Hi Team,
Regarding the versions of the Gemini APIs, I see that both v1 and v1beta are available, and the documentation also refers to the v1beta version [1] as well. However, in document [2], it’s clearly mentioned that breaking changes may occur.
I’d like to better understand the versioning strategy. There are many useful APIs in the v1beta version, so after testing v1beta, will these APIs be moved to v1, or will new features be released as part of v2 instead?
Could you please provide some clarity on this?
[1] All methods | Gemini API | Google AI for Developers
[2] API versions explained | Gemini API | Google AI for Developers
Thank you,
Dilshan
Welcome to the forums!
Typically, in Google’s API naming pattern, backwards compatible changes will (eventually) roll into the base version. the “v1beta” API only introduces new settings, so it likely to (eventually) make it into “v1”. Only if they changed the fundamental requirements would they go to a new version.
In this case, they’ve kinda boxed themselves into a corner. The new models need the beta API to access some new features, but the old models will reject them in some cases. So they are in this state where a production model with production features needs to use a “beta” API. And all the libraries have to use the “beta” API to support those features.
And since the libraries are coded to use “v1beta” now, it seems unlikely that they’ll move away from using it at any point soon.
1 Like