Amazon Bedrock Managed Knowledge Base simplifies enterprise AI data integration

A new managed capability that abstracts the complexity of building and managing retrieval-augmented generation (RAG) pipelines for enterprise AI applications. It addresses challenges in connecting to diverse data sources, optimizing RAG accuracy through automatic parsing and model selection, and managing infrastructure at scale.

Includes six pre-built ingestion connectors for Amazon S3, SharePoint, Confluence, Web Crawler, Google Drive, and OneDrive, natively pulling enterprise data and permissions. This eliminates the overhead of managing application-specific requirements for data integration.

Automatically selects the right parsing strategy for different content types and sources, including connector-specific data models, multimodal processing, and optimized chunking. This ensures higher accuracy for AI agent responses without manual configuration.

Optimized for complex queries requiring reasoning and recursive retrieval across single or multiple knowledge bases. It automatically infers user intent and creates a step-by-step query plan to gather relevant context from distributed data.

Managed Knowledge Base integrates seamlessly with Amazon Bedrock AgentCore Gateway as a native target type, reducing integration to a few lines of code. It also provides auto-generated role-based permissions and observability metrics.

Creating a Managed Knowledge Base is straightforward via the Amazon Bedrock AgentCore or Bedrock console. Users can select connectors from a dropdown, and optimized defaults for embeddings, re-rankers, and foundation models are presented for quick setup.