import gradio as gr
import math
import numpy as np
from joblib import load

# Load model once
model = load('model2.joblib')

# Prediction function
def model2(data):
    input_array = np.array([data]).reshape(1, -1)
    prediction = model.predict(input_array)[0]
    return prediction

# Gradio interface function
def process_input(num_str):
    if len(num_str) != 6 or not num_str.isdigit():
        return ["Error: Input must be a 6-digit number"]

    windows = [int(num_str[i:i+2]) for i in range(len(num_str)-1)]  # [12, 23, 34, 45, 56]
    divisor = 3 * math.pi
    normalized = [x / divisor for x in windows]

    # Get predictions
    pred1 = model2(normalized[0])
    pred2 = model2(normalized[1])
    pred3 = model2(normalized[2])
    pred4 = model2(normalized[3])
    pred5 = model2(normalized[4])

    # Errors
    err1 = pred1 - windows[1]
    err2 = pred2 - windows[2]
    err3 = pred3 - windows[3]
    err4 = pred4 - windows[4]
    errors = [err1, err2, err3, err4]

    combined_error = sum(errors)
    avg_error = combined_error / len(errors)
    distances = [abs(e - avg_error) for e in errors]
    nearest_indices = sorted(range(len(distances)), key=lambda i: distances[i])[:2]
    nearest_values = [errors[i] for i in nearest_indices]
    all_three = nearest_values + [avg_error]
    mean_val = np.mean(all_three)

    # Adjust predictions
    ads_list = [(pred4 + err if avg_error > 0 else pred4 - err) for err in errors]
    ads_array = np.array(ads_list, dtype=np.float32)
    ads2 = ads_array * (3 * math.pi)

    # Extract digit before decimal
    digit_before_decimal = [int(str(int(x))[-1]) for x in ads2]

    return digit_before_decimal

# Gradio UI
iface = gr.Interface(
    fn=process_input,
    inputs=gr.Textbox(label="Enter a 6-digit number"),
    outputs=gr.Textbox(label="Digit Before Decimal Array"),
    title="ML Prediction Error Adjustment",
    description="Input a 6-digit number. Returns processed digit array after model predictions and transformations."
)

iface.launch()