look here, and at this thread in general:
I did this for a while and couldn’t really abide it. I use Omnivore.app now and prefer this workflow to zotero. Basically I read and highlight in Omnivore and do all the reference management in Logseq. The Omnivore plugin manages synchronization of the hightlights for me and each highlight can be linked back to the pdf. Your mileage will vary.
My take is this: staying 100% in logseq for everything including reading/annotating is the best. But this doesn’t work with, say, an iOS device as the reader/highlighter, because the pdf viewer only works on desktop. So if you can’t read with Logseq, then I think Omnivore has the best workflow right now. Ideally the logseq mobile client will gain the pdf reader functionality at some point but it seems like it is not currently on the roadmap, which is pretty irritating – because if we had iOS/Android pdf reading Logseq would be the magic solution to many problems that have not been solved anywhere.