DynamoDB native vector search launched: I tested it—three pros, one pitfall

I compared DynamoDB's native vector search with my previous setup using OpenSearch and zero-ETL, actually running through both. Bottom line first: this feature is great news for those doing RAG (Retrieval-Augmented Generation), but beginners might fall into a trap. My lab has been working on a scheduling optimization project combining reinforcement learning with knowledge bases, requiring us to convert state descriptions into vectors for storage and quickly find similar historical scenarios. Previously, we either had to set up a separate OpenSearch cluster or use third-party services like Pinecone. Now that DynamoDB supports this directly, it means one less component to manage.