Elasticsearch Vector Search

Elasticsearch has added dense vector search capabilities, enabling semantic search alongside its traditional full-text search strengths. This allows teams already using Elasticsearch to add AI-powered search without new infrastructure.

Key features include: kNN search with HNSW indexing, hybrid search combining BM25 and vector similarity, filtering during vector search, and integration with Elasticsearch's mature ecosystem (Kibana, observability, etc.). The Elastic Learned Sparse Encoder (ELSER) provides built-in semantic capabilities.

For AI engineers, Elasticsearch is compelling when: hybrid search (keyword + semantic) is needed, there's existing Elasticsearch infrastructure, or enterprise features like security and scalability are required. It's particularly strong for search applications transitioning from traditional to semantic search. Evaluate whether built-in features suffice versus specialized vector databases for pure semantic search use cases.