Contributing
Where the behaviour actually comes from
Almost everything you might want to change in a client is decided somewhere else:
| What you want to change | Where it is decided |
|---|---|
| A method name | the operation’s operationId in the spec |
| A field name or type | the schema in the spec |
| A missing endpoint | the spec |
| Doc comments on a model | the spec’s descriptions |
| Retries, pagination, helpers | nowhere yet. See what you build yourself |
The specification lives in a private repository and is not published, so there is currently no way for you to read it, let alone send a change. That is a real limitation and we are not going to describe it as anything else.
What is useful to us
Reporting a problem, precisely. It is worth more than a patch here, because a good report changes the spec and every one of the ten clients at once, where a patch would have changed one until the next release.
For a wrong or missing field, tell us the endpoint, what the client produced and what the API actually accepted or returned. That is a spec bug and the fix propagates everywhere.
For a confusing method name, name the operation and the method you
expected. Most odd names are route-derived because the operation has no
operationId, and those are being filled in; knowing which ones people
actually hit helps us order the work.
For something broken in one language only, say which version and which registry. That is usually a generator-template or publishing problem rather than a spec one, and it is the category most likely to be genuinely unnoticed, because our own checks run per repository.
Send any of it to support, with the request_id from a failing
response where you have one.
What would make contribution possible
Said plainly, since you may be reading this because you wanted to help:
- Publishing the specification. Nothing can be contributed to an input nobody can see. It is deliberately unpublished today because it is still being corrected, and serving a stale contract would hand people generated clients built from it.
- A public issue tracker on something that is not generated output.
Neither exists yet. Until they do, a careful bug report is the whole of what is available, and it is genuinely useful.
The ergonomics layer, if you were going to offer one
If you were about to write a retry wrapper and contribute it: a tested one already exists internally for Go, TypeScript and Python and is simply not published. The gap is in our release process, not in the code, so an implementation from you would not close it.
Telling us it is blocking you is the thing that moves it.