o-valo/TuxBook ? reverse-engineered prompt

Build me a local Python CLI tool that can take either a big PDF book or a ZIP of page images, translate it page by page into another language, and then turn it back into a clean PDF.

I want it to handle really large books, like 1000 plus pages, without crashing, and if I stop it halfway it should be able to pick up where it left off. It should work with an embedded text layer when possible, but fall back to OCR for scanned pages. Use Tesseract locally by default, and if needed support a vision model fallback through Ollama. The app should also save a Markdown version of the translated book, including headings and linked images, so I can edit the text and then re build the final PDF from that Markdown.

Please make it support both a simple page by page mode and a nicer book layout mode with chapter breaks, page numbers, and a title page. If you need to look up current docs online for the PDF or OCR libraries, do that.