import gradio as gr
import pandas as pd
import joblib

# โหลดโมเดลและ label encoder
model = joblib.load("biochar_lightgbm_model.pkl")
le = joblib.load("biochar_label_encoder.pkl")

# รายชื่อฟีเจอร์
feature_names = ['SurfaceArea', 'Fixed Carbon', 'pH', 'Pore Volume', 'Pore Size', 'Energy Content']

# ฟังก์ชันทำนาย
def recommend_with_probs(surface_area, fixed_carbon, pH, pore_volume, pore_size, energy_content):
    input_data = pd.DataFrame([[surface_area, fixed_carbon, pH, pore_volume, pore_size, energy_content]],
                              columns=feature_names)
    pred_class = model.predict(input_data)[0]
    pred_label = le.inverse_transform([pred_class])[0]

    # ทำนายความน่าจะเป็น
    probs = model.predict_proba(input_data)[0]
    prob_labels = le.inverse_transform(range(len(probs)))
    prob_df = pd.DataFrame({
        'Application': prob_labels,
        'Probability (%)': (probs * 100).round(2)
    }).sort_values(by='Probability (%)', ascending=False).reset_index(drop=True)

    return pred_label, prob_df

# Custom CSS
css = """
.gradio-container {background-color: #f4f9f4; font-family: 'Arial', sans-serif;}
h1, h2, h3, h4 {text-align: center;}
.output-class {font-size: 1.5em; color: #1b5e20; font-weight: bold;}
.gr-button {background-color: #66bb6a !important; color: white !important; border: none;}
"""

with gr.Blocks(css=css, title="Biochar Application Recommender") as demo:
    gr.Markdown("# 🌱 Biochar Application Recommender")
    gr.Markdown("**ระบบแนะนำการประยุกต์ใช้งานไบโอชาร์โดยอิงจากคุณสมบัติทางเคมีและกายภาพ**")

    with gr.Row():
        with gr.Column():
            surface_area = gr.Number(label="🧪 Surface Area (m²/g)", interactive=True)
            fixed_carbon = gr.Number(label="🌑 Fixed Carbon (%)", interactive=True)
            pH = gr.Number(label="⚗️ pH", interactive=True)

        with gr.Column():
            pore_volume = gr.Number(label="🔬 Pore Volume (cm³/g)", interactive=True)
            pore_size = gr.Number(label="📏 Pore Size (nm)", interactive=True)
            energy_content = gr.Number(label="🔥 Energy Content (kJ/kg)", interactive=True)

    submit_btn = gr.Button("🚀 Submit")

    with gr.Row():
        with gr.Column(scale=1):
            gr.Markdown("")
        with gr.Column(scale=6):
            output_text = gr.Textbox(label="✅ Recommended Application", elem_classes="output-class", interactive=False)
            output_table = gr.Dataframe(label="📊 Prediction Probabilities", interactive=False)
        with gr.Column(scale=1):
            gr.Markdown("")

    submit_btn.click(
        fn=recommend_with_probs,
        inputs=[surface_area, fixed_carbon, pH, pore_volume, pore_size, energy_content],
        outputs=[output_text, output_table]
    )

demo.launch()