Amazing - AWS finally embedded vector support directly into S3. No external databases, no ops overhead. It's serverless, supports cosine similarity, filterable metadata, up to 10k indexes per bucket, and millions of vectors per index - all with low latency.
It's not a full replacement for FAISS or Weaviate, but for typical use cases, it removes a ton of complexity. You get simpler architecture: data and vectors co-exist in the same store.
🔗Link: https://aws.amazon.com/blogs/aws/introducing-amazon-s3-vectors-first-cloud-storage-with-native-vector-support-at-scale/