I would much prefer BugBot in the UI as an extention. This allows us to run it when we want to with more control (sometimes I want to run it twice on a PR, but usually not on every commit), allows for better UI (e.g. the checkbox list), allows for easier immediate edits (i.e. the ‘edit in cursor’ link doesn’t work for me in Linux), doesn’t require us to get integration approval for our github organizations, doesn’t force coworkers to use a tool I am using, removes clutter from the PR, allows us to scan more possible/lower-probability bugs (or risk comments becoming too long), etc.
I’ve really enjoyed this feature, so thanks to the team for making it! Though I would appreciate it moving back to the UI.
I’ll add my voice to the calls to allow BugBot to run locally before I open my PRs. That would reduce PR churn and also allow me to privately fix my dumb mistakes.
I will add my opinion to this, because i cant us it as-is.
I usually make sure my code is functional and has no issues before committing it.
It is only working with GitHub
Would save time testing before commit - if I find a bug towards the end of testing i will have to rerun all tests, this way I could catch as many bugs before testing, replicate them and reduce the loops of testing.
As an update, I’ve migrated to the local version of coderabbit. This has been easier, is free (though I would pay for it), doesn’t force coworkers to use my tooling, and doesn’t have any configuration issues (i.e. I don’t need to get permission from my organization to enable it, etc).
The link issue is probably specific to my computer/distro (it’s a problem with all “Open link in specific app and not a brower” links I use).
Hi friends we’ve added Agent Review, it’s available in the latest Nightly build (still beta, please feel free to report bugs to me). There are two entrypoints: in the source control panel there’s a new “Agent Review” tab, and when you’re in the “Review Changes” UI from a chat, there’s a button at the top.
I agree with Martin. I have tried using prompts to make the cursor act like Bugbot, but it seems they do not encounter the same issues. Kind of like playing ping-pong, your code is refactored, working, and tested, but bugbot finds issues, the cursor does not. It would be nice to fix these BEFORE I create a PR.