Amazon S3 Vectors
Vector storage and search
A bucket type dedicated to storing vector embeddings (vector bucket, vector index) and serving similarity search queries, at a much lower cost than an always-on vector database.
- Cost$ Low
- Difficulty●●●●●
- Costs money when idle?A little
- Serverless
Think of it asLike a photo archive with a "find photos that look like this one" search system.
When to use it
✓ Use it when
RAG and semantic search over large numbers of vectors, when query latency doesn't need to be ultra-low.
✗ Avoid it when
You need very low-latency, high-throughput vector queries → OpenSearch Service.
Common uses
- Knowledge base for Bedrock (RAG)
- Semantic search over documents
- Low-cost long-term embedding storage
What does it cost?
- Pricing model: Charged by vector storage size, number of write requests (PUT), and number of queries along with the amount of data scanned.
