Update app.py

app.py

@@ -9,91 +9,104 @@ DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
 class BasicAgent:
     def __init__(self):
-        # InferenceClientModel is the latest standard for smolagents v1.7+
-        # It automatically uses the HF_TOKEN secret if added to your Space
+        # 1. Initialize the Model (the 'brain')
+        # This wrapper is the most stable version for HF Inference API in 2026.
+        # It will automatically use your HF_TOKEN secret if added to the Space.
         self.model = InferenceClientModel(model_id="Qwen/Qwen2.5-72B-Instruct")
         
-        # add_base_tools=True enables the Python interpreter tool
+        # 2. Initialize the CodeAgent (the 'body')
+        # - tools=[]: We start with no external tools.
+        # - add_base_tools=False: This prevents the 'ddgs' / DuckDuckGo error.
+        # Note: CodeAgent still has a built-in Python interpreter to solve math/logic!
         self.agent = CodeAgent(
             tools=[], 
             model=self.model,
-            add_base_tools=True 
+            add_base_tools=False 
         )
-        print("Agent initialized with InferenceClientModel and Python Interpreter.")
+        print("Agent successfully initialized with Python Interpreter (No ddgs needed).")
 
     def __call__(self, question: str) -> str:
-        # Strict prompt to ensure "Exact Match" scoring works
+        # 3. Prompting for Exact Match scoring
+        # We tell the agent to be as direct as possible.
         clean_prompt = (
-            f"Solve the following task: {question}\n\n"
-            "Final Answer Instructions: Provide ONLY the final result value. "
-            "No extra words, no units, no 'The answer is...', and no 'FINAL ANSWER' text."
+            f"Question: {question}\n\n"
+            "Final Answer Requirement: Provide ONLY the numeric or text value of the answer. "
+            "Do not include any explanation, units, or 'The answer is' text. "
+            "Do not include 'FINAL ANSWER' in your output."
         )
         
         try:
-            # The agent will use its Python tool if it needs to calculate or process data
+            # The agent will write and run Python code if the question requires it.
             result = self.agent.run(clean_prompt)
             return str(result).strip()
         except Exception as e:
-            print(f"Agent execution error: {e}")
-            return "Error"
+            print(f"Error during agent execution: {e}")
+            return "Error solving question"
 
 def run_and_submit_all(profile: gr.OAuthProfile | None):
-    # Determine the Space ID for the code link
     space_id = os.getenv("SPACE_ID")
 
     if not profile:
-        return "Please Login to Hugging Face first.", None
+        return "Please Login to Hugging Face with the button.", None
 
-    username = profile.username
-    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" if space_id else ""
+    username = f"{profile.username}"
+    api_url = DEFAULT_API_URL
+    questions_url = f"{api_url}/questions"
+    submit_url = f"{api_url}/submit"
+    
+    # URL to your code for verification
+    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" if space_id else "https://huggingface.co/spaces"
 
-    # 1. Fetch the evaluation questions
+    # 1. Fetch Questions
     try:
-        response = requests.get(f"{DEFAULT_API_URL}/questions", timeout=15)
+        response = requests.get(questions_url, timeout=15)
         response.raise_for_status()
         questions_data = response.json()
     except Exception as e:
         return f"Error fetching questions: {e}", None
 
-    # 2. Instantiate and run the agent
-    agent = BasicAgent()
-    answers_payload = []
+    # 2. Run Agent
+    # Instantiate inside the function to ensure a fresh session
+    try:
+        agent = BasicAgent()
+    except Exception as e:
+        return f"Error initializing agent: {e}", None
+
     results_log = []
+    answers_payload = []
     
-    print(f"Processing {len(questions_data)} tasks...")
+    print(f"Starting evaluation for {len(questions_data)} questions...")
 
     for item in questions_data:
         task_id = item.get("task_id")
         question_text = item.get("question")
         
-        if not task_id: continue
+        if not task_id or question_text is None:
+            continue
 
-        # Generate answer
-        submitted_answer = agent(question_text)
-        
-        answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
-        results_log.append({
-            "Task ID": task_id, 
-            "Question": question_text[:100] + "...", 
-            "Agent Answer": submitted_answer
-        })
-
-    # 3. Submit to the leaderboard
+        try:
+            submitted_answer = agent(question_text)
+            answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
+            results_log.append({"Task ID": task_id, "Question": question_text[:80], "Submitted Answer": submitted_answer})
+        except Exception as e:
+            results_log.append({"Task ID": task_id, "Question": question_text[:80], "Submitted Answer": f"ERROR: {e}"})
+
+    # 3. Submit Results
     submission_data = {
-        "username": username, 
+        "username": username.strip(), 
         "agent_code": agent_code, 
         "answers": answers_payload
     }
 
     try:
-        response = requests.post(f"{DEFAULT_API_URL}/submit", json=submission_data, timeout=60)
+        response = requests.post(submit_url, json=submission_data, timeout=60)
         response.raise_for_status()
         result_data = response.json()
         
         final_status = (
             f"Submission Successful!\n"
             f"User: {result_data.get('username')}\n"
-            f"Score: {result_data.get('score', 0)}% "
+            f"Overall Score: {result_data.get('score', 0)}% "
             f"({result_data.get('correct_count', 0)}/{result_data.get('total_attempted', 0)} correct)"
         )
         return final_status, pd.DataFrame(results_log)
@@ -102,14 +115,14 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
 
 # --- Gradio UI ---
 with gr.Blocks() as demo:
-    gr.Markdown("# GAIA Solver (v2026 Optimized)")
-    gr.Markdown("Click Login, then Run to evaluate your agent on the GAIA dataset.")
+    gr.Markdown("# GAIA Final Evaluation Solver")
+    gr.Markdown("Click 'Login' then 'Run' to solve all questions and submit your score.")
 
     gr.LoginButton()
-    run_button = gr.Button("Run Evaluation & Submit", variant="primary")
+    run_button = gr.Button("Run Evaluation & Submit All Answers", variant="primary")
 
-    status_output = gr.Textbox(label="Status/Score", lines=4)
-    results_table = gr.DataFrame(label="Task Log", wrap=True)
+    status_output = gr.Textbox(label="Submission Status", lines=5)
+    results_table = gr.DataFrame(label="Agent Answers Log", wrap=True)
 
     run_button.click(
         fn=run_and_submit_all,