Next iteration of automated refinement process
Summary of feedback and improvement ideas on current automated refinement process
Context
This issue concludes feedback and discussions from #677 (closed) and introduces few improvements to how our refinement process works.
Requirements
-
Allow manual transitions to workflowrefinement by ensuring that triage bot will apply refinement thread to such issues (not moved by the bot but by user). Suggested solution: apply reactive rule on transition to workflowrefinement by anyone other than triage bot and add refinement thread to such issues (alternative solution: apply it during daily run of the bot, but this will introduce latency and manually transitioned issues usually are time senstivie). (discussion) -
Add a way to block bot from moving issues out of refinement when we deem the issue still has to be further discussed. This is done by adding ❌ reaction to the refinement thread and the bot should check for this reaction when deciding whether the issue is refined or not - if the❌ is applied it skips this issue even if it meets other conditions (like number of estimates) (discussion)
Coming in next iteration...
- Skip refinement on ~"experiment-rollout" issues
- Throttle bot transitioning issues to refinement according to number of issues in workflowscheduling
- Lower evaluating threshold from
3
to2
due to recent team reduction
Edited by Kamil Niechajewicz