The way Keymate "reads" your PDFs
Neither ChatGPT nor your Keymate Memory knows anything about your PDF's metadata. This means you should not use specific pages, titles, or subsections as a reference while talking to ChatGPT. Your queries from your Keymate Memory should be topic based.
Keymate does semantic lookup on your Keymate Memory and it only reads the text in the PDF, not seeing page numbers or dividing the text according to section names.
Images, specific titles, page numbers, table of contents in PDF files are not used to train Keymate Memory.