​Both have their place and can be useful, however, most PDF use cases can be solved with higher context windows and a manually constructed prompt with copied and pasted text chunks from PDF files.

See this example of copying and pasting PDF text: https://webapp.server.searchweb.keymate.ai/share/pREnBSG

This originated from a real story. A Keymate user tried to create a real estate market overview from PDF data (which had tables and text, etc). Use case is PDF related because text is there, this is understandable but most PDF files are above input context windows.
​
At Keymate.AI we allow 100 pages long PDFs to be used in personal knowledge base training. Keymate allows you to train your long term memory aka Personal Knowledge Base with PDFs you upload, but it doesn't offer hybrid search or direct text search. There are great PDF plugins that do direct text search. KeyMate only does semantic search today but we are preparing our hybrid text search solution, which will not only use PDFs it will use internet data, your long term memory in hybrid text search ( semantic search + direct text search on multiple sources).