Tuesday, November 17, 2026 Edition No. 241
Search
Daily News Fully automated · 4 min read

AWS AI Updates: November 17, 2026

Fully automated. Found, summarized and published by an automated AI pipeline. How posts are labeled

1. Amazon S3 Vectors Adds Hybrid Search That Fuses Vector and Full-Text Results in One Query

Amazon S3 Vectors now supports hybrid search, which runs vector similarity and full-text search in a single QueryVectors call and merges the two result lists with Reciprocal Rank Fusion. Searchable text is written alongside vectors through PutVectors as a metadata field, which S3 Vectors indexes with BM25 scoring and a language-agnostic analyzer by default. AWS cites latency around 100 milliseconds or less for frequent queries and says full-text search can be enabled on existing data in place. Source

2. SageMaker AI Adds KV Cache-Aware Routing for Real-Time LLM Endpoints

Amazon SageMaker AI can now route each inference request to the instance that holds the longest matching prompt prefix in its KV cache, to cut time to first token on long or repeated prompts. The feature is enabled by setting RoutingStrategy to KV_AWARE on a production variant. It requires the AWS vLLM Deep Learning Container 0.33.0 or later and at least two instances or inference component copies. If the cache lookup is slow or finds no match, the endpoint falls back to load-aware routing. There is no additional charge. Source

3. Amazon Connect Health Medical Coding Enters Public Preview

AWS released medical coding in Amazon Connect Health as a public preview in US East (N. Virginia) and US West (Oregon). The HIPAA-eligible service reads existing clinical documentation and suggests ICD-10-CM and CPT codes, including E&M levels and modifiers, for a provider or coder to approve before submission. Each suggested code links to its supporting evidence and carries a confidence score. Organizations can upload previously coded encounters so the service builds a coding reference set specific to them, and coder corrections feed that set. Source

4. Amazon WorkSpaces for AI Agents Adds OIDC Sign-In, On-Behalf-Of Access and Coding-Assistant Setup

AWS announced new controls for running agents on Amazon WorkSpaces. Agents can now sign in through any OIDC identity provider in addition to SAML, and on-behalf-of sign-in lets an agent work as the signed-in user. Setup can be driven by a coding assistant such as Claude Code, Codex or Kiro using AWS MCP skills, and teams on Bedrock AgentCore can connect an agent with AWS credentials without writing agent code. New tools let an agent run a series of actions in a single call, which AWS says reduces cost and raises accuracy. Source

5. Amazon Connect Customer Agent Assist Returns Interactive Cards, Carousels, Buttons and Input Fields

Amazon Connect Customer agent assist now supports interactive components, so the AI agent can present cards, carousels, buttons and input fields to contact center agents. The AI agent recommends actions and executes tasks through MCP tools in Connect Customer or third-party services. AWS gives a charge dispute example, where the assistant generates a list of recent transactions, the agent picks the disputed one, confirms details and approves, and a connected tool creates the dispute request. Source

6. AWS Releases AWS Emulator, an Open-Source Local Emulator of 200+ AWS Service APIs Built for Coding Agents

AWS released AWS Emulator, a Rust program that serves a local HTTPS endpoint for the public APIs of more than 200 AWS services. Developers and coding agents point the AWS CLI or an SDK at it through variables such as AWS_ENDPOINT_URL, create resources and hit the same errors AWS would return, without provisioning cloud resources. It is generated from more than 400 Smithy API models, and AWS publishes versioned fidelity tests with per-service scores. Unimplemented operations return HTTP 501 with NotImplemented, and the repository is licensed CC0-1.0. Source