View a markdown version of this page

GenAI Accelerator Starter Package - Modern Data Architecture Accelerator
Shared Domain ComponentsGenAI Domain ComponentsDeployment InstructionsUsage Instructions

GenAI Accelerator Starter Package

The GenAI Accelerator Starter Package delivers a production-ready foundation for building sophisticated AI agents using Amazon Bedrock. Out of the box, it demonstrates a customer support assistant that can understand queries, search knowledge bases, and take actions - but this is just the beginning.

Built on AWS best practices, this package provides the building blocks to create any type of agentic AI application. Whether you need sales assistants, technical troubleshooters, or domain-specific experts, the modular architecture adapts to your unique requirements.

This architecture is particularly effective when:

  1. You need to deploy intelligent agents with autonomous decision-making capabilities that can perform complex tasks.

  2. Your organization requires governed access to AI resources with appropriate content safety measures.

Deploy this package when you need a scalable, secure foundation that supports your organization’s agentic AI initiatives with enterprise-grade controls.

The GenAI Accelerator Starter Package provides a complete environment for developing and deploying agentic AI applications. This package is organized into multiple domains that work together to create a comprehensive AI platform:

Shared Domain Components

  • IAM Roles - Comprehensive roles including data-admin with Bedrock permissions, agent execution roles, and Lambda execution roles

  • Data Lake - KMS-encrypted S3 buckets for knowledge base data sources and document storage

  • Access Policies - Granular permissions for AI resource management and data access

GenAI Domain Components

  • Bedrock Agents - Intelligent agents with custom instructions and action groups

  • Knowledge Bases - RAG-enabled knowledge bases with vector stores for efficient retrieval

  • Lambda Functions - Custom functions for agent action groups and document processing

  • Guardrails - Content safety measures to ensure appropriate AI responses

  • Vector Stores - Efficient knowledge retrieval using Amazon OpenSearch Serverless

This package accelerates your agentic AI initiatives by providing a ready-to-use environment with AWS best practices built in. It’s ideal for organizations looking to establish or enhance their AI capabilities with autonomous agents, intelligent task execution, and enterprise-grade AI governance.

Deployment Instructions

Step-by-step guide for deploying the GenAI Accelerator Starter Package

You can deploy the GenAI Accelerator Starter Package using Manual CLI Deploy Method

Manual CLI Deploy Method

Prerequisites

Before deploying the GenAI Accelerator Starter Package using the CLI method, ensure you have:

  1. AWS CLI configured with appropriate credentials

  2. Node.js 16.x or later installed

  3. AWS CDK installed (npm install -g aws-cdk)

  4. CDK bootstrapped in your target account and region

  5. A VPC with at least two private subnets (required for OpenSearch Serverless)

  6. Access to Bedrock foundation models in your region

Deployment Steps

Step 1: Clone the MDAA repository 

git clone https://github.com/aws/modern-data-architecture-accelerator.git &&
cd modern-data-architecture-accelerator

Step 2: Configure your deployment

  • Copy the sample configuration files:

cp -r sample_configs/genai_accelerator my_genai_config
cd my_genai_config

  • Edit the mdaa.yaml file to set your organization name and VPC/subnet information:

organization: <your-unique-org-name>
context:
  vpc_id: <your vpc id>
  subnet_id_1:  <your subnet id 1>
  subnet_id_2:  <your subnet id 2>
  subnet_id_3:  <your subnet id 3>
  llm_model: <your_model> # e.g. anthropic.claude-3-5-sonnet-20240620-v1:0 or an ARN
  kb_embedding_model: <kb_embedding_model> # e.g. amazon.titan-embed-text-v2:0
  kb_parsing_model: <kb_parsing_model> # e.g. anthropic.claude-3-5-sonnet-20240620-v1:0

Important Notes:

  • Cross-region inference: Use inference profile ARN in llm_model (e.g., arn:aws:bedrock:us-east-1:<account_id>:inference-profile/anthropic.claude-3-7-sonnet-20250219-v1:0)

  • Single region: Use model ID directly

Step 3: Deploy the solution * Ensure you are authenticated to your target AWS account.

  • Optionally, run the following command to understand what stacks will be deployed:

../bin/mdaa ls

  • Optionally, run the following command to review the produced templates:

../bin/mdaa synth

  • Run the following command to deploy all modules:

../bin/mdaa deploy

Step 4: Verify deployment * Check the AWS CloudFormation console to ensure all stacks have been created successfully * Verify the Bedrock agents, knowledge bases, S3 buckets, and other resources have been created

Usage Instructions

How to effectively use the GenAI Accelerator Starter Package after deployment

Once the MDAA deployment is complete, follow these steps to interact with the GenAI platform:

Initial Setup and Document Upload

  1. Assume the data-user role

    • This role is configured with AssumeRole trust to the local account by default

    • It has permissions to upload documents to the knowledge base S3 buckets

  2. Upload documents to knowledge base buckets

    • Upload documents to the support-docs prefix for customer support materials

    • Upload documents to the product-docs prefix for product documentation

    • Ensure documents are uploaded with KMS encryption

Using Bedrock Agents

  1. Assume the data-admin role

    • This role is configured with AssumeRole trust to the local account by default

    • It has comprehensive permissions for managing Bedrock resources

  2. Sync knowledge bases

    • Navigate to the Amazon Bedrock console

    • Go to Knowledge Bases and select your deployed knowledge base

    • Check if your documents are automatically synced to the knowledge base

  3. Test the Bedrock Agent

    • In the Amazon Bedrock console, go to Agents

    • Select the customer-support-agent

    • Use the Test Agent interface to interact with the agent

    • Ask questions related to the documents you uploaded

  4. Monitor agent performance

    • Use CloudWatch logs to monitor agent interactions

    • Review Bedrock traces for detailed execution information

    • Monitor Lambda function logs for action group executions

Advanced Usage

  1. Customize agent behavior

    • Modify the agent instructions in the bedrock-builder.yaml configuration

    • Update Lambda functions for custom action groups

    • Adjust guardrails for content safety requirements

  2. Expand knowledge bases

    • Add additional data sources to existing knowledge bases

    • Create new knowledge bases for different domains

    • Configure custom transformation Lambda functions

For more detailed information about the configuration files and their purposes, refer to the README.md file in the sample_configs/genai_accelerator directory of the MDAA repository.