Amazon Bedrock AgentCore Memory announces metadata for long-term memory
Amazon Web Services has enhanced its Bedrock AgentCore Memory service with metadata support for long-term memory records, allowing AI agents to tag, filter, and retrieve stored memories using structured attributes in addition to semantic search capabilities. The new feature supports up to ten indexed keys per memory resource across STRING, NUMBER, and STRING_LIST data types, with various operator types available for filtering retrieval results. The metadata can be applied in two ways: attached directly to events during ingestion or automatically inferred by the underlying large language model based on extraction instructions defined within the memory resource. Developers define a metadata schema that includes indexed key definitions and extraction instructions to guide the LLM in generating metadata from conversation content. This enhancement enables agents to retrieve records using specific structured attributes such as ticket numbers, priority levels, or dates, potentially eliminating irrelevant context and improving response accuracy. The feature is now available across all AWS regions where Amazon Bedrock AgentCore Memory is supported.
Why It Matters
This enhancement addresses a critical limitation in AI agent memory systems by adding structured data retrieval capabilities alongside semantic search. The ability to filter memories using metadata reduces noise in AI responses and enables more precise contextual recall, which is essential for enterprise applications like customer service, technical support, and knowledge management systems where agents need to access specific historical information efficiently.
This summary is generated using AI analysis of the original press release. Always refer to the original source for complete details.