Update app.py

app.py

@@ -5,110 +5,149 @@ import pandas as pd
 import math
 
 from smolagents import ToolCallingAgent, tool
-from smolagents.models import OpenAIServerModel
 from duckduckgo_search import DDGS
+from openai import OpenAI
+
+# Load OpenAI API key
+client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))
+
+# ------------------------
+# Define Tools
+# ------------------------
 
-# --- Tools ---
 @tool
 def web_search(query: str) -> str:
-    """
-    Perform a web search using DuckDuckGo.
+    """Search the web using DuckDuckGo.
 
     Args:
-        query (str): The search query string.
+        query: The search query to look up.
 
     Returns:
-        str: A formatted string with search results.
+        A summary of the top web results.
     """
     try:
         with DDGS() as ddgs:
             results = ddgs.text(query, max_results=3)
+            if not results:
+                return "No results found."
             return "\n\n".join(
-                f"Title: {r['title']}\nContent: {r['body']}\nURL: {r['href']}"
+                f"Title: {r['title']}\nSnippet: {r['body']}\nURL: {r['href']}"
                 for r in results
-            ) if results else "No results found."
+            )
     except Exception as e:
         return f"Search error: {str(e)}"
 
-
 @tool
 def calculate(expression: str) -> str:
-    """
-    Evaluate a mathematical expression safely.
+    """Evaluate a mathematical expression.
 
     Args:
-        expression (str): The math expression to evaluate.
+        expression: The math expression to evaluate (e.g. '2 + 3 * 5').
 
     Returns:
-        str: The result as a string, or an error message.
+        Result of the calculation.
     """
     try:
-        safe_dict = {k: v for k, v in math.__dict__.items() if not k.startswith("__")}
-        safe_dict.update({'abs': abs, 'round': round})
-        result = eval(expression, {"__builtins__": None}, safe_dict)
+        safe_math = {k: v for k, v in math.__dict__.items() if not k.startswith("__")}
+        result = eval(expression, {"__builtins__": None}, safe_math)
         return str(result)
     except Exception as e:
         return f"Calculation error: {str(e)}"
 
-# --- Agent ---
+# ------------------------
+# Define Agent
+# ------------------------
+
 class GAIAAgent:
     def __init__(self):
-        self.system_prompt = """You are an AI assistant that answers questions using tools:
-- Use web_search for factual queries
-- Use calculate for math problems
-- Be concise and accurate."""
-
-        try:
-            self.agent = ToolCallingAgent(
-                name="GAIA_Agent",
-                description=self.system_prompt,  # this sets the prompt
-                tools=[web_search, calculate],
-                model=client.chat.completions  # callable, no kwargs
-            )
-            print("✅ Agent initialized successfully")
-        except Exception as e:
-            raise RuntimeError(f"Agent init failed: {str(e)}")
+        self.agent = ToolCallingAgent(
+            name="GAIA Agent",
+            description="""You are an AI assistant that answers questions using tools:
+- Use 'web_search' for looking up facts and recent information.
+- Use 'calculate' for evaluating math expressions.
+Be accurate and concise.""",
+            tools=[web_search, calculate],
+            model=client.chat.completions
+        )
 
     def __call__(self, question: str) -> str:
         try:
-            response = self.agent.run(question)  # <-- no system_prompt here
-            return str(response) if response else "No answer generated"
+            response = self.agent.run(question)
+            return str(response)
         except Exception as e:
-            return f"Error: {str(e)}"
+            return f"Agent error: {str(e)}"
+
+# ------------------------
+# Gradio App Logic
+# ------------------------
 
-# --- Gradio UI + Submission ---
-def submit_answers(profile: gr.OAuthProfile | None):
+def run_agent_and_submit(profile: gr.OAuthProfile | None):
     if not profile:
-        return "Please login to Hugging Face", None
+        return "⚠️ Please log in to Hugging Face.", None
+
+    try:
+        agent = GAIAAgent()
+        response = requests.get("https://agents-course-unit4-scoring.hf.space/questions", timeout=20)
+        questions = response.json()
+    except Exception as e:
+        return f"❌ Error fetching questions: {e}", None
 
-    agent = GAIAAgent()
-    resp = requests.get("https://agents-course-unit4-scoring.hf.space/questions", timeout=20)
-    questions = resp.json() or []
+    results = []
+    answers = []
 
-    answers, rows = [], []
-    for item in questions[:15]:
-        tid, q = item.get("task_id"), item.get("question")
-        ans = agent(q)
-        answers.append({"task_id": tid, "submitted_answer": ans[:1000]})
-        rows.append({"Task ID": tid, "Question": q[:100], "Answer": ans[:200]})
+    for q in questions:
+        task_id = q.get("task_id")
+        question_text = q.get("question")
+        if not task_id or not question_text:
+            continue
+        try:
+            answer = agent(question_text)
+        except Exception as e:
+            answer = f"Agent error: {e}"
+
+        answers.append({
+            "task_id": task_id,
+            "submitted_answer": answer[:1000]
+        })
+        results.append({
+            "Task ID": task_id,
+            "Question": question_text,
+            "Answer": answer
+        })
+
+    # Submit answers
+    try:
+        submit_url = "https://agents-course-unit4-scoring.hf.space/submit"
+        payload = {
+            "username": profile.username,
+            "agent_code": f"https://huggingface.co/spaces/{os.getenv('SPACE_ID')}/tree/main",
+            "answers": answers
+        }
+        submit_resp = requests.post(submit_url, json=payload, timeout=60)
+        result_data = submit_resp.json()
+        summary = (
+            f"✅ Submitted {len(answers)} answers\n"
+            f"📊 Score: {result_data.get('score', 'N/A')}%\n"
+            f"✔️ Correct: {result_data.get('correct_count', '?')}/{len(answers)}"
+        )
+    except Exception as e:
+        summary = f"❌ Submission error: {e}"
 
-    post = requests.post("https://agents-course-unit4-scoring.hf.space/submit", json={
-        "username": profile.username,
-        "agent_code": f"https://huggingface.co/spaces/{os.getenv('SPACE_ID')}",
-        "answers": answers
-    }, timeout=60)
-    data = post.json()
+    return summary, pd.DataFrame(results)
 
-    out = f"Submitted {len(answers)} answers\nScore: {data.get('score', 'N/A')}%\nCorrect: {data.get('correct_count',0)}/{len(answers)}"
-    return out, pd.DataFrame(rows)
+# ------------------------
+# Gradio Interface
+# ------------------------
 
 with gr.Blocks() as demo:
-    gr.Markdown("# GAIA Agent")
+    gr.Markdown("# 🤖 GAIA Tool Agent")
+    gr.Markdown("This agent answers GAIA benchmark questions using tool-calling with search and math.")
     gr.LoginButton()
-    btn = gr.Button("Run Evaluation", variant="primary")
-    status = gr.Textbox(label="Results")
-    df = gr.DataFrame(label="Details")
-    btn.click(submit_answers, outputs=[status, df])
+    run_btn = gr.Button("🔍 Run Agent & Submit")
+    status = gr.Textbox(label="Status", lines=4)
+    results_df = gr.DataFrame(label="Results")
+
+    run_btn.click(fn=run_agent_and_submit, outputs=[status, results_df])
 
 if __name__ == "__main__":
     demo.launch()