At this point it seems like just learning the library is easier than trying to cram the documentation into an LLM compatible format.
I'm a library author myself, so publishing LLM-enhanced versions of the docs to help other people use my library more effectively feels like a sensible use of my time.