Deploy GAIA agent

app.py

@@ -4,24 +4,24 @@ import os
 import gradio as gr
 import requests
 import pandas as pd
-from bs4 import BeautifulSoup
 
-from smolagents import CodeAgent, tool, InferenceClientModel
+from smolagents import CodeAgent, tool
+from litellm import LiteLLMModel
 
 # --- Constants ---
 DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 
-# --- Custom Web Search Tool (avoiding proxies) ---
+# --- Tool: simple DuckDuckGo search via HTML scraping ---
 @tool
 def simple_search(query: str) -> str:
     """
-    Search DuckDuckGo via HTML parsing.
+    Perform a basic DuckDuckGo search.
 
     Args:
-        query (str): search query string.
+        query (str): Search query.
 
     Returns:
-        str: top 3 titles and URLs.
+        str: Top 3 titles and URLs from search.
     """
     try:
         resp = requests.get(
@@ -30,57 +30,64 @@ def simple_search(query: str) -> str:
             timeout=10
         )
         resp.raise_for_status()
+        from bs4 import BeautifulSoup
         soup = BeautifulSoup(resp.text, "html.parser")
-        results = []
-        for a in soup.select("a.result__a")[:3]:
-            title = a.get_text()
-            url = a["href"]
-            results.append(f"- {title}\n  {url}")
-        return "\n".join(results) if results else "No results."
+        items = soup.select("a.result__a")[:3]
+        results = [f"{a.get_text()}\n{a['href']}" for a in items]
+        return "\n\n".join(results) if results else "No results found."
     except Exception as e:
         return f"Search error: {e}"
 
-# --- Enhanced Agent ---
+# --- Enhanced Agent using a public model ---
 class BasicAgent:
     def __init__(self):
-        print("BasicAgent initialized.")
-        self.model = InferenceClientModel()
+        print("BasicAgent initialized with Falcon 7B Instruct model.")
+        self.model = LiteLLMModel(
+            model_id="tiiuae/falcon-7b-instruct",
+            max_tokens=512,
+            temperature=0.1
+        )
         self.agent = CodeAgent(
             model=self.model,
             tools=[simple_search]
         )
 
     def __call__(self, question: str) -> str:
-        print("Received:", question[:50], "...")
+        print(f"Received question: {question[:50]}...")
         try:
             return self.agent.run(question)
         except Exception as e:
-            return f"Agent run error: {e}"
+            return f"Agent error: {e}"
 
 def run_and_submit_all(profile: gr.OAuthProfile | None):
     space_id = os.getenv("SPACE_ID")
     if not profile:
-        return "Please login to Hugging Face", None
+        return "Please log in to Hugging Face to submit.", None
     username = profile.username
+
     questions_url = f"{DEFAULT_API_URL}/questions"
     submit_url = f"{DEFAULT_API_URL}/submit"
 
-    try: agent = BasicAgent()
+    try:
+        agent = BasicAgent()
     except Exception as e:
-        return f"Agent init error: {e}", None
+        return f"Agent init failed: {e}", None
 
     agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"
+
     try:
-        r = requests.get(questions_url, timeout=15)
-        r.raise_for_status()
-        qs = r.json()
+        resp = requests.get(questions_url, timeout=15)
+        resp.raise_for_status()
+        qs = resp.json()
     except Exception as e:
-        return f"Fetch questions error: {e}", None
+        return f"Failed to fetch questions: {e}", None
 
     logs, payload = [], []
     for item in qs:
-        tid, q = item.get("task_id"), item.get("question")
-        if not tid or q is None: continue
+        tid = item.get("task_id")
+        q = item.get("question")
+        if not tid or q is None:
+            continue
         ans = agent(q)
         payload.append({"task_id": tid, "submitted_answer": ans})
         logs.append({"Task ID": tid, "Question": q, "Submitted Answer": ans})
@@ -90,26 +97,29 @@ def run_and_submit_all(profile: gr.OAuthProfile | None):
 
     submit = {"username": username, "agent_code": agent_code, "answers": payload}
     try:
-        resp = requests.post(submit_url, json=submit, timeout=60)
-        resp.raise_for_status()
-        data = resp.json()
+        r2 = requests.post(submit_url, json=submit, timeout=60)
+        r2.raise_for_status()
+        data = r2.json()
         status = (
-            f"✅ Score: {data.get('score','N/A')}% "
-            f"({data.get('correct_count','?')}/{data.get('total_attempted','?')} correct)\n"
-            f"{data.get('message','No message')}"
+            f"✅ Submission Successful!\n"
+            f"Score: {data.get('score','N/A')}% "
+            f"({data.get('correct_count','?')}/{data.get('total_attempted','?')})\n"
+            f"{data.get('message','')}"
         )
         return status, pd.DataFrame(logs)
     except Exception as e:
-        return f"Submit error: {e}", pd.DataFrame(logs)
+        return f"Submit failed: {e}", pd.DataFrame(logs)
 
 # --- Gradio UI ---
 with gr.Blocks() as demo:
-    gr.Markdown("# GAIA Evaluation Runner")
+    gr.Markdown("# GAIA Agent Evaluation Runner")
     gr.LoginButton()
-    run_btn = gr.Button("Run Evaluation & Submit")
-    status = gr.Textbox(lines=5, interactive=False)
-    results = gr.DataFrame(wrap=True)
+    run_btn = gr.Button("Run Evaluation & Submit All Answers")
+    status = gr.Textbox(label="Status / Submission Result", lines=5, interactive=False)
+    results = gr.DataFrame(label="Questions & Submitted Answers", wrap=True)
+
     run_btn.click(run_and_submit_all, outputs=[status, results])
 
 if __name__ == "__main__":
+    print("Launching Gradio app...")
     demo.launch(debug=True, share=False)