|
|
import os |
|
|
import gradio as gr |
|
|
import requests |
|
|
import pandas as pd |
|
|
from smolagents import ToolCallingAgent, tool |
|
|
from duckduckgo_search import DDGS |
|
|
from smolagents.schema import Tool |
|
|
|
|
|
|
|
|
|
|
|
@Tool |
|
|
def duck_search(query: str) -> str: |
|
|
"""Searches the web using DuckDuckGo for the given query and returns the top results.""" |
|
|
with DDGS() as ddgs: |
|
|
results = ddgs.text(query, max_results=3) |
|
|
return "\n".join([r["body"] for r in results]) |
|
|
|
|
|
|
|
|
|
|
|
class GAIAAgent: |
|
|
def __init__(self): |
|
|
self.agent = ToolCallingAgent( |
|
|
name="GAIAWebToolAgent", |
|
|
description="An agent using DuckDuckGo and calculator tools.", |
|
|
tools=[duck_search, calculator], |
|
|
model="gpt-3.5-turbo" |
|
|
) |
|
|
|
|
|
def __call__(self, question: str) -> str: |
|
|
print(f"π Question: {question}") |
|
|
try: |
|
|
return self.agent.run( |
|
|
question, |
|
|
step_limit=5, |
|
|
system_prompt=( |
|
|
"You are a helpful reasoning agent. You can answer questions using search and calculation tools. " |
|
|
"Be concise and accurate. Think step by step when needed. Use DuckDuckGo for recent or factual queries." |
|
|
) |
|
|
) |
|
|
except Exception as e: |
|
|
print(f"β Agent error: {e}") |
|
|
return f"Error: {e}" |
|
|
|
|
|
|
|
|
|
|
|
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space" |
|
|
|
|
|
def run_and_submit_all(profile: gr.OAuthProfile | None): |
|
|
space_id = os.getenv("SPACE_ID") |
|
|
api_url = DEFAULT_API_URL |
|
|
questions_url = f"{api_url}/questions" |
|
|
submit_url = f"{api_url}/submit" |
|
|
|
|
|
if profile: |
|
|
username = profile.username |
|
|
print(f"User: {username}") |
|
|
else: |
|
|
return "Please login to Hugging Face.", None |
|
|
|
|
|
try: |
|
|
agent = GAIAAgent() |
|
|
except Exception as e: |
|
|
return f"Agent initialization error: {e}", None |
|
|
|
|
|
agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main" |
|
|
|
|
|
try: |
|
|
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 |
|
|
|
|
|
answers_payload = [] |
|
|
results_log = [] |
|
|
|
|
|
for item in questions_data: |
|
|
task_id = item.get("task_id") |
|
|
question = item.get("question") |
|
|
if not task_id or not question: |
|
|
continue |
|
|
try: |
|
|
answer = agent(question) |
|
|
answers_payload.append({"task_id": task_id, "submitted_answer": answer}) |
|
|
results_log.append({"Task ID": task_id, "Question": question, "Submitted Answer": answer}) |
|
|
except Exception as e: |
|
|
results_log.append({"Task ID": task_id, "Question": question, "Submitted Answer": f"Error: {e}"}) |
|
|
|
|
|
if not answers_payload: |
|
|
return "No answers generated.", pd.DataFrame(results_log) |
|
|
|
|
|
submission_data = { |
|
|
"username": username, |
|
|
"agent_code": agent_code, |
|
|
"answers": answers_payload |
|
|
} |
|
|
|
|
|
try: |
|
|
response = requests.post(submit_url, json=submission_data, timeout=60) |
|
|
response.raise_for_status() |
|
|
result_data = response.json() |
|
|
final_status = ( |
|
|
f"β
Submission Successful!\n" |
|
|
f"User: {result_data.get('username')}\n" |
|
|
f"Score: {result_data.get('score')}% " |
|
|
f"({result_data.get('correct_count')}/{result_data.get('total_attempted')})\n" |
|
|
f"Message: {result_data.get('message', '')}" |
|
|
) |
|
|
return final_status, pd.DataFrame(results_log) |
|
|
except Exception as e: |
|
|
return f"Submission error: {e}", pd.DataFrame(results_log) |
|
|
|
|
|
|
|
|
|
|
|
with gr.Blocks() as demo: |
|
|
gr.Markdown("## π GAIA Agent with Web Search & Calculator Tools") |
|
|
gr.Markdown( |
|
|
"1. Log in with your Hugging Face account.\n" |
|
|
"2. Click the button to evaluate the agent on GAIA questions.\n" |
|
|
"3. Results and score will appear below." |
|
|
) |
|
|
|
|
|
gr.LoginButton() |
|
|
run_button = gr.Button("π Run & Submit to GAIA") |
|
|
status_output = gr.Textbox(label="Status", lines=5) |
|
|
results_table = gr.DataFrame(label="Results") |
|
|
|
|
|
run_button.click(fn=run_and_submit_all, outputs=[status_output, results_table]) |
|
|
|
|
|
|
|
|
if __name__ == "__main__": |
|
|
print("π Launching app...") |
|
|
demo.launch() |
|
|
|