Using DocuPanda Review and redacting everything that comes up

gistfile1.txt

import os

import dotenv
import requests

dotenv.load_dotenv()


class PiiBoundingBox(object):
    def __init__(self, x1: float, y1: float, x2: float, y2: float, page: int, pii_type: str):
        self.x1 = x1  # upper left (fractional coordinate)
        self.y1 = y1  # upper left (fractional coordinate)
        self.x2 = x2  # lower right (fractional coordinate)
        self.y2 = y2  # lower right (fractional coordinate)
        self.page = page - 1  # make page 0 indexed
        self.pii_type = pii_type  # type of PII (account number, etc.)


def get_redactions(review_id) -> list[PiiBoundingBox]:
    api_key = os.environ['DOCUPANDA_API_KEY']
    url = f"https://app.docupanda.io/review?review_id={review_id}"
    headers = {
        "accept": "application/json",
        "X-API-Key": api_key
    }

    response = requests.get(url, headers=headers)
    payload = response.json()

    redaction_list = []
    for pii_payload in payload['data']['pii']:
        for occurrence in pii_payload['occurrences']:
            bboxes = occurrence['textSpan']['review']['boundingBoxes']
            page = occurrence['textSpan']['review']['page']
            for bbox in bboxes:
                x1y1x2y2 = bbox
                redaction_list.append(PiiBoundingBox(
                    x1=x1y1x2y2[0],
                    y1=x1y1x2y2[1],
                    x2=x1y1x2y2[2],
                    y2=x1y1x2y2[3],
                    page=page,
                    pii_type=pii_payload['type']['value']
                ))

    return redaction_list


def redact_pdf(pdf_content: bytes, redactions: list[PiiBoundingBox]) -> bytes:
    from pdf2image import convert_from_bytes
    from PIL import ImageDraw
    import io

    if not redactions:
        return pdf_content

    # Convert PDF to images
    images = convert_from_bytes(pdf_content)

    # Group redactions by page
    redactions_by_page = {}
    for redaction in redactions:
        page = redaction.page
        if page not in redactions_by_page:
            redactions_by_page[page] = []
        redactions_by_page[page].append(redaction)

    # Process each page image
    processed_images = []
    for page_number, image in enumerate(images):
        # Get redactions for this page
        page_redactions = redactions_by_page.get(page_number, [])

        if page_redactions:
            draw = ImageDraw.Draw(image)
            width, height = image.size
            for redaction in page_redactions:
                # Coordinates are fractional; calculate actual pixel positions
                x1 = int(redaction.x1 * width)
                y1 = int(redaction.y1 * height)
                x2 = int(redaction.x2 * width)
                y2 = int(redaction.y2 * height)
                # Draw black rectangle
                draw.rectangle([x1, y1, x2, y2], fill='black')
        processed_images.append(image)

    # Convert images back to PDF
    output_stream = io.BytesIO()
    if processed_images:
        first_image = processed_images[0]
        first_image.save(output_stream, format='PDF', save_all=True, append_images=processed_images[1:])
        pdf_bytes = output_stream.getvalue()
        output_stream.close()
        return pdf_bytes
    else:
        return b''


def main():
    file_path = '/var/data/docupanda/data/PII (uri)/bofa_statement.pdf'
    redactions = get_redactions(review_id='3ec69589')
    with open(file_path, 'rb') as file:
        pdf_content = file.read()
    redacted_pdf = redact_pdf(pdf_content, redactions)

    with open('redacted.pdf', 'wb') as file:
        file.write(redacted_pdf)


if __name__ == '__main__':
    main()