| Maybe there are ways to customize coderabbitai, like to not show nit-picks, not be overly friendly and spam replies like "oh thank you". Even customize by users. For some maintainers, only show super critical findings, more nip-picking. This is more useful for new contributors than it is for veterans. Let's see, let's have it on demand but use it diligently, for any large PR with many moving parts and complexity for the reviewers, it can save time.
We use the copilot reviewer at my work, and it used to give quite good feedback. It has gotten noticeably worse over the past few months, though, to the point where it's just flat-out wrong on very obvious things now. At best, it just corrects our spelling in comments. They're always tinkering with the models, though, so who knows what it will say next week. That's part of the fun, haha!
Scott
Sounds good; agree on all points, especially copilot. :P (I suspect
that will get better pretty rapidly, but coderabbit is much much
better at the moment)
On Sat, Oct 25, 2025 at 1:05 PM Victor Lazzarini
<000010b17ddd988e-dmarc-request@listserv.heanet.ie> wrote:
>
> That's fine, we just have to trigger it as necessary. And yes it flagged up stuff, most of it typos, but useful anyway. Some of the analysis involving more complex things were spurious in the cases I looked at. It needs careful revision so we don't end up doing the wrong thing. Certainly it's better than co pilot, which gave us very little. Since we don't have static analysis, this may help find incorrect stuff.
>
> best
> Prof. Victor Lazzarini
> Maynooth University
> Ireland
>
> > On 25 Oct 2025, at 17:57, Steven Yi <stevenyi@gmail.com> wrote:
> >
> > Well, I'll probably end up triggering it on a number of PRs I review,
> > especially if the PR is very big in scope. I see it's already caught a
> > number of issues so it's useful. It also does learn from previous
> > mistakes so replying to it helps it perform better later.
> >
> >> On Sat, Oct 25, 2025 at 12:48 PM Victor Lazzarini
> >> <000010b17ddd988e-dmarc-request@listserv.heanet.ie> wrote:
> >>
> >> It's very disruptive to me, I was trying to do a simple job and it made the whole thing a 3-hour job as it kept distracting me.
> >>
> >> I don't mind if it is run like co pilot, invoked manually. Otherwise it'll drive me nuts everytime I make a PR.
> >>
> >>>> On 25 Oct 2025, at 17:42, Steven Yi <stevenyi@gmail.com> wrote:
> >>>
> >>> Coderabbit can be a bit noisy but I generally prefer it on by default
> >>> to get a first pass before doing human PR review.
> >>>
> >>>> On Sat, Oct 25, 2025 at 11:25 AM Victor Lazzarini
> >>>> <000010b17ddd988e-dmarc-request@listserv.heanet.ie> wrote:
> >>>>
> >>>> Never mind, I’ve done it myself. No auto reviews please.
> >>>>
> >>>>>> On 25 Oct 2025, at 15:41, Victor Lazzarini <000010b17ddd988e-dmarc-request@LISTSERV.HEANET.IE> wrote:
> >>>>>
> >>>>> *Warning*
> >>>>>
> >>>>> This email originated from outside of Maynooth University's Mail System. Do not reply, click links or open attachments unless you recognise the sender and know the content is safe.
> >>>>>
> >>>>> Can we make code rabbit on request? I am getting tons of review stuff that’s sometimes irrelevant.
> >>>>> It’s very annoying to make a PR and get this stuff thrown at you on email.
> >>>>>
> >>>>> Thanks
> >>>>>
> >>>>> Victor
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>
|