engram
Dashboard Graph Inbox Context Preview Wisdom
Design Mockup 中文 ← All pages

Wisdom Metrics

Four quantitative curves showing whether the system improves over time.

Regression alert: Context Efficiency dropped 11% week-over-week (2026-04-11 → 2026-04-18). Possible cause: new large reference assets loaded without budget adjustment. Check the Relevance Gate's dynamic budget allocation.
Workflow Mastery
pass_rate per workflow over autolearn rounds
0.60 0.75 0.85 0.95 Apr 1 Apr 8 Apr 15 Apr 22 Apr 30 release-checklist deploy-pipeline incident-response
Task Recurrence Efficiency
weekly cost ratio: repeated tasks vs. first-time cost
0% 30% 60% 90% Apr 1–7 Apr 8–14 Apr 15–21 Apr 22–28 cheaper than first time same cost first-time (higher cost)
Memory Curation Ratio
stacked area: active / deprecated / archived over time
0 100 300 500 Apr 1 Apr 8 Apr 15 Apr 22 Apr 30 active deprecated archived
Context Efficiency
rolling 7-session avg: fraction of packed tokens that LLM actually used
Regression: -11% WoW
40% 60% 80% 95% Apr 1 Apr 8 Apr 15 Apr 22 Apr 30 ↓ -11% WoW

Drill-down: Workflow Mastery

Source data breakdown — 30d
Workflow pass_rate Δ
release-checklist 0.94 +0.20
deploy-pipeline 0.88 +0.13
incident-response 0.76 +0.08
Contributing events (sample)
2026-04-18 · release-checklist r8 accepted
pass_rate +0.030 · event-id: ae-2026-04-18-042
2026-04-16 · release-checklist r7 accepted
pass_rate +0.030 · event-id: ae-2026-04-16-031
2026-04-15 · release-checklist r6 rejected
pass_rate -0.040 (auto-reverted) · event-id: ae-2026-04-15-028
2026-04-13 · deploy-pipeline r4 accepted
pass_rate +0.045 · event-id: ae-2026-04-13-021