You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I've been skimming the codebase and also the OpenAPI spec. Just curious (and this might help me understand the codebase better) - what was the main design and considerations you took while developing this library to use the OpenAI API? Did you mostly use the OpenAPI spec or look at the design of their official python library?
The primary source was OpenAPI spec and async-stripe was source of inspiration, you might find evolution of this library if you look at early commits. When I started I couldnt find good resources to generate code from spec, and json spec is not always up to date, sometimes has human errors. Moreover there are complex situations like streaming where an input field like "stream": true determines SSE response.
So I decided to build this library with hand instead of relying on auto generation (although auto generation would be ideal for Rust in long term) because my focus was on correctness, auto retries on rate limit and full API coverage - all the things to make library production ready.
I had skimmed through the offcial python library too, and then went with spec as source of truth and grouping ("images", "fine-tunes" etc.) based on OpenAPI spec + RESTful spec.
Just like
examples/function-call
but using streaming call.Note: There is open PR at the moment but no way to test it #83
The text was updated successfully, but these errors were encountered: