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
Fix: in a Codex branch or subagent thread, the parent conversation's transcript is injected as a user message — its assistant replies were being shown as if you had typed them, and were indexed for search. They are now treated as injected context, like the other <environment_context>-style blocks
Update: sessions sorted by newest creation or update time are grouped into Pinned, Today, Yesterday, Earlier this week, and month sections
Fix: the session list now loads additional pages near the bottom instead of stopping after the first batch, and search hits beyond that batch are loaded and selected correctly
Fix: long mixed-language titles now end with a Unicode-width-aware ellipsis in the session list while wrapping in full in the detail header
Update: the project badge in session detail opens its folder, while empty, HEAD, and detached branch labels stay hidden
Fix: session detail now shows the concrete missing-adapter or transcript parsing error instead of a blank reader, with an action to reveal the source file
Update: conversation Markdown now has clearer h1–h4 hierarchy and framed code blocks with language labels, copy actions, and immediate light/dark syntax colors
Update: conversation text has roomier line spacing and window-level selection, including Shift extension across messages and edge auto-scroll while dragging a selection
Update: the destructive session action in the detail overflow menu now uses the danger color for both its trash icon and label
Fix: Codex code-review results now render as a readable review with findings, locations, and confidence instead of exposing the structured JSON payload
Update: Thinking attached to assistant replies is now a collapsible panel with a one-line summary and full expanded content, independently of tool-call expansion
Update: tool calls now use expandable cards with Unicode-width-aware summaries, complete available inputs, successful and failed outputs, and full-content copy actions beyond the 600-character preview
Update: Insights uses 28px overview values, 24px leaderboard rows, 32px section spacing, and one shared 2px radius for chart and heatmap cells
Credits: Several UI and interaction improvements were inspired by PR #12. Thanks to @Tespera for the thoughtful exploration and contribution; the final implementations were independently rebuilt on the current codebase