import gradio as gr
import numpy as np
import pandas as pd
import tensorflow as tf
from IPython.display import HTML
from tensorflow.keras.metrics import MeanSquaredError


# Load the trained model
model = tf.keras.models.load_model('real_estate_price_prediction_model.h5')

# Load the original dataset to get unique categories for 'secteur' and 'city'
original_df = pd.read_excel('Moroccan Real Estate Price Clean Dataset .xlsx')  # Replace with your dataset path

# Get unique categories for 'secteur' and 'city'
unique_secteurs = original_df['secteur'].unique()
unique_cities = original_df['city'].unique()

# Define the column names
columns = ['surface', 'pieces', 'chambres', 'sdb', 'age', 'etage', 'etat_Bon état', 'etat_Nouveau', 'etat_À rénover', 'secteur', 'city']

# Function to preprocess user input
def preprocess_input(user_input, columns, unique_secteurs, unique_cities):
    # Define the total number of features expected by the model
    total_features = 1015

    # Initialize all features to 0
    input_array = np.zeros((1, total_features), dtype=np.float64)

    # Update numerical features
    numerical_features = ['surface', 'pieces', 'chambres', 'sdb', 'age', 'etage', 'etat_Bon état', 'etat_À rénover']
    for feature in numerical_features:
        input_array[0, columns.index(feature)] = user_input[feature]

    # Update categorical features
    for feature in ['secteur', 'city']:
        if user_input[feature] in unique_secteurs or user_input[feature] in unique_cities:
            input_array[0, columns.index(user_input[feature])] = 1

    return input_array

# Function to predict price based on user input
def predict_price(surface, pieces, chambres, sdb, age, etage, etat_Bon_état, etat_Nouveau, etat_À_rénover, secteur, city):
    # Preprocess the user input
    user_input = {
        'surface': surface,
        'pieces': pieces,
        'chambres': chambres,
        'sdb': sdb,
        'age': age,
        'etage': etage,
        'etat_Bon état': etat_Bon_état,
        'etat_Nouveau': etat_Nouveau,
        'etat_À rénover': etat_À_rénover,
        'secteur': secteur,
        'city': city
    }
    input_array = preprocess_input(user_input, columns, unique_secteurs, unique_cities)

    # Make prediction using the model
    predicted_price = model.predict(input_array)

    return f"Predicted price: {predicted_price[0][0]}"

# Create HTML code to display an image
image_html = "<img src='/content/Capture d’écran 2024-01-28 155359.jpg' style='max-width:100%;'>"

# Gradio interface setup
interface = gr.Interface(
    fn=predict_price,  # The function to be called with user input
    inputs=[
        gr.Slider(label=f"Enter value for 'surface(m²)'", minimum=0, maximum=500, step=1),
        gr.Slider(label=f"Enter value for 'pieces'", minimum=0, maximum=15, step=1),
        gr.Slider(label=f"Enter value for 'chambres'", minimum=0, maximum=10, step=1),
        gr.Slider(label=f"Enter value for 'sdb'", minimum=0, maximum=5, step=1),
        gr.Slider(label=f"Enter value for 'age'", minimum=0, maximum=115, step=1),
        gr.Slider(label=f"Enter value for 'etage'", minimum=0, maximum=20, step=1),
        gr.Slider(label=f"Enter value for 'etat_Bon état'", minimum=0, maximum=1, step=1),
        gr.Slider(label=f"Enter value for 'etat_Nouveau'", minimum=0, maximum=1, step=1),
        gr.Slider(label=f"Enter value for 'etat_À rénover'", minimum=0, maximum=1, step=1),
        gr.Textbox(label=f"Enter value for 'secteur'", type="text"),
        gr.Textbox(label=f"Enter value for 'city'", type="text")
    ],
    outputs=gr.Textbox(label="Predicted Price(Dh):", interactive=False),
    title="Real Estate Price Prediction",
    description="Enter property details to predict its price.",
    examples=[
        [250, 5, 3, 2, 10, 3, 1, 0, 0, "'Secteur_A'", "'City_X'"],
        [150, 4, 2, 1, 5, 2, 1, 0, 0, "'Secteur_B'", "'City_Y'"]
    ],
    theme="compact",  # Compact theme for a cleaner look
)

# Launch the Gradio interface
interface.launch(share=False, debug=False)