You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Improved streaming refresh performance during long conversations, reducing UI re-render pressure when models output quickly.
Fixed memory growth after heavy Subagent use: completed task execution history could be duplicated and kept accumulating.
Smoother typing in long sessions by cutting unnecessary re-renders of the history area and input editor.
Fixed Agent runtime state lingering in memory after deleting long conversations.
Faster timeline rendering for long chats and Agent streaming: history now uses a more stable row-level layout, with less churn on older messages while a reply is generating; Chat and Quick Ask history paths are also improved for input, retry, branching, and tool results.
🚀 Startup
Reduced initialization work at plugin startup (#425).
✨ Experience Improvements
Removed misleading skill-read prompts when all Skills are disabled, so the model is no longer nudged to read non-existent skill paths.
Improved Tab completion context so the model better understands cursor position and surrounding text; more leading context is kept so suggestions fit headings, paragraphs, and prior context more naturally.
Removed the empty-state card shadow that showed as clipped gray edges on narrow mobile screens (#430).
Fixed some OpenAI-compatible models rejecting plain-text-only messages: text-only requests now fall back to a more compatible legacy string format without affecting image or PDF inputs.
Better Obsidian pop-out window behavior: fixed the Agent settings dropdown and opening chat in a right split incorrectly jumping back to the main window (#432) (#434).