SvectorDB

Storing and searching vector embeddings on AWS without running any servers is the job SvectorDB takes on.

This serverless vector database has no provisioning or scaling to manage, and updates show up instantly.

You define it with CloudFormation, so it fits into infrastructure you already describe as code.

Built-in vectorizers mean you can hand SvectorDB your raw content instead of generating embeddings yourself first.

Code examples and tutorials help you get started, and the selling point is lower cost than comparable vector databases.

That claim is worth testing against your own workload.

Databases default to a limit of 1 million records, though support can raise it.

Database snapshots are not available.

The company is very small, which may matter if you need long-term vendor stability, and security measures are not documented in much detail.

The team does communicate openly with developers.

For an AWS-centered project that needs a simple AI search backend, SvectorDB is a reasonable option to evaluate.