Glossary
vector database
A datastore optimized for approximate nearest-neighbor search over high-dimensional embedding vectors, the storage substrate for most RAG and recommendation pipelines.
A specialized index over embedding vectors that answers “find the k
vectors most similar to this query vector” in sublinear time. The
canonical algorithms are HNSW (graph-based, fast and accurate) and IVF
(inverted-file with product quantizationweightsStoring or computing model weights in lower-precision number formats (FP8, INT8, INT4) to reduce memory and bandwidth, accepting small quality loss.
Open full entry , memory-efficient).
The choice space stratifies by where the index lives. Embedded-in-the- process (sqlite-vec, LanceDB, DuckDB-vss) suits local-firstsovereignty-decentralizationAn architecture stance where inference (and increasingly memory and agent state) runs on the user's own device rather than a remote API, prioritizing privacy, latency, and offline operation. Open full entry and small deployments. Standalone servers (Qdrant, Weaviate, Milvus) suit production RAGretrieval-memoryA pattern where a model retrieves relevant documents from an external store at query time and conditions its answer on them, instead of relying only on parametric knowledge. Open full entry . Cloud-hosted (Pinecone, MongoDB Atlas Vector Search) suits teams wanting a managed offering. Postgres extensions (pgvector) suit teams already on Postgres.
In open-source AI the trend through 2024 to 2026 has been toward embedded vector indexes (LanceDB, sqlite-vec) for local-firstsovereignty-decentralizationAn architecture stance where inference (and increasingly memory and agent state) runs on the user's own device rather than a remote API, prioritizing privacy, latency, and offline operation. Open full entry and edge deployments, with the standalone servers consolidating on enterprise features (multi-tenancy, hybrid search, observability).