import gradio as gr
import fitz
import os
import zipfile

def process(input_pdf):
    # Conversion of PDF to JPG images
    pdf = fitz.open(input_pdf)
    first_page = pdf[0]
    pix = first_page.get_pixmap()
    image_bytes = pix.tobytes("jpg")
    pdf.close()

    temp_dir = "images"
    basename = os.path.basename(input_pdf).split('.')[0]
    image_name = basename + "jpg"
    os.makedirs(temp_dir, exist_ok=True)
    with open(os.path.join(temp_dir, image_name), "wb") as f:
        f.write(image_bytes)
    
    image_path = os.path.join(temp_dir, image_name)
    output = model.inference(image=image_path, prompt=task_prompt)["predictions"][0]

    return output

task_name = "SGSInvoice"
task_prompt = f"<s_{task_name}>"
model = DonutModel.from_pretrained("uartimcs/donut-invoice-extract")
model.eval()
demo = gr.Interface(fn=process,inputs=gr.File(file_types=['.pdf']),outputs="json", title=f"Donut 🍩 demonstration for `{task_name}` task",)
demo.launch()