import gradio as gr
import numpy as np
import pandas as pd
from sklearn.ensemble import GradientBoostingRegressor
import pickle

def create_gui():
    """
    Creates and launches the Gradio user interface for the biomass adsorption prediction.
    """
    # Define the input and output components of the GUI
    inputs = [
        gr.Dropdown(
            ["Benzocaine", "Ciprofloxacin", "Citalopram", "Diclofenac", "Dimetridazole", "Floxentine", "Ibuprofen", "Metronidazole", "Nitroimidazole", "Norfloxacin", "Oxytetracycline", "Salicylic Acid", "Sulfadiazine", "Sulfamethazine", "Sulfamethoxazole", "Tetracycline", "Triclosan"], label="Select Pharmaceutical"
        ),
        gr.Slider(minimum=0, maximum=950, step=0.01, label="TemP (K)"),
        gr.Slider(minimum=0, maximum=480, step=0.01, label="Time (min)"),
        gr.Slider(minimum=0, maximum=250, step=0.01, label="PS (mm)"),
        gr.Slider(minimum=0, maximum=2000, step=0.01, label="BET surface area (m2/g)"),
        gr.Slider(minimum=0, maximum=1, step=0.01, label="Pore Volume (cm3)"),
        gr.Slider(minimum=0, maximum=100, step=0.01, label="C (%)"),
        gr.Slider(minimum=0, maximum=100, step=0.01, label="H (%)"),
        gr.Slider(minimum=0, maximum=100, step=0.01, label="N (%)"),
        gr.Slider(minimum=0, maximum=100, step=0.01, label="O (%)"),
    ]

    outputs = [
        gr.Textbox(label="Predicted Maximum Adsorption Capacity, Qm (mg/g)")
    ]
    
    # Define the title and description of the GUI
    title = "GUI for Pharmaceutical Removal via Biochar Adsorption"
    description = "This GUI uses machine learning to predict maximum absorption capacity (Qm) for the selected pharmaceutical based on various input parameters."

    gr.Interface(fn=predict_qm, inputs=inputs, outputs=outputs, title=title, description=description).launch()

def predict_qm(pharmaceutical, temp, time, ps, bet, pv, c, h, n, o):
    """
    Predicts maximum adsorption capacity (Qm) based on input features for the selected pharmaceutical.
    
    Parameters:
    pharmaceutical (str): Pharmaceutical type.
    TemP (float): Temperature K of the biomass.
    Time (float): Time it takes for biochar to remove pharmaceutical in minutes .
    PS (float): Pore space in mm.
    BET (float): Area of powder in m2/g being poured into the solution.
    PV (float): Pore volume in cm3.
    C (float): C (%) content of the biomass.
    H (float): H (%) content of the biomass.
    N (float): N (%) content of the biomass.
    O (float): O (%) content of the biomass.
    
    Returns:
    float: The predicted maximum adsorption capacity (Qm).
    """

    # Concatenate the inputs into a numpy array
    input_data = pd.DataFrame([[temp, time, ps, bet, pv, c, h, n, o]], columns=["Temp", "Time", "PS", "BET", "PV", "C", "H", "N", "O"])

    # Load the trained model from the pickled file
    with open('xgb_best_params.pkl', 'rb') as file:
            loaded_model = pickle.load(file)

    # Make predictions using the loaded machine learning model
    prediction = loaded_model.predict(input_data)
    qm = np.round(prediction[0], 2)

    return qm

if __name__ == '__main__': 
    # Create GUI
    create_gui()