DynamoDB vector search is now generally available, in every commercial AWS region including the Middle East. The feature indexes vector embeddings stored inside DynamoDB tables and runs similarity searches over them, with single-digit millisecond latency at 99%+ recall, at any scale up to trillions of vectors.
Until now, adding semantic search to a DynamoDB-backed application meant copying data into a dedicated vector database and running a sync pipeline between the two. The new vector index type keeps embeddings and operational data on the same serverless infrastructure and the same pay-per-request pricing, with no servers to provision, no versions and no maintenance windows.
How DynamoDB vector search works
Embeddings are generated with a model of your choice — Amazon Bedrock Titan Text Embeddings, Cohere Embed or OpenAI text embedding models — and stored as a list of floats in a standard PutItem call. A vector index is then created on that attribute, with support for up to 4096 dimensions, Euclidean, Cosine and Dot product distance functions, and inline filters that narrow results at query time. The SearchVectors API takes a query vector and returns up to 100 results ranked by similarity.
The announced use cases lean on agentic AI: storing agent memory in DynamoDB and retrieving it semantically for grounding, alongside retrieval augmented generation, product similarity search, personalised advertising and anomaly detection. For teams already on DynamoDB, vector search now lives where the data lives. The sync pipeline was the part nobody enjoyed.
What it means for the UAE
DynamoDB vector search is live in all commercial regions, including AWS’s UAE (me-central-1) and Bahrain (me-south-1) regions, so local teams can use it without moving data out of the region. Developers and enterprises building AI features on AWS — semantic search, RAG, agent memory — can add them to the database they already run instead of provisioning a second one.
What is vector search in DynamoDB?
It is a native DynamoDB feature that indexes vector embeddings stored in a table and runs similarity searches over them, so applications can find items by meaning rather than by exact key or text match.
How much does DynamoDB vector search cost?
It uses DynamoDB’s existing pay-per-request model — you pay for the data processed and returned by searches, with no separate infrastructure to provision. Exact rates are on the AWS DynamoDB pricing page.
Which embedding models work with DynamoDB vector search?
Any model you like, stored as a list of floats: AWS points to Amazon Bedrock Titan Text Embeddings, Cohere Embed and OpenAI text embedding models.
