Update app.py

app.py

@@ -25,6 +25,7 @@ class LLMManager:
         self.executor = ThreadPoolExecutor(max_workers=2)
         self.generation_lock = threading.Lock()
         self.load_all_models()
+        self.session = requests.Session()
 
     def load_all_models(self):
         """Cargar todos los modelos en RAM"""
@@ -78,7 +79,7 @@ class LLMManager:
 
         logging.info("📥 Descargando modelo…")
         
-        response = requests.get(model_url, stream=True, timeout=300)
+        response = self.session.get(model_url, stream=True, timeout=300)
         response.raise_for_status()
 
         downloaded = 0