AWS Weekly Roundup: Web Search, DynamoDB Vector Search, Lambda Bandwidth, and AI Agent Tools

Amazon Bedrock now allows OpenAI models (GPT-5.4, GPT-5.5, and GPT-5.6 Sol/Terra/Luna) to browse and retrieve real-time information from the internet. This capability supports building AI applications that access up-to-date web content while maintaining data residency within a secured AWS environment.

AI agents on Amazon Bedrock AgentCore can now be deployed and run on dedicated runtime instances. This provides more control over agent execution environments, offering predictable performance and cost management.

DynamoDB now supports storing and querying vector embeddings directly alongside existing data, eliminating the need for a separate vector database. This enables semantic retrieval for AI agentic grounding over agent memory with predictable performance.

AWS Transform, a capability for analyzing and remediating technical debt across source code repositories, is now generally available. It supports modernizing mainframe and legacy workloads through an ongoing, automated approach.

AWS open-sourced Dogwood, a governance language designed for AI agents that integrates Cedar policies and adds temporal conditions. Powering Dogwood, Amazon Bedrock AgentCore introduced temporal policies that make decisions based on an agent’s historical actions within a session.

AWS Lambda functions now support increased network bandwidth, up to 3,000 Mbps for functions configured with 10 GB of memory. This enhancement enables data-intensive workloads and faster communication with other AWS services for functions outside a VPC.