Yes it's certainly a first draft!! :-) Thanks for trying it out.

With the disambig pages, can you suggest how to detect them?

Not sure.
Could you detect the presence of the Q6148868 templates ? (and same thing for Q15701815 ).
Or else maybe with the categories.
Ah, there's a couple of other bugs here:

The page actually
already has a WIkidata ID, but the ws-search database didn't know about
it :-( probably because it was failing for a while on some weird
problems. I've re-run the scraper, and now that work is showing up with
it's proper Q-number:

The idea with the quickstatements is that it'll only show it for works
that are *not yet* linked to wikidata. This is where the disambig
problem comes in, because there doesn't seem to be a simple way to
determine what's an edition and what's a work without resorting to
Wikidata. We could look at categories? Is it a truth universally
acknowledged that pages in the categories defined as are all disambiguation pages?
That could work...

The truth (and I guess it is universal but could someone confirm?) is that pages with 'multiples editions' are 'works' (Q571, this is what I do for fr.wikisource at least).

Thank you for all the work!

Cdlt, ~nicolas