import gradio as gr
import pickle
import numpy as np
import requests

# Load new model
with open('fire_rf_model_v2.pkl', 'rb') as f:
    model = pickle.load(f)

def get_real_weather(lat, lon):
    """Fetch real ERA5 weather from Open-Meteo"""
    url = (
        f"https://archive-api.open-meteo.com/v1/archive"
        f"?latitude={lat}&longitude={lon}"
        f"&start_date=2023-03-08&end_date=2023-03-08"
        f"&daily=temperature_2m_max,precipitation_sum,"
        f"windspeed_10m_max,dewpoint_2m_mean"
        f"&timezone=Asia/Kolkata"
    )
    try:
        r = requests.get(url, timeout=10)
        d = r.json().get('daily', {})
        temp = d.get('temperature_2m_max', [None])[0]
        precip = d.get('precipitation_sum', [None])[0]
        wind = d.get('windspeed_10m_max', [None])[0]
        dew = d.get('dewpoint_2m_mean', [None])[0]
        humidity = round(100 * np.exp(17.625*dew/(243.04+dew)) /
                        np.exp(17.625*temp/(243.04+temp)), 1) if dew and temp else 60.0
        return temp or 25.0, precip or 5.0, wind or 12.0, humidity
    except:
        return 25.0, 5.0, 12.0, 60.0

def predict_fire_risk(latitude, longitude, ndvi, aod, elevation,
                      drought_index, land_cover):

    land_cover_map = {'Forest': 0, 'Grassland': 1,
                      'Agricultural': 2, 'Shrubland': 3}

    # Auto-fetch real ERA5 weather
    temperature, precipitation, wind_speed, humidity = get_real_weather(
        round(latitude, 2), round(longitude, 2)
    )

    features = np.array([[
        temperature, precipitation, wind_speed, humidity,
        ndvi, aod, elevation, drought_index,
        land_cover_map[land_cover]
    ]])

    prob = model.predict_proba(features)[0][1]
    risk_pct = prob * 100

    if prob >= 0.80:
        level = "🔴 CRITICAL FIRE RISK"
        color = "background-color: #ff4444; color: white; padding: 20px; border-radius: 10px; font-size: 20px; font-weight: bold;"
    elif prob >= 0.60:
        level = "🟠 HIGH FIRE RISK"
        color = "background-color: #ff8800; color: white; padding: 20px; border-radius: 10px; font-size: 20px; font-weight: bold;"
    elif prob >= 0.40:
        level = "🟡 MODERATE FIRE RISK"
        color = "background-color: #ffcc00; color: black; padding: 20px; border-radius: 10px; font-size: 20px; font-weight: bold;"
    else:
        level = "🟢 LOW FIRE RISK"
        color = "background-color: #44bb44; color: white; padding: 20px; border-radius: 10px; font-size: 20px; font-weight: bold;"

    result = f"""
**{level}**
**Fire Probability: {risk_pct:.1f}%**

---
### 🌤️ ERA5 Weather (Auto-fetched via Open-Meteo)
| Parameter | Value |
|-----------|-------|
| 🌡️ Temperature | {temperature:.1f}°C |
| 🌧️ Precipitation | {precipitation:.1f} mm |
| 💨 Wind Speed | {wind_speed:.1f} km/h |
| 💧 Humidity | {humidity:.1f}% |

---
### 🌿 Vegetation & Terrain Inputs
| Parameter | Value |
|-----------|-------|
| NDVI | {ndvi:.2f} |
| AOD | {aod:.2f} |
| Elevation | {elevation:.0f} m |
| Drought Index | {drought_index:.2f} |
| Land Cover | {land_cover} |

---
*Model: Random Forest | AUC: 0.9583 | Weather: ERA5 Reanalysis via Open-Meteo*
*Data fusion: NASA FIRMS MODIS + ERA5 atmospheric reanalysis*
"""
    return result

# Build UI
with gr.Blocks(title="AirSense Fire Risk — ERA5 Fusion", theme=gr.themes.Soft()) as app:

    gr.Markdown("""
    # 🔥 AirSense — Forest Fire Risk Prediction
    ### Northeast India | ERA5 Weather Data Fusion
    **Enter coordinates and vegetation parameters — real atmospheric data is fetched automatically from ERA5 reanalysis.**
    """)

    with gr.Row():
        with gr.Column(scale=1):
            gr.Markdown("### 📍 Location")
            latitude = gr.Slider(21.5, 29.5, value=26.5, step=0.1, label="Latitude")
            longitude = gr.Slider(88.5, 95.5, value=93.5, step=0.1, label="Longitude")

            gr.Markdown("### 🌿 Vegetation & Terrain")
            ndvi = gr.Slider(0.0, 1.0, value=0.4, step=0.01, label="NDVI (Vegetation Index)")
            aod = gr.Slider(0.0, 1.5, value=0.4, step=0.01, label="AOD (Aerosol Optical Depth)")
            elevation = gr.Slider(0, 3000, value=400, step=10, label="Elevation (m)")
            drought_index = gr.Slider(-2.0, 2.0, value=0.5, step=0.1, label="Drought Index")
            land_cover = gr.Dropdown(
                ['Forest', 'Grassland', 'Agricultural', 'Shrubland'],
                value='Forest', label="Land Cover Type"
            )

            predict_btn = gr.Button("🔍 Predict Fire Risk", variant="primary", size="lg")

        with gr.Column(scale=1):
            gr.Markdown("### 📊 Prediction Results")
            output = gr.Markdown(value="*Adjust parameters and click Predict*")

    predict_btn.click(
        fn=predict_fire_risk,
        inputs=[latitude, longitude, ndvi, aod, elevation, drought_index, land_cover],
        outputs=output
    )

    gr.Markdown("""
    ---
    **Data Sources:** NASA FIRMS MODIS (fire detections) • ERA5 Reanalysis via Open-Meteo (weather)
    **Model:** Random Forest Classifier | Trained on 60 presence-absence points | AUC: 0.9583
    **Region:** Northeast India (21.5°N–29.5°N, 88.5°E–95.5°E)
    """)

app.launch()