Skip to content
AI 24 UpdatesAI 24UpdatesWhat changed. Why it matters to you. What to do.

AWS makes OpenSearch Memory generally available for AI agents, plus 2 more updates

Version 1

Share on X

In Brief

  1. AWS — Amazon OpenSearch Memory is generally available as serverless long-term memory for AI agents
  2. Amazon SageMaker AI — cross-Region inference routing at no extra charge
  3. AWS Glue — Apache Spark Optimization Agent recommends cost and performance fixes
Abstract editorial illustration on a dark canvas: a glowing memory core linked to agent nodes, captioned OpenSearch Memory

Details

AWS — Amazon OpenSearch Memory is generally available as serverless long-term memory for AI agents

Amazon

AWS made Amazon OpenSearch Memory generally available, a serverless capability in Amazon OpenSearch Service that adds persistent memory to AI agents.

  • OpenSearch Memory stores agent interactions as searchable events, derives memories from them and retrieves relevant context across sessions without replaying the full conversation history to the model.
  • Original events are stored verbatim and become searchable within seconds; structured memories are derived asynchronously.
  • It manages extraction, embedding, indexing, consolidation and retrieval, with keyword, semantic and hybrid search.
  • Access is controlled with IAM per memory store, with actor and session scoping and expiration periods for events and memories; there are no clusters, instance types or indexes to manage.
  • It is available in US East (N. Virginia, Ohio), US West (Oregon), Europe (Frankfurt) and Asia Pacific (Tokyo).
Sources:AWS What's New✓ official

Amazon SageMaker AI — cross-Region inference routing at no extra charge

Amazon

Amazon SageMaker AI added cross-Region inference routing, which sends requests to endpoints in other AWS Regions and accounts when a Region runs short of capacity.

  • Developers create an inference router, register existing real-time endpoints, asynchronous endpoints and inference components, and point applications at the router; endpoints, models and containers do not change.
  • The router prefers endpoints in its own Region, shifts new requests to healthy destinations elsewhere as errors rise, and returns traffic gradually after recovery.
  • An endpoint policy controls which routers in other accounts can use an endpoint, and an optional cross-Region traffic cap reserves capacity for local callers.
  • Each real-time response identifies the serving endpoint and the reason; router metrics appear in Amazon CloudWatch.
  • It is available for real-time inference in 32 commercial Regions and asynchronous inference in 25; there is no additional charge beyond standard SageMaker AI and CloudWatch pricing.
Sources:AWS What's New✓ official

AWS Glue — Apache Spark Optimization Agent recommends cost and performance fixes

Amazon

AWS Glue ETL added the Apache Spark Optimization Agent, which analyzes completed Spark job runs and returns optimization recommendations within minutes.

  • Users pick a cost or performance goal; the agent returns specific configuration changes with the telemetry evidence and reasoning behind each one.
  • It reads the job's configuration, logs and Spark telemetry and uses a deterministic recommendation engine to find issues such as low parallelism, skewed join keys, memory spill, idle executors or over-provisioned capacity.
  • Recommendations are labeled by area: infrastructure, Spark configuration, code or data layout.
  • It runs from the AWS Glue console or from other tools through the Amazon SageMaker Unified Studio MCP Server.
  • It is free for AWS Glue 4.0 and later in all Glue Regions except AWS GovCloud (US) and China.
Sources:AWS What's New✓ official

Subscribe for updates

Subscribe to get AI 24 Updates by email. You'll receive a confirmation email first.

We use your email address only to send the AI 24 Updates newsletter. You will get a confirmation email first, and every newsletter has an unsubscribe link. Subscriptions are managed by Buttondown (USA). See our Privacy Policy.