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

from smolagents import ToolCallingAgent, OpenAIServerModel
from smolagents import DuckDuckGoSearchTool, PythonInterpreterTool
from agentsTools.toolVisitWebpage import visit_webpage
from agentsTools.tool_fetch_task_file import fetch_task_file
from agentsTools.tool_read_excel_as_json import read_excel_as_json

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

class GeminiReActAgent:
    def __init__(self):
        self.model = OpenAIServerModel(
            model_id="gemini-2.5-pro-preview-03-25",
            api_base="https://generativelanguage.googleapis.com/v1beta/",
            api_key=os.getenv("GEMINI_API_KEY_1")
        )
        self.agent = ToolCallingAgent(
            tools=[
                DuckDuckGoSearchTool(),
                PythonInterpreterTool(),
                visit_webpage,
                fetch_task_file,
                read_excel_as_json,
            ],
            model=self.model,
            max_steps=8
        )

    def __call__(self, question: str, taskid: str) -> str:
        prompt = """
        You are an intelligent assistant answering questions step-by-step.
        Use tools only when necessary. You can take up to 8 steps.
        You have access to tools such as a web search, Python execution, and file reading.
        Use the task ID {tid} if you need to fetch additional data.

        Think aloud, justify each step. Conclude your process with:
        FINAL ANSWER = [your answer here]  

        Question: {q}
        """.format(tid=taskid, q=question)

        result = self.agent.run(prompt)
        return result

def submit_eval_run(profile: gr.OAuthProfile | None):
    space_id = os.getenv("SPACE_ID")
    if profile:
        username = profile.username
    else:
        return "Please login to Hugging Face first.", None

    agent = GeminiReActAgent()
    agent_code = f"https://huggingface.co/spaces/{space_id}/tree/main"

    try:
        response = requests.get(f"{DEFAULT_API_URL}/questions", timeout=15)
        response.raise_for_status()
        questions_data = response.json()
    except Exception as e:
        return f"Failed to fetch questions: {e}", None

    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, task_id)
            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 "Agent did not return any answers.", pd.DataFrame(results_log)

    submission_data = {
        "username": username,
        "agent_code": agent_code,
        "answers": answers_payload
    }

    try:
        #esponse = requests.post(f"{DEFAULT_API_URL}/submit", json=submission_data, timeout=60)
        #response.raise_for_status()
        #result = response.json()
        #summary = (
        #    f"Submission Successful!\n"
        #    f"User: {result.get('username')}\n"
        #    f"Score: {result.get('score')}% ({result.get('correct_count')}/{result.get('total_attempted')})\n"
        #    f"Message: {result.get('message', 'No message received.')}"
        #)
        #return summary, pd.DataFrame(results_log)
        # TEST MODE: Skip submission
        summary = f"🧪 TEST MODE: Completed {len(answers_payload)} questions. No results submitted."
        return summary, pd.DataFrame(results_log)
    except Exception as e:
        return f"Submission failed: {e}", pd.DataFrame(results_log)

with gr.Blocks() as demo:
    gr.Markdown("# Gemini Agent Submission Tool")
    gr.LoginButton()
    btn = gr.Button("Submit Answers")
    status = gr.Textbox(label="Submission Status")
    table = gr.DataFrame(label="Answers Log")
    btn.click(fn=submit_eval_run, outputs=[status, table])

demo.launch(debug=True, share=False)