Fix: Batch watcher re-queue loop
Fixed a bug where the batch watcher reprocessed the same conversations indefinitely due to an invalid parameter in the state-tracking call.
What happened
_trigger_batch()calledadd_imported_file(metadata=...)but the method doesn't acceptmetadata- The resulting TypeError was caught by a generic exception handler that re-queued all files
- This created a loop: batch runs successfully, state tracking fails, files re-queued, repeat every ~15 minutes
What changed
- Use valid
importer="batch"parameter instead of invalidmetadatakwarg - Separated batch execution from state tracking — a state-tracking failure after a successful batch no longer re-queues files
- Added explicit
TimeoutExpiredhandling
Files
src/runtime/batch_watcher.py
Full Changelog: v7.1.14...v7.1.15