@ruvector/core

To build a vector database in Node.js, this engine provides Rust-based HNSW indexing with NAPI bindings, allowing fast creation of in-memory vector stores for similarity search applications. It supports batch operations and memory-mapped I/O for persistence.

HNSW indexing builds Hierarchical Navigable Small Worlds indexes to perform fast approximate nearest neighbor search. This approach navigates layered graphs to quickly locate vectors similar to a query, enabling efficient real-time similarity matching within large datasets.

Yes, you can integrate vector search into a Node.js backend system using the provided NAPI bindings. These bindings connect the underlying Rust engine to your JavaScript environment, enabling real-time similarity search for recommendation or anomaly detection systems.

The best way to perform high-throughput batch insertion of vectors is using the engine's native batch operations feature. This allows you to insert and delete large sets of vectors efficiently, optimizing the performance of your HNSW index construction.

Yes, the vector indexing engine supports persistence for in-memory databases by utilizing efficient memory-mapped I/O. This mechanism ensures your vector data and HNSW indexes are saved to disk and can be reloaded quickly without rebuilding.

A Rust-based NAPI engine provides higher performance for vector search compared to pure Node.js libraries. Rust executes computationally heavy HNSW indexing tasks at native speeds, minimizing overhead and maximizing throughput for similarity applications.