import gradio as gr
import pickle
import pandas as pd

# Load the trained model
filename = 'knn_model.sav'
loaded_model = pickle.load(open(filename, 'rb'))

# Helper functions for input transformations
def hptension(hp):
    return 1 if hp == 'yes' else 0

def ht_dis(ht):
    return 1 if ht == 'yes' else 0

def gender_select(gen):
    return 1 if gen == 'male' else 0

def age_group_selector(age_grp):
    age_map = {'0-16': 0, '17-32': 1, '33-48': 2, '49-64': 3, '64+': 4}
    return age_map.get(age_grp, 0)

def smoker_cat(smoke):
    smoke_map = {'formerly smoked': 0, 'never smoked': 1, 'smokes': 2, 'Prefer not to say': 3}
    return smoke_map.get(smoke, 3)

# Prediction function
def predict_insurance(input_gender, input_age_group, input_hypertension, input_heart_disease, input_avg_glucose_level, input_bmi, input_smoking_status):
    # Prepare the input data
    series = {
        'gender': [gender_select(input_gender)],
        'age_band': [age_group_selector(input_age_group)],
        'hypertension': [hptension(input_hypertension)],
        'heart_disease': [ht_dis(input_heart_disease)],
        'avg_glucose_level': [input_avg_glucose_level / 272],
        'bmi': [input_bmi / 49],
        'smoking_status': [smoker_cat(input_smoking_status)],
    }

    vector = pd.DataFrame(series)

    # Perform prediction
    result = loaded_model.predict(vector)
    return "Risk of having stroke is high" if result[0] == 1 else "Risk of having stroke is low"

# CSS to hide footer and markdown elements
css = """
footer {display:none !important}
.output-markdown{display:none !important}
footer {visibility: hidden} 
.gr-button-lg {
    z-index: 14;
    width: 113px;
    height: 30px;
    left: 0px;
    top: 0px;
    padding: 0px;
    cursor: pointer !important; 
    background: none rgb(17, 20, 45) !important;
    border: none !important;
    text-align: center !important;
    font-size: 14px !important;
    font-weight: 500 !important;
    color: rgb(255, 255, 255) !important;
    line-height: 1 !important;
    border-radius: 6px !important;
    transition: box-shadow 200ms ease 0s, background 200ms ease 0s !important;
    box-shadow: none !important;
}
.gr-button-lg:hover{
    z-index: 14;
    width: 113px;
    height: 30px;
    left: 0px;
    top: 0px;
    padding: 0px;
    cursor: pointer !important; 
    background: none rgb(66, 133, 244) !important;
    border: none !important;
    text-align: center !important;
    font-size: 14px !important;
    font-weight: 500 !important;
    color: rgb(255, 255, 255) !important;
    line-height: 1 !important;
    border-radius: 6px !important;
    transition: box-shadow 200ms ease 0s, background 200ms ease 0s !important;
    box-shadow: rgb(0 0 0 / 23%) 0px 1px 7px 0px !important;
}
"""

# Gradio app layout
with gr.Blocks(title="Brain Stroke Prediction | Data Science Dojo", css=css) as demo:
    with gr.Row():
        input_gender = gr.Radio(["male", "female"], label="Gender")
        input_hypertension = gr.Radio(["yes", "no"], label="Hypertension")
        input_heart_disease = gr.Radio(["yes", "no"], label="Heart Disease")
    
    with gr.Row():
        input_age_group = gr.Dropdown(['0-16', '17-32', '33-48', '49-64', '64+'], label='Age Group')
        input_smoking_status = gr.Dropdown(['formerly smoked', 'never smoked', 'smokes', 'Prefer not to say'], label='Smoker')

    with gr.Row():
        input_avg_glucose_level = gr.Slider(0, 270, label='Average Glucose Level')

    with gr.Row():
        input_bmi = gr.Slider(0, 45, label='BMI Range')

    with gr.Row():
        stroke = gr.Textbox(label='Chances of Stroke', interactive=False)

    btn_ins = gr.Button(value="Submit")
    btn_ins.click(
        fn=predict_insurance,
        inputs=[
            input_gender,
            input_age_group,
            input_hypertension,
            input_heart_disease,
            input_avg_glucose_level,
            input_bmi,
            input_smoking_status,
        ],
        outputs=[stroke]
    )

demo.launch(debug=True)