16 sources. One unified corpus.
Every Cosmos knowledge source you would search separately, indexed and exposed through33 MCP tools. The numbers below refresh every minute, straight from the database, and a source that has not been synced yet shows zero.
How the data lands here
Each source is polled on its own cadence - the forum every 5 min, the SDK index every 6 hr, whitepapers every 24 hr. Anything changed since the last sync is upserted and re-embedded.
Why one corpus matters
Cross-source queries (“which release fixed the breakage discussed in this forum thread?”) only work when forum, docs and GitHub share a single semantic index.
Verify the numbers
The widget calls /api/v1/sources. Same endpoint anyone can hit - no caching, no embellishment.