Update app.py

app.py

@@ -5,10 +5,8 @@ import pandas as pd
 from smolagents import ToolCallingAgent, tool
 from duckduckgo_search import DDGS
 import math
-import re
-from datetime import datetime
 
-# --- Tools with Proper Docstrings ---
+# --- Tools ---
 @tool
 def web_search(query: str) -> str:
     """Performs a web search using DuckDuckGo.
@@ -17,7 +15,7 @@ def web_search(query: str) -> str:
         query: The search query string.
         
     Returns:
-        A formatted string with search results.
+        A formatted string with search results or error message.
     """
     try:
         with DDGS() as ddgs:
@@ -31,7 +29,7 @@ def web_search(query: str) -> str:
 
 @tool
 def calculate(expression: str) -> str:
-    """Evaluates mathematical expressions.
+    """Evaluates mathematical expressions safely.
     
     Args:
         expression: The math expression to evaluate.
@@ -40,54 +38,48 @@ def calculate(expression: str) -> str:
         The result as a string or error message.
     """
     try:
-        # Safe evaluation environment
         safe_dict = {k: v for k, v in math.__dict__.items() 
                     if not k.startswith("__")}
-        safe_dict.update({
-            '__builtins__': None,
-            'abs': abs,
-            'round': round
-        })
-        result = eval(expression, {'__builtins__': None}, safe_dict)
-        return str(result)
+        safe_dict.update({'__builtins__': None, 'abs': abs, 'round': round})
+        return str(eval(expression, {'__builtins__': None}, safe_dict))
     except Exception as e:
         return f"Calculation error: {str(e)}"
 
-# --- Robust Agent Class ---
+# --- Agent with System Prompt ---
 class GAIAAgent:
     def __init__(self):
-        """Initialize the agent with proper error handling."""
+        self.system_prompt = """You are an AI assistant designed to answer questions accurately using these tools:
+
+1. web_search - For finding current information
+2. calculate - For mathematical calculations
+
+Guidelines:
+- For factual questions, always use web_search
+- For math problems, use calculate
+- Be concise but thorough
+- Verify information when possible
+- If unsure, say so rather than guessing"""
+
         try:
             self.agent = ToolCallingAgent(
-                name="GAIA_Submission_Agent",
-                description="Agent for GAIA benchmark tasks",
+                name="GAIA_Agent",
+                description=self.system_prompt,
                 tools=[web_search, calculate],
                 model="gpt-3.5-turbo",
-                planning_interval=3  # Changed from max_iterations
+                planning_interval=3
             )
-            print("✅ Agent initialized successfully")
         except Exception as e:
-            print(f"❌ Agent initialization failed: {str(e)}")
-            raise
+            raise RuntimeError(f"Agent initialization failed: {str(e)}")
 
     def __call__(self, question: str) -> str:
         """Process a question with proper error handling."""
         try:
-            if not question or not isinstance(question, str):
-                return "Invalid question format"
-                
-            # Run the agent
             response = self.agent.run(question)
-            
-            # Ensure we return a string
-            if response is None:
-                return "No response generated"
-            return str(response)
+            return str(response) if response else "No answer generated"
         except Exception as e:
-            print(f"⚠️ Processing error: {str(e)}")
-            return f"Error processing question: {str(e)}"
+            return f"Error: {str(e)}"
 
-# --- Submission Logic ---
+# --- Gradio Interface ---
 def submit_answers(profile: gr.OAuthProfile | None):
     if not profile:
         return "Please login to Hugging Face", None
@@ -98,25 +90,19 @@ def submit_answers(profile: gr.OAuthProfile | None):
             "https://agents-course-unit4-scoring.hf.space/questions",
             timeout=20
         )
-        questions = response.json()
+        questions = response.json() or []
         
-        if not questions:
-            return "No questions received", None
-            
         answers = []
         results = []
         
         for item in questions[:15]:  # Process first 15 for testing
-            task_id = item.get("task_id")
-            question = item.get("question")
+            task_id = item.get("task_id", "")
+            question = item.get("question", "")
             
-            if not task_id or not question:
-                continue
-                
             answer = agent(question)
             answers.append({
                 "task_id": task_id,
-                "submitted_answer": answer[:1000]  # Limit answer length
+                "submitted_answer": answer[:1000]
             })
             results.append({
                 "Task ID": task_id,
@@ -146,18 +132,14 @@ def submit_answers(profile: gr.OAuthProfile | None):
     except Exception as e:
         return f"Error: {str(e)}", None
 
-# --- Gradio Interface ---
 with gr.Blocks() as demo:
-    gr.Markdown("# GAIA Submission Agent")
+    gr.Markdown("# GAIA Agent")
     gr.LoginButton()
     submit_btn = gr.Button("Run Evaluation", variant="primary")
     output = gr.Textbox(label="Results")
     table = gr.DataFrame(label="Details")
     
-    submit_btn.click(
-        fn=submit_answers,
-        outputs=[output, table]
-    )
+    submit_btn.click(submit_answers, outputs=[output, table])
 
 if __name__ == "__main__":
     demo.launch()