Belated forum post for 0.4.0, but here are the things that changed since our last major release:
- Bring your own key option. Users can input their own Anthropic and OpenAI API keys to code in Rao without ever signing in to Lotas. We’ve also added functionality to connect to on-premise models for sensitive data analysis (contact us if interested). One-click Lotas sign-in still works as before with 50 free queries per month.
- New models. Claude Sonnet 4.5 and GPT 5 are now available.
- New context. Users can attach previous chats and function documentation as context, and models can now choose to retrieve documentation and plots from the plots pane as needed.
- Auto-accept. All model actions can now be put on auto-accept mode for faster editing or code running. Each type of action can have its automation configured individually, and users can specify lists of functions to allow (or deny) the model to run automatically.
- In-line change tracking. On auto-accept mode, changes made by the model are tracked in-line for individual acceptance or rejection. Chat checkpoints allow users to undo any undesired changes or change and resend queries.
- Dark mode. Dark mode and standard themes all work, providing an easy transition from other IDEs.
- Code searching, file editing, and overall speed have been improved.
Let us know what you want to see in the next version!