Update app.py

app.py

@@ -7,7 +7,7 @@ import requests
 import tempfile
 import json
 from concurrent.futures import ThreadPoolExecutor
-import signal
+import threading
 
 app = Flask(__name__)
 logging.basicConfig(level=logging.INFO)
@@ -96,37 +96,42 @@ class LLMManager:
 
     def chat_completion(self, model_name, messages, **kwargs):
         """Generar respuesta con modelo específico"""
+        # Intentar obtener el lock (no bloqueante)
         if not self.generation_lock.acquire(blocking=False):
             return {"error": "Servidor ocupado - Generación en progreso"}
-    
+        
         try:
             model_data = self.get_model(model_name)
-        
+            
             if not model_data or not model_data["loaded"]:
                 error_msg = f"Modelo {model_name} no cargado"
                 if model_data and "error" in model_data:
                     error_msg += f": {model_data['error']}"
                 return {"error": error_msg}
-        
+            
+            import signal
+            
             def timeout_handler(signum, frame):
                 raise TimeoutError("Timeout en generación (120 segundos)")
-        
+            
+            # Configurar timeout (120 segundos)
             signal.signal(signal.SIGALRM, timeout_handler)
             signal.alarm(120)
-        
+            
             try:
                 response = model_data["instance"].create_chat_completion(
                     messages=messages,
                     **kwargs
                 )
             finally:
-                signal.alarm(0)
+                signal.alarm(0)  # Cancelar timeout
 
             response["provider"] = "telechars-ai"
             response["model"] = model_name
             return response
-        
+            
         finally:
+            # Siempre liberar el lock
             self.generation_lock.release()
 
     def get_loaded_models(self):