import gradio as gr
import pickle
import numpy as np

# Load the trained models
with open('rf_crop.pkl', 'rb') as file:
    crop_model = pickle.load(file, errors='ignore')

with open('knn_fertilizer.pkl', 'rb') as file:
    fertilizer_model = pickle.load(file, errors='ignore')
    
with open('Scaler_fertilizer.pkl', 'rb') as file:  # Assuming you saved the scaler during model training
    scaler = pickle.load(file, errors='ignore')

# Label Encoders for the models
crop_label_encoder = {
    0: "Sugarcane", 1: "Wheat", 2: "Cotton", 3: "Jowar", 4: "Rice",
    5: "Maize", 6: "Groundnut", 7: "Grapes", 8: "Tur", 9: "Ginger",
    10: "Turmeric", 11: "Urad", 12: "Gram", 13: "Moong", 14: "Soybean", 15: "Masoor"
}

fertilizer_label_encoder = {
    0: "Urea", 1: "DAP", 2: "MOP", 3: "SSP", 4: "19:19:19 NPK",
    5: "Chilated Micronutrient", 6: "50:26:26 NPK", 7: "Magnesium Sulphate", 
    8: "10:26:26 NPK", 9: "Ferrous Sulphate", 10: "13:32:26 NPK", 
    11: "10:10:10 NPK", 12: "Ammonium Sulphate", 13: "12:32:16 NPK", 
    14: "White Potash", 15: "Hydrated Lime", 16: "20:20:20 NPK", 
    17: "18:46:00 NPK", 18: "Sulphur"
}

# Prediction functions
def predict_crop(Nitrogen, Phosphorus, Potassium, pH, Rainfall, Temperature):
    # Prepare the input data and scale it
    crop_input = np.array([[Nitrogen, Phosphorus, Potassium, pH, Rainfall, Temperature]])
    crop_input_scaled = scaler.transform(crop_input)
    
    # Predict the crop
    crop_prediction = crop_model.predict(crop_input_scaled)
    return crop_prediction[0]

def predict_fertilizer(Nitrogen, Phosphorus, Potassium, pH, Rainfall, Temperature, Crop):
    # Prepare the input data and scale it
    crop_index = list(crop_label_encoder.keys())[list(crop_label_encoder.values()).index(Crop)]
    fertilizer_input = np.array([[Nitrogen, Phosphorus, Potassium, pH, Rainfall, Temperature, crop_index]])
    fertilizer_input_scaled = scaler.transform(fertilizer_input[:, :-1])
    
    # Add crop index back to the scaled input
    fertilizer_input_scaled = np.hstack([fertilizer_input_scaled, [[crop_index]]])
    
    # Predict the fertilizer
    fertilizer_prediction = fertilizer_model.predict(fertilizer_input_scaled)

    return fertilizer_label_encoder[int(fertilizer_prediction[0])]

# Gradio Interface for Crop Prediction
crop_interface = gr.Interface(
    fn=predict_crop,
    inputs=[
        gr.Number(label="Nitrogen"),
        gr.Number(label="Phosphorus"),
        gr.Number(label="Potassium"),
        gr.Number(label="pH"),
        gr.Number(label="Rainfall"),
        gr.Number(label="Temperature")
    ],
    outputs=gr.Label(num_top_classes=1),
    title="Crop Prediction",
    allow_flagging='never'
)

# Gradio Interface for Fertilizer Prediction
fertilizer_interface = gr.Interface(
    fn=predict_fertilizer,
    inputs=[
        gr.Number(label="Nitrogen"),
        gr.Number(label="Phosphorus"),
        gr.Number(label="Potassium"),
        gr.Number(label="pH"),
        gr.Number(label="Rainfall"),
        gr.Number(label="Temperature"),
        gr.Dropdown(label="Crop", choices=list(crop_label_encoder.values()))
    ],
    outputs=gr.Label(num_top_classes=1),
    title="Fertilizer Prediction",
    allow_flagging='never'
)

# Create a Tabbed Interface in Gradio
app = gr.TabbedInterface([crop_interface, fertilizer_interface], ["Crop Prediction", "Fertilizer Prediction"])

# Launch the app
app.launch()