Upload app.py with huggingface_hub

app.py

@@ -18,7 +18,7 @@ import logging
 from threading import Thread, Lock
 from huggingface_hub import snapshot_download
 
-# 🛡️ 1. SILENCE LOGS & WARNINGS (v100: Absolute Silence)
+# 🛡️ 1. SILENCE LOGS & WARNINGS (v101: Pure Power & Stability)
 logging.getLogger("transformers").setLevel(logging.ERROR)
 logging.getLogger("TTS").setLevel(logging.ERROR)
 logging.getLogger("onnxruntime").setLevel(logging.ERROR)
@@ -81,8 +81,8 @@ except ImportError:
             if f is None: return lambda x: x
             return f
 
-# FORCE BUILD TRIGGER: 13:00:00 Jan 21 2026
-# v100: Centennial Stability Update. 1-Worker Lockdown. Lifespan Events.
+# FORCE BUILD TRIGGER: 13:10:00 Jan 21 2026
+# v101: Docker SDK Transition. Absolute Port Isolation.
 
 os.environ["COQUI_TOS_AGREED"] = "1"
 MODELS = {"stt": None, "translate": None, "tts": None, "denoiser": None}
@@ -91,21 +91,19 @@ WARMUP_STATUS = {"complete": False, "in_progress": False}
 WARMUP_LOCK = Lock()
 
 def activate_gpu_models(action):
-    """v100: Stability-First GPU Activation"""
+    """v101: Mission-Critical GPU Mode"""
     global MODELS, WARMUP_STATUS
     local_only = WARMUP_STATUS["complete"]
     
-    # 1. Faster-Whisper: Lockdown to 1 worker for stability on the H200 MIG
     if action in ["stt", "s2st"]:
         stt_on_gpu = False
         try: stt_on_gpu = MODELS["stt"] is not None and MODELS["stt"].model.device == "cuda"
         except: pass
         if not stt_on_gpu:
-            print(f"🎙️ [v100] Activating Whisper (GPU: Stability Protocol)...")
+            print(f"🎙️ [v101] Activating Whisper (GPU)...")
             try:
                 if MODELS["stt"]: del MODELS["stt"]
                 gc.collect(); torch.cuda.empty_cache()
-                # 🛡️ v100: 1-Worker to prevent CUDA deadlocks observed in v99
                 MODELS["stt"] = WhisperModel(
                     "large-v3", 
                     device="cuda", 
@@ -114,10 +112,9 @@ def activate_gpu_models(action):
                     local_files_only=local_only
                 )
             except Exception as e:
-                print(f"⚠️ Stability Init failed: {e}. Falling back to CPU.")
+                print(f"⚠️ GPU Init failed: {e}. Falling back to CPU in-RAM.")
                 MODELS["stt"] = WhisperModel("large-v3", device="cpu", compute_type="int8", local_files_only=True)
 
-    # 2. XTTS-v2
     if action in ["tts", "s2st"]:
         tts_on_gpu = False
         try:
@@ -125,26 +122,24 @@ def activate_gpu_models(action):
             tts_on_gpu = "cuda" in curr
         except: pass
         if MODELS["tts"] is None or not tts_on_gpu:
-            print(f"🔊 [v100] Activating XTTS-v2 (GPU)...")
+            print(f"🔊 [v101] Activating XTTS-v2 (GPU)...")
             try:
                 if MODELS["tts"] is None:
                     MODELS["tts"] = TTS(model_name="tts_models/multilingual/multi-dataset/xtts_v2", gpu=True)
                 else: MODELS["tts"].to("cuda")
             except: pass
 
-    # 3. Chatterbox GPU-Mode (Zero-Latency)
     chatterbox_utils.load_chatterbox(device="cuda" if torch.cuda.is_available() else "cpu")
 
-    # 4. Helpers
     if MODELS["denoiser"] is None:
         try: MODELS["denoiser"] = init_df()
         except: pass
     if MODELS["translate"] is None: MODELS["translate"] = "active"
 
 def release_gpu_models():
-    """v100: RAM-Resident Cleanup"""
+    """v101: Clean Resident State"""
     global MODELS
-    print("🧹 [v100] Releasing GPU. Engines resident in RAM.")
+    print("🧹 [v101] Releasing GPU resources.")
     try:
         if MODELS["stt"] and MODELS["stt"].model.device == "cuda":
             del MODELS["stt"]
@@ -158,19 +153,20 @@ def release_gpu_models():
     if torch.cuda.is_available(): torch.cuda.empty_cache()
 
 def warmup_task():
-    """Silent Warmup (v100)"""
+    """Silent Pre-loading (v101)"""
     global WARMUP_STATUS
     with WARMUP_LOCK:
         if WARMUP_STATUS["complete"] or WARMUP_STATUS["in_progress"]: return
         WARMUP_STATUS["in_progress"] = True
-    print("\n🔥 --- V100: STABILITY WARMUP STARTED ---")
+    print("\n🔥 --- V101: DOCKER DEPLOY WARMUP STARTED ---")
     try:
         MODELS["stt"] = WhisperModel("large-v3", device="cpu", compute_type="int8")
         MODELS["tts"] = TTS(model_name="tts_models/multilingual/multi-dataset/xtts_v2", gpu=False)
         chatterbox_utils.warmup_chatterbox()
         WARMUP_STATUS["complete"] = True
-        print(f"✅ --- SYSTEM READY: v100 --- \n")
-    except: pass
+        print(f"✅ --- SYSTEM WARM: v101 --- \n")
+    except Exception as e:
+        print(f"❌ Warmup fail: {e}")
     finally: WARMUP_STATUS["in_progress"] = False
 
 def _stt_logic(request_dict):
@@ -221,7 +217,7 @@ def _tts_logic(text, lang, speaker_wav_b64):
 def core_process(request_dict):
     action = request_dict.get("action")
     t1 = time.time()
-    print(f"--- [v100] 🚀 GPU SESSION: {action} ---")
+    print(f"--- [v101] 🚀 GPU SESSION: {action} ---")
     activate_gpu_models(action)
     try:
         if action == "stt": res = _stt_logic(request_dict)
@@ -234,16 +230,15 @@ def core_process(request_dict):
             res = {"text": stt_res.get("text"), "translated": translated, "audio": tts_res.get("audio")}
         else: res = {"error": f"Unknown action: {action}"}
     finally:
-        print(f"--- [v100] ✨ END: {action} ({time.time()-t1:.2f}s) ---")
+        print(f"--- [v101] ✨ SUCCESS: {action} ({time.time()-t1:.2f}s) ---")
         release_gpu_models()
     return res
 
 @asynccontextmanager
 async def lifespan(app: FastAPI):
-    # Startup
+    # DOCKER ENTRYPOINT TRIGGER
     Thread(target=warmup_task, daemon=True).start()
     yield
-    # Shutdown
 
 app = FastAPI(lifespan=lifespan)
 
@@ -251,14 +246,13 @@ app = FastAPI(lifespan=lifespan)
 async def api_process(request: Request):
     try:
         req_data = await request.json()
-        # 🏥 v100: LIGHTWEIGHT CPU HEALTH (Prevent Queue Bloat)
         if req_data.get("action") == "health":
             return {"status": "awake", "warm": WARMUP_STATUS["complete"]}
         return core_process(req_data)
     except Exception as e: return {"error": str(e)}
 
 @app.get("/health")
-def health(): return {"status": "ok", "warm": WARMUP_STATUS["complete"], "version": "v100"}
+def health(): return {"status": "ok", "warm": WARMUP_STATUS["complete"], "v": "101"}
 
 @app.post("/api/v1/clear_cache")
 async def clear_cache():
@@ -272,8 +266,10 @@ async def clear_cache():
         return {"status": "success"}
     except: return {"status": "error"}
 
+# GRADIO INTERFACE (v101)
 demo = gr.Interface(fn=lambda x: json.dumps(core_process(json.loads(x))), inputs="text", outputs="text")
 app = gr.mount_gradio_app(app, demo, path="/")
 
 if __name__ == "__main__":
-    uvicorn.run(app, host="0.0.0.0", port=7860, log_level="error")
+    print("🚀 [v101] DOCKER SERVER STARTING ON 7860...")
+    uvicorn.run(app, host="0.0.0.0", port=7860, log_level="error", loop="asyncio")