Upload app.py with huggingface_hub

app.py

@@ -61,8 +61,8 @@ if not hasattr(torchaudio, "info"):
 
 from df.enhance import enhance, init_df, load_audio, save_audio
 
-# FORCE BUILD TRIGGER: 09:10:00 Jan 21 2026
-# v81: Stability Optimizations (Memory Management + Cache Clearing)
+# FORCE BUILD TRIGGER: 09:40:00 Jan 21 2026
+# v84: Fixed SyntaxError (Missing try block in core_process)
 
 # 🛠️ Monkeypatch torchaudio.load
 try:
@@ -186,12 +186,12 @@ def _tts_logic(text, lang, speaker_wav_b64):
         lang_key = lang.strip().lower()
         mapped_lang = XTTS_MAP.get(lang_key) or XTTS_MAP.get(lang_key.split('-')[0])
         
-    print(f"[v81] TTS Request - Original: {lang}, Mapped: {mapped_lang}")
+    print(f"[v84] TTS Request - Original: {lang}, Mapped: {mapped_lang}")
     
     # 🛣️ INTELLIGENT ROUTING
     # Case A: XTTS Support (Voice Cloning)
     if mapped_lang and mapped_lang in XTTS_LANG_CODES:
-        print(f"[v81] Using XTTS-v2 for '{mapped_lang}'")
+        print(f"[v84] Using XTTS-v2 for '{mapped_lang}'")
         speaker_wav_path = None
         if speaker_wav_b64:
             sb = base64.b64decode(speaker_wav_b64)
@@ -217,7 +217,7 @@ def _tts_logic(text, lang, speaker_wav_b64):
             if 'output_path' in locals() and os.path.exists(output_path): os.unlink(output_path)
             
     # Case B: Chatterbox ONNX Support (High-Quality Fast Fallback)
-    print(f"[v81] Using Chatterbox ONNX Fallback for '{lang}'")
+    print(f"[v84] Using Chatterbox ONNX Fallback for '{lang}'")
     try:
         # Use local file if available for cloning in Chatterbox too
         temp_ref = None
@@ -241,38 +241,39 @@ def _tts_logic(text, lang, speaker_wav_b64):
 @spaces.GPU
 def core_process(request_dict):
     """
-    Unified GPU Entry Point (v81). 
+    Unified GPU Entry Point (v84). 
     This function handles all high-speed tasks inside a single GPU allocation.
     The container stays resident on CPU but triggers GPU on demand.
     """
     action = request_dict.get("action")
     t0 = time.time()
-    print(f"--- [v81] 🚀 GPU SESSION START: {action} at {time.ctime()} ---")
+    print(f"--- [v84] 🚀 GPU SESSION START: {action} at {time.ctime()} ---")
     load_models() 
     
-    if action == "stt":
-        res = _stt_logic(request_dict)
-    elif action == "translate":
-        res = {"translated": _translate_logic(request_dict.get("text"), request_dict.get("target_lang", "en"))}
-    elif action == "tts":
-        res = _tts_logic(request_dict.get("text"), request_dict.get("lang"), request_dict.get("speaker_wav"))
-    elif action == "s2st":
-        # 🔗 FULL PIPELINE (Single GPU Call)
-        stt_res = _stt_logic({"file": request_dict.get("file"), "lang": request_dict.get("source_lang")})
-        text = stt_res.get("text", "")
-        if not text: return {"error": "No speech detected"}
-        
-        translated = _translate_logic(text, request_dict.get("target_lang"))
-        
-        tts_res = _tts_logic(translated, request_dict.get("target_lang"), request_dict.get("speaker_wav"))
-        res = {"text": text, "translated": translated, "audio": tts_res.get("audio")}
-    elif action == "health":
-        res = {"status": "awake", "time": time.ctime()}
-    else:
-        res = {"error": f"Unknown action: {action}"}
-        
+    try:
+        if action == "stt":
+            res = _stt_logic(request_dict)
+        elif action == "translate":
+            res = {"translated": _translate_logic(request_dict.get("text"), request_dict.get("target_lang", "en"))}
+        elif action == "tts":
+            res = _tts_logic(request_dict.get("text"), request_dict.get("lang"), request_dict.get("speaker_wav"))
+        elif action == "s2st":
+            # 🔗 FULL PIPELINE (Single GPU Call)
+            stt_res = _stt_logic({"file": request_dict.get("file"), "lang": request_dict.get("source_lang")})
+            text = stt_res.get("text", "")
+            if not text: return {"error": "No speech detected"}
+            
+            translated = _translate_logic(text, request_dict.get("target_lang"))
+            
+            tts_res = _tts_logic(translated, request_dict.get("target_lang"), request_dict.get("speaker_wav"))
+            res = {"text": text, "translated": translated, "audio": tts_res.get("audio")}
+        elif action == "health":
+            res = {"status": "awake", "time": time.ctime()}
+        else:
+            res = {"error": f"Unknown action: {action}"}
+            
     finally:
-        print(f"--- [v81] ✨ SESSION END: {action} ---")
+        print(f"--- [v84] ✨ SESSION END: {action} (Total: {time.time()-t0:.2f}s) ---")
         gc.collect()
         if torch.cuda.is_available():
             torch.cuda.empty_cache()