Stale bot 30 days

[marsh cosmos]

The action itself doesn't appear to have that option.

[viscid field]

Yeah I think 30 days is enough 🙂

Ask claude to add it 😛

[little frigate]

I would like to know what our average time to get back to PRs is, because if we put it to 30, then we also need to make sure we are able to consistently get back to PRs within 30 days

[viscid field]

Offtopic: It would help if github would be working reliable 😛 So we could bring the number down faster and without reloadsing

[timid locust]

lets change it to 30. then if we see ourselfs "unstaling" many PRs, it means its too low and we increase it again 😛

[little frigate]

Is there a way to only do 30 for draft PRs?

[timid locust]

also, the bot commenting will pull the PR back to the top of the PR queue

[little frigate]

So one of the things that would help air up the queue a bit was Frenck going through draft PRs and closing them, which I think is okay. (side note, I think we can make stalebot nicer). Would it also help if we figure out/document "guidelines" on which we think we can close PRs? Because I personally always feel bad closing PRs because I feel like I am demotivating people

So like, what if there is a crap AI PR that we pushed back on and that hasn't been returned to in 2 weeks, let's just close it

or if CLA hasn't been signed after 2 weeks and no pushes have been done

(I think we can even automate some of this, by using labels (like the needs-more-information label on issues), which might make it easier)

[timid locust]

before we dive into more suggestions... can we decide on Martin's suggestion? 😅

[little frigate]

I mean we can already gather this data beforehand, because if it is indeed close or higher than 30 days, then it's a no from me 🙂

[timid locust]

not trying to kill your suggestions @little frigate ! sorry if it sounded like that! just want to closee one before opening more 🙂

[little frigate]

Oh don't worry, I just thought about how we could maybe close more efficiently, other than just a 30 day limit

[timid locust]

is it an hard thing to do? changing the bot time? because if not, then why not try it and see how we do, instead of wasting time gathering data -> discussing it again -> etc?

its kinda low-impact anyway

[little frigate]

wasting time gathering data
We can always ask AI to make a report on it. Like I wasn't aiming to do that manually, rather use a data aggregation like ossinsights

[timid locust]

we already seen how good those github data aggregation tools are... 😆

but then again, why not try it? what would be the big downside of it?

[little frigate]

I'd be pretty annoyed if I opened a PR, which didn't get a response for 30 days and then I get a bot asking me if I am still there

[timid locust]

at the same time, it will bump your PR to the top and you may get a review 😄

[crimson jay]

@viscid field could we stale it after 30 days only if it is in draft? I think that approach would be best

[marsh cosmos]

We need some custom action code for that. The stale action we use now doesn't support that by default I gather from reading the readme.

They just have the opposite (ignore draft PRs).

[timid locust]

there are 159 PRs not updated in the last 30 days.
only 22 of those are not draft (I haven't checked if they are pending changes but not drafted).

https://github.com/home-assistant/core/pulls?q=sort%3Aupdated-desc+is%3Apr+is%3Aopen+updated%3A<%3D2026-03-02
https://github.com/home-assistant/core/pulls?q=sort%3Aupdated-desc+is%3Apr+is%3Aopen+updated%3A<%3D2026-03-02+-is%3Adraft

since 22 is quite small in the universe of PRs we have, I would go for it

[viscid field]

That's not something a bot can do. The bot will be triggered by github webhooks/events and scanning for stale PR should be done via a workflow on a schedule. I recommend that we add the missing feature to the stale action