When activity spreads across GitHub, the forum, and content, real contribution is easy to lose in the noise. A community program only works when people trust the numbers behind the thank-you, so the leaderboard has to be accurate and explainable. For our Community Leaderboard we built scheduled indexers for those sources, Percona Server for MongoDB (PSMDB) as a flexible store for messy multi-source data, private dashboards to verify staff vs community and identity maps, and a daily JSON feed that powers a static Hugo widget. AI agents took care of much of the boilerplate: API clients, first-pass field mappings, early widget prototypes. That let us spend the week on rules, trust, and the public experience. The same pattern should transfer if you run a community or like database-shaped side projects.
This is a companion discussion topic for the original entry at https://percona.community/blog/2026/08/31/community-leaderboard/