AI Agent System

An AI Agent system that functions as a "digital twin" for team leaders within an organization, providing intelligent assistance, cross-team impact analysis, and automated artifact verification.

Project Structure

ai-agent-system/
├── frontend/     # Next.js chat application with strands integration
├── knowledge_base/    # Knowledge base service
├── strands_agents/    # StrandsAgents Python scripts

Getting Started

Typical prompts

• I want to build a Next.js App on Amazon Bedrock
• How can I use Amazon Bedrock and generative AI with LLMs to build an Internal Wiki
• Next.js で SEO 対策をする方法を教えてください。

Prerequisites

• Node.js >= 18.0.0
• npm >= 9.0.0

Installation

# Install dependencies for all workspaces
npm install

# Install frontend dependencies
npm install --workspace=frontend

# Install backend dependencies
npm install --workspace=backend

Development

# Start frontend development server
npm run dev

# Run tests across all workspaces
npm test

# Lint code across all workspaces
npm run lint

# Format code
npm run format

Build

# Build all workspaces
npm run build

# Type check all workspaces
npm run type-check

Architecture

The system follows a serverless architecture on AWS with:

• Frontend: React 18 + Vite SPA with Tailwind CSS
• Backend: AWS Lambda functions with API Gateway
• Database: DynamoDB + RDS PostgreSQL
• Search: Amazon Kendra
• Storage: S3 for documents and artifacts
• Authentication: IAM Identity Center/SAML
• AI Agents: StrandsAgents Python service for intelligent chat

StrandsAgents Integration

The system includes a Python FastAPI microservice built using strands-agents (sdk-python) that powers a three-agent pipeline:

• InfoCollector: extracts search keywords from user input
• PeopleFinder: enriches mock search results with the best person to consult based on an editable people influence graph and preferred contact method
• ResponseBuilder: tailors the final answer to the user's role and skills profile

Running StrandsAgents Service

1. Start the Python service:

./scripts/start-strands-service.sh

2. Configure the frontend:

cd frontend
cp .env.local.example .env.local
# Edit .env.local and set STRANDS_SERVICE_URL=http://localhost:8001

3. Start the Next.js chat application:

cd frontend
npm install
npm run dev

Open http://localhost:3000/chat and interact with the AI agents.

API Endpoints

• POST /api/strands - Chat with strands agents
• POST /agents/run - Direct strands service endpoint
• POST /search - Search functionality

Resource Tagging Strategy

All AWS resources are automatically tagged with a comprehensive tagging strategy that includes:

• Mandatory Tags: Project, Stage, ManagedBy, Component, Owner, CostCenter, Environment, CreatedDate, CreatedBy
• Resource-Specific Tags: Component-specific tags based on resource type (e.g., FunctionPurpose for Lambda, TablePurpose for DynamoDB)
• Environment-Specific Tags: Environment-appropriate values for cost allocation and lifecycle management
• Compliance Tags: DataClassification for data storage resources, ComplianceScope for production resources

For detailed information about the tagging strategy, see infrastructure/TAGGING_GOVERNANCE_POLICY.md.

Deployment

Infrastructure Deployment

Deploy infrastructure using the deployment scripts:

# Deploy to staging
./scripts/deploy-infrastructure.sh staging

# Deploy to production (requires confirmation)
./scripts/deploy-infrastructure.sh production

# Show diff only (no deployment)
./scripts/deploy-infrastructure.sh staging --diff-only

The deployment process includes:

1. Dependency installation and testing
2. Tag validation to ensure compliance
3. Security checks and CloudFormation synthesis
4. Infrastructure deployment with progress tracking
5. Post-deployment validation including tag verification
6. Documentation generation for tags and resources

Tag Validation

Before deployment, all resources are validated for:

• Mandatory tags (Project, Stage, Component, Owner, etc.)
• Resource-specific tags (FunctionPurpose, DataClassification, etc.)
• Environment-appropriate values
• Tag format and length constraints

Deployment will fail if tag validation does not pass.

Development Guidelines

• Follow TypeScript strict mode
• Maintain 80% test coverage minimum
• Use Prettier for code formatting
• Follow ESLint rules for code quality
• Write meaningful commit messages
• Follow infrastructure code review checklist for AWS resources
• Ensure all AWS resources have proper tags before deployment

Code Review Process

For infrastructure changes, follow the comprehensive checklist at infrastructure/CODE_REVIEW_CHECKLIST.md, which includes:

• General code quality requirements
• AWS CDK specific guidelines
• Resource tagging requirements (mandatory for all AWS resources)
• Security and compliance validation
• Deployment and operations checks

Resource Tagging Requirements

All AWS resources must have:

• Mandatory tags: Project, Stage, Component, Owner, CostCenter, etc.
• Resource-specific tags: Based on resource type (Lambda, DynamoDB, S3, etc.)
• Data classification tags: For all data storage resources
• Environment-specific tags: Appropriate for dev/staging/production

Tag validation runs automatically during deployment and will fail if required tags are missing.

AIAgentSample