import pickle
import numpy as np
from PIL import Image
import gradio as gr

# load your pickled RandomForest (make sure mnist_model.pkl lives in /app)
model = pickle.load(open("mnist_model.pkl", "rb"))

def classify_digit(img: Image.Image) -> str:
    # convert to 28×28 grayscale array
    gray = img.convert("L").resize((28, 28))
    arr = np.array(gray).reshape(1, -1)
    pred = model.predict(arr)[0]
    return f"Predicted digit: {pred}"

demo = gr.Interface(
    fn=classify_digit,
    inputs=gr.Image(type="pil", label="Upload a 28×28 digit"),
    outputs=gr.Textbox(label="Prediction"),
    title="MNIST Digit Classifier",
    description="Upload a handwritten digit image (28×28) to get a live prediction!"
)

if __name__ == "__main__":
    demo.launch()