import gradio as gr
import pandas as pd
import plotly.graph_objects as go
from datetime import datetime
import os

from preprocessing import get_latest_sequence
from predictor import MineROIPredictor
from fetch_blockchain_data import get_latest_blockchain_data

# internal dummy miner used only for age_days etc. (not shown in UI)
DEFAULT_MINER_NAME = "s19pro"


MODEL_PATH = "best_model_weights.pth"

# Predictor (global)
predictor = None


def init_predictor():
    """Initialize predictor once"""
    global predictor
    if predictor is None:
        predictor = MineROIPredictor(MODEL_PATH)


def init_app():
    """Initialize app (no need for local blockchain_data_complete.csv)."""
    print("\n" + "="*80)
    print("🚀 INITIALIZING MINEROI-NET APP")
    print("="*80)
    print("\nUsing live blockchain.com data (last 90 days).")
    print("Model will use the latest 30 days for ROI prediction.")
    print("="*80 + "\n")
    init_predictor()



def predict_roi(machine_price, machine_hashrate, machine_power, machine_efficiency, electricity_rate,machine_release_date):
    """
    Real-time prediction:
    - Uses latest 90 days from blockchain.com
    - Model uses last 30 days
    - Scaler chosen based on electricity_rate:
        < 0.05  -> ethiopia scaler
        0.05-0.09 -> china scaler
        > 0.09 -> texas scaler
    """
    try:
        window_size = 30

        # -------- parse user inputs --------
        miner_price = float(machine_price)
        miner_hashrate = float(machine_hashrate)
        machine_power = float(machine_power)
        machine_efficiency = float(machine_efficiency)
        user_electricity_rate = float(electricity_rate)

        # ----- parse release date -----
        release_str = None
        if machine_release_date is not None:
            release_str = str(machine_release_date).strip()
            if release_str:
                try:
                    # validate format YYYY-MM-DD
                    datetime.strptime(release_str, "%Y-%m-%d")
                except ValueError:
                    error_msg = """
                    <div style='background: #e74c3c; color: white; padding: 20px; border-radius: 10px;'>
                        <h3 style='margin: 0;'>❌ Invalid release date</h3>
                        <p style='margin: 10px 0 0 0;'>
                            Please enter the machine release date in the format <b>YYYY-MM-DD</b>,
                            for example <code>2020-05-01</code>.
                        </p>
                    </div>
                    """
                    return error_msg, error_msg, None, None
            else:
                # empty box -> fall back to S19 Pro default
                release_str = None


        # ---------------------------------------------------------
        # Bucket electricity rate ONLY to choose scaler
        # The actual feature value will be the user input, repeated
        # for all 30 days in the window.
        # ---------------------------------------------------------
        if user_electricity_rate < 0.05:
            scaler_region = "ethiopia"
            region_bucket = "Low-cost (< $0.05/kWh)"
        elif user_electricity_rate < 1.0:
            scaler_region = "china"
            region_bucket = "Medium-cost ($0.05–$1.00/kWh)"
        else:
            scaler_region = "texas"
            region_bucket = "High-cost (≥ $1.00/kWh)"

        # Region for the pipeline (used by get_latest_sequence & predictor)
        region = scaler_region

        # This is the value that will be used for all 30 days in the window
        # (prepare_miner_features repeats it across time).
        electricity_rate_used = user_electricity_rate

        

        print("User machine specs:")
        print(f"  Price:           {miner_price}")
        print(f"  Hashrate (TH/s): {miner_hashrate}")
        print(f"  Power (W):       {machine_power}")
        print(f"  Efficiency:      {machine_efficiency}")
        print(f"  User elec rate:  {user_electricity_rate} USD/kWh")
        print(f"  Bucket region:   {region}")
        print(f"  Elec used in features (all 30 days):  {electricity_rate_used} USD/kWh")
        print("=" * 80 + "\n")


        # # -------- choose scaler region from electricity_rate --------
        # if electricity_rate < 0.05:
        #     scaler_region = "texas"
        #     region_bucket = "Low-cost (< $0.05/kWh)"
        # elif electricity_rate <= 0.09:
        #     scaler_region = "texas"
        #     region_bucket = "Medium-cost ($0.05–$0.09/kWh)"
        # else:
        #     scaler_region = "texas"
        #     region_bucket = "High-cost (> $0.09/kWh)"

        # print("\n" + "=" * 80)
        # print("PREDICTION REQUEST")
        # print("=" * 80)
        # print(f"Scaler region (from electricity rate): {scaler_region}")
        print("=" * 80 + "\n")

        # -------- fetch latest blockchain data (no date input) --------
        print("📡 Fetching latest blockchain data (last 90 days)...")
        blockchain_df = get_latest_blockchain_data(days=90)

        if blockchain_df is None or len(blockchain_df) < window_size:
            error_msg = f"""
            <div style='background: #e74c3c; color: white; padding: 20px; border-radius: 10px;'>
                <h3 style='margin: 0;'>❌ Error: Insufficient Data</h3>
                <p style='margin: 10px 0 0 0;'>
                    Not enough blockchain data available.
                    Need at least {window_size} days of historical data.
                </p>
            </div>
            """
            return error_msg, error_msg, None, None

        print(f"✅ Got {len(blockchain_df)} days of data")
        print(f"   Date range: {blockchain_df['date'].min().date()} to {blockchain_df['date'].max().date()}")

        price_source = "User input"
        print(f"   Using user-provided price: ${miner_price:,.2f}")

        # -------- build sequence with user machine specs --------
        print("\n🔧 Preparing features...")
        sequence, df_window, pred_date = get_latest_sequence(
            blockchain_df,
            DEFAULT_MINER_NAME,          # internal dummy miner, not shown to user
            miner_price,
            scaler_region,
            window_size,
            machine_hashrate=miner_hashrate,
            power=machine_power,
            efficiency=machine_efficiency,
            electricity_rate=electricity_rate_used,
            release_date=release_str
        )
        print(f"✅ Sequence prepared: {sequence.shape}")

        # -------- model prediction --------
        print("\n🤖 Running prediction...")
        result = predictor.predict(sequence, scaler_region)
        print(f"✅ Prediction: {result['predicted_label']} ({result['confidence']:.1%})")

        # -------- build UI outputs --------
        miner_info = create_miner_info(
            miner_price,
            price_source,
            pred_date,
            miner_hashrate,
            machine_power,
            machine_efficiency,
            electricity_rate_used,
            region_bucket,
        )
        prediction_html = create_prediction_html(result, pred_date, window_size)
        confidence_chart = create_confidence_chart(result["probabilities"])
        price_chart = create_price_chart(blockchain_df, window_size)

        print("=" * 80 + "\n")

        return miner_info, prediction_html, confidence_chart, price_chart

    except Exception as e:
        import traceback

        error_details = traceback.format_exc()
        print("\n❌ ERROR:")
        print(error_details)

        error = f"""
        <div style='background: #e74c3c; color: white; padding: 20px; border-radius: 10px;'>
            <h3 style='margin: 0;'>❌ Prediction Error</h3>
            <p style='margin: 10px 0 0 0;'>{str(e)}</p>
        </div>
        """
        return error, error, None, None



def create_miner_info(
    price,
    source,
    pred_date,
    machine_hashrate,
    machine_power,
    machine_efficiency,
    electricity_rate,
    region_bucket,
):
    """
    Display miner info for a user-specified machine (no ASIC dropdown).
    """
    elec_rate = float(electricity_rate)
    daily_cost = (float(machine_power) * 24.0 / 1000.0) * elec_rate

    # Color coding for price source
    if source == "API":
        badge_color = "#27ae60"  # Green
    elif source == "User input":
        badge_color = "#3498db"  # Blue
    else:
        badge_color = "#e74c3c"  # Red

    return f"""
    <div style="
        background:#111111;
        padding:20px;
        border-radius:10px;
        border:1px solid #333333;
        color:#ffffff;
        font-size:14px;
    ">
        <h3 style="color:#F7931A; margin-top:0; margin-bottom:10px;">
            Custom ASIC Miner
        </h3>

        <div style="display:grid; grid-template-columns:1fr 1fr; gap:15px;">
            <div style="color:#f5f5f5;">
                <p style="font-size: 1.2em; margin: 10px 0; color: #ffffff;"><strong style="color: #ffffff;">Hashrate:{machine_hashrate:.2f} TH/s</strong></p>
                <p style="font-size: 1.2em; margin: 10px 0; color: #ffffff;"><strong style="color: #ffffff;">Power:{machine_power:.1f} W</strong></p>
                <p style="font-size: 1.2em; margin: 10px 0; color: #ffffff;"><strong style="color: #ffffff;">Efficiency:{machine_efficiency:.2f} W/TH</strong></p>
            </div>
            <div style="color:#f5f5f5;">
                <p style="margin:4px 0;">
                    <strong style="color: #ffffff;">Price (as of {pred_date.date()}):${price:,.2f}</strong>
                    <span style="
                        background:{badge_color};
                        color:#ffffff;
                        padding:2px 8px;
                        border-radius:4px;
                        font-size:11px;
                        margin-left:6px;
                    ">
                        {source}
                    </span>
                </p>
                <p style="font-size: 1.2em; margin: 10px 0; color: #ffffff;"><strong style="color: #ffffff;">Electricity rate:{elec_rate:.4f} USD/kWh</strong></p>
                <p style="font-size: 1.2em; margin: 10px 0; color: #ffffff;"><strong style="color: #ffffff;">Cost bucket:{region_bucket}</strong></p>
                <p style="font-size: 1.2em; margin: 10px 0; color: #ffffff;"><strong style="color: #ffffff;">Estimated daily elec cost:${daily_cost:,.2f}</strong></p>
            </div>
        </div>
    </div>
    """




def create_prediction_html(result, date, window):
    label = result['predicted_label']
    conf = result['confidence']
    
    if 'Unprofitable' in label:
        color, emoji, rec = '#e74c3c', '🔴', 'NOT RECOMMENDED'
    elif 'Marginal' in label:
        color, emoji, rec = '#f39c12', '🟡', 'PROCEED WITH CAUTION'
    else:
        color, emoji, rec = '#27ae60', '🟢', 'GOOD OPPORTUNITY'
    
    return f"""
    <div style="background: #1e1e1e; padding: 30px; border-radius: 10px; border: 2px solid {color}; text-align: center; color: #ffffff;">
        <h2 style="color: {color}; margin: 0 0 10px 0;">{emoji} {label}</h2>
        <p style="font-size: 1.2em; margin: 10px 0; color: #ffffff;"><strong style="color: #ffffff;">Confidence: {conf:.1%}</strong></p>
        <p style="font-size: 1.5em; color: {color}; margin: 20px 0;"><strong style="color: {color};">{rec}</strong></p>
        <p style="color: #cccccc; margin: 10px 0 0 0; font-size: 0.9em;">
            Prediction based on data up to: {date.strftime('%Y-%m-%d')}<br>Window: {window} days
        </p>
    </div>
    """


def create_confidence_chart(probs):
    categories = ['Unprofitable', 'Marginal', 'Profitable']
    values = [probs['unprofitable'], probs['marginal'], probs['profitable']]
    colors = ['#e74c3c', '#f39c12', '#27ae60']
    
    fig = go.Figure()
    fig.add_trace(go.Bar(x=categories, y=values, marker_color=colors, text=[f'{v:.1%}' for v in values], textposition='auto'))
    fig.update_layout(title='Prediction Confidence', yaxis_title='Probability', yaxis=dict(range=[0, 1], tickformat='.0%'),
                       template='plotly_dark', height=300, margin=dict(l=0, r=0, t=40, b=0))
    return fig


def create_price_chart(df, window):
    # Show more context if available
    display_days = min(len(df), window * 2)
    df_display = df.tail(display_days)
    
    fig = go.Figure()
    fig.add_trace(go.Scatter(x=df_display['date'], y=df_display['bitcoin_price'], mode='lines', name='Bitcoin Price', line=dict(color='#F7931A', width=2)))
    fig.update_layout(title=f'Bitcoin Price ({len(df_display)} Days)', xaxis_title='Date', yaxis_title='Price (USD)',
                       template='plotly_dark', height=300, margin=dict(l=0, r=0, t=40, b=0))
    return fig


def create_interface():
    with gr.Blocks(title="MineROI-Net") as app:
        gr.Markdown("# 🪙 MineROI-Net: Bitcoin Mining Hardware ROI Predictor")
        gr.Markdown(
            "Uses the **latest 30 days** of Bitcoin network data from blockchain.com "
            "to classify your miner as Unprofitable / Marginal / Profitable."
        )

        with gr.Row():
            # ---- Left: inputs ----
            with gr.Column(scale=1):
                gr.Markdown("### Configuration")

                machine_price = gr.Number(
                    label="Machine price (USD)",
                    value=2500.0,
                    precision=2,
                )
                machine_hashrate = gr.Number(
                    label="Machine hashrate (TH/s)",
                    value=100.0,
                    precision=2,
                )
                machine_power = gr.Number(
                    label="Power (W)",
                    value=3000.0,
                    precision=1,
                )
                machine_efficiency = gr.Number(
                    label="Efficiency (W/TH)",
                    value=30.0,
                    precision=2,
                )
                machine_release_date = gr.Textbox(
                    label="Release date (YYYY-MM-DD)",
                    value="2020-05-01",
                    placeholder="e.g. 2020-05-01",
                    lines=1,
                    scale=1,
                    container=True,
                    show_label=True,
                )
                electricity_rate = gr.Number(
                    label="Electricity rate (USD/kWh)",
                    value=0.07,   # neutral default
                    precision=4,
                )

                btn = gr.Button("🔮 Predict ROI", variant="primary", size="lg")

                gr.Markdown(
                    """
                    ### About

                    - 🔴 **Unprofitable** (ROI ≤ 0)  
                    - 🟡 **Marginal** (0 < ROI < 1)  
                    - 🟢 **Profitable** (ROI ≥ 1)  

                    **Model:** trained on 30-day windows of Bitcoin network and miner features.  
                    **Live mode:** whenever you click *Predict*, the app pulls the latest blockchain data.
                    """
                )

            # ---- Right: outputs ----
            with gr.Column(scale=2):
                gr.Markdown("### Results")
                miner_info = gr.HTML()
                prediction = gr.HTML()
                with gr.Row():
                    conf_plot = gr.Plot()
                    price_plot = gr.Plot()

        # Connect button to prediction function
        btn.click(
            fn=predict_roi,
            inputs=[
                machine_price,
                machine_hashrate,
                machine_power,
                machine_efficiency,
                electricity_rate,
                machine_release_date
            ],
            outputs=[miner_info, prediction, conf_plot, price_plot],
        )

    return app



if __name__ == "__main__":
    # Initialize app (loads complete data into memory)
    init_app()
    
    # Launch
    app = create_interface()
    app.launch()
    # app.launch(server_name="0.0.0.0", server_port=7860, share=True)