ReliefSignal

Methodology changelog

Versioned notes on how scoring and governance evolved.

v1.1.1
  • Added source health hardening (timeouts, headers) and improved observability.
  • Improved dedup fingerprint to reduce false collisions.
v1.1.0
  • Introduced review governance layer (pending/approved/rejected).
  • Rejected events are excluded from scoring.
  • Added neutral language safeguards and review queue APIs.
v1
  • Initial public-source ingestion and risk scoring (0–100).
  • Safe-by-design constraints: no precise coordinates; choropleth-only map.