Hi! We've been experiencing some difficulties with the

pr-review-bot

, so I made this *umbrella issue *to track all pain points. Once it stabilizes we can break it down into sub-issues and assign them. Please give it a look and dont hesitate to post comments if you have anything to add 🙂 @modern-fireman-99755 @limited-student-10747

Thanks, I noticed this too yesterday and started trying to fix it, taking a closer look now.

Can I also propose to divide the logic we use for OSS projects and the PR rview bot? Because we probably want to extend the PR review bot also to other repos which are not open source.

Could you elaborate a little bit Vasco? Why do we need to divide the logic?

For example @fierce-account-55183 expressed interest to hav the review bot in some other repos. But perhaps they don't use the HUMAN section as we do. So was thinking to divide the logic. But actually i think that it is better to restrict the options we have in the review bot (see the following paragraph) to a set of selected repos. FYI this is the logic now: 1. If all-hands-bot is a requested reviewer, run a full AI review unconditionally. This overrides every gate below. 2. Else if the description has no filled-in HUMAN section, post a plain comment, edited in place rather than reposted. 3. Else if CI is red on the head commit, post a plain comment, once per head SHA. 4. Else if the author is an org member, collaborator, or has at least three merged PRs, run a full AI review. 5. Else ping the repo's Slack channel that the PR needs a human, at most once per PR. @strong-ice-84145 I will start to fix some point of your investigation 😉

Could we use GitSync to “export” the Automations to git, and then

enterprise

can pick it up and change it to its needs?

Yeah, we’re exporting automations to git, but not publicly at the moment, because often the automations are vibe-coded and not high quality. Vasco has been moving them over to OpenHands/extensions

I'd add also that the core Automations here were PR Review/etc that used to run in GitHub Actions in Open Source repos, which we converted to Automations.

@modern-fireman-99755 :

For example @fierce-account-55183 expressed interest to hav the review bot in some other repos. But perhaps they don’t use the HUMAN section as we do. So was thinking to divide the logic.

I think we should try to make the core automation extensible so that it can take this into account. Already, I’m trying to use the custom review guide that is included in the repo to give the guidelines

^^ The custom review guide is already supported, yes

@calm-artist-7206 They don't have the same PR template summary that we are using. So they don't need to fill an HUMAN section

But then CI fails for us, and doesn’t for them, so isn’t that the criterium?

sorry I don't get it.

If the rule is: • the automation runs the conversation if CI is green Then • CI is green on Open Source repos if HUMAN: is filled, along with other criteria • CI is green on

enterprise

repos regardless of HUMAN:

Sorry, I’m not sure what the big deal is here? Could you elaborate?

Of course it’s important , I’m saying they can always run a copy, like everyone else. A clone. 🙏 Instead of directly editing the same Automation run running on Open Source repos. Sorry, directly editing is not the same with just applying, it seems to me. Then if they need changes, they are always welcome to make a PR, like everyone else. So we can see the proposals.

We're editing the automation to make it more general though (make it respect the repo specific review criteria stored in the repo, instead of baking in repo specific criteria directly into the automation code). That modification just seems like good software engineering practice, it seems to me?