Update app.py

app.py

@@ -272,7 +272,7 @@ def generate_endpoint(user_message):
         system_instruction = request.args.get('system', '')
         temperature = float(request.args.get('temperature', 0.7))
         top_p = float(request.args.get('top_p', 0.95))
-        top_k = int(request.args.get('top_k', 40))
+        top_k = int(request.args.get('top_k', 0))
         model_name = request.args.get('model', MODELS[0]["name"])
         max_tokens = int(request.args.get('max_tokens', MAX_GENERATION_TOKENS))
         
@@ -291,12 +291,6 @@ def generate_endpoint(user_message):
                 mimetype='text/plain'
             )
 
-        if top_k:
-            try:
-                kwargs["top_k"] = int(top_k)
-            except ValueError:
-                return Response("Error: top_k debe ser número entero", status=400)
-            
         # Limitar max_tokens a la configuración máxima
         if max_tokens > MAX_GENERATION_TOKENS:
             max_tokens = MAX_GENERATION_TOKENS
@@ -322,7 +316,13 @@ def generate_endpoint(user_message):
             "max_tokens": max_tokens,
             "stream": False
         }
-        
+
+        if top_k:
+            try:
+                kwargs["top_k"] = int(top_k)
+            except ValueError:
+                return Response("Error: top_k debe ser número entero", status=400)
+            
         # Generar respuesta
         result = llm_manager.chat_completion(model_name, messages, **kwargs)