Agents for Amazon Bedrock now support memory retention and code interpretation (preview)

This article announces new capabilities added to Agents for Amazon Bedrock, which allow generative AI applications to perform multistep tasks across different systems and data sources.

Specifically, the article covers:

• Memory retention across multiple interactions, enabling agents to maintain a persistent memory and provide a smooth flow of conversations and workflows for complex, multistep tasks
• Support for code interpretation, allowing agents to dynamically generate and run code snippets within a secure, sandboxed environment for tasks like data analysis, visualization, text processing, equation solving, and optimization
• Ability to upload documents directly to an agent for the agent to process and extract information from various file formats
• Conclusion on the preview availability and pricing details for these new capabilities