import gradio as gr
import numpy as np
import pandas as pd
import joblib

# Load model and preprocessing objects
model = joblib.load("model.pkl")
scaler = joblib.load("scaler.pkl")
label_encoder = joblib.load("label_encoder.pkl")

# Define prediction function
def predict_class(COUNT, COUNT_CV, STRENGTH, CSP, U_PERCENT, THIN, THICK, NEPS, IPI):
    # Input as a DataFrame with correct feature names
    input_df = pd.DataFrame([{
        "COUNT": COUNT,
        "COUNT_CV": COUNT_CV,
        "STRENGTH": STRENGTH,
        "CSP": CSP,
        "U_PERCENT": U_PERCENT,
        "THIN": THIN,
        "THICK": THICK,
        "NEPS": NEPS,
        "IPI": IPI
    }])
    
    # Scale input
    scaled_input = scaler.transform(input_df)
    
    # Predict
    pred_index = model.predict(scaled_input)[0]
    
    # Decode class label
    predicted_label = label_encoder.inverse_transform([pred_index])[0]
    
    return predicted_label

# Build Gradio Interface
with gr.Blocks() as demo:
    gr.Markdown("## 🧵 Textile Mixing Classifier (Scikit-learn Model)")
    gr.Markdown("Use the sliders to set values and click **Predict** to classify the textile mix.")

    with gr.Column():
        with gr.Row():
            COUNT = gr.Slider(7, 50, step=0.5, label="COUNT")
            COUNT_CV = gr.Slider(0.48, 1.87, step=0.1, label="COUNT_CV")
            STRENGTH = gr.Slider(62.71, 384.2, step=0.1, label="STRENGTH")
        with gr.Row():
            CSP = gr.Slider(1500, 4755, step=1, label="CSP")
            U_PERCENT = gr.Slider(6.38, 12.12, step=0.1, label="U_PERCENT")
            THIN = gr.Slider(0.0, 19, step=0.1, label="THIN")
        with gr.Row():
            THICK = gr.Slider(2.0, 150, step=1, label="THICK")
            NEPS = gr.Slider(1, 494, step=1, label="NEPS")
            IPI = gr.Slider(6.0, 646, step=1, label="IPI")

    output = gr.Textbox(label="Predicted Class")

    btn = gr.Button("Predict")
    btn.click(fn=predict_class,
              inputs=[COUNT, COUNT_CV, STRENGTH, CSP, U_PERCENT, THIN, THICK, NEPS, IPI],
              outputs=output)

demo.launch()




# import tensorflow as tf

# import gradio as gr
# import numpy as np
# import joblib

# # Load your trained model
# model = joblib.load("mixing_prediction_model.pkl")

# # Prediction function
# def predict(COUNT, COUNT_CV, STRENGTH, CSP, U_PERCENT, THIN, THICK, NEPS, IPI):
#     features = np.array([[COUNT, COUNT_CV, STRENGTH, CSP, U_PERCENT, THIN, THICK, NEPS, IPI]])
#     # pred = model.predict(features)[0]
#     # mixing_name = le.inverse_transform([pred])[0]
#     prediction = model.predict(features)[0][0]
#     return prediction

# # Gradio app with custom sliders
# with gr.Blocks() as demo:
#     gr.Markdown("## 🧵 Textile Mixing Predictor (TensorFlow Model)")
#     gr.Markdown("Adjust the sliders for each feature and get the predicted output.")

#     with gr.Row():
#         with gr.Column():
#             COUNT = gr.Slider(minimum=7, maximum=50, step=0.5, label="COUNT")
#             COUNT_CV = gr.Slider(minimum=0.48, maximum=1.87, step=0.1, label="COUNT_CV")
#             STRENGTH = gr.Slider(minimum=62.71, maximum=384.2, step=0.1, label="STRENGTH")
#         with gr.Column():
#             CSP = gr.Slider(minimum=1500, maximum=4755, step=1, label="CSP")
#             U_PERCENT = gr.Slider(minimum=6.38, maximum=12.12, step=0.1, label="U_PERCENT")
#             THIN = gr.Slider(minimum=0.0, maximum=19, step=0.1, label="THIN")
#         with gr.Column():
#             THICK = gr.Slider(minimum=2.0, maximum=150, step=1, label="THICK")
#             NEPS = gr.Slider(minimum=1, maximum=494, step=1, label="NEPS")
#             IPI = gr.Slider(minimum=6.0, maximum=646, step=1, label="IPI")

#     output = gr.Number(label="Predicted Output")

#     predict_button = gr.Button("Predict")

#     # Bind inputs to function
#     predict_button.click(
#         fn=predict,
#         inputs=[COUNT, COUNT_CV, STRENGTH, CSP, U_PERCENT, THIN, THICK, NEPS, IPI],
#         outputs=output
#     )

# demo.launch()