amazon-bedrock-agentcore-samples/02-use-cases/customer-support-assistant/README.md

Customer Support Agent

Important

The examples provided in this repository are for experimental and educational purposes only. They demonstrate concepts and techniques but are not intended for direct use in production environments.

This is a customer support agent implementation using AWS Bedrock AgentCore framework. The system provides an AI-powered customer support interface with capabilities for warranty checking, customer profile management, Google calendar integration, and Amazon Bedrock Knowledge Base retrieval.

Table of Contents

• Customer Support Agent
  • Table of Contents
  • Prerequisites
    • AWS Account Setup
  • Deploy
  • Sample Queries
  • Scripts
    • Amazon Bedrock AgentCore Gateway
      • Create Amazon Bedrock AgentCore Gateway
      • Delete Amazon Bedrock AgentCore Gateway
    • Amazon Bedrock AgentCore Memory
      • Create Amazon Bedrock AgentCore Memory
      • Delete Amazon Bedrock AgentCore Memory
    • Cognito Credentials Provider
      • Create Cognito Credentials Provider
      • Delete Cognito Credentials Provider
    • Google Credentials Provider
      • Create Credentials Provider
      • Delete Credentials Provider
    • Agent Runtime
      • Delete Agent Runtime
  • Cleanup
  • 🤝 Contributing
  • 📄 License
  • 🆘 Support
  • 🔄 Updates

Prerequisites

AWS Account Setup

1. AWS Account: You need an active AWS account with appropriate permissions

   • Create AWS Account
   • AWS Console Access

2. AWS CLI: Install and configure AWS CLI with your credentials

   • Install AWS CLI
   • Configure AWS CLI

   aws configure
   

3. Bedrock Model Access: Enable access to Amazon Bedrock Anthropic Claude 4.0 models in your AWS region

   • Navigate to Amazon Bedrock Console
   • Go to "Model access" and request access to:
     • Anthropic Claude 4.0 Sonnet model
     • Anthropic Claude 3.5 Haiku model
   • Bedrock Model Access Guide

4. Python 3.10+: Required for running the application

   • Python Downloads

5. Create OAuth 2.0 credentials for calendar access : For Google Calendar integration

   • Follow Google OAuth Setup

Deploy

1. Create infrastructure

   python -m venv .venv
   source .venv/bin/activate
   pip install -r dev-requirements.txt
   
   chmod +x scripts/prereq.sh
   ./scripts/prereq.sh
   
   chmod +x scripts/list_ssm_parameters.sh
   ./scripts/list_ssm_parameters.sh
   

   Caution

   Please prefix all the resource name with customersupport.

2. Create Agentcore Gateway

   python scripts/agentcore_gateway.py create --name customersupport-gw
   

3. Setup Agentcore Identity

   • Setup Cognito Credential Provider

   python scripts/cognito_credentials_provider.py create --name customersupport-gateways
   
   python test/test_gateway.py --prompt "Check warranty with serial number MNO33333333"
   

   • Setup Google Credential Provider

   Follow instructions to setup Google Credentials.

   python scripts/google_credentials_provider.py create --name customersupport-google-calendar
   
   python test/test_google_tool.py
   

4. Create Memory

   python scripts/agentcore_memory.py create --name customersupport
   
   python test/test_memory.py load-conversation
   python test/test_memory.py load-prompt "My preference of gaming console is V5 Pro"
   python test/test_memory.py list-memory
   

5. Setup Agent Runtime

Caution

Please ensure the name of the agent starts with customersupport.

agentcore configure --entrypoint main.py -er arn:aws:iam::<Account-Id>:role/<Role> --name customersupport<AgentName>

Use ./scripts/list_ssm_parameters.sh to fill:

• Role = ValueOf(/app/customersupport/agentcore/runtime_iam_role)
• OAuth Discovery URL = ValueOf(/app/customersupport/agentcore/cognito_discovery_url)
• OAuth client id = ValueOf(/app/customersupport/agentcore/web_client_id).

Caution

Please make sure to delete .agentcore.yaml before running agentcore launch.

rm .agentcore.yaml

agentcore launch

python test/test_agent.py customersupport<AgentName> -p "Hi"

6. Local Host Streamlit UI

Caution

Streamlit app should only run on port 8501.

streamlit run app.py --server.port 8501 -- --agent=customersupport<AgentName>

Sample Queries

1. I have a Gaming Console Pro device , I want to check my warranty status, warranty serial number is MNO33333333.

2. What are the warranty support guidelines ?

3. What’s my agenda for today?

4. Can you create an event to setup call to renew warranty?

5. I have overheating issues with my device, help me debug.

Scripts

Amazon Bedrock AgentCore Gateway

Create Amazon Bedrock AgentCore Gateway

python scripts/agentcore_gateway.py create --name my-gateway
python scripts/agentcore_gateway.py create --name my-gateway --api-spec-file custom/path.json

Delete Amazon Bedrock AgentCore Gateway

# Delete gateway (reads from gateway.config automatically)
python scripts/agentcore_gateway.py delete

# Delete with confirmation skip
python scripts/agentcore_gateway.py delete --confirm

Amazon Bedrock AgentCore Memory

Create Amazon Bedrock AgentCore Memory

python scripts/agentcore_memory.py create --name MyMemory
python scripts/agentcore_memory.py create --name MyMemory --event-expiry-days 60

Delete Amazon Bedrock AgentCore Memory

# Delete memory (reads from SSM automatically)
python scripts/agentcore_memory.py delete

# Delete with confirmation skip
python scripts/agentcore_memory.py delete --confirm

Cognito Credentials Provider

Create Cognito Credentials Provider

python scripts/cognito_credentials_provider.py create --name customersupport-gateways

Delete Cognito Credentials Provider

# Delete provider (reads name from SSM automatically)
python scripts/cognito_credentials_provider.py delete

# Delete specific provider by name
python scripts/cognito_credentials_provider.py delete --name customersupport-gateways

# Delete with confirmation skip
python scripts/cognito_credentials_provider.py delete --confirm

Google Credentials Provider

Create Credentials Provider

python scripts/google_credentials_provider.py create --name customersupport-google-calendar
python scripts/google_credentials_provider.py create --name my-provider --credentials-file /path/to/credentials.json

Delete Credentials Provider

# Delete provider (reads name from SSM automatically)
python scripts/google_credentials_provider.py delete

# Delete specific provider by name
python scripts/google_credentials_provider.py delete --name customersupport-google-calendar

# Delete with confirmation skip
python scripts/google_credentials_provider.py delete --confirm

Agent Runtime

Delete Agent Runtime

# Delete specific agent runtime by name
python scripts/agentcore_agent_runtime.py customersupport

# Preview what would be deleted without actually deleting
python scripts/agentcore_agent_runtime.py --dry-run customersupport

# Delete any agent runtime by name
python scripts/agentcore_agent_runtime.py <agent-name>

Cleanup

chmod +x scripts/cleanup.sh
./scripts/cleanup.sh

python scripts/google_credentials_provider.py delete
python scripts/cognito_credentials_provider.py delete
python scripts/agentcore_memory.py delete
python scripts/agentcore_gateway.py delete
python scripts/agentcore_agent_runtime.py customersupport<AgentName>

rm .agentcore.yaml
rm .bedrock_agentcore.yaml

🤝 Contributing

We welcome contributions! Please see our Contributing Guidelines for details on:

• Adding new samples
• Improving existing examples
• Reporting issues
• Suggesting enhancements

📄 License

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

🆘 Support

• Issues: Report bugs or request features via GitHub Issues
• Documentation: Check individual folder READMEs for specific guidance

🔄 Updates

This repository is actively maintained and updated with new capabilities and examples. Watch the repository to stay updated with the latest additions.