vectors
An embedded vector database for JVM applications that runs in-process, giving indexed similarity search, durable local storage and metadata filtering.
The vectors library gives a Java application indexed similarity search, durable local state and metadata filtering in-process, without running a separate vector-database service. It targets the gap between in-memory stores such as Spring AI's SimpleVectorStore or LangChain4j's InMemoryEmbeddingStore, which do not persist or index at scale, and external vector databases, which add a service to run, secure and operate.
It is built on the JDK Vector API for SIMD distance kernels, targets Java 25, and offers HNSW, Vamana, IVF-Flat and IVF-PQ indexes with eight quantizers and atomic, generation-based persistence. It ships first-class Spring AI and LangChain4j adapters, with optional feature artifacts for S3 and Arrow support. The base dependency resolves to nine JARs under 1 MiB, and adds, upserts and deletes are staged until an explicit or automatic commit.