OpenHands

BTW I notice that when all-hands-bot reviews, the PR no longer requires a human reviewer. Is that intentional?

Yes, that's intentional.

However, I believe there’s an unwritten rule that we should wait for a human to approve the PR, if it involves a major change. :slightly_smiling_face:

Good question.

My rule of thumb is: if I’m making a significant architectural change, i.e., changing internal or exposed APIs, I wait for a human reviewer.

I’ve noticed the `all-hands-bot` is quite good at catching:
• bugs that often slip into large changes.
• Inaccurate or missing docstrings.
• leftovers

If unsure, please feel free to ask for human review. We'll figure it out

The way I think of it, is the other way around: does anyone else need to load into memory and spend time on a workflow tweak, or a model config change used only by benchmarks, or five lines of code that you verified manually? Probably not.

Maybe it makes sense to think of it as an experiment, and if we don't like the results, we can always fix/undo it.