Search and vector store
Amazon OpenSearch Serverless
Serverless OpenSearch option for search and vector workloads with auto-scaling, pay-per-use capacity and native Bedrock Knowledge Bases integration.
Key points
- Automatically scales capacity for search and vector workloads without cluster capacity planning.
- Uses pay-per-use serverless capacity rather than provisioned cluster sizing.
- Is the default vector store used by Amazon Bedrock Knowledge Bases for RAG.
- Fits spiky or unpredictable RAG retrieval workloads that should avoid manual capacity management.
When to use it
- Use as the default vector store for Amazon Bedrock Knowledge Bases.
- Support RAG vector search workloads with unpredictable demand.
- Avoid managing OpenSearch cluster capacity for application search or vector retrieval.
Exam tips
- OpenSearch Serverless is the default Bedrock Knowledge Bases vector store.
- Serverless auto-scales; OpenSearch Service uses provisioned clusters.
- Choose it for RAG vector workloads that need native Bedrock integration and no capacity planning.