Turbovec Brings Google TurboQuant Vector Search to Rust, Python
TL;DR. Turbovec introduces a new Rust vector index with Python bindings based on Google Research's TurboQuant algorithm for efficient AI search. - The tool significantly reduces memory footprint for vector indexes, fitting 10 million documents in 4 GB RAM. - It offers faster search performance than FAISS, with average speeds 3.4x higher at 4-bit quantization. - Turbovec supports online ingest, incremental saves, and filtered search for local, privacy-focused RAG stacks.
- Turbovec implements Google's TurboQuant algorithm in Rust for vector search.
- It drastically reduces memory usage, enabling 10M document indexes in 4 GB RAM.
- Search performance surpasses FAISS IndexPQFastScan across various configurations.
- Features include online ingestion, incremental saves, and direct filter support.
- Designed for privacy-sensitive RAG applications, running purely local with no external services.