github alexgreensh/token-optimizer v2.4.5
v2.4.5 - Plan Mode + Context Fill Fixes

latest releases: openclaw-v2.4.1, v5.6.4, v5.6.2...
one month ago

Includes v2.4.4 + v2.4.5 fixes:

v2.4.5: Plan Mode Reset

  • ExitPlanMode now treated as a context reset boundary, just like compact_boundary
  • Quality signals reset after plan mode exit so scores reflect the clean post-plan context
  • Previously, stale signals from before plan mode dragged scores down

v2.4.4: Per-Session Quality Scoring

  • Signal accumulators reset on compact_boundary markers (score reflects current window)
  • Fill estimation: live-fill sidecar bridge from statusline.js (ground truth) with char-length fallback
  • PostCompact hook forces immediate quality cache refresh (uses Claude Code v2.1.76)
  • Removed dead total_input_tokens code path
  • Fixed potential UnboundLocalError for density_ratio

Don't miss a new token-optimizer release

NewReleases is sending notifications on new releases.