Methodology changelog
Versioned notes on how scoring and governance evolved.
Trust-impacting change path
Use this changelog together with methodology and corrections when a revision may change how users interpret trust, method or provenance.
Public revision-accountability contract
The changelog should help users see not only what changed, but whether the change belongs to routine maintenance, methodology evolution or trust-sensitive revision.
Public accountability milestone
The changelog should sit inside a wider public-accountability milestone, not as an isolated technical log.
Benchmark proof package
See the surfaces where ReliefSignal asks to be inspected as a reference-grade product.
Open this accountability surfaceMethod and revision accountability
Inspect methodology and methodology changelog together when trust-sensitive changes matter.
Open this accountability surfaceCorrections and trust path
Check how corrections remain visible, classifiable and connected to trust interpretation.
Open this accountability surfaceProvenance and sources
Inspect provenance mix, source role and transparency surfaces behind live reads.
Open this accountability surfacev1.2.0
methodology-update
This change affects how the system forms or governs outputs, even if it is not a direct correction.
- Promoted configurable global watchlists with weighted priority boosts (environment-managed).
- Added production notification preference center and subscription APIs (channel, language, severity, watchlists, quiet hours).
- Added dispatch engine with preview/execute modes, dedupe and cooldown controls, and dispatch logging for auditability.
- Expanded admin visibility: watchlist governance, watch-share trends, action inbox status workflow, and SLA breach indicators.
v1.1.1
trust-impacting
This change may affect how users interpret confidence, provenance or trustworthiness.
- Added source health hardening (timeouts, headers) and improved observability.
- Improved dedup fingerprint to reduce false collisions.
v1.1.0
correction-log
This entry records a correction or revision without implying the strongest trust-impacting change class.
- Introduced review governance layer (pending/approved/rejected).
- Rejected events are excluded from scoring.
- Added neutral language safeguards and review queue APIs.
v1
trust-impacting
This change may affect how users interpret confidence, provenance or trustworthiness.
- Initial public-source ingestion and risk scoring (0–100).
- Safe-by-design constraints: no precise coordinates; choropleth-only map.