2.19.0
This is a significant release that implements LLM response streaming in Jupyter AI along with several other enhancements & fixes listed below. Special thanks to @krassowski for his generous contributions this release!
Enhancements made
- Upgrade to
langchain~=0.2.0andlangchain_community~=0.2.0#897 (@dlqqq) - Rework selection replacement #895 (@dlqqq)
- Ensure all slash commands support
-h/--help#878 (@krassowski) - Add keyboard shortcut command to focus chat input #876 (@krassowski)
- Implement LLM response streaming #859 (@dlqqq)
- Add Ollama #646 (@jtpio)
Bugs fixed
- Fix streaming in
HuggingFaceHubprovider #894 (@krassowski) - Fix removal of pending messages on error #888 (@krassowski)
- Ensuring restricted access to the
/learnindex directory #887 (@krassowski) - Make preferred-dir the default read/write directory for slash commands #881 (@andrewfulton9)
- Fix prefix removal when streaming inline completions #879 (@krassowski)
- Limit chat input height to 20 lines #877 (@krassowski)
- Do not redefine
refreshCompleterStateon each render #875 (@krassowski) - Remove unused toolbars/menus from schema #873 (@krassowski)
- Fix plugin ID format #872 (@krassowski)
- Address error on
/learnafter change of embedding model #870 (@srdas) - Fix pending message overlapping text #857 (@michaelchia)
- Fixes error when allowed or blocked model list is passed in config #855 (@3coins)
- Fixed
/exportfor timestamp, agent name #854 (@srdas)
Maintenance and upkeep improvements
- Update to
actions/checkout@v4#893 (@jtpio) - Upload
jupyter-releaserbuilt distributions #892 (@jtpio) - Updated integration tests workflow #890 (@krassowski)
Contributors to this release
(GitHub contributors page for this release)
@3coins | @andrewfulton9 | @brichet | @dannongruver | @dlqqq | @JasonWeill | @jtpio | @krassowski | @lalanikarim | @michaelchia | @pedrogutobjj | @srdas