IPFS issue triage

1

Hey everyone!

I’m John and me and my co-creator Mike have built an automated GitHub Issue assignment service called Heupr. It helps very active repositories like yours (with 600+ open issues and 100+ contributors) coordinate software maintenance and feature creation.

We reached out to Juan and Victor, who actually raised this point before, and they recommended that we reach out to the community to get your input about using our platform to help your development.

We’re actively looking to help more projects improve how they coordinate bug fixes and develop new features. We’d love to help and get your feedback!

Best,
John

1 Like
2

Thanks for sharing this with us @forstmeier.

Had a small discussion over at https://github.com/ipfs/go-ipfs/issues/4625 but better to continue here.

I think we should try out the automatic labeling feature from Heapr for ipfs/go-ipfs and see how it works.

@Kubuxu seems to have figured out we don’t need write access to our repositories, which is nice.

I don’t see any blockers in at least trying out Heapr and then write any feedback here after a month or so.

1 Like
3

My only worry is: as they use ML for training on existing data, with the mess we have currently in go-ipfs it might not be very helpful but I think it might be worth a shot.

1 Like
4

Hey all!

We’re in the process of preparing to launch the automatic labeler feature - things are looking good and we just need to make a few tweaks to have a working beta available (we’ll also be reworking the site so things will look a bit different).

And quickly touching on the ML (great point, by the way): for the labeler, we only process each individual issue as it’s raised so it would be independent of prior activity taken in the repository issues. The contributor assignment does look at history (since we try to get an idea of who can best handle what); we’re evaluating some changes to this feature to make it more flexible so we’d definitely love input on that as well.

We’ll reach out once we have the labeler ready for use. Thanks so much you guys!

Best,
John

P.S. To answer your question from the GitHub issue comment chain, Victor: currently we do not utilize cross-training in our models but since the labeler takes each issue as an independent event, you can use as many repositories as you feel are necessary with Heupr.

1 Like
5

Hey all!

Our new beta is LIVE! Feel free to check it out at our newly renovated website; you’ll see on the “Console” page (after authenticating) the various options to select the labels you want to indicate bug/improvement/feature. Right now, we’re focusing on just classifying the issues with these categories but we’ll expand the options there as we receive feedback. Let me know if you have any questions! :smile:

Best,
John

P.S. We’ll also be making a few minor tweaks in the coming weeks that we’ll be sure to update you on as well.