AI Reviews and use of AI for JBang code

[maxandersen]

Trying out AI to help our code, but let's keep the noise dialed down!

Yesterday, just for fun and giggles, I enabled Sourcery.ai on the JBang repositories to see how well it would perform.

I have to say—I actually like it! It found things I or others didn’t spot, and its summaries are surprisingly helpful too. 🥳

BUT... then we hit cases like this pull request, which was implemented using Brokk AI, and suddenly we get walls of text and questions. 🙈

In this case, everything worked out well, and I’m still positive overall of having the option to aid by AI —but I can also see how this could get annoying for contributors if it’s on by default.

Unfortunately, Sourcery.ai doesn’t seem to have a way to be added as a reviewer explicitly, but it does honor labels. So, for now, I’ve set it up so it won’t start doing reviews unless you add the ai-review label (or use one of the @sourcery-ai commands).

That way, we get the best of both worlds—some AI help without overwhelming contributors with noise.