Decision Adaptations
Overview
Decision adaptations let a matched decision control whether global Router Learning can adjust its proposed model.
Most decisions inherit global learning behavior. Add adaptations only when a
decision needs a hard boundary, observe-only rollout, or small protection
tuning.
Key Advantages
- Keeps policy boundaries close to the decision that owns them.
- Lets sensitive routes bypass learning with one small block.
- Supports observe-only rollouts before adaptation or protection can affect traffic.
- Lets one decision use a narrower or broader adaptation candidate set than the global default.
- Lets one decision tune the stability trade-off without changing global defaults.
What Problem Does It Solve?
Global learning is convenient, but not every decision should be adjusted by online state. Privacy, local-only, security, compliance, and operational routes often need hard boundaries. Decision adaptations give the matched decision the final say on whether learning may apply, observe, or bypass.