Add Schema parameter to API endpoint requests

[tp]

I just noticed errors like

2026/02/27 17:20:40 ERROR error running API route error="error listing job kinds: ERROR: relation \"river_job\" does not exist (SQLSTATE 42P01)"

when using the autocomplete in the job list.

As this was my first hour using River, I just gave the error to Claude who immediately pointed out the inconsistencies between the 2 fixed methods (which did not pass the schema) and the others which do.

Depending on if you're fine with this, I could take more time later on to align this more with the repo style or add tests. Whatever is desired :)

[bgentry]

Hi @tp! Thanks for reporting this and attempting a fix. Two requests for you before we proceed further:

1. Can you add yourself to the CLA? https://github.com/riverqueue/rivercla
2. Can you also work with Claude to add some test coverage that illustrates the issue (with a verified failure prior to fixing it)? While we don't have a CLAUDE.md in the repo, just tell it to pay attention to AGENTS.md and adhere to the repo's existing testing conventions & code patterns 😄 🤖

[tp]

I signed the CLA.

Adding the test seems a bit tricky, because the test setup here already uses schemas for test isolation, and therefore the test driver behaves differently than a production setup.

Hence the test code Claude came up with it not very straightforward/production like.

While I also often use the "schema per test case" pattern for "normal" test isolation, I wonder if the specific schema testing would warrant something like https://testcontainers.com/guides/getting-started-with-testcontainers-for-go?
(We have also used this in other projects as the default, instead of simple schema switches, and performance wise it was totally fine.)

So while the tests now pass, they also did so before the fix (due to the implicitly applied schema)…

Context / Claude explanation

[tp]

@bgentry Are my updates what you had in mind, or do you need anything else from me here to get this fixed?

I think the underlying tensions I describe (schema switches being a legitimate feature vs. also being used for tests) is a real problem for now, maybe something to tackle separately, and for now just rely on expert judgement (since this is basically just an alignment with how other River UI APIs work) and end-user integration test runs.