LlamaParse is a document parser that turns complex PDFs - tables, multi-column layouts, scanned pages - into clean Markdown or text that downstream AI and RAG pipelines can use. The typical flow is: upload a PDF, LlamaParse extracts structured text, and you get back Markdown.
TigZig publishes an open-source Flask reference that wraps LlamaParse as a simple upload-and-extract service, so you can self-host it or adapt it. The repo is at https://github.com/amararun/shared-llama-parse-flask.
TigZig's other document tools (including Markdown-to-PDF) are listed at https://www.tigzig.com/apis.
Building something like this? How I work covers the rates, the availability and what I take on.