Manufacturing & Automotive AI Toolkit (MA3T)

A collection of sample AI agents for Automotive and Manufacturing use cases.

Prerequisites

• AWS CLI. See the official installation guide for your operating system.

  # Verify installation and configure
  aws --version
  aws configure list
  # Verify AWS access
  aws sts get-caller-identity

• Node.js >=22 and npm installed

  # Verify installation
  node --version

• Python >=3.12

  # Ensure Python v3.12+ is installed
  python --version

• CDK CLI

  # Install AWS CDK CLI globally
  npm install -g aws-cdk
  # Verify installation
  cdk --version

Bootstrap (One-time Setup)

1. CDK Bootstrap (if needed)

If this is your first time using CDK in this AWS account/region:

cd cdk
cdk bootstrap aws://AWS-ACCOUNT-NUMBER/us-west-2

You can find your account number by running aws sts get-caller-identity.

2. Bedrock Model Access

Enable access to required models in your AWS account:

• Go to AWS Bedrock Console → Model access
• Request access to:
  • anthropic.claude-3-haiku-20240307-v1:0 (Used by VISTA agents, default for AgentCore)
  • us.anthropic.claude-3-7-sonnet-20250219-v1:0 (Used by the frontend UI)

3. Setup Environment

# Clone repository
git clone [email protected]:aws-samples/sample-manufacturing-automotive-ai-toolkit.git
cd sample-manufacturing-automotive-ai-toolkit

# Create Python virtual environment
python3 -m venv .venv
source .venv/bin/activate
pip install -r cdk/requirements.txt

Deployment

Supported Regions

Due to AgentCore regional restrictions, only these regions are supported:

• us-east-1 (US East - N. Virginia)
• us-west-2 (US West - Oregon) - default
• eu-central-1 (Europe - Frankfurt)
• ap-southeast-2 (Asia Pacific - Sydney)

Deploy Command

# Interactive deployment (prompts for credentials)
./deploy_cdk.sh

# Command line with parameters
./deploy_cdk.sh --auth-user admin --auth-password yourpassword

# With custom region and account
./deploy_cdk.sh --region us-east-1 --account 123456789012 --auth-user admin --auth-password secretpass

Deployment Options

• --auth-user: Username for UI basic authentication
• --auth-password: Password for UI basic authentication
• --region: AWS region (must be one of the supported regions)
• --account: AWS account ID to deploy to (optional, uses current AWS CLI account if not specified)
• --stack-name: Custom CDK stack name (default: MA3TMainStack)
• --skip-nag: Skip CDK security checks

Note: The CDK stack automatically triggers agent deployments via CodeBuild after successful deployment. The deployment creates an internet-accessible UI secured with basic authentication using your provided credentials.

Cleanup

To destroy the deployed resources:

cdk destroy --context region=us-east-1

Replace us-east-1 with the region where you deployed the stack.

Architecture

The MA3T architecture consists of:

1. Agent Catalog: A collection of agents implemented using various frameworks

   • Standalone Agents: Individual agents for specific tasks
   • Multi-Agent Collaborations: Groups of agents that work together

2. Agent Frameworks:

   • AWS Bedrock Agents: Native, managed Bedrock agents
   • AgentCore Agents: Container-based agents using the Bedrock AgentCore framework
     • Support for Strands, LangGraph, CrewAI, and LlamaIndex

3. UI Framework: A Next.js+React-based user interface for interacting with agents

   • Single pane of glass for all agents
   • Automatic agent discovery and registration
   • Basic authentication for secure access

4. Deployment Framework: CDK for deploying agents to AWS

Contributing

We welcome additional agent examples to the framework. Please see CONTRIBUTING.md for details.

To contribute an idea for an agent, please open an issue. If you'd like to contribute your own example, follow the instructions in developing.md.

License

This project is licensed under MIT - see the LICENSE file for details.