[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"glossary-vector-database::en":3,"gloss-cluster-vector-database::en":23,"gloss-next-vector-database::en":9},{"slug":4,"category":5,"name":6,"definition":7,"meta_desc":8,"faq":9,"schema_markup":9,"related":10},"vector-database","core-ai","Vector Database","A vector database is a data store purpose-built to hold high-dimensional numeric vectors (embeddings) and answer \"find me the vectors most similar to this one\" queries at scale — usually via approximate nearest neighbor (ANN) algorithms like HNSW or IVF, which trade a small amount of accuracy for massive speed gains over brute-force comparison. This matters for SaaS\u002FAI builders because it's the infrastructure that makes RAG, semantic search, deduplication, and recommendation features possible: instead of matching on exact keywords, you match on meaning. Popular options span a spectrum: dedicated vector databases like Pinecone, Weaviate, Qdrant, and Milvus are built vector-first with rich filtering and horizontal scaling; pgvector adds vector columns and similarity operators directly to PostgreSQL, letting teams avoid running a separate database when their scale doesn't demand it; and Redis, Elasticsearch, and MongoDB have all bolted on vector search to their existing engines. The mechanics: each document (or chunk of a document) is converted to a vector via an embedding model, stored alongside metadata (source URL, timestamp, tags), and indexed. A query is embedded the same way, and the database returns the nearest vectors by a distance metric — usually cosine similarity or dot product. A concrete example using pgvector in Postgres: `CREATE TABLE docs (id serial, content text, embedding vector(1536)); CREATE INDEX ON docs USING hnsw (embedding vector_cosine_ops); SELECT content FROM docs ORDER BY embedding \u003C=> '[0.012, -0.045, ...]' LIMIT 5;` — that single query returns the 5 most semantically similar documents to the input vector in milliseconds, even across millions of rows. Choosing between a managed vector database and pgvector-in-Postgres is a real architectural decision: managed services scale further with less ops burden but add a new system and cost line; pgvector keeps your stack simple if you're already on Postgres and your corpus is in the low millions of vectors. Beyond raw similarity search, most production vector databases support metadata filtering (e.g., \"only search documents tagged `region: EU` and `status: published`\"), which is essential for multi-tenant SaaS products where one customer's RAG queries must never surface another customer's private documents — filtering happens at the database query level, not as a post-hoc check, to avoid ever transmitting the wrong tenant's data to the LLM in the first place. Index rebuild strategy also matters operationally: as documents are added, updated, or deleted, the vector index needs to stay in sync, and some databases handle this with near-real-time upserts while others require periodic full re-indexing — a meaningful difference for a product where content changes frequently (like a live support-ticket knowledge base) versus rarely (like a static product manual).","A vector database stores embeddings and enables fast similarity search — the retrieval backbone behind RAG, semantic search, and recommendation systems.",null,[11,14,17,20],{"slug":12,"name":13},"embedding","Embedding",{"slug":15,"name":16},"retrieval","Retrieval",{"slug":18,"name":19},"retrieval-augmented-generation","Retrieval-Augmented Generation (RAG)",{"slug":21,"name":22},"semantic-search","Semantic Search",[24,28,32,36,39,42,45,48,51,54,57,60],{"slug":25,"category":5,"name":26,"updated_at":27},"agentic","Agentic AI","2026-08-24T02:46:36+00:00",{"slug":29,"category":5,"name":30,"updated_at":31},"alignment-tax","Alignment Tax","2026-08-24T02:46:37+00:00",{"slug":33,"category":5,"name":34,"updated_at":35},"artificial-intelligence","Artificial Intelligence (AI)","2026-08-24T02:46:38+00:00",{"slug":37,"category":5,"name":38,"updated_at":27},"attention","Attention",{"slug":40,"category":5,"name":41,"updated_at":35},"beam-search","Beam Search",{"slug":43,"category":5,"name":44,"updated_at":31},"benchmark-contamination","Benchmark Contamination",{"slug":46,"category":5,"name":47,"updated_at":31},"catastrophic-forgetting","Catastrophic Forgetting",{"slug":49,"category":5,"name":50,"updated_at":35},"computer-vision","Computer Vision",{"slug":52,"category":5,"name":53,"updated_at":31},"constitutional-ai","Constitutional AI",{"slug":55,"category":5,"name":56,"updated_at":27},"context-window","Context Window",{"slug":58,"category":5,"name":59,"updated_at":35},"deep-learning","Deep Learning",{"slug":61,"category":5,"name":62,"updated_at":27},"diffusion-model","Diffusion Model"]