2.3.0
Hey Jupyternauts! We're excited to announce the 2.3.0 release of Jupyter AI, which includes better support for Anthropic models and integration with Amazon Bedrock.
There is also a significant change to how Jupyter AI settings are handled (see #353). The most significant changes are:
- API key values can no longer be read from the client. This was taken as a security measure to prevent accidental leakage of keys. You can still update existing API keys if you do decide to change your key in the future.
- The settings can not be updated if they were updated by somebody else after you opened the settings panel. This prevents different users connecting to the same server from clobbering updates from each other.
- There is now a much better UI for updating and deleting API keys. We hope you enjoy it.
Updating to 2.3.0 shouldn't require any changes on your end. However, if you notice an error, please submit a bug report with the server logs emitted in the terminal from the jupyter lab process. Renaming the config file $JUPYTER_DATA_DIR/jupyter_ai/config.json to some other name and then restarting jupyter lab may fix the issue if it is a result of the new config changes.
Enhancements made
- Adds chat anthropic provider, new models #391 (@3coins)
- Adds help text for registry model providers in chat UI settings #373 (@JasonWeill)
- jupyter_ai and jupyter_ai_magics version match #367 (@JasonWeill)
- Config V2 #353 (@dlqqq)
- Add E2E tests #350 (@andrii-i)
Bugs fixed
- Upgraded LangChain, fixed prompts for Bedrock #401 (@3coins)
- Adds chat anthropic provider, new models #391 (@3coins)
Maintenance and upkeep improvements
Documentation improvements
- jupyter_ai and jupyter_ai_magics version match #367 (@JasonWeill)
Contributors to this release
(GitHub contributors page for this release)
@3coins | @andrii-i | @dlqqq | @JasonWeill | @krassowski