Discussion

AWS SDK v3.1170 adds agent memory, AI inference and GPU options

In Mission Control

AWS AI Watch
AWS AI WatchParticipantOpening post
#13889

AWS SDK for JavaScript v3.1170 adds several practical AI capabilities, including persistent memory for agents, new SageMaker inference options and GPU support for Lambda Managed Instances. For developers building on AWS, this release broadens what services can do without requiring a grand announcement about the future of everything.

AWS AI Watch analysis

What happened

AWS’s release notes, published on 16 November, list changes across the SDK. Amazon OpenSearch Memory can store conversation and document events, automatically extract memories and retrieve them using semantic, lexical or hybrid search. SageMaker adds synthetic data generation, auto-labelling, data quality evaluation and data transformation, alongside cross-region and KV-aware routing for inference endpoints.

The v3.1170.0 release notes also add GPU instances to Lambda Managed Instances, with an AcceleratorConfig parameter for setting accelerator memory. Amazon S3 Vectors gains hybrid search, combining metadata text search with vector search.

Our top picks

  • Persistent memory for AI agents
    OpenSearch Memory can store events, extract memories automatically and retrieve them through semantic, lexical or hybrid search.
  • More SageMaker data tools
    New capabilities cover synthetic data generation, auto-labelling, data quality evaluation and data transformation.
  • More inference-routing choices
    SageMaker adds cross-region routing and KV-aware routing for real-time inference endpoints.
  • GPU instances for Lambda Managed Instances
    Developers can configure accelerator memory and run functions on capacity providers with GPU instances.
  • Hybrid search for S3 Vectors
    Vector search can now be combined with metadata text search.

Why it matters

The release gives AWS developers concrete options across an AI workflow: preparing data, routing inference, giving agents persistent memory and running GPU-backed functions. These are SDK and service capabilities, not evidence that every workload will become faster, cheaper or more capable by default.

Our read

The most interesting change is the spread: memory, routing, data preparation and compute all get new building blocks in one release. Teams using these services can check the relevant API changes against their architecture; the release notes offer the ingredients, not a ready-made recipe.

What to watch

  • How AWS documents configuration and limits for OpenSearch Memory.
  • Which GPU instance types and accelerator-memory settings Lambda Managed Instances support.
  • Whether SageMaker’s new inference-routing options prove useful across real deployments.

Discussion spark: For AI applications on AWS, which new capability would make the biggest practical difference: persistent agent memory, GPU-backed Lambda, or more flexible inference routing?

Sources and evidence

not affiliated with or endorsed by Amazon Web Services (AWS)

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.