Update app.py

app.py

@@ -11,21 +11,27 @@ DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
 # --- Basic Agent Definition ---
 # ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
+import os
+from huggingface_hub import InferenceClient
+
 class BasicAgent:
     def __init__(self):
-        print("BasicAgent initialized.")
+        print("BasicAgent initialized with Llama 3.1-8B Instruct.")
+        # Make sure you set HF_TOKEN in your Space secrets
         self.client = InferenceClient(
-            model="mistralai/Mistral-7B-Instruct-v0.2",
+            model="meta-llama/Llama-3.1-8B-Instruct",
             token=os.getenv("HF_TOKEN")
         )
 
     def __call__(self, question: str) -> str:
         print(f"Agent received question (first 50 chars): {question[:50]}...")
 
+        # GAIA exact match prompt rules
         prompt = f"""
-Answer the question EXACTLY.
+You are answering a benchmark question.
 
 Rules:
+- Answer EXACTLY
 - No explanations
 - No extra words
 - No punctuation unless required
@@ -35,21 +41,25 @@ Question:
 {question}
 """
 
-        response = self.client.chat.completions.create(
-            messages=[
-                {"role": "system", "content": "You return only the final answer."},
-                {"role": "user", "content": prompt}
-            ],
-            max_tokens=64,
-            temperature=0.0,
-        )
+        # Use chat-completion for instruction-tuned Llama 3.1
+        try:
+            response = self.client.chat.completions.create(
+                model="meta-llama/Llama-3.1-8B-Instruct",
+                messages=[{"role": "user", "content": prompt}],
+                max_tokens=128,
+                temperature=0.0,
+            )
+            # Extract the assistant's message text
+            answer = response.choices[0].message.content.strip().replace("\n", " ")
+        except Exception as e:
+            print(f"Error generating answer: {e}")
+            answer = "AGENT ERROR"
 
-        answer = response.choices[0].message.content
-        answer = answer.strip().replace("\n", "")
         print(f"Agent answer: {answer}")
         return answer
 
 
+
 def run_and_submit_all( profile: gr.OAuthProfile | None):
     """
     Fetches all questions, runs the BasicAgent on them, submits all answers,