This was something I was hoping to bring up in today’s office hours. If there was a way Pickaxe could convert PDFs to .MD files before passing them to the models, that would be a massive token saver. There are a lot of packages that can do this like this one. So it is just a matter of making it an option.
It should be an option though. Sometimes the images are important for the pickaxe. I have a web page scanner that looks over a pdf screenshot. That would break if converted to a markdown file.