import gradio as gr 
import pandas as pd 
import numpy as np 
from xgboost import XGBRegressor 
import joblib 
import os 
import logging 

# -----------------------------
# 1️⃣ File paths
# -----------------------------
MODEL_PATH = "xgb_model_corn_final.json"
SCALER_PATH = "minmax_scaler.pkl"
EXCEL_PATH = "microalgae_pot_experiment_corrected_doses.xlsx"

# Check files exist
for path in [MODEL_PATH, SCALER_PATH, EXCEL_PATH]:
    if not os.path.exists(path):
        raise FileNotFoundError(f"File not found: {path}")

# -----------------------------
# 2️⃣ Load model, scaler, and data
# -----------------------------
xgb_model = XGBRegressor()
xgb_model.load_model(MODEL_PATH)
scaler = joblib.load(SCALER_PATH)
df = pd.read_excel(EXCEL_PATH)

# Only one strain exists
strain_names = df['Microalgae_Strain'].unique().tolist()
single_strain_value = 0  # training used LabelEncoder -> 0

# -----------------------------
# 3️⃣ Prediction function
# -----------------------------
def predict_dose(crop, microalgae_strain, soil_n, soil_p, soil_k, soil_ec, soil_moisture,
                 chlorophyll, shoot_length, root_length, yield_g, relative_yield,
                 actual_dose):
    logs = []  # Capture all debug info

    # 1️⃣ Validate inputs
    required = [soil_n, soil_p, soil_k, soil_ec, soil_moisture,
                chlorophyll, shoot_length, root_length, yield_g, relative_yield]
    if any(v is None for v in required):
        logs.append("[DEBUG] Missing numeric inputs!")
        return "⚠️ Please fill all numeric input fields.", None, None, "\n".join(logs)
    logs.append("[DEBUG] All inputs received.")

    # 2️⃣ Encode strain
    strain_encoded = single_strain_value
    logs.append(f"[DEBUG] Using Microalgae_Strain encoded value: {strain_encoded}")

    # 3️⃣ Create DataFrame
    feature_cols = ['Soil_N_ppm','Soil_P_ppm','Soil_K_ppm','Soil_EC_dS_m',
                    'Soil_Moisture_%','Chlorophyll_SPAD','Shoot_Length_cm',
                    'Root_Length_cm','Yield_g_per_pot','Relative_Yield_%','Microalgae_Strain']
    X_input_df = pd.DataFrame([[soil_n, soil_p, soil_k, soil_ec, soil_moisture,
                                chlorophyll, shoot_length, root_length, yield_g,
                                relative_yield, strain_encoded]],
                              columns=feature_cols).astype(float)
    logs.append(f"[DEBUG] Input DataFrame:\n{X_input_df}")

    # 4️⃣ Scale
    X_input_scaled = scaler.transform(X_input_df)
    logs.append(f"[DEBUG] Scaled Input:\n{X_input_scaled}")

    # 5️⃣ Predict
    predicted_dose = xgb_model.predict(X_input_scaled)[0]
    logs.append(f"[DEBUG] Predicted dose (raw): {predicted_dose}")

    # 6️⃣ Compute error
    if actual_dose is not None:
        abs_error = abs(predicted_dose - actual_dose)
        logs.append(f"[DEBUG] Actual dose: {actual_dose}")
        logs.append(f"[DEBUG] Absolute Error: {abs_error}")
        return (f"🌱 **Predicted Dose:** {predicted_dose:.2f} g/pot",
                f"📏 **Actual Dose:** {actual_dose:.2f} g/pot",
                f"❌ **Absolute Error:** {abs_error:.2f} g/pot",
                "\n".join(logs))
    else:
        logs.append("[DEBUG] Actual dose not provided.")
        return (f"🌱 **Predicted Dose:** {predicted_dose:.2f} g/pot",
                "📏 Actual Dose: Not provided",
                "❌ Absolute Error: N/A",
                "\n".join(logs))

# -----------------------------
# 4️⃣ Gradio Interface
# -----------------------------
with gr.Blocks(
    theme=gr.themes.Soft(primary_hue="green", secondary_hue="lime", neutral_hue="gray"),
    title="EcoGrowAI — Microalgae Dose Inference"
) as demo:
    gr.Markdown(
        """
        <div style='text-align:center'>
        <h1 style='color:#00b16a; margin-bottom:8px;'>🌿 EcoGrowAI — Inference App</h1>
        <p style='color:#555;'>Predict the optimal <b>microalgae dose (g/pot)</b> for Corn based on soil and growth parameters.</p>
        </div>
        <hr style='margin: 10px 0;'>
        """
    )
    with gr.Row():
        with gr.Column(scale=1):
            gr.Markdown("### 🌾 Crop & Microalgae Selection")
            crop = gr.Dropdown(["Corn"], label="Select Crop", value="Corn")
            strain = gr.Dropdown(strain_names, label="Microalgae Strain", value=strain_names[0])
            gr.Markdown("### 🌱 Soil Parameters")
            soil_n = gr.Number(label="Soil N (ppm)")
            soil_p = gr.Number(label="Soil P (ppm)")
            soil_k = gr.Number(label="Soil K (ppm)")
            soil_ec = gr.Number(label="Soil EC (dS/m)")
            soil_moisture = gr.Number(label="Soil Moisture (%)")
            gr.Markdown("### 🌾 Plant Growth Parameters")
            chlorophyll = gr.Number(label="Chlorophyll (SPAD)")
            shoot_length = gr.Number(label="Shoot Length (cm)")
            root_length = gr.Number(label="Root Length (cm)")
            yield_g = gr.Number(label="Yield (g/pot)")
            relative_yield = gr.Number(label="Relative Yield (%)")
            gr.Markdown("### 📏 Actual Dose (optional)")
            actual_dose = gr.Number(label="Actual Dose (g/pot)")
            predict_btn = gr.Button("🔍 Predict Dose", variant="primary")
        with gr.Column(scale=1):
            gr.Markdown("### 📊 Inference Result")
            pred_box = gr.Markdown("Awaiting prediction...")
            actual_box = gr.Markdown("")
            abs_box = gr.Markdown("")
            log_box = gr.Textbox(label="Debug Logs", lines=15)  # <-- Add this
            predict_btn.click(
                fn=predict_dose,
                inputs=[crop, strain, soil_n, soil_p, soil_k, soil_ec, soil_moisture,
                        chlorophyll, shoot_length, root_length, yield_g, relative_yield,
                        actual_dose],
                outputs=[pred_box, actual_box, abs_box, log_box]
            )

# -----------------------------
# 5️⃣ Launch
# -----------------------------
if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0", server_port=7860)