import os
import gradio as gr
import requests
import pandas as pd
from huggingface_hub import InferenceClient
from duckduckgo_search import DDGS
from datasets import load_dataset

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

# Hugging Face Token (set in environment)
HF_TOKEN = os.environ.get("HF_TOKEN")
deepseek_model = "deepseek-ai/DeepSeek-R1"
hf_client = InferenceClient(model=deepseek_model, token=HF_TOKEN)

# Load Wikipedia dataset (small subset for efficient retrieval)
wiki_dataset = load_dataset("wikipedia", "20220301.en", split="train[:10000]")

def search_wikipedia(question):
    results = wiki_dataset.filter(lambda x: question.lower() in x["text"].lower())
    if len(results):
        return results[0]["text"][:1000]  # limit to first 1000 chars
    return "No relevant information found on Wikipedia."

def duckduckgo_search(query):
    with DDGS() as ddgs:
        results = [r["body"] for r in ddgs.text(query, max_results=3)]
        return "\n".join(results) if results else "No results found."

def ask_deepseek(prompt, max_tokens=512):
    try:
        response = hf_client.text_generation(
            prompt, max_new_tokens=max_tokens, temperature=0.2, repetition_penalty=1.1
        )
        return response
    except Exception as e:
        return f"DeepSeek Error: {e}"

class SmartAgent:
    def __call__(self, question: str) -> str:
        q_lower = question.lower()
        if any(term in q_lower for term in ["current", "latest", "2024", "2025", "recent", "live", "today", "now"]):
            return duckduckgo_search(question)
        deepseek_response = ask_deepseek(question)
        if "DeepSeek Error" not in deepseek_response and deepseek_response.strip():
            return deepseek_response
        # fallback to Wikipedia if DeepSeek fails
        return search_wikipedia(question)

def run_and_submit_all(profile: gr.OAuthProfile | None):
    if not profile:
        return "Please Login to Hugging Face with the button.", None
    username = profile.username
    questions_url = f"{DEFAULT_API_URL}/questions"
    submit_url = f"{DEFAULT_API_URL}/submit"
    agent_code = f"https://huggingface.co/spaces/{os.getenv('SPACE_ID')}/tree/main"
    
    try:
        agent = SmartAgent()
    except Exception as e:
        return f"Agent Error: {e}", None

    questions_data = requests.get(questions_url).json()
    results_log, answers_payload = [], []

    for item in questions_data:
        task_id = item.get("task_id")
        question_text = item.get("question")
        if task_id and question_text:
            answer = agent(question_text)
            answers_payload.append({"task_id": task_id, "submitted_answer": answer})
            results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": answer})

    submission_data = {"username": username, "agent_code": agent_code, "answers": answers_payload}
    response = requests.post(submit_url, json=submission_data).json()

    final_status = (
        f"Submission Successful!\n"
        f"User: {response.get('username')}\n"
        f"Overall Score: {response.get('score', 'N/A')}%\n"
        f"({response.get('correct_count', '?')}/{response.get('total_attempted', '?')} correct)\n"
        f"Message: {response.get('message', 'No message received.')}"
    )

    return final_status, pd.DataFrame(results_log)

with gr.Blocks() as demo:
    gr.Markdown("# Smart Agent Evaluation Runner")
    gr.LoginButton()
    run_button = gr.Button("Run Evaluation & Submit All Answers")
    status_output = gr.Textbox(label="Run Status", lines=5, interactive=False)
    results_table = gr.DataFrame(label="Questions and Answers")

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

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