Amazon DynamoDB adds general availability for real-time vector search

DynamoDB now supports native vector search, allowing for indexing and real-time searching of vector embeddings with high recall and low latency. This feature is designed to scale to trillions of vectors and operates with the same serverless benefits as other DynamoDB features.

The addition of vector search enables new use cases such as semantic retrieval for AI agent grounding, product similarity search, personalized advertising, retrieval augmented generation, and recommendation systems. Developers can store vector embeddings alongside other attributes and generate them using various models, including those from Amazon Bedrock.

Vector search in DynamoDB inherits the existing serverless advantages, including zero infrastructure management, zero downtime, zero maintenance windows, and a pay-for-use model.