Build multi-tenant agentic chat apps on enterprise data with Bedrock
AWS ML Blog · rss · 2026-09-01
This post details building a multi-tenant agentic document chat application using Amazon Bedrock Managed Knowledge Base.
Core Challenge: In multi-tenant environments, strict document isolation is required, enforced by server-side verified identity rather than client values. Agentic retrieval involves multi-hop queries where every step must carry the tenant filter.
Solution Architecture:
- Ingestion: Users upload docs via API Gateway; Lambda sends messages to SQS. A worker Lambda tags docs with the Cognito sub and calls the IngestKnowledgeBaseDocuments API for async chunking, embedding, and indexing.
- Retrieval: When users ask questions, the app calls the AgenticRetrieveStream API with a server-side constructed userid filter. The knowledge base performs multi-hop retrieval and returns a cited response.
Key Components: Bedrock Knowledge Base (managed retrieval infra), API Gateway/Lambda (logic), Cognito (auth/isolation), SQS (decoupling), DynamoDB (status tracking), S3 (large file hosting).
More from coding & agent
- Checklist: How to Clean Up AI Slop in Your Codebase — blelbach · 2026-09-01
- Apple PIM: Open Source macOS Native PIM MCP Tools — steipete · 2026-09-01
- Chrome releases WebMCP tool security guide to prevent prompt injection — prd_008 · 2026-09-01
- AWS launches Agent Registry for scalable agent and tool governance — AWS ML Blog · 2026-09-01
- Monid: Unified API for 1700+ Tools with Pay-Per-Call — SimplyAnnisa · 2026-09-01
- Small Models Can Power Agents Effectively, Offering Better Value Than Large Ones — thechrisperry · 2026-09-01