[!WARNING]
This is a beta release
Improvements
- Additional AI metadata — Added more metadata for AI-related function runs (#3629)
- Reset
die after parallelism ends — Ensures the die signal is properly reset when parallel execution completes (#3717)
- Trigger compaction on delete by ID — Compaction is now triggered when items are deleted by ID, and compaction is no longer subject to context cancellation (#3720, #3722)
- Shard lease acquisition on startup — Attempt to grab a shard lease before kicking off the ticker to avoid unnecessary work (#3719)
- Exclude stale custom concurrency keys — Custom concurrency keys that are no longer present in the function config are now excluded (#3706)
- Peeked item metrics — Started collecting metrics for peeked items in the queue (#3727)
- Capacity lease optimization — Disabled capacity lease acquisition on backlog refill for improved efficiency (#3736)
- Reduced logging noise — Removed excessive logging, debug-level pause deletion messages, and noisy requeue messages for connect worker capacity errors (#3734, #3708, #3730)
- New Run Details view toggle — Added a toggle to switch between old and new Run Details views (#3711)
Bug Fixes
- Fix throttle constraints — Resolved an issue where throttle constraints were not applied correctly (#3740)
- Fix peeked item metrics — Corrected successive count logic, division by zero errors, and clock consistency issues in peeked item metrics (#3728)
- Fix constraint check source — Added the correct source for constraint checks (#3705)