Feedback & Approval
Operators review automation drafts, approve, reject, or modify them. Each decision adjusts confidence scores and drives the pattern lifecycle.
Pattern Lifecycle
graph LR
D["Discovered"] --> S["Suggested"]
S --> A["Approved"]
A --> DEP["Deployed"]
S --> R["Rejected"]
R -->|"3× = permanent"| EX["Excluded"]
Feedback Decisions
| Decision | Confidence Effect | Status Change |
|---|---|---|
| Approved | × 1.2 (capped at 1.0) | → Approved, rejection_count reset to 0 |
| Modified | × 1.2 (counts as approval) | → Approved, operator edits tracked |
| Rejected | × 0.5 | → Discovered (or Rejected if count ≥ 3) |
Confidence Formula
confidence = (support / total_sessions) × recency_decay × completion_ratio
- support: Number of sessions where pattern appeared
- total_sessions: Total sessions in the mining window
- recency_decay: exp(-age_ms × ln(2) / 7_days_ms) — 7-day half-life
- completion_ratio: Fraction of pattern steps completed in sessions
Permanent Exclusion
After 3 consecutive rejections (MAX_REJECTIONS), a pattern is permanently excluded from future suggestions. Its status moves to Rejected and it will not be re-surfaced.
Questions?
Reach out for help with integration, deployment, or custom domain codecs.