import gradio as gr
import os
import json
from huggingface_hub import upload_file
import pandas as pd
from datasets import load_dataset

HF_TOKEN = os.getenv("HF_TOKEN")
SUBMISSIONS_REPO = "NAMAA-Space/ocr-competition-submissions"
RESULTS_REPO = "NAMAA-Space/ocr-competition-results"

def validate_fields(team_name, email, model_name, hf_model_id, hf_token, code):
    if not team_name or not email or not model_name or not hf_model_id or not hf_token or not code:
        return "All fields are required. Please fill in all fields."
    return submit(team_name, email, model_name, hf_model_id, hf_token, code)

def submit(team_name, email, model_name, hf_model_id, hf_token, code):
    # entry = {
    #     "team_name": team_name,
    #     "email": email,
    #     "model_name": model_name,
    #     "hf_model_id": hf_model_id,
    #     "hf_token": hf_token,
    #     "code": code
    # }
    # filename = f"{team_name}_{model_name}_{pd.Timestamp.now().strftime('%Y%m%d%H%M%S')}.json"
    # filename = filename.replace("/", "-")
    # with open(filename, "w") as f:
    #     json.dump(entry, f)
    # upload_file(path_or_fileobj=filename,
    #             path_in_repo=filename,
    #             repo_id=SUBMISSIONS_REPO,
    #             repo_type="dataset",
    #             token=HF_TOKEN)
    return "Submission is closed"

def show_results():
    try:
        ds = load_dataset(RESULTS_REPO, split="train")
        df = ds.to_pandas()[["team_name", "model_name", "WER", "CER", "BLEU"]]
            
        # Calculate composite score
        # Formula: (100 - WER) × 0.35 + (100 - CER) × 0.35 + BLEU × 0.30
        df['Score'] = (100 - df['WER']) * 0.35 + (100 - df['CER']) * 0.35 + df['BLEU'] * 0.30
            
        # Round score to 2 decimal places
        df['Score'] = df['Score'].round(2)
        
        # Reorder columns to show Score first
        df = df[['Score', 'team_name', 'model_name', 'WER', 'CER', 'BLEU']]
        
        # Sort by Score (descending - highest is best)
        df = df.sort_values(by='Score', ascending=False).reset_index(drop=True)
        
        # Add rank column
        df.insert(0, 'Rank', range(1, len(df) + 1))
        
        return df
        
    except Exception as e:
        return f"An error occurred while loading the results: {e}"

with gr.Blocks() as demo:
    # Welcome message
    gr.Markdown("""
    <h2 style="font-size:28px;">👋 Welcome to the VLM OCR Competition!</h2>
    <p style="font-size:18px;">
    This competition aims to improve **open-source Arabic OCR models**.  
    It's part of the NAMAA Community mission to strengthen the Arabic presence in the ML space.  
    This competition is designed to **push the boundaries** of OCR performance on diverse Arabic documents.
    </p>
    """)

    with gr.Tabs():
        with gr.Tab("📜 Rules"):
            # Text instructions
            gr.Markdown("""
        <h3 style="font-size:22px;">QARI OCR Competition Rules</h3>
        <p style="font-size:18px;">
        Welcome to the <b>QARI OCR Competition</b> organized by the <b>NAMAA Community</b> and sponsored by <b>KANDCA</b>!  
        The competition runs from <b>September 15 to October 15</b>.  
        Join the <a href="https://discord.gg/GDTpeHZt" target="_blank">Discord server</a> for support and discussion.  
        Full rules and submission portal: <a href="https://huggingface.co/spaces/NAMAA-Space/QARI-Competition" target="_blank">Hugging Face Space</a>.
        </p>
        
        <h4 style="font-size:20px;">📜 Submission Rules</h4>
        <ul style="font-size:18px;">
            <li>Each team can submit <b>one model evaluation per week</b>.</li>
            <li>Provide with your submission:
                <ol>
                    <li>Team name (must stay consistent across submissions)</li>
                    <li>Model name & Hugging Face Model ID</li>
                    <li>A valid Hugging Face token with access</li>
                    <li><b>The inference code</b> and any <b>dependency installation instructions</b></li>
                    <li><b>The OCR output must be a single-page structured HTML</b> using the following tags:
                        <ul>
                            <li>&lt;header&gt;, &lt;footer&gt;, &lt;main&gt;, &lt;section id="1"&gt;, &lt;section id="2"&gt;</li>
                            <li>&lt;p&gt;, &lt;h1&gt;-&lt;h5&gt;, &lt;b&gt;, &lt;i&gt;, &lt;u&gt;</li>
                            <li>&lt;img&gt;, &lt;table&gt;, &lt;hr&gt;, &lt;ul&gt;, &lt;ol&gt;</li>
                        </ul>
                    </li>
                    <li><b>Submitting only unstructured output will result in a 5-point deduction</b> from your final score.</li>
                </ol>
            </li>
            <li><b>The submitted code is the responsibility of the submitting team.</b></li>
            <li>Ideally, provide a working <b>Google Colab link</b> with all details and dependencies.</li>
        </ul>
        
        <h4 style="font-size:20px;">📆 Evaluation Schedule</h4>
        <ul style="font-size:18px;">
            <li>Submissions received by <strong>Sunday at midnight</strong> will be evaluated on <strong>Monday</strong>.</li>
            <li>The leaderboard will be updated by <strong>Wednesday or Thursday</strong> of the same week.</li>
        </ul>
        """)
        

        with gr.Tab("🎁 Prizes"):
            gr.Markdown("""
            <h3 style="font-size:22px;">Prize Distribution ((bank transfer or API credits))</h3>
            <ul style="font-size:18px;">
                <li>1st Place: 🥇 250 USD</li>
                <li>2nd Place: 🥈 125 USD</li>
                <li>3rd Place: 🥉 75 USD</li>
                <li>4th Place: 🎖️ 50 USD</li>
                <li>5th Place: 🎖️ 25 USD</li>
            </ul>
            """)

        with gr.Tab("📊 Evaluation"):
            gr.Markdown("""
            <h3 style="font-size:22px;">Evaluation Details</h3>
            <ul style="font-size:18px;">
                <li>The evaluation dataset will remain <b>private</b> and is not shared with participants.</li>
                <li>It will include:
                    <ul>
                        <li>Historical documents</li>
                        <li>Scanned pages</li>
                        <li>Different layouts</li>
                        <li>Handwritten pages</li>
                    </ul>
                </li>
                <li>Models will be evaluated on <b>accuracy metrics</b> such as:
                    <ul>
                        <li>Word Error Rate (WER)</li>
                        <li>Character Error Rate (CER)</li>
                        <li>BLEU score</li>
                    </ul>
                </li>
                <li><b>Evaluation schedule:</b>
                    <ul>
                        <li>Submissions received by <b>Sunday at midnight</b> will be evaluated on <b>Monday</b>.</li>
                        <li>The leaderboard will be updated by <b>Wednesday or Thursday</b> of the same week.</li>
                    </ul>
                </li>
            </ul>
            """)

        with gr.Tab("🚀 Submit & Leaderboard"):
            gr.Markdown("<h3 style='font-size:22px;'>Submit Your Model</h3>")
            with gr.Row():
                team = gr.Textbox(label="Team Name", placeholder="Enter your team name")
                email = gr.Textbox(label="Email", placeholder="Enter your email")
            model = gr.Textbox(label="Model Name", placeholder="Enter your model name")
            hf_model = gr.Textbox(label="Hugging Face Model ID", placeholder="Enter your HF Model ID")
            hf_token = gr.Textbox(label="Hugging Face Access Token", type="password", placeholder="Enter your HF token")
            code = gr.Textbox(label="Code (instructions to run your model) or colab link", lines=6, placeholder="Paste your run code here...")

            submit_btn = gr.Button("Submit")
            status = gr.Textbox(label="Status")

            submit_btn.click(fn=validate_fields,
                             inputs=[team, email, model, hf_model, hf_token, code],
                             outputs=status)

            gr.Markdown("<h3 style='font-size:22px;'>Leaderboard Results</h3>")
            results = gr.Dataframe(headers=["model_name", "WER", "CER", "BLEU", "team_name"])
            demo.load(fn=show_results, outputs=results)

    demo.launch()