcribeSpeechFile(capture.file, registerAbort);
 
-export const synthesizeSpeech = (text: string, voiceProfile?: SpeechVoiceProfile) =>
+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 },
+    data: { text: text.trim().slice(0, 300), voiceProfile, practiceContext },
     timeout: 60000
   });
 
@@ -133,7 +133,8 @@ export const createSpeechPlaybackController = (
   let generation = 0;
   const sourceCache = new Map<string, string>();
   const maxCachedSources = 4;
-  const cacheKey = (text: string, profile?: SpeechVoiceProfile) => JSON.stringify([text, profile || null]);
+  const cacheKey = (text: string, profile?: SpeechVoiceProfile, practiceContext?: PracticeTtsContext) =>
+    JSON.stringify([text, profile || null, practiceContext || null]);
 
   const publish = (key = '', status: SpeechPlaybackStatus = 'idle') => {
     activeKey = key;
@@ -212,7 +213,7 @@ export const createSpeechPlaybackController = (
     context.play();
   };
 
-  const toggle = async (key: string, text: string, voiceProfile?: SpeechVoiceProfile) => {
+  const toggle = async (key: string, text: string, voiceProfile?: SpeechVoiceProfile, practiceContext?: PracticeTtsContext) => {
     if (activeKey === key) {
       stop();
       return;
@@ -221,7 +222,7 @@ export const createSpeechPlaybackController = (
     const value = text.trim();
     if (!value) return;
     const profile = voiceProfile || defaultVoiceProfile;
-    const sourceKey = cacheKey(value, profile);
+    const sourceKey = cacheKey(value, profile, practiceContext);
     const currentGeneration = generation;
     publish(key, 'loading');
     try {
@@ -232,7 +233,7 @@ export const createSpeechPlaybackController = (
         playAudioSource(key, cachedSource, sourceKey, currentGeneration);
         return;
       }
-      const result = await synthesizeSpeech(value, profile);
+      const result = await synthesizeSpeech(value, profile, 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');
@@ -257,15 +258,15 @@ export const createSpeechPlaybackController = (
    * WeChat-style voice bubble can show duration before the first tap.
    * Returns the audio source URL, or null when remote TTS is unavailable.
    */
-  const preload = async (text: string, voiceProfile?: SpeechVoiceProfile): Promise<string | null> => {
+  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);
+    const sourceKey = cacheKey(value, profile, practiceContext);
     const cached = sourceCache.get(sourceKey);
     if (cached) return cached;
     try {
-      const result = await synthesizeSpeech(value, profile);
+      const result = await synthesizeSpeech(value, profile, practiceContext);
       const source = result.inlineAudioUrl || result.audioUrl;
       if (!source) return null;
       sourceCache.set(sourceKey, source);
diff --git a/mobile-uni/src/types/api.ts b/mobile-uni/src/types/api.ts
index cd6cdd49..2a289563 100644
--- a/mobile-uni/src/types/api.ts
+++ b/mobile-uni/src/types/api.ts
@@ -31,6 +31,13 @@ export interface TtsResponse {
   ossId?: number | string;
 }
 
+/** 可选的服务端训练回合上下文；仅用于关联本次新生成的业主 TTS 录音。 */
+export interface PracticeTtsContext {
+  sessionId: string;
+  turnIndex: number;
+  role: 'customer';
+}
+
 export type SpeechPlaybackStatus = 'idle' | 'loading' | 'playing';
 
 export interface SpeechVoiceProfile {