Kangas

New in this release:
1. Fixed multiple issues in horizontal scrolling: boundary updates, throttle events
2. Suspension boundaries are correctly installed
3. Fetches now use cache
4. Requires flask 2.2 or greater
5. Mask class for easily manipulating masks
6. Importing kangas now loads much faster
7. New download datagrid for hosted services
8. Much faster segmentation mask rendering
9. JsonViewer now has copy button, and shows element sizes
10. Bounding box annotations now show labels
11. Labels now show layer name, if given
12. Added DataGrid.add_iou_columns() for automatically added Intersection Over Union columns
13. Many bug fixes and enhancements

New in this release:
* New annotation types:
* `Image.add_marker()` - "raindrop" and "circle" types
* `Image.add_line()` - draw a line
* `Image.add_mask()` - segmentation-style masks (every pixel is a class id)
* `Image.add_mask_metric()` - heatmap-style mask (every value is a measurement of a metric, 0 - 1)
* Bounding boxes now have labels showing label name and score
* Annotations now have layers
* `DataGrid.add_iou_columns()` - adds "Intersection Over Union" visualization and measurement columns
* New JSON Viewer
* Fixed zoom in expanded image viewer
* New flags for server: `--backend-port` and `--backend-proxy`
* Many bug fixes

Major changes:

* complete frontend rewrite
* many bug fixes
* many UI enhancements
* backend now uses flask (with waitress to serve) rather than tornado by default
* better caching
* backend port is no longer used directly by the browser (eliminates special colab code)
* new integrations with huggingface and Comet
* can now use remotely-stored images
* new annotation format to support layers
* use `kangas upgrade *.datagrid` to update datagrids
* enhanced Python-to-SQL: more syntax supported, new sqlite_query command for any db
* new URL variables, including rows
* better filter command completion, and error indicator

Fixes in this release:
* Fixed numpy deprecated np.float to np.float_
* fixed hyper-link error
* Don't allow thumbnails to get too small
* Strip spaces off of whereExpr in handlers
* Refactor completions; added support for columns of strings, more general
* Query language: handle NOT IN
* Added completions: .values() and .keys(); fixed unprotected {'column'}.split()

New in this release:
1. Updated JSON metadata format
2. New completions (endpoint, UI): Entering "." after a JSON {"column name"} will give JSON object keys
3. Added more test coverage for queries: Python and SQL `in` operator
4. Added .keys() and .values()
5. Deprecate `.contains()`: use `item in item` instead
6. Fixed bug with `.startswith()`
7. Optimization for `dg.select()` (removed unused count query)
8. Don't crash when a HuggingFace dataset can't `.take()`
9. Added `dg.append_columns(dict)`

Fixed in this release:
* Fixed a regression with numpy numbers--they were interpreted as VECTORS rather than floats or ints