feat: Add voice design support for voice cloning and enhance cache management

qwen3-tts-backend/core/cache_manager.py

@@ -72,6 +72,36 @@ class VoiceCacheManager:
             logger.error(f"Cache retrieval error: {e}", exc_info=True)
             return None
 
+    async def get_cache_by_id(self, cache_id: int, db: Session) -> Optional[Dict[str, Any]]:
+        try:
+            cache_entry = db.query(VoiceCache).filter(VoiceCache.id == cache_id).first()
+            if not cache_entry:
+                logger.debug(f"Cache not found: id={cache_id}")
+                return None
+
+            cache_file = Path(cache_entry.cache_path)
+            if not cache_file.exists():
+                logger.warning(f"Cache file missing: {cache_file}")
+                return None
+
+            with open(cache_file, 'rb') as f:
+                cache_data = pickle.load(f)
+
+            cache_entry.last_accessed = datetime.utcnow()
+            cache_entry.access_count += 1
+            db.commit()
+
+            logger.info(f"Cache loaded by id: cache_id={cache_id}, access_count={cache_entry.access_count}")
+            return {
+                'cache_id': cache_entry.id,
+                'data': cache_data,
+                'metadata': cache_entry.meta_data
+            }
+
+        except Exception as e:
+            logger.error(f"Cache retrieval by id error: {e}", exc_info=True)
+            return None
+
     async def set_cache(
         self,
         user_id: int,

qwen3-tts-backend/core/tts_service.py

@@ -83,19 +83,23 @@ class LocalTTSBackend(TTSBackend):
             audio_data = np.array(audio_data)
         return self._numpy_to_bytes(audio_data), 24000
 
-    async def generate_voice_clone(self, params: dict, ref_audio_bytes: bytes) -> Tuple[bytes, int]:
+    async def generate_voice_clone(self, params: dict, ref_audio_bytes: bytes = None, x_vector=None) -> Tuple[bytes, int]:
         from utils.audio import process_ref_audio
 
-        ref_audio_array, ref_sr = process_ref_audio(ref_audio_bytes)
-
         await self.model_manager.load_model("base")
         _, tts = await self.model_manager.get_current_model()
 
-        x_vector = tts.create_voice_clone_prompt(
-            ref_audio=(ref_audio_array, ref_sr),
-            ref_text=params.get('ref_text', ''),
-            x_vector_only_mode=False
-        )
+        if x_vector is None:
+            if ref_audio_bytes is None:
+                raise ValueError("Either ref_audio_bytes or x_vector must be provided")
+
+            ref_audio_array, ref_sr = process_ref_audio(ref_audio_bytes)
+
+            x_vector = tts.create_voice_clone_prompt(
+                ref_audio=(ref_audio_array, ref_sr),
+                ref_text=params.get('ref_text', ''),
+                x_vector_only_mode=False
+            )
 
         wavs, sample_rate = tts.generate_voice_clone(
             text=params['text'],