app.py

import os
import gradio as gr
import requests
import pandas as pd
from smolagents import ToolCallingAgent, InferenceClientModel, DuckDuckGoSearchTool, VisitWebpageTool

# --- Constants ---
DEFAULT_API_URL = "https://hf.space"

# --- Robust AI Agent Definition ---
class BasicAgent:
    def __init__(self):
        print("Initializing robust ToolCallingAgent...")
        # Fetch the Hugging Face token from the Space variables
        self.token = os.getenv("HF_TOKEN")
        
        # Use a highly accurate, reliable serverless model
        self.model = InferenceClientModel(
            model_id="Qwen/Qwen2.5-72B-Instruct",
            token=self.token
        )
        
        self.search_tool = DuckDuckGoSearchTool()
        self.web_tool = VisitWebpageTool()
        
        self.agent = ToolCallingAgent(
            tools=[self.search_tool, self.web_tool],
            model=self.model,
            max_steps=5
        )

    def __call__(self, question: str) -> str:
        print(f"Agent executing task: {question[:60]}...")
        
        clean_instruction = (
            f"{question}\n\n"
            "CRITICAL: Output ONLY the final raw answer string or numeric value. "
            "Do NOT include conversational filler like 'The answer is', do not use punctuation, "
            "and do not write full sentences. Output just the clean value itself."
        )
        
        try:
            # Attempt to solve using the autonomous agent loop
            result = self.agent.run(clean_instruction)
            return str(result).strip()
            
        except Exception as agent_error:
            print(f"Agent loop failed, engaging direct LLM fallback. Error: {agent_error}")
            
            # FALLBACK: Direct serverless call to ensure an exact-match answer is provided
            try:
                headers = {"Authorization": f"Bearer {self.token}"} if self.token else {}
                api_url = f"https://huggingface.co"
                
                payload = {
                    "inputs": f"<|im_start||user\n{clean_instruction}<|im_end|>\n<|im_start|>assistant\n",
                    "parameters": {"max_new_tokens": 50, "temperature": 0.1}
                }
                
                response = requests.post(api_url, json=payload, headers=headers, timeout=10)
                if response.status_code == 200:
                    output_text = response.json()[0]['generated_text']
                    # Clean up assistant token formatting if present
                    if "assistant" in output_text:
                        output_text = output_text.split("assistant")[-1]
                    return output_text.strip()
            except Exception as fallback_error:
                print(f"Fallback failed: {fallback_error}")
            
            return "Unknown"


def run_and_submit_all(profile: gr.OAuthProfile | None):
    """
    Fetches all questions, runs the AI Agent on them, submits all answers, and displays the results.
    """
    space_id = os.getenv("SPACE_ID") 
    if profile:
        username = f"{profile.username}"
        print(f"User logged in: {username}")
    else:
        print("User not logged in.")
        return "Please Login to Hugging Face with the button.", None

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

    try:
        agent = BasicAgent()
    except Exception as e:
        print(f"Error instantiating agent: {e}")
        return f"Error initializing agent: {e}", None

    agent_code = f"https://huggingface.co{space_id}/tree/main"

    # Fetch Questions
    try:
        response = requests.get(questions_url, timeout=15)
        response.raise_for_status()
        questions_data = response.json()
        if not questions_data:
            return "Fetched questions list is empty.", None
    except Exception as e:
        return f"Error fetching questions: {e}", None

    # Run Agent Loop
    results_log = []
    answers_payload = []
    
    for item in questions_data:
        task_id = item.get("task_id")
        question_text = item.get("question")
        if not task_id or question_text is None:
            continue
        try:
            submitted_answer = agent(question_text)
            answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
            results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
        except Exception as e:
            answers_payload.append({"task_id": task_id, "submitted_answer": "Unknown"})
            results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": "Unknown"})

    if not answers_payload:
        return "Agent did not produce any answers.", pd.DataFrame(results_log)

    # Submit Results
    try:
        submission_data = {"username": username.strip(), "agent_code": agent_code, "answers": answers_payload}
        response = requests.post(submit_url, json=submission_data, timeout=60)
        response.raise_for_status()
        result_data = response.json()
        
        final_status = (
            f"Submission Successful!\n"
            f"User: {result_data.get('username')}\n"
            f"Overall Score: {result_data.get('score', 'N/A')}% "
            f"({result_data.get('correct_count', '?')}/{result_data.get('total_attempted', '?')} correct)\n"
            f"Message: {result_data.get('message', 'No message received.')}"
        )
        return final_status, pd.DataFrame(results_log)
    except Exception as e:
        return f"Submission Failed: {e}", pd.DataFrame(results_log)


# --- Build Gradio Interface ---
with gr.Blocks() as demo:
    gr.Markdown("# Verified Agent Evaluation Runner")
    gr.LoginButton()
    run_button = gr.Button("Run Evaluation & Submit All Answers")
    status_output = gr.Textbox(label="Run Status / Submission Result", lines=5, interactive=False)
    results_table = gr.DataFrame(label="Questions and Agent Answers", wrap=True)
    
    run_button.click(
        fn=run_and_submit_all,
        outputs=[status_output, results_table]
    )

if __name__ == "__main__":
    demo.launch(debug=True, share=False)