You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Administrators can request learning runs explicitly, but repositories with new review feedback or unfinished work otherwise wait until someone notices and starts another run. Operators need an optional daily path that keeps learning current without taking capacity ahead of user requests.
What changes
Hub can now opt in to a daily 02:00 UTC learning discovery schedule. It runs only while the saved Copilot connection is usable, finds accessible repositories with review changes or unfinished comments and suggestions, and queues the least recently run repositories first.
Automatic work stays on the scheduled queue. Explicit requests remain on the on-demand queue, can promote a queued automatic run for the same repository, and take precedence before automatic execution claims the existing global learning lock. Recovery returns interrupted runs to their original queue, and enqueue or snapshot-drift failures remain visible in run history.
Completed run history is pruned after 90 days while retaining each repository's latest run and latest successful scheduling watermark. Learning consent, evidence import, labeling, consolidation, and package policy remain in the existing run processor. Operators enable discovery with REPLICA_LEARNING_AUTOMATIC_RUNS=true; disabled discovery still performs run retention without initializing learning or Copilot integrations.
The reason will be displayed to describe this comment to others. Learn more.
🟡 Changes recommended
A queue-ownership race can send promoted user requests to the wrong worker, while repository-specific failures can repeatedly block scheduling for later repositories.
While a long user-requested run holds the global lock, every queued automatic job retries at the fixed 30-second interval. With a full batch, normal priority deferral repeatedly opens database connections, logs failed attempts, and consumes each job's retry budget. Use a non-failing delay for expected contention, or dispatch only one automatic run at a time.
Reflect automatic-run availability in the Learnings status
With the new switch left at its default false, the Learnings page still says “Hub looks for learnings once a day” whenever Copilot is connected (client/application/hub/view-models/settings-learnings/index.ts:52–56). Expose automatic-run availability in the Learnings response and use it to choose the status message, so default deployments do not promise work they never schedule.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Why
Administrators can request learning runs explicitly, but repositories with new review feedback or unfinished work otherwise wait until someone notices and starts another run. Operators need an optional daily path that keeps learning current without taking capacity ahead of user requests.
What changes
Hub can now opt in to a daily 02:00 UTC learning discovery schedule. It runs only while the saved Copilot connection is usable, finds accessible repositories with review changes or unfinished comments and suggestions, and queues the least recently run repositories first.
Automatic work stays on the scheduled queue. Explicit requests remain on the on-demand queue, can promote a queued automatic run for the same repository, and take precedence before automatic execution claims the existing global learning lock. Recovery returns interrupted runs to their original queue, and enqueue or snapshot-drift failures remain visible in run history.
Completed run history is pruned after 90 days while retaining each repository's latest run and latest successful scheduling watermark. Learning consent, evidence import, labeling, consolidation, and package policy remain in the existing run processor. Operators enable discovery with
REPLICA_LEARNING_AUTOMATIC_RUNS=true; disabled discovery still performs run retention without initializing learning or Copilot integrations.