The OpenAPI autogenerated clients kinda suck though.
My preferred approach for doing this is to have a hand-rolled SDK generator that reads the request, response and error models out of the microservice project and emits the same in each language targeted by the SDK, along with a minimal stub that calls the API.
You then spend 15 minutes at most, customizing the stub if needed, if you need custom behaviours like streaming.
The OpenAPI autogenerated clients kinda suck though.
My preferred approach for doing this is to have a hand-rolled SDK generator that reads the request, response and error models out of the microservice project and emits the same in each language targeted by the SDK, along with a minimal stub that calls the API.
You then spend 15 minutes at most, customizing the stub if needed, if you need custom behaviours like streaming.