hmm: it's much clunkier than I'd like, but if I go on Photoprism to "Edit/Files", I can get a "Filename" from the physical store. I could make that available to the web via symlinks, then use it for hotlinks on #Wordpress.
but images online these days are complicated (what isn't?), what's with srcset and everything. perhaps at the cost of some backing store duplication, I should just import my images into wordpress; thankfully, a mass import CLI tool is provided.
the one snag in my workflow is that, contrary to what I assumed, #photoprism does not write metadata back into the exif! there's a large thread about it on github but it's clearly the kind of thing that users care about but devs shy away from. the state of acessibility in open source is abysmal, there's basically no tools at any point (photo app, cellphone gallery apps, online gallery apps, blogging, #mastodon....) that will let me write image descriptions that stay with the image when the image is copied somewhere.
I would like to emphasize that the exif ImageDescription field has existed for 28 motherfucking years.
For now I think the least effort, albeit significant effort, would be to mouse-click the filepath of each photo I want and copy it into a list on Nextcloud, and that list is automatically wp media import'ed on cron.
I could conceive of writing a script to scrap the Photoprism database (inside goddamn Docker!) to find the $#!@# image description SQL column and paste it into Wordpress' image description SQL column and hope neither database schema changes next week, but at this point it sounds like a better deal to just manually copy-paste image descriptions everywhere images are exposed.
"fuck you visually impaired people"—FOSSbros