Deploy the Strands Agent Data Analyst Assistant for Video Game Sales using AWS Bedrock AgentCore's fully managed service for scalable agent applications with Runtime and Memory capabilities.

Overview

This tutorial guides you through deploying a video game sales data analyst agent using Amazon Bedrock AgentCore's managed infrastructure, includes the following modular services:

Amazon Bedrock AgentCore Runtime: Provides the managed execution environment with invocation endpoints (/invocations) and health monitoring (/ping) for your agent instances
Amazon Bedrock AgentCore Memory: A fully managed service that gives AI agents the ability to remember, learn, and evolve through interactions by capturing events, transforming them into memories, and retrieving relevant context when needed

Don't forget to review the Amazon Bedrock AgentCore documentation.

Important

This sample application is meant for demo purposes and is not production ready. Please make sure to validate the code with your organizations security best practices.

Remember to clean up resources after testing to avoid unnecessary costs by following the clean-up steps provided.

Environment Setup and Requirements

Before you begin, ensure you have:

Back-End Deployment - Data Source and Configuration Management Deployment with CDK
Docker
Required Packages:
- Install the Amazon Bedrock AgentCore CLI:
```
pip install bedrock-agentcore
```
- Install all project dependencies:
```
pip install -r requirements.txt
```

Create Short-Term AgentCore Memory

Before deploying your agent, you need to create a short-term memory store that will help your agent maintain conversation context:

Create a memory store with a 7-day default expiry period:

python3 resources/memory_manager.py create DataAnalystAssistantMemory ${MEMORY_ID_SSM_PARAMETER}

Validate that your memory store was created successfully:

# List all available memory stores to confirm creation
python3 resources/memory_manager.py list

This memory store enables your agent to remember previous interactions within the same session, providing a more coherent and contextual conversation experience.

Local Testing

Before deploying to AWS, you can test the Data Analyst Agent locally to verify functionality:

Start the local agent server:

python3 app.py

This launches a local server on port 8080 that simulates the AgentCore runtime environment:

Processes natural language queries about video game sales data
Uses AgentCore Memory (Short-Term Memory) to maintain conversation context
Maintains conversation history through the last_k_turns parameter

Test the agent with example queries using curl:

curl -X POST http://localhost:8080/invocations \
 -H "Content-Type: application/json" \
 -d '{"prompt": "Hello world!", "session_id": "b2e9c4f7-3a1d-8e5b-6c2f-9d4e7a8b5c3f", "last_k_turns": 20}'

curl -X POST http://localhost:8080/invocations \
 -H "Content-Type: application/json" \
 -d '{"prompt": "what is the structure of your data available?!", "session_id": "b2e9c4f7-3a1d-8e5b-6c2f-9d4e7a8b5c3f", "last_k_turns": 20}'

curl -X POST http://localhost:8080/invocations \
 -H "Content-Type: application/json" \
 -d '{"prompt": "Which developers tend to get the best reviews?", "session_id": "b2e9c4f7-3a1d-8e5b-6c2f-9d4e7a8b5c3f", "last_k_turns": 20}'

curl -X POST http://localhost:8080/invocations \
 -H "Content-Type: application/json" \
 -d '{"prompt": "Give me a summary of our conversation", "session_id": "b2e9c4f7-3a1d-8e5b-6c2f-9d4e7a8b5c3f", "last_k_turns": 20}'

Deploy the Strands Agent with Amazon Bedrock AgentCore

Deploy your agent to AWS with these simple steps:

Configure the agent deployment accepting default values when prompted:

agentcore configure --entrypoint app.py --name agentcoredataanalystassistant -er $AGENT_CORE_ROLE_EXECUTION

Launch the agent infrastructure:

agentcore launch

Testing the Deployed Agent

Test your deployed agent with these example queries:

agentcore invoke '{"prompt": "Hello world!", "session_id": "c5b8f1e4-9a2d-4c7f-8e1b-5a9c3f6d2e8a", "last_k_turns": 20}'

agentcore invoke '{"prompt": "what is the structure of your data available?!", "session_id": "c5b8f1e4-9a2d-4c7f-8e1b-5a9c3f6d2e8a", "last_k_turns": 20}'

agentcore invoke '{"prompt": "Which developers tend to get the best reviews?", "session_id": "c5b8f1e4-9a2d-4c7f-8e1b-5a9c3f6d2e8a", "last_k_turns": 20}'

agentcore invoke '{"prompt": "Give me a summary of our conversation", "session_id": "c5b8f1e4-9a2d-4c7f-8e1b-5a9c3f6d2e8a", "last_k_turns": 20}'

Expected Behavior: The agent responds as "Gus," a video game sales data analyst assistant who provides information about the video_games_sales_units database (64,016 game titles from 1971-2024), analyzes developer review scores, and maintains conversation context across interactions.

Next Step

You can now proceed to the Front-End Implementation - Integrating AgentCore with a Ready-to-Use Data Analyst Assistant Application).

Cleaning-up Resources (Optional)

To avoid unnecessary charges, delete the AgentCore run environment from the AWS Console.

Thank You

License

This project is licensed under the Apache-2.0 License.