Vector search built for every AI application
Vectors turn your data into AI context. Choose from a comprehensive vector portfolio of AWS services optimized for latency, cost, or scale.
Bring semantic insight to your data
Add vector search to the AWS services and databases you already run. You can quickly extract semantic insight from data in place, with no migration and no second system to operate. This is a great starting point and the right approach for most AI applications.
Some applications have one requirement that stands out: microsecond latency, billion-scale storage at the lowest cost, a specialty capability such as graph traversal, or speed to production. When one requirement dominates, choose the service purpose-built for it.
Purpose-built vector engines matched to your dominant requirements
AWS delivers a comprehensive vector portfolio with each service built for a distinct requirement: rapid queries without performance compromise, infrequent access of large vector datasets at the lowest cost, or general purpose, which allows you to just build fast without overthinking it. Pick the right fit for your applications. Keep vectors where your data already lives when there's no specialized need, or match a purpose-built service to the requirement that dominates.
Agentic RAG
Retrieve the right passages at query time and ground model responses in your own data, improving accuracy and reducing hallucinations. Combine vector, lexical, and hybrid retrieval for improved relevance.
Semantic search
Retrieve by meaning and intent across billions of embeddings. Index millions of hours of video to surface the right scene in seconds, or match a medical image against similar cases for clinical review.
Agent memory
Build lasting memory instead of forcing agents to forget. Retain interactions, documents, and insights as vectors at petabyte scale, and retrieve the right context for continual learning and reasoning.
Multimodal search
Put AI to work on datasets that span images, documents, audio, and video. Retrieve the photos, policy documents, and recorded call behind an insurance claim in a single query, or describe a scene in text and get the video segments that match.
Real-time recommendation
Serve personalized results at query time with microsecond similarity matching against a continuously updated catalog. Match users to content, products, or actions without retraining a model.
How it works
Service highlights
- lower cost to store and infrequently query vectors with Amazon S3 Vectors
- Up to 90%
- monthly requests from 100,000+ active customers of Amazon OpenSearch Service
- 10T
- cost savings by scaling to zero when idle with the next generation of Amazon OpenSearch Serverless
- Up to 60%
- recall and single-digit millisecond latency for vector search at any scale with Amazon DynamoDB
- 99%
Key services
Additional resources
Get guidance on choosing the right vector engine
Did you find what you were looking for today?
Let us know so we can improve the quality of the content on our pages