v0.1.6

v0.1.6 (2026-04-06)

Removals, Deprecations and changes

Bug Fixes

• Fixed embargoed dandiset support by using asset.client instead of creating a new unauthenticated DandiAPIClient(), and by preserving pre-signed S3 query parameters (strip_query=False). PR #18

Features

• Added local_example_notebook.ipynb and appropriate mock nwb file functions to demonstrate usage of the widgets in a local Jupyter environment without requiring DANDI access. PR #21
• Added codec validation for local video and pose widgets. Videos using codecs not supported by browsers (e.g. MJPEG, mp4v, FFV1) now raise a clear ValueError with the detected codec and an ffmpeg command to re-encode to H.264. Detection is pure Python with no new dependencies. PR #24

Improvements

• Replaced OpenCV-generated synthetic test videos with committed stub videos from DANDI (H.264, MJPEG, mp4v) at 160x120 resolution. PR #24
• Added comprehensive unit tests for PoseEstimation widgets covering discovery, widget creation, lazy loading, keypoint colors, error handling, and video mapping. PR #16
• Added support for Pose Estimation Objects anywhere in the NWB file PR #17
• Updated readme to describe supported video codecs PR #25
• Added support for starting time and rate PR #27
• Optimized init-time metadata loading to use indexed timestamp access instead of loading full arrays, improving widget creation speed for DANDI streaming with large datasets PR #32
• Vectorized pose coordinate conversion from row-by-row Python loop to numpy bulk operations, reducing processing time for large datasets PR #32