Rust core · Python bindings

A vector index built on TurboQuant.

Approximate nearest-neighbour search with a quantization core in Rust, exposed through a Python API you can import in one line.

pip install turbovec Read the design notes →

search.py

from turbovec import Index

# build over your embeddings
idx = Index(dim=768, metric="cosine")
idx.add(vectors)

# query in Rust, return in Python
hits = idx.search(query, k=10)

warm query, k=10 sub-millisecond — illustrative

Core

Quantization, not guesswork

TurboQuant compresses vectors so the index stays small and the recall stays honest.

Surface

Python where you work

The fast path lives in Rust. The API lives where your pipeline already is.

Footprint

One dependency

No server to run, no cluster to babysit. An index you embed inside your process.