Upload app.py with huggingface_hub

app.py

@@ -10,8 +10,8 @@ from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import HTMLResponse
 import uvicorn
 
-# --- [v156] 🚀 PRO STABLE ENGINE (GPU-STT-Base + CPU-TTS-XTTS) ---
-print(f"--- [v156] 📡 BOOTING PRO STABLE ENGINE ---")
+# --- [v157] 🚀 EXTREME STABILITY ENGINE ---
+print(f"--- [v157] 📡 BOOTING ENGINE ---")
 
 from transformers import pipeline, AutoModelForSpeechSeq2Seq, AutoProcessor
 from TTS.api import TTS
@@ -46,15 +46,15 @@ MODELS = {"stt": None, "tts": None}
 def load_tts_cpu():
     global MODELS
     if MODELS.get("tts") is None:
-        print("--- [v156] 📥 LOADING XTTS V2 (CPU) ---")
+        print("--- [v157] 📥 LOADING XTTS V2 (CPU) ---")
         MODELS["tts"] = TTS("tts_models/multilingual/multi-dataset/xtts_v2").to("cpu")
-        print("--- [v156] ✅ XTTS READY ---")
+        print("--- [v157] ✅ XTTS READY (CPU) ---")
 
 @spaces.GPU(duration=60)
 def gpu_stt_base(temp_path, lang):
     global MODELS
     if MODELS.get("stt") is None:
-        print("--- [v156] 📥 LOADING WHISPER (Base) ON GPU ---")
+        print("--- [v157] 📥 LOADING WHISPER (Base) ON GPU ---")
         model_id = "openai/whisper-base"
         model = AutoModelForSpeechSeq2Seq.from_pretrained(model_id, torch_dtype=torch.float32).to("cuda")
         processor = AutoProcessor.from_pretrained(model_id)
@@ -65,17 +65,18 @@ def gpu_stt_base(temp_path, lang):
             feature_extractor=processor.feature_extractor,
             device="cuda"
         )
+    # Whisper Base is extremely robust and fast on H200
     res = MODELS["stt"](temp_path, generate_kwargs={"language": lang if lang and len(lang) <= 3 else None})
     return res["text"].strip()
 
 async def handle_process(request: Request):
-    t1 = time.time()  # Initialize t1 at the very beginning to avoid UnboundLocalError
+    t1 = time.time()
     try:
         data = await request.json()
         action = data.get("action")
-        if action == "health": return {"status": "awake", "v": "156"}
+        if action == "health": return {"status": "awake", "v": "157"}
         
-        print(f"--- [v156] 🛠️ {action} ---")
+        print(f"--- [v157] 🛠️ {action} ---")
         
         stt_text = ""
         if action in ["stt", "s2st"]:
@@ -129,20 +130,20 @@ async def handle_process(request: Request):
             return {"text": stt_text, "translated": trans_text, "audio": audio_b64}
 
     except Exception as e:
-        print(f"❌ [v156] ERROR: {traceback.format_exc()}")
+        print(f"❌ [v157] ERROR: {traceback.format_exc()}")
         return {"error": str(e)}
     finally:
-        print(f"--- [v156] ✨ DONE ({time.time()-t1:.1f}s) ---")
+        print(f"--- [v157] ✨ DONE ({time.time()-t1:.1f}s) ---")
 
 @app.post("/process")
 @app.post("/api/v1/process")
 async def api_process(request: Request): return await handle_process(request)
 
 @app.get("/health")
-def health(): return {"status": "ok", "v": "156", "gpu": torch.cuda.is_available()}
+def health(): return {"status": "ok", "v": "157", "gpu": torch.cuda.is_available()}
 
 @app.get("/", response_class=HTMLResponse)
-def root(): return "<h1>🚀 AI Engine v156 (PRO STABLE)</h1>"
+def root(): return "<h1>🚀 AI Engine v157 (EXTREME STABILITY)</h1>"
 
 if __name__ == "__main__":
     uvicorn.run(app, host="0.0.0.0", port=7860)