Update app.py

app.py

@@ -2,79 +2,83 @@ import os
 import gradio as gr
 import requests
 import pandas as pd
-from smolagents import CodeAgent, HfApiModel
+from smolagents import CodeAgent, InferenceClientModel
 
 # --- Constants ---
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
 class BasicAgent:
     def __init__(self):
-        # HfApiModel is native and fast within the HF ecosystem
-        # It uses your HF_TOKEN automatically if set in Secrets
-        self.model = HfApiModel(model_id="Qwen/Qwen2.5-72B-Instruct")
+        # InferenceClientModel is the latest standard for smolagents v1.7+
+        # It automatically uses the HF_TOKEN secret if added to your Space
+        self.model = InferenceClientModel(model_id="Qwen/Qwen2.5-72B-Instruct")
         
-        # We use CodeAgent WITHOUT external tools like Search.
-        # add_base_tools=True provides the Python Interpreter for logic/math.
+        # add_base_tools=True enables the Python interpreter tool
         self.agent = CodeAgent(
             tools=[], 
             model=self.model,
             add_base_tools=True 
         )
-        print("Clean Agent initialized (Python Interpreter enabled).")
+        print("Agent initialized with InferenceClientModel and Python Interpreter.")
 
     def __call__(self, question: str) -> str:
-        # Prompt specifically designed for GAIA Exact Match scoring
+        # Strict prompt to ensure "Exact Match" scoring works
         clean_prompt = (
-            f"Solve this task: {question}\n\n"
-            "Final Answer Requirement: Provide ONLY the numeric or text value. "
-            "Do not include units, symbols, or conversational filler. "
-            "No 'The answer is...', no 'FINAL ANSWER' text. Just the raw value."
+            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."
         )
         
         try:
-            # The agent will write code to solve if the question is complex
+            # The agent will use its Python tool if it needs to calculate or process data
             result = self.agent.run(clean_prompt)
             return str(result).strip()
         except Exception as e:
-            print(f"Agent Error: {e}")
+            print(f"Agent execution error: {e}")
             return "Error"
 
 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 with the button.", None
+        return "Please Login to Hugging Face first.", None
 
     username = profile.username
     agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" if space_id else ""
 
-    # 1. Fetch
+    # 1. Fetch the evaluation questions
     try:
         response = requests.get(f"{DEFAULT_API_URL}/questions", timeout=15)
+        response.raise_for_status()
         questions_data = response.json()
     except Exception as e:
-        return f"Fetch Error: {e}", None
+        return f"Error fetching questions: {e}", None
 
-    # 2. Run
+    # 2. Instantiate and run the agent
     agent = BasicAgent()
     answers_payload = []
     results_log = []
     
-    print(f"Starting evaluation for {len(questions_data)} questions...")
-    
+    print(f"Processing {len(questions_data)} tasks...")
+
     for item in questions_data:
         task_id = item.get("task_id")
         question_text = item.get("question")
         
         if not task_id: continue
 
-        # Run agent
+        # 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], "Answer": submitted_answer})
+        results_log.append({
+            "Task ID": task_id, 
+            "Question": question_text[:100] + "...", 
+            "Agent Answer": submitted_answer
+        })
 
-    # 3. Submit
+    # 3. Submit to the leaderboard
     submission_data = {
         "username": username, 
         "agent_code": agent_code, 
@@ -83,27 +87,29 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
 
     try:
         response = requests.post(f"{DEFAULT_API_URL}/submit", json=submission_data, timeout=60)
+        response.raise_for_status()
         result_data = response.json()
         
         final_status = (
             f"Submission Successful!\n"
-            f"Score: {result_data.get('score')}% "
-            f"({result_data.get('correct_count')}/{result_data.get('total_attempted')} correct)"
+            f"User: {result_data.get('username')}\n"
+            f"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)
     except Exception as e:
         return f"Submission Failed: {e}", pd.DataFrame(results_log)
 
-# --- Gradio Interface ---
+# --- Gradio UI ---
 with gr.Blocks() as demo:
-    gr.Markdown("# GAIA Agent Evaluation Runner (Clean Version)")
-    gr.Markdown("Uses `smolagents` with a built-in Python Interpreter to solve tasks.")
+    gr.Markdown("# GAIA Solver (v2026 Optimized)")
+    gr.Markdown("Click Login, then Run to evaluate your agent on the GAIA dataset.")
 
     gr.LoginButton()
-    run_button = gr.Button("Run Evaluation & Submit All", variant="primary")
+    run_button = gr.Button("Run Evaluation & Submit", variant="primary")
 
-    status_output = gr.Textbox(label="Status", lines=4)
-    results_table = gr.DataFrame(label="Generated Answers", wrap=True)
+    status_output = gr.Textbox(label="Status/Score", lines=4)
+    results_table = gr.DataFrame(label="Task Log", wrap=True)
 
     run_button.click(
         fn=run_and_submit_all,