We're now generating the proper openapi spec via FastAPI. The problem is, if we point the Speakeasy clients to the new spec, the generated code has too much churn. All good stuff if we were releasing the clients for the first time, but now it will break compatibility. We'll need to maintain a separate openapi spec for now, and add new params here whenever they go into the server.
We're now generating the proper openapi spec via FastAPI. The problem is, if we point the Speakeasy clients to the new spec, the generated code has too much churn. All good stuff if we were releasing the clients for the first time, but now it will break compatibility. We'll need to maintain a separate openapi spec for now, and add new params here whenever they go into the server.
Example speakeasy churn: https://github.com/Unstructured-IO/unstructured-python-client/pull/42