PDF -> Markdown looks like a pretty great use case
Just added box detection support -- maybe I'll start from here https://github.com/junhoyeo/BetterOCR#-box-detection