Build Agent Memory with NeMo and S3 Vectors

This post details how to implement Amazon S3 Vectors as a persistent memory backend for the NVIDIA NeMo Agent Toolkit, deployed on Amazon EKS. S3 Vectors supports semantic retrieval, filterable metadata, strong write consistency, and scaling to two billion vectors per index. The guide covers three steps: creating S3 Vectors infrastructure, implementing a custom MemoryEditor plugin using NAT's plugin interface, and configuring the agent workflow. A multi-agent investment research scenario serves as the running example throughout the implementation.