BibTeX export via the web API now returns stored citation keys instead of generated ones
We maintain a manuscript whose bibliography is generated rather than hand-written: a script fetches our group library from the web API as BibTeX (`/groups//items?format=bibtex`) and writes `references.bib`. The manuscript cites by citation key, so for us those keys act as an interface.
Sometime between 2026-07-30 and 2026-08-06, the keys that endpoint returns changed format:
```
before: @article{wilkinson_fair_2016,
after: @article{wilkinsonFAIRGuidingPrinciples2016,
```
After digging a bit, it looks like the export now returns each item's stored `citationKey` field, where previously it generated a key at export time from the `%a_%t_%y` format. So effectively every latex \cite key in the manuscript needs to change.
As far as I can tell the data didn't change, only the export. Nothing is lost, so this is a rename rather than a breakage, and I'm migrating the new keys into the manuscript.
I am wondering if this is safe to build on now?
Sometime between 2026-07-30 and 2026-08-06, the keys that endpoint returns changed format:
```
before: @article{wilkinson_fair_2016,
after: @article{wilkinsonFAIRGuidingPrinciples2016,
```
After digging a bit, it looks like the export now returns each item's stored `citationKey` field, where previously it generated a key at export time from the `%a_%t_%y` format. So effectively every latex \cite key in the manuscript needs to change.
As far as I can tell the data didn't change, only the export. Nothing is lost, so this is a rename rather than a breakage, and I'm migrating the new keys into the manuscript.
I am wondering if this is safe to build on now?
Upgrade Storage
could there be a way to change in our library/collection the generation pattern for CitationKey? or at least have a set of common patterns to choose from?
IMHO the current one (which I simply didn't see since was not exported to bibtex before) is way too long and with its camelCasing is too distracting from the text when used within article text in .tex, .md etc. Sure thing we could likely code up some service which would curate our records via API and change to desired format, but that would introduce a delay between import/use which would introduce friction.
The most likely way this would have been added is if one of your collaborators has the BetterBibTeX plugin installed and set to that format -- BetterBibTeX is very flexible in how Citation Keys are generated but working in a group can pose some challenges if different users have different preferences for the format. In that case, with default set up otherwise, you would, I believe, get the format of whoever opens the library first after an item is added (I understand the dev is interested in improving that)