What changed

Amazon S3 Vectors can now resolve metadata filters before similarity search in ENHANCED indexes, adding prefix matching, up to 100 filter constraints and as much as 2 KB of filterable metadata per vector. The feature is available without re-ingestion or an added feature charge.

Why it matters

Multi-tenant RAG and agents often need retrieval scoped to one customer, matter or document set. Evaluating that scope before similarity search can improve recall while making the authorization boundary explicit in the query path.

What to do

Canary an ENHANCED index with known relevant and cross-tenant documents. Test highly selective filters, forbidden prefixes, more than 100 constraints and missing GetVectors permission; compare recall, latency, cost and isolation with the existing CLASSIC index before changing the bucket default.

Caveat

AWS's up-to-five-times recall result is provider-run and workload-specific. Metadata filters are not a substitute for IAM or tenant isolation, existing indexes remain CLASSIC until changed, and enhanced-index rollback and performance need environment-specific testing.