Recommendation Engine

Stable

Applies to: SmartBlocks recommendation pipeline

Introduced: Shared registry + governed apply model

Updated: Documentation v1

The Recommendation Engine translates measured signals into deterministic, evidence-backed actions using a shared key registry and governed apply handlers.

Summary

Recommendations are defined in smartblocks_recommendation_registry, surfaced with evidence, and optionally applied through explicit safe handlers.

Use cases

Prioritize optimization by impact and confidence
Keep actions explainable and auditable
Avoid ad-hoc recommendation engines

How it works

Signals are mapped to recommendation keys, registry definitions provide labels/categories/handlers, and governed apply executes allowed changes with logging and snapshots.

Inputs

Readiness section scores
Telemetry trends and anomalies
Feature-specific evidence payloads

Outputs

Recommendation cards and queue rows
Apply status and response details
Ledger events with before/after states

Admin UI location

Rendered across SmartBlocks optimization and campaign surfaces.

Telemetry metrics help score confidence and expected gain for recommendation keys.

All governed applies should log to the Visibility Ledger.

Security

Handlers must enforce nonce and capability checks; unsafe auto-edits are blocked by policy and default safe mode.