Voyager is an approximate nearest-neighbor search library for Python and Java that enables fast vector similarity queries on in-memory collections.
The tool solves the problem of efficiently finding similar vectors in high-dimensional spaces. It implements the HNSW algorithm, building on the open-source hnswlib package with additional features for convenience and performance. Voyager maintains indexes in memory and supports querying through bindings in both Python and Java, with full feature parity and index compatibility between the two languages.
Voyager suits projects that need to perform vector similarity searches at scale, particularly those already using Python or Java in their stack. The README positions it as offering higher recall than Annoy while providing a simpler interface than raw hnswlib. It is production-tested at Spotify, where it handles hundreds of millions of queries daily. The library supports Linux, macOS, and Windows across both Intel and ARM architectures, with broad Python version coverage.
The project maintains active engagement with external contributions through a documented contribution process. Development activity shows consistent attention to platform support across multiple operating systems and processor architectures. The tool has been refined through extensive production use, indicating mature handling of real-world deployment scenarios.
