guidance for multi agent orchestration on aws

Report Abuse

Basic Information

This repository provides guidance, reference code, and a starter kit for building and deploying coordinated multi-agent systems on AWS using Amazon Bedrock Agents. It focuses on a demonstrator intelligent customer support solution that uses a Supervisor Agent to orchestrate specialized sub agents (Order Management, Product Recommendation, Troubleshooting, Personalization) to handle complex, real-world support scenarios. The project includes a React-based runtime chatbot that communicates via WebSocket to AWS Lambda which calls Bedrock Converse and uses Action Groups to perform text-to-SQL against Amazon Athena, plus vector search against Bedrock knowledge bases. The repo supplies architecture diagrams, deployment instructions using AWS CDK, configuration files, environment variable handling for agent IDs, and guidance for running and testing locally and in AWS with Cognito, CloudFront, WAF, Amplify, S3, DynamoDB and Glue components.

Links

Categorization

App Details

Features
Clear multi-agent architecture with a Supervisor Agent and multiple specialized sub agents for order management, recommendations, troubleshooting and personalization. A React runtime chatbot that uses WebSocket and an AWS Lambda backend which calls Amazon Bedrock Converse and Action Groups for generating and executing SQL against Amazon Athena. Integration with Bedrock knowledge bases and vector search for document retrieval, use of AWS Glue Data Catalog to interpret Athena schemas, and persistent session storage in Amazon DynamoDB. Deployment automation via AWS CDK, frontend hosting with AWS Amplify and CloudFront, authentication with Amazon Cognito, container/ECR authentication guidance, and a CLI-driven starter kit. Documentation includes cost estimates, model enablement guidance for Bedrock models, and step-by-step setup and local development instructions.
Use Cases
The project helps developers and architects prototype and deploy coordinated multi-agent solutions that combine LLM reasoning, semantic search, and database querying to solve customer support tasks. It demonstrates task decomposition and delegation patterns, persistent customer profiles for personalization, dynamic tool selection, and end-to-end flows for order tracking, recommendations and troubleshooting. The included infrastructure as code and starter kit reduce setup time and ensure repeatable deployments with CDK and Amplify. Operational details such as environment variable injection for agent IDs, Athena query configuration, and cost estimates support planning and local testing. Overall, it accelerates building production-like multi-agent orchestration on AWS while illustrating best practices for security, scalability and service integration.

Please fill the required fields*