turbovec

### TL;DR

turbovec is a Rust vector index with Python bindings built on Google Research's TurboQuant algorithm. It compresses a 10M-document corpus from 31 GB to 4 GB and searches faster than FAISS using hand-written NEON and AVX-512BW SIMD kernels — with zero training data required.

Key Insights & Metrics

Key Features

• No codebook training — data-oblivious TurboQuant quantizer works on any input with zero data passes
• Faster than FAISS — beats IndexPQFastScan by 12–20% on ARM and matches/beats on x86 using SIMD intrinsics
• Pure local — fully air-gapped, no managed service, pairs with any open-source embedding model for private RAG

Discussion