DynamoDB Vector Search Is Now Generally Available

Amazon DynamoDB has made vector search generally available, letting developers store vector embeddings alongside operational data and run similarity searches directly in the database without copying data to a separate vector store.

The feature provides real-time vector search at any scale, supporting up to trillions of vectors with single-digit millisecond latency at 99% recall or better. It is fully serverless: no servers to provision, no software to install, no versions or maintenance windows, and no downtime for updates.

Previously, adding vector search to a DynamoDB-backed application meant maintaining a separate vector database and a synchronization pipeline, adding operational overhead, data movement costs, and licensing costs. With the new built-in index type, teams use the same pay-per-request pricing model and can store embeddings as a list of floats with a standard PutItem call. The SearchVectors API accepts a query vector, returns up to 100 results, and supports optional filter conditions.

Vector indexes have no storage limits and scale horizontally as data grows. They support up to 4,096 dimensions, Euclidean, Cosine and Dot product distance functions, and inline filtering. Use cases include agentic memory, retrieval augmented generation, recommendation engines, personalized experiences, and anomaly detection. Vector search is available in all commercial AWS Regions, including AWS GovCloud (US) Regions.

Source: AWS