Update app.py

app.py

@@ -17,26 +17,41 @@ from huggingface_hub import InferenceClient, login
 
 class BasicAgent:
     def __init__(self, hf_token: str):
-        # Validate authentication
+        # Required for gated models
         login(token=hf_token, add_to_git_credential=False)
         
-        # Initialize client with production endpoint
+        # Verify model access
+        try:
+            model_info = HfApi().model_info(
+                repo_id="meta-llama/Meta-Llama-3-70B-Instruct",
+                token=hf_token
+            )
+            if model_info.gated:
+                print("Verified model access")
+        except Exception as e:
+            raise RuntimeError(f"Model access denied: {str(e)}")
+
+        # Initialize client
         self.client = InferenceClient(
             model="meta-llama/Meta-Llama-3-70B-Instruct",
             token=hf_token,
-            timeout=45
+            timeout=60
         )
         
     def __call__(self, question: str) -> str:
         try:
             response = self.client.text_generation(
-                prompt=f"GAIA Question: {question}\nAnswer:",
-                temperature=0.3,
-                max_new_tokens=512
+                prompt=f"""GAIA RESPONSE PROTOCOL
+                        Question: {question}
+                        Answer: [EXACT ANSWER ONLY, NO EXPLANATION]""",
+                temperature=0.1,
+                max_new_tokens=50
             )
-            return response.split("Answer:")[-1].strip()
-        except Exception as e:
-            return f"Model Error: {str(e)}"
+            # Extract only the answer portion
+            return response.split("Answer:")[-1].strip().split("\n")[0]
+        except Exception:
+            # Return empty string instead of error message
+            return ""
 
 
 def run_and_submit_all( profile: gr.OAuthProfile | None):