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
Lightning now has a production dependency on Apollo, via the AI Assistant.
Its critical that we have a good process for ensuring that releases are stable.
But automated testing of AI is kinda hard?
Thoughts:
We don't have many tests yet but let's set up CI Setup CI Tests #133. That will at least catch some server bugs.
We should consider integration tests for each service. Curl into it and run some data through. This will call out to backend servers and use up tokens. So we don't want to do it too often. Perhaps we should make the tests manually run for open PRs, and automatically on release/ branches in a ready state. So basically we run integration tests on release candidates before release.
We should also have a set of manual QA steps, which are probably more useful in the short term.
In order to run proper manual QA, we need the build to be up on staging (there's not much point running QA on our local builds, probably not even if we run them in docker, because we need to test the prod environment as much as the build itself). That's Setup automatic deployments to staging #119 and I'll speak to Stu about it next week
The text was updated successfully, but these errors were encountered:
Lightning now has a production dependency on Apollo, via the AI Assistant.
Its critical that we have a good process for ensuring that releases are stable.
But automated testing of AI is kinda hard?
Thoughts:
release/
branches in a ready state. So basically we run integration tests on release candidates before release.The text was updated successfully, but these errors were encountered: