v0.0.1Latest
27 May 2026
Sovereign Search at Scale
A detailed benchmark report
Our first public benchmark of the sovereign retrieval engine: tested on a 1M-chunk synthetic corpus and real AWS S3. ~13 KB read per query at scale, sub-millisecond warm latency, competitive recall on FiQA, plus the cache bug we caught and the permission claim we tested and retired.
BenchmarkRetrieval engine
Read the full report