Suggested Reviewers - Internal Beta Feedback
Hi everyone, thanks for all your feedback on this internal beta feature, it's helped give us specific examples of where things go wrong.
We'll be transitioning existing internal projects to our new UI integrated public beta which also has an updated model which addresses much of the feedback shared here. The biggest change with this new UI integration is recommendations will no longer be made via comment and will show up in the reviewer dropdown.
- Release announcement: https://about.gitlab.com/releases/2022/09/22/gitlab-15-4-released/#suggested-reviewers-open-beta
- Internal docs: https://docs.gitlab.com/ee/user/project/merge_requests/reviews/#suggested-reviewers
===== OLD CONTENT ======
Please use this issue for any feedback related to Suggested Reviewers
You may have noticed a new beta feature on some internal projects: Suggested Reviewers.
Feature screenshot |
---|
We've enabled it to dogfood on internal merge requests for the handbook (www-gitlab-com) and soon to be enabled for Gitlab source (gitlab-org). This feature uses machine learning
This feature is currently in INTERNAL beta, though customers can express interest here. We'll be iterating on this feature before expanding to customers during the beta.
Currently, this feature automatically adds a comment with Suggested Reviewers. In a future release, this will integrate with the MR Reviewer UI.
If you have feedback we welcome thoughts/suggestions on this feedback issue. If you'd like it enabled on other internal projects drop us a note in #g_applied_ml
Known Issues
- Occasionally the bot will recommend a reviewer who is no longer an active user (left the company, etc). We plan to have the bot factor in suspended accounts and inactive users in an upcoming release
- Suggested Reviewers does not have context of GitLab specific development practices (team assignments, workflows, etc) We are trying to build a generic recommender, not something specific to GitLab corporate as we use GitLab differently than our customers.
- Suggestions can be slow. We're in early days of building our model and triggering it. We plan to make this faster.
- MR author or assignees can be suggested as reviewers