Update modules/llm/hybrid_llm.py

modules/llm/hybrid_llm.py

@@ -17,35 +17,40 @@ class HybridLLM:
         2) Fallback NVIDIA Llama
 
         Prioridad:
-        API ingresada por usuario > .env
+        clave enviada > variable de entorno
         """
 
-        # 🔐 Prioridad a claves dinámicas
         self.gemini_api_key = gemini_key or os.getenv("GOOGLE_API_KEY")
         self.nvidia_api_key = nvidia_key or os.getenv("NVIDIA_API_KEY")
 
         if not self.gemini_api_key:
-            raise ValueError("No se encontró API Key de Gemini.")
+            raise ValueError("❌ No se encontró GOOGLE_API_KEY")
 
         if not self.nvidia_api_key:
-            raise ValueError("No se encontró API Key de NVIDIA.")
+            raise ValueError("❌ No se encontró NVIDIA_API_KEY")
 
-        # 🔹 Inicializar Gemini
+        # Gemini principal
         self.gemini = ChatGoogleGenerativeAI(
-            model="gemini-3-flash-preview",
+            model="gemini-1.5-flash",
             temperature=0.2,
             google_api_key=self.gemini_api_key,
             convert_system_message_to_human=True
         )
 
-    # ---------------- GEMINI ----------------
+    # ----------------------------------------
+    # GEMINI
+    # ----------------------------------------
     def _invoke_gemini(self, prompt):
         print("🟢 Usando Gemini...")
         return self.gemini.invoke(prompt)
 
-    # ---------------- NVIDIA ----------------
+    # ----------------------------------------
+    # NVIDIA FALLBACK
+    # ----------------------------------------
     def _invoke_nvidia(self, prompt):
-        print("🟡 Usando NVIDIA 🤖✅👁️‍🗨️ Llama (fallback)...")
+        print("🟡 Usando NVIDIA fallback...")
+
+        url = "https://integrate.api.nvidia.com/v1/chat/completions"
 
         headers = {
             "Authorization": f"Bearer {self.nvidia_api_key}",
@@ -57,21 +62,25 @@ class HybridLLM:
             "messages": [
                 {
                     "role": "user",
-                    "content": prompt
+                    "content": str(prompt)
                 }
             ],
             "temperature": 0.6,
             "top_p": 0.95,
-            "max_tokens": 2048
+            "max_tokens": 1024,
+            "stream": False
         }
 
         response = requests.post(
-            "https://integrate.api.nvidia.com/v1",
+            url,
             headers=headers,
             json=payload,
             timeout=60
         )
 
+        print("🟠 NVIDIA status:", response.status_code)
+        print("🟠 NVIDIA response:", response.text)
+
         response.raise_for_status()
 
         result = response.json()
@@ -80,42 +89,27 @@ class HybridLLM:
 
         return SimpleResponse(content)
 
-    # ---------------- INVOKE PRINCIPAL ----------------
+    # ----------------------------------------
+    # INVOKE PRINCIPAL
+    # ----------------------------------------
     def invoke(self, prompt):
         """
-        Flujo híbrido:
-        1) Gemini
-        2) si falla o no tiene cuota → NVIDIA
+        Intenta Gemini primero.
+        Si falla por cuota o error -> NVIDIA.
         """
 
         try:
             return self._invoke_gemini(prompt)
 
-        except Exception as e:
-            error = str(e).lower()
-
-            print("⚠️ Error Gemini:", e)
+        except Exception as gemini_error:
+            print("⚠️ Error Gemini:", gemini_error)
 
-            # 🔥 fallback inteligente
-            quota_errors = [
-                "429",
-                "quota",
-                "resource exhausted",
-                "rate limit"
-            ]
-
-            if any(err in error for err in quota_errors):
-                try:
-                    return self._invoke_nvidia(prompt)
-
-                except Exception as nvidia_error:
-                    print("❌ Error NVIDIA:", nvidia_error)
-                    raise nvidia_error
-
-            # fallback por error general
             try:
                 return self._invoke_nvidia(prompt)
 
             except Exception as nvidia_error:
-                print("❌ Error total del sistema:", nvidia_error)
-                raise nvidia_error
+                print("❌ Error NVIDIA:", nvidia_error)
+
+                return SimpleResponse(
+                    f"Error general del sistema: {str(nvidia_error)}"
+                )