Upload app.py with huggingface_hub

app.py

@@ -1,4 +1,4 @@
-# 🚀 V109: ZEROGPU STERILE STARTUP
+# 🚀 V110: ZEROGPU RESILIENT STARTUP
 # Must be first to patch environment correctly
 try:
     import spaces
@@ -26,10 +26,11 @@ import gc
 import sys
 import types
 import logging
+import traceback
 from threading import Thread, Lock
 from huggingface_hub import snapshot_download, hf_hub_download
 
-# 🛡️ 1. SILENCE & ENV (v109)
+# 🛡️ 1. SILENCE & ENV (v110)
 logging.getLogger("transformers").setLevel(logging.ERROR)
 logging.getLogger("TTS").setLevel(logging.ERROR)
 os.environ["CT2_VERBOSE"] = "0"
@@ -37,7 +38,7 @@ os.environ["ORT_LOGGING_LEVEL"] = "3"
 os.environ["GRADIO_SERVER_NAME"] = "0.0.0.0"
 os.environ["GRADIO_SERVER_PORT"] = "7860"
 
-# 🛠️ 2. COMPATIBILITY PATCHES (v109)
+# 🛠️ 2. COMPATIBILITY PATCHES (v110)
 if "torchaudio.backend" not in sys.modules:
     backend = types.ModuleType("torchaudio.backend")
     common = types.ModuleType("torchaudio.backend.common")
@@ -77,14 +78,13 @@ except Exception: pass
 
 # 📦 3. AI LIBRARIES
 import chatterbox_utils
-# We import types/classes but do NOT instantiate them at top-level
 from faster_whisper import WhisperModel
 from TTS.api import TTS
 from df.enhance import init_df
 import deep_translator
 
-# FORCE BUILD TRIGGER: 17:35:00 Jan 21 2026
-# v109: Sterile Startup. defer all AI init to GPU session.
+# FORCE BUILD TRIGGER: 17:40:00 Jan 21 2026
+# v110: Resilient Warmup. No blocks.
 
 os.environ["COQUI_TOS_AGREED"] = "1"
 MODELS = {"stt": None, "translate": None, "tts": None, "denoiser": None}
@@ -93,7 +93,7 @@ WARMUP_STATUS = {"complete": False, "in_progress": False}
 WARMUP_LOCK = Lock()
 
 def activate_gpu_models(action):
-    """v109: Deferred Activation (Inside GPU)"""
+    """v110: Safe Deferred Activation"""
     global MODELS, WARMUP_STATUS
     local_only = WARMUP_STATUS["complete"]
     
@@ -102,7 +102,7 @@ def activate_gpu_models(action):
         try: stt_on_gpu = MODELS["stt"] is not None and MODELS["stt"].model.device == "cuda"
         except: pass
         if not stt_on_gpu:
-            print(f"🎙️ [v109] Initializing Whisper (GPU: int8_float16)...")
+            print(f"🎙️ [v110] Init Whisper (MIG-Ready: int8_float16)...")
             try:
                 if MODELS["stt"]: del MODELS["stt"]; gc.collect(); torch.cuda.empty_cache()
                 MODELS["stt"] = WhisperModel(
@@ -123,14 +123,13 @@ 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"🔊 [v109] Initializing XTTS-v2 (GPU)...")
+            print(f"🔊 [v110] Init 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
 
-    # In v109, we only load chatterbox into CUDA when needed
     chatterbox_utils.load_chatterbox(device="cuda")
 
     if MODELS["denoiser"] is None:
@@ -139,9 +138,9 @@ def activate_gpu_models(action):
     if MODELS["translate"] is None: MODELS["translate"] = "active"
 
 def release_gpu_models():
-    """v109: GPU Release"""
+    """v110: Quiet Release"""
     global MODELS
-    print("🧹 [v109] Releasing GPU.")
+    print("🧹 [v110] Releasing.")
     try:
         if MODELS["stt"] and MODELS["stt"].model.device == "cuda":
             del MODELS["stt"]
@@ -156,41 +155,37 @@ def release_gpu_models():
     time.sleep(0.5)
 
 def warmup_task():
-    """v109: Sterile Warmup (Downloads only)"""
+    """v110: Resilient Warmup (Independent Downloads)"""
     global WARMUP_STATUS
     if WARMUP_STATUS["complete"] or WARMUP_STATUS["in_progress"]: return
     WARMUP_STATUS["in_progress"] = True
-    print("\n🔥 --- V109: ZEROGPU STERILE WARMUP ---")
-    try:
-        # Pre-fetch files WITHOUT loading them into memory/CUDA
-        print("⏬ Pre-fetching Whisper Large-v3...")
-        snapshot_download("Systran/faster-whisper-large-v3")
-        
-        print("⏬ Pre-fetching XTTS-v2...")
-        # TTS internal download trigger (folder structure varies)
-        snapshot_download("coqui/XTTS-v2") 
-        
-        print("⏬ Pre-fetching Chatterbox...")
-        chatterbox_utils.warmup_chatterbox()
-        
-        # DeepFilterNet
-        hf_hub_download(repo_id="R_S/DeepFilterNet3", filename="config.json")
-        
-        WARMUP_STATUS["complete"] = True
-        print(f"✅ --- CACHE SYNCED: v109 --- \n")
-    except Exception as e:
-        print(f"❌ Warmup Warning (might be fine): {e}")
-    finally: WARMUP_STATUS["in_progress"] = False
+    print("\n🔥 --- V110: ZEROGPU RESILIENT WARMUP ---")
+    
+    # 1. Faster-Whisper
+    try: snapshot_download("Systran/faster-whisper-large-v3")
+    except Exception as e: print(f"⚠️ Whisper download: {e}")
+    
+    # 2. XTTS-v2
+    try: snapshot_download("coqui/XTTS-v2") 
+    except Exception as e: print(f"⚠️ XTTS download: {e}")
+    
+    # 3. Chatterbox
+    try: chatterbox_utils.warmup_chatterbox()
+    except Exception as e: print(f"⚠️ Chatterbox download: {e}")
+    
+    WARMUP_STATUS["complete"] = True
+    print(f"✅ --- SYSTEM READY: v110 --- \n")
+    WARMUP_STATUS["in_progress"] = False
 
 @spaces.GPU(duration=150)
 def core_process(request_dict):
     action = request_dict.get("action")
-    print(f"--- [v109] 🚀 REQUEST: {action} ---")
+    print(f"--- [v110] 🚀 PROCESSING: {action} ---")
     
-    # Wait for downloads if necessary
+    # Wait for thermal preparation
     waited = 0
-    while not WARMUP_STATUS["complete"] and waited < 300: # 5 min limit
-        if waited % 10 == 0: print(f"⏳ Downloading models... ({waited}s)")
+    while not WARMUP_STATUS["complete"] and waited < 60:
+        if waited % 10 == 0: print(f"⏳ Syncing assets... ({waited}s)")
         time.sleep(1)
         waited += 1
 
@@ -212,9 +207,8 @@ def core_process(request_dict):
         
         elif action == "tts":
             text = request_dict.get("text")
-            lang = request_dict.get("lang", "en")
             XTTS_MAP = {"en": "en", "de": "de", "fr": "fr", "es": "es", "it": "it", "pl": "pl", "pt": "pt", "tr": "tr", "ru": "ru", "nl": "nl", "cs": "cs", "ar": "ar", "hu": "hu", "ko": "ko", "hi": "hi", "zh": "zh-cn"}
-            clean_lang = lang.strip().lower().split('-')[0]
+            clean_lang = (request_dict.get("lang") or "en").strip().lower().split('-')[0]
             mapped_lang = XTTS_MAP.get(clean_lang) or ("zh-cn" if clean_lang == "zh" else None)
             
             if mapped_lang:
@@ -247,10 +241,10 @@ 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}"}
     except Exception as e:
-        print(f"❌ Core Error: {traceback.format_exc()}")
+        print(f"❌ Processing Fault: {traceback.format_exc()}")
         res = {"error": str(e)}
     finally:
-        print(f"--- [v109] ✨ DONE ({time.time()-t1:.2f}s) ---")
+        print(f"--- [v110] ✨ FINISHED ---")
         release_gpu_models()
     return res
 
@@ -259,7 +253,7 @@ async def lifespan(app: FastAPI):
     Thread(target=warmup_task, daemon=True).start()
     yield
 
-# 🚀 4. FastAPI & Gradio Unified
+# 🚀 FastAPI & Gradio Unified
 app = FastAPI(lifespan=lifespan)
 app.add_middleware(CORSMiddleware, allow_origins=["*"], allow_methods=["*"], allow_headers=["*"])
 
@@ -268,21 +262,21 @@ async def api_process(request: Request):
     try:
         req_data = await request.json()
         if req_data.get("action") == "health":
-            return {"status": "awake", "warm": WARMUP_STATUS["complete"], "v": "109"}
+            return {"status": "awake", "warm": WARMUP_STATUS["complete"], "v": "110"}
         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"], "v": "109"}
+def health(): return {"status": "ok", "warm": WARMUP_STATUS["complete"], "v": "110"}
 
 def gradio_fn(req_json):
     try: return json.dumps(core_process(json.loads(req_json)))
     except Exception as e: return json.dumps({"error": str(e)})
 
-demo = gr.Interface(fn=gradio_fn, inputs="text", outputs="text", title="🚀 AI Engine v109")
+demo = gr.Interface(fn=gradio_fn, inputs="text", outputs="text", title="🚀 AI Engine v110")
 demo.queue()
 app = gr.mount_gradio_app(app, demo, path="/")
 
 if __name__ == "__main__":
-    print("🚀 [v109] Starting Sterile Server...")
+    print("🚀 [v110] Starting Resilient Server...")
     uvicorn.run(app, host="0.0.0.0", port=7860, log_level="error")