YannByron opened a new pull request, #9665: URL: https://github.com/apache/paimon/pull/9665
## Summary Add a LeRobot recording-loop compatible writer that sends captured frames directly to a Paimon multimodal table without materializing an intermediate LeRobot dataset. ## Changes - Add `PaimonLeRobotWriter` with `add_frame`, episode save/discard, explicit flush, and finalize lifecycle methods. - Batch completed episodes into one Paimon commit (10 by default) while keeping episode/global/task indices deterministic. - Encode image features as PNG BLOBs; MP4/video features remain unsupported in this first version. - Document the direct-capture API and validate frames, images, schemas, failure states, and tail-batch commits. ## Testing - [x] `python3 -m unittest pypaimon.tests.multimodal_lerobot_writer_test pypaimon.tests.multimodal_lerobot_test` (56 run, 24 optional LeRobot integration tests skipped) - [x] `python3 -m ruff check pypaimon/multimodal/lerobot/writer.py pypaimon/multimodal/lerobot/loader.py pypaimon/tests/multimodal_lerobot_writer_test.py` - [x] `git diff --check` ## Notes This draft intentionally starts with PNG image frames, synchronous encoding, and new or empty target tables. MP4 encoding, restart/resume state, and streaming-checkpoint semantics are deferred. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
