I run a very crude shell-based scanning app (termux has an option to add shell scripts as "apps" on the homescreen), amongst others.

termux supports native GUI dialogues (input/checkboxes), and the scan-shellscript uses that for input (name, greyscale/color, jpg/pdf output and number of pages for pdf) then launches a couple of remote scanimage + ocr postproc + pdf generation commands and transfers + opens that in the local pdf app