Highlights:
- Support for video input
- Updated logo file
- Improved support for arcanas
- Improved markdown and LaTeX rendering
- Added support for more file type attachments
- Added support for PDF processing using docling
- Added image and text attachment previews
- Smoother response display
- Improved support for multiple conversations
- Added edit button for responses
Highlights:
- Support for arcanas
- Support for multiple conversations.
- Added profile window
- Improved LaTeX rendering with simplified code and markdown fixes.
- UI Support for scale-to-zero models.
- Model status popups for unavailable models.
- Code copy button added for responses.
- UI fixes: Responsive width, improved profile pop-ups, and settings flow.
Fixes:
- Retry button path fixed.
- LaTeX bugs resolved and libraries optimized.
- Markdown handling improved.
- First name typo corrected.
Features:
- Token limit removed: No more restrictions on the number of tokens.
- Image upload feature: You can now upload images with an exportable option. Images can be added by copy-pasting from the clipboard or via drag-and-drop functionality.
- CSV upload: Users can now upload CSV files directly.
- Model status indicator: Added a point system that shows the status of models (e.g., active, loading).
- Temporary model execution: Users can run available models temporarily, even if they are not active.
UI Updates:
- Clipart addition: Added clipart of an image near the model selection section to indicate where image uploads are allowed.
- Updated models API endpoint to /models
Features:
- Sharing model and settings with base64-encoded URL parameter
- Importing external settings, i.e., personas
UI Updates:
- Made scrollbar for model selection more visible
- Fixed header issue on tablet-size screens
- Improved overall design of options section
- Memory leak resolved: Issue caused by using
rehype-highlightwas fixed by switching toreact-syntax-highlighter. - Removed unwanted libraries: Unnecessary libraries were removed to improve performance and reduce bundle size.
- Zoom down to 90%: Adjusted view for better user experience and interface visibility.
- Minor UI changes:
- Resolved clear history toast issue.
- Removed edit button from the system prompt UI.
- Added "Do not show again" option to clear history.
- Top_p and Temperature slider added:
- A slider for
top_pallows users to adjust the total probability distribution for next token selection, similar to temperature. - Fixed and improved sliders for
temperatureandtop_pto allow for easier customization.
- A slider for
- Custom configuration page integrated: Separate page for custom configuration of
temperatureandtop_pmerged into the main page for a seamless experience. - Export and Import functionality: Users can export the settings for
temperatureandtop_pas a JSON file. Imported settings will automatically configure the interface.
- PDF beautification completed: Improved the look and layout of exported PDF files.
- Endnotes with export: PDF and TXT exports now include
temperature,top_pand model name in the endnotes.
- Prompt resending/editing: Users can now resend or edit prompts for better control.
- Unwanted announcement bar removed: Cleaned up the interface by removing unnecessary elements.
- Bug fixes
- Polished UI
- Decoupled proxies
- Service announcement
- Bug fixes and cleanup
- Security patch for proxy
- Conversation save/load bug fixes
- Improved error messages
- Bug fixes
- Minor UI cleanup
- Website now running in production mode (much faster)
- Better error handling on frontend and backend
- Bug fixes
- UI changes, progress loading
- Model names fixed
- Bug fixes
- Added initial DB file
- Ask for mic permission
- Cleaner UI
- Readable model names
- Rebranded as Chat AI
- Responsive design
- Light/dark mode
- English/German interface
- Change system prompt for internal models
- Change temperature for internal models
- Load/save chat history
- API service
- Admin visualization with Prometheus
- Dump/restore database