Our intelligence model blends multiple repository and ecosystem signals to estimate momentum and reliability. We intentionally do not publish exact weighting rules to protect model integrity and reduce score gaming.
This is the broadest surface. It includes all non-rejected records we have indexed, including records that are still under review for public visibility.
This is the curated public subset. Records here passed the public eligibility gate and are safe to expose as the external-facing shortlist.
This is a published-only momentum surface. It is optimized for discovery of breakouts and risers, not total inventory coverage.
We evaluate short-term and medium-term growth trends to detect sustained momentum versus one-time spikes.
We prioritize projects that show recent and consistent maintenance activity across repository and release lifecycle.
Contributor participation and collaboration patterns are used as quality-of-engagement signals.
Health, anomaly checks, and historical consistency improve confidence in score reliability.
Public visibility is stricter than indexing. A server can be indexed but still remain outside the public subset until it clears quality, health, and review guardrails. Legacy health labels are normalized into a single automated status set: healthy, warning, at_risk, abandoned.