Gitika's AgentCore Testing :)

Testing Agentcore Runtime primitive

Files

Agent Implementation:

• src/agents/agent_runtime.py - FastAPI server with basic agent functionality
• src/agents/agent_native_memory.py - Agent with native memory integration using AgentCoreMemorySessionManager
• src/agents/agent_with_built_in_tools.py - Flask app with built-in code interpreter tool
• src/docker/Dockerfile.runtime - Docker configuration for runtime agent
• src/docker/Dockerfile.native - Docker configuration for native memory agent
• src/docker/pyproject.toml - Python dependencies

Build & Deploy:

• scripts/build_and_push.sh - Build and push Docker image to ECR
• Usage:
  • sh scripts/build_and_push.sh runtime - Deploy basic agent
  • sh scripts/build_and_push.sh native - Deploy memory agent

Testing:

• tests/test_agent_runtime.py - Test script for deployed agent
• test_memory_primitive.py - Test script for memory functionality
• tests/test_code_interpreter.py - Test script for code interpreter functionality
• Usage: python tests/test_agent_runtime.py

Quick Start

Basic Runtime Agent

1. Build and deploy:

   sh scripts/build_and_push.sh runtime

2. Update Bedrock AgentCore with new image URI

3. Test the deployed agent:

   python tests/test_agent_runtime.py

Memory Agent

1. Build and deploy:

   sh scripts/build_and_push.sh native

2. Update Bedrock AgentCore with new image URI

3. Test memory functionality:

   python test_memory_primitive.py

Code Interpreter Agent

1. Run locally:

   cd src/agents
   python agent_with_built_in_tools.py

2. Test code interpreter functionality:

   python tests/test_code_interpreter.py
   # Use http://localhost:8080 when prompted for endpoint

Memory Integration

The native memory agent uses AgentCoreMemorySessionManager which:

• Automatically saves conversation history to AgentCore Memory
• Retrieves relevant context for each interaction
• Provides transparent memory management without manual tool calls

Code Interpreter Integration

The code interpreter agent uses CodeInterpreter from bedrock_agentcore.tools which:

• Executes Python code in a secure sandbox environment
• Handles portfolio calculations and data analysis
• Maintains separate code execution sessions per user

Configuration

• ECR Repository: 998846730471.dkr.ecr.us-east-1.amazonaws.com/agentcore-test
• Agent ARN: arn:aws:bedrock-agentcore:us-east-1:998846730471:runtime/first_testing_agent-qt0qXTEqu4
• Memory ID: TestMemory_20250926064545-57A2248i7b