DmitriyAlergant-T1A
commented
2 months ago - Fix disappearing chat header (use absolute layout) on scrolling
- Persist new message drafts per-chat
- Refactoring of some confusing code (related to message submission, token calculations, etc), improved error handling
- Fix chat header on narrow screens and prioritize showing of Model
- GPT-4 (non-turbo) is made obsolete (price/performance) and removed from selection. Changed model descriptions verbiage.
- Improve handling of the "Generating" state in the UI: actually disable buttons that should not work in that state
- Show toast warning when chat context was truncated for submission
- Hidden Presence and Frequency Penalty settings to improve visibility of the Confirm button on smaller screens
- Chat model settings window: Improved Input/Completions Tokens Slider behavior, added numeric input
- Auto chat title generation: limit submitted content to conserve tokens and avoid issues with models (context length)
- Clear error box on selecting (or starting) another chat
- Other smaller bug fixes and enhancements