FerresDB is a vector search engine designed for semantic search, retrieval-augmented generation pipelines, and recommendation systems. It provides approximate nearest-neighbor search using HNSW indexes, hybrid search combining vector similarity with BM25 full-text scoring, and a REST API. The system includes write-ahead log persistence, periodic snapshots, crash recovery, and observability metrics.
- Website
- https://www.ferres.io[01]
- Source Code
- https://github.com/ferres-db/ferres-db[02]
- Tech Docs
- https://www.ferres.io/docs[03]
- Developer
- Country of Origin
- BR
- Start Year
- 2026
- Project Types
- Hobby, Open Source
- Written in
- Rust
- Licenses
- Apache v2, MIT License
FerresDB is a vector search engine designed for semantic search, retrieval-augmented generation pipelines, and recommendation systems. It provides approximate nearest-neighbor search using HNSW indexes, hybrid search combining vector similarity with BM25 full-text scoring, and a REST API. The system includes write-ahead log persistence, periodic snapshots, crash recovery, and observability metrics.
Citations
5 sources- FerresDB — High-Performance Vector Search in Rust ferres.io
- GitHub - ferres-db/ferres-db · GitHub github.com
- FerresDB — High-Performance Vector Search in Rust ferres.io
- https://github.com/ferres-db/ferres-db/commit/aaa4cd382cee78fa82d4b0c8c3cb9a54810110c2 github.com
- https://github.com/ferres-db/ferres-db/commit/7a34b640f757fb3a458ee24445af3d6205819471 github.com