AWS bedrock documentation change
Summary
Restructured knowledge base documentation to introduce Managed and Self-managed knowledge bases, removed outdated sections about query conversion and agent workflows, and updated next-step links.
Security assessment
The change adds documentation about Managed Knowledge Base's document-level permission filtering using Access Control Lists at retrieval time, which is a security feature. However, there's no evidence of addressing a specific security vulnerability or incident.
Diff
diff --git a/bedrock/latest/userguide/knowledge-base.md b/bedrock/latest/userguide/knowledge-base.md index 5ea70c9b1..2ebd15ec6 100644 --- a//bedrock/latest/userguide/knowledge-base.md +++ b//bedrock/latest/userguide/knowledge-base.md @@ -9 +9,10 @@ -While foundation models have general knowledge, you can further improve their responses by using Retrieval Augmented Generation (RAG). RAG is a technique that uses information from data sources to improve the relevancy and accuracy of generated responses. With Amazon Bedrock Knowledge Bases, you can integrate proprietary information into your generative-AI applications. When a query is made, a knowledge base searches your data to find relevant information to answer the query. The retrieved information can then be used to improve generated responses. You can build your own RAG-based application by using the capabilities of Amazon Bedrock Knowledge Bases. +While foundation models have general knowledge, you can further improve their responses by using Retrieval Augmented Generation (RAG). RAG is a technique that uses information from data sources to improve the relevancy and accuracy of generated responses. With Amazon Bedrock Knowledge Bases, you can integrate proprietary information into your generative-AI applications. When a query is made, a knowledge base searches your data to find relevant information to answer the query. The retrieved information can then be used to improve generated responses. Amazon Bedrock Knowledge Bases helps build secure, enterprise-grade agentic AI applications by leveraging retrieval securely over large-scale datasets. + +Amazon Bedrock Knowledge Bases offers two types of knowledge bases: + + * **Managed Knowledge Base** – Amazon Bedrock manages the underlying data ingestion, indexing, storage, and retrieval infrastructure so you can focus on your application and agent logic. Managed Knowledge Base offers advanced capabilities including ingestion of multi-modal data, storage auto-scaling, agentic retrieval for multi-hop reasoning and more while continuing to offer key customizations so you can tailor agents to your use case. You connect your data sources and Amazon Bedrock manages embedding, re-ranking, and reasoning with service-managed models by default (with option to choose your own models). Managed Knowledge Base integrates natively with AgentCore Gateway so any MCP-compatible agent framework can discover and invoke your Knowledge Base as a tool without custom code. Managed Knowledge offers connectors for Amazon S3, SharePoint, Confluence, Google Drive, OneDrive, Web Crawler, along with document-level permission filtering using Access Control Lists (except for Web Crawler) at retrieval time. The Smart Parsing feature auto-selects parsing strategy per document type including PDFs, PPTX, DOCX, documents with embedded visuals, audio, video, and scanned documents. Agentic Retrieval supports multi-hop reasoning, decomposes complex queries into sub-queries, retrieves iteratively across multiple knowledge bases, and evaluates sufficiency of responses. Native integration with AgentCore Observability offers built-in monitoring, retrieval traces, agentic traces, and per-knowledge base metrics. + + * **Self-managed Knowledge Base** – This enables you set up and manage your own RAG pipeline, including the vector store (such as Amazon OpenSearch Serverless, Amazon Aurora, and Amazon Neptune), and have full control over data ingestion, parsing, indexing and storage configurations. You will need to setup and manage related infrastructure and configurations such as parsing for multimodal data. Note that several capabilities such as third-party connectors, document-level permissions and native AgentCore Gateway integration are only available for Managed Knowledge Bases. + + + @@ -25,4 +33,0 @@ With Amazon Bedrock Knowledge Bases, you can: - * Convert natural language into queries (such as SQL queries) that are customized for structured databases. These queries are used to retrieve data from structured data stores. - - * Update your data sources and ingest the changes into the knowledge base directly so they can be immediately accessed. - @@ -31,21 +35,0 @@ With Amazon Bedrock Knowledge Bases, you can: - * Include the knowledge base in an [Amazon Bedrock Agents](./agents.html) workflow. - - - - -To set up a knowledge base, you must complete the following general steps: - - 1. (Optional) If you connect your knowledge base to an unstructured data source, set up your own [supported vector store](https://docs.aws.amazon.com/bedrock/latest/userguide/knowledge-base-setup.html) to index the vector embeddings representation of your data. You can skip this step if you plan to use the Amazon Bedrock console to create an Amazon OpenSearch Serverless vector store for you. - - 2. Connect your knowledge base to an unstructured or structured data source. - - 3. Sync your data source with your knowledge base. - - 4. Set up your application or agent to do the following: - - * Query the knowledge base and return relevant sources. - - * Query the knowledge base and generate natural language responses based on the retrieved results. - - * (If you query a knowledge base connected to a structured data store) Transform a query into a structured data language-specific query (such as an SQL query). - @@ -65 +49 @@ To set up a knowledge base, you must complete the following general steps: - * [Build a knowledge base by connecting to a data source](./knowledge-base-build.html) + * [Build a managed knowledge base](./kb-build-managed.html) @@ -67 +51 @@ To set up a knowledge base, you must complete the following general steps: - * [Build a knowledge base for multimodal content](./kb-multimodal.html) + * [Build a knowledge base with vector stores](./knowledge-base-build.html)