Open source teams build great software but often struggle with a simple question: What should we build next? GitMesh began as a community experiment inside LF Decentralized Trust to answer that using search, not guesswork. We monitor conversations across GitHub, Reddit, X, Discord, Stack Overflow, HN, and Product Hunt, then use OpenSearch to classify signals, rank what matters, and turn them into actionable GitHub issues with real context.

This session shares how we built this in the open: a multi-tenant search architecture that can serve many OSS projects, nested OpenSearch documents that track both market signals and contributor history, and the event-driven pipeline that keeps everything in sync. I’ll show the queries we use to connect “what the community wants” to “which contributors can build it,” how we handle unpredictable schema changes through versioned indices, and the lessons learned building this transparently within LFDT. The goal is simple: give open source maintainers a practical, reproducible way to understand community demand and route work to the right contributors using OpenSearch.