This is the polish-and-visibility follow-up to v5.28.0: better model
selection, a sidebar that stays legible under real load, and faster ways to
get back to recent work.
Added
- The model catalog now includes current OpenAI Codex pricing and plan limits,
alongside the Claude, OpenCode, Factory Droid, and Devin information added
across the recent model-catalog work. - CCC self-checks the dashboard every five minutes and surfaces its health,
so a silent local failure is easier to catch.
Changed
- Sidebar row styles are now intentionally adjustable: choose tighter or more
spacious rows, wrap long titles when they matter, and use divider, zebra, or
card delineation. The selected row now has a distinct purple tint instead
of looking like an ordinary hover state.
Fixed
- Sidebar search no longer launches Total Recall on every keystroke. It scans
recent local transcripts in-process across Claude, Codex, Kimi, Gemini, and
Cursor, and keeps results stable while you type. - Compact, grouped, and Cards rows now reserve the same icon gutter; long
titles and outcome text no longer overlap or drift right of the content. - The worker automatically retires stale or incompatible code once idle,
instead of remaining degraded until the next full dashboard restart. - Claude auto-titles work again after the CLI changed its
--mcp-config
handling; Kimi titles also drop the injected continuation boilerplate.
Full changelog: v5.28.0...v5.29.0