opendatalab/PDF-Extract-Kit ? reverse-engineered prompt

I want to build a Python toolkit that can really dig into complex PDF files and pull out all the content in a structured way. This is for things like dense academic papers, financial reports, or textbooks, not just simple text documents.

The tool should be able to perform layout analysis to identify different elements like titles, paragraphs, lists, and images. It's really important that it can find tables and accurately convert them into a structured format like HTML or Markdown. It also needs to be great with math—it should detect mathematical formulas in the document and convert them into LaTeX code. Basically, I want a set of high-quality models for layout detection, OCR, table recognition, and formula recognition that I can use together. The whole thing should be modular so I can use these pieces to build other applications on top of it later.