Skip to content

v2.1.0

Choose a tag to compare

View all tags
@changhansung changhansung released this 27 Jul 07:20
v2.1.0
0db4cde
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Highlights

  • Auto-Pause & Wake — Instances automatically pause when idle and wake on incoming messages. Configure via defaults.auto_pause_after or Settings.
  • Live Instance Status — Idle / Working / Stuck detected in real-time with event-driven pane monitoring (92% fewer CPU calls).
  • Grok Build backend — Full support for Grok Build CLI. Auth via x.ai OAuth, crash recovery, context display.
  • /model command — Switch AI models interactively (TG keyboard / DC menu). Reads live model list from each CLI. Updates fleet.yaml and restarts.
  • Settings Redesign — 3-level UI (Basic/Advanced/Developer). ClassicBot channels editable. Unified Apply bar.
  • /pause /wake commands — Manually pause/wake any instance from TG or DC (admin only).

Improvements

  • Shared Logger: fleet process memory -78%, threads -80%
  • Warm cap LRU eviction: auto-evict oldest idle instance when warm count exceeds limit
  • Adaptive startup concurrency: scales to available RAM on fleet start
  • Startup CLI-env probe: version and model list cached on startup
  • One-shot schedules (at: datetime) and silent schedules (no channel message)
  • Kiro UI mode: kiro_ui: legacy/tui/v3 per-instance
  • Error notifications for login expired and model unavailable
  • Cross-instance messages wait for target idle before delivery
  • General knowledge skills updated and trimmed

Fixes

  • Kiro timeout: no longer silently stops notifying after first occurrence
  • Claude resume: now selects full session (no context compression)
  • Classic chat log timezone: fixed midnight freeze (00:00-08:00 UTC+8)
  • Classic chat log dedup: current message no longer duplicated in context
  • Dashboard, export, daily-summary: all timestamps now local timezone
  • Error monitor recovery gate: 5-minute timeout prevents permanent suppression
  • restart_instance / start_instance support for ClassicBot instances
  • Dead process no longer shows false Idle state
  • agy deprecated model: auto fresh-restart on 'unknown model key'
  • agend-agent reply supports file attachments
  • General instance exempt from auto-pause

Breaking Changes

  • /start requires explicit backend selection (no more 60s timeout fallback)
  • /api/fleet includes ClassicBot instances (additive, filter classic: false if needed)
  • fleet.log timestamp format: [YYYY-MM-DD HH:MM:SS] (was time-only)

Full Changelog: v2.0.12...v2.1.0

Assets 2
Loading