Hey PH 👋 I'm Aditya, a backend engineer who kept hitting the same wall on side projects: I wanted real search, typo tolerance, relevance ranking, facets - not LIKE '%query%', but I didn't want to stand up a JVM or a multi-node cluster just to search a few hundred thousand records. So I built Tachyon: an open-source, self-hosted full-text search engine in Rust that ships as a single 8.7MB binary. BM25 relevance, typo tolerance, filters, facets, sorting, autocomplete, a REST API - the Typesense/Meilisearch feature set, minus the operational weight. On a 1M-document benchmark: 155k docs/sec indexing, 33.6ms p95 search latency, ~625MiB steady-state memory. docker run -p 8108:8108 adikeshri/tachyon:latest It's early (13 stars, actively developed) and I'd genuinely value feedback - especially from anyone who's run Typesense or Meilisearch in production and can tell me where Tachyon's query planner or indexing would fall over at scale. Happy to answer anything.

Tachyon

Full-text search in an 8.7MB Rust binary

Full-text search in an 8.7MB Rust binary