New Providers

NVIDIA API support — connect to z-ai/glm4.7 (free endpoint) via /connect nvidia
Model picker (/models) now shows only models relevant to the active provider
Markdown Rendering
Fixed headings rendering properly (no more raw ## / ### text)
Tables now render correctly with ENABLE_TABLES parser option
Dynamic word-wrapping based on terminal width — no more overflow
Code blocks and horizontal rules auto-fit to available space
Scrolling
Replaced ratatui scrollbar with tui-scrollbar for fractional thumb rendering
Smooth auto-scroll: follows bottom on new messages, stays put when you scroll up
Re-enables auto-scroll when you scroll back to the bottom
File & Editor
@ mention file picker with badges in chat messages
File tools: read, create, and edit files in EDIT mode
Built-in vim-style editor with syntax highlighting and file tree (ctrl+e / tab)
Other
NVIDIA connection status now notes endpoint speed difference vs Cerebras
Improved layout with equal left/right padding
Website updated with screenshots and correct links
Usage

Connect to NVIDIA (free)

/connect nvidia nvapi-your-key

/connect cerebras your-key

/models