Degradation Intelligence
Your AI is getting dumber and you can't see it. v2.4.0 tracks the quality decay.
New Commands
| Command | What You Get |
|---|---|
quick
| 10-second health check: context fill, degradation risk, top offenders, model recommendation |
doctor
| Installation score out of 10 with exact fix commands |
drift
| Side-by-side comparison vs last snapshot, catches config creep |
quality
| 7-signal analysis of your live conversation quality |
Quality Scoring (7 signals)
Context fill, stale reads, bloated results, compaction depth, duplicates, decision density, agent efficiency. Weighted score maps to degradation bands:
- Green (<50% fill): peak quality zone
- Yellow (50-70%): degradation starting
- Orange (70-80%): quality dropping
- Red (80%+): severe, consider /clear
1M Context Support
Auto-detects 1M context windows. At 1M, startup overhead is only 5-7%, but the MRCR degradation curve (93% at 256K, dropping to 76% at 1M) is what gets you. Token Optimizer tracks this.
Visual-First README
SVG diagrams replace code blocks. Session lifecycle, quality breakdown, status bar progression, all rendered as terminal-style visuals.
Smart Compaction
Checkpoints decisions, errors, and agent state before auto-compact fires. Restores what the summary dropped.
Zero context tokens consumed. Runs as external Python. No dependencies.
python3 $MEASURE_PY quick # start here