import os
import gradio as gr
import requests
import pandas as pd
from huggingface_hub import login
from dotenv import load_dotenv

# Cargar variables de entorno
load_dotenv()
API_BASE_URL = "https://my-custom-api.hf.space"
MODEL_NAME = os.getenv("MODEL_NAME", "meta-llama/Meta-Llama-3-8B-Instruct")  # o usa 'google/gemma-2b-it'

def call_model(prompt):
    headers = {
        "Authorization": f"Bearer {os.environ['HF_TOKEN']}",
        "Content-Type": "application/json"
    }
    payload = {
        "inputs": prompt,
        "parameters": {"max_new_tokens": 512}
    }
    response = requests.post(
        f"https://api-inference.huggingface.co/models/{MODEL_NAME}",
        headers=headers,
        json=payload,
        timeout=60
    )
    result = response.json()
    if isinstance(result, dict) and result.get("error"):
        return f"ERROR: {result['error']}"
    return result[0]["generated_text"] if isinstance(result, list) else result

def execute_agent_operations(profile: gr.OAuthProfile | None):
    space_id = os.getenv("MY_SPACE_ID")

    if profile:
        username = f"{profile.username}"
        print(f"Usuario conectado: {username}")
    else:
        print("No has iniciado sesión.")
        return "Inicia sesión en Hugging Face.", None

    questions_url = f"{API_BASE_URL}/questions"
    attachments_url = f"{API_BASE_URL}/files/"
    submit_url = f"{API_BASE_URL}/submit"

    try:
        response = requests.get(questions_url, timeout=15)
        response.raise_for_status()
        questions = response.json()
        if not questions:
            return "No se encontraron preguntas.", None
        
        for q in questions:
            task_id = q.get("task_id")
            file_name = q.get("file_name", "")
            if task_id and file_name:
                try:
                    att_response = requests.get(f"{attachments_url}{task_id}", timeout=15)
                    att_response.raise_for_status()
                    q["attachment_b64"] = att_response.text
                except Exception as e:
                    print(f"Error al obtener adjunto: {e}")
                    q["attachment_b64"] = None
    except Exception as e:
        return f"Error al obtener preguntas: {e}", None

    results_log = []
    answers_payload = []

    for item in questions:
        task_id = item.get("task_id")
        question_text = item.get("question", "")
        attachment = item.get("attachment_b64", "")
        full_prompt = f"{question_text}\n\nArchivo adjunto:\n{attachment}" if attachment else question_text

        if not task_id or not full_prompt:
            continue

        try:
            submitted_answer = call_model(full_prompt)
            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:
            results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": f"ERROR: {e}"})

    if not answers_payload:
        return "El agente no produjo respuestas.", pd.DataFrame(results_log)

    submission_data = {
        "username": username.strip(),
        "agent_code": f"https://huggingface.co/spaces/{space_id}",
        "answers": answers_payload
    }

    try:
        post_response = requests.post(submit_url, json=submission_data, timeout=60)
        post_response.raise_for_status()
        result_data = post_response.json()
        score = result_data.get("score", "N/A")
        return f"¡Envío exitoso!\nPuntuación: {score}", pd.DataFrame(results_log)
    except Exception as e:
        return f"Error al enviar: {e}", pd.DataFrame(results_log)

# Interfaz Gradio
with gr.Blocks() as demo:
    gr.Markdown("# Evaluador de Agente (versión personalizada)")
    gr.LoginButton()
    run_button = gr.Button("Ejecutar Evaluación y Enviar Respuestas")
    status_output = gr.Textbox(label="Estado", lines=3)
    results_table = gr.DataFrame(label="Respuestas del agente")

    run_button.click(
        fn=execute_agent_operations,
        outputs=[status_output, results_table]
    )

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