app.py

import os
import gradio as gr
import requests
import pandas as pd

from smolagents import CodeAgent, HfApiModel, DuckDuckGoSearchTool

# --- Constants ---
DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"

# --- Model + Agent ---
model = HfApiModel(model_id="Qwen/Qwen2.5-Coder-32B-Instruct")

real_agent = CodeAgent(
    model=model,
    tools=[DuckDuckGoSearchTool()]
)

class BasicAgent:
    def __init__(self):
        print("Agent initialized")

    def __call__(self, question: str) -> str:
        try:
            result = real_agent.run(question)
            return str(result).strip()
        except Exception as e:
            print(f"Agent error: {e}")
            return "unknown"


# --- Main runner ---
def run_and_submit_all(profile: gr.OAuthProfile | None):

    if profile:
        username = profile.username
    else:
        return "Please login first", pd.DataFrame([])

    space_id = os.getenv("SPACE_ID", "unknown")
    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"

    questions_url = f"{DEFAULT_API_URL}/questions"
    submit_url = f"{DEFAULT_API_URL}/submit"

    agent = BasicAgent()

    # --- Fetch questions ---
    try:
        response = requests.get(questions_url, timeout=30)
        response.raise_for_status()
        questions_data = response.json()
    except Exception as e:
        return f"Failed to fetch questions: {e}", pd.DataFrame([])

    results_log = []
    answers_payload = []

    # --- Run agent ---
    for item in questions_data:
        task_id = item.get("task_id")
        question = item.get("question")

        if not task_id or not question:
            continue

        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
        })

    # --- Submit ---
    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 = response.json()

        return (
            f"Score: {result.get('score', 0)}%",
            pd.DataFrame(results_log)
        )

    except Exception as e:
        return f"Submission failed: {e}", pd.DataFrame(results_log)


# --- UI ---
with gr.Blocks() as demo:
    gr.Markdown("# Basic Agent Evaluation Runner")

    gr.LoginButton()

    btn = gr.Button("Run Evaluation & Submit All Answers")

    out1 = gr.Textbox(label="Score")
    out2 = gr.DataFrame(label="Results")

    btn.click(run_and_submit_all, outputs=[out1, out2])


if __name__ == "__main__":
    demo.launch()