Orbit Playground

A comprehensive monorepo of beta samples and templates for AI integrations, SaaS API development, and Model Context Protocol (MCP) implementations. Built for sister companies to learn, adapt, and deploy advanced AI systems with real-world patterns designed for local development, sandbox testing, and training.

Quick Navigation

git clone <repository-url>
cd orbit-ai-resources
docker-compose up -d

Overview

Orbit Playground provides comprehensive, self-contained examples that demonstrate modern AI development patterns including MCP server creation, conversational agents, structured UI components, and seamless integration architectures.

Each sample is designed for development and training, well-documented, and designed for adaptation across different business domains.

Sample Projects

4 production-ready samples + 1 experimental for learning and adaptation:

Repository Structure

orbit-ai-resources/
├── README.md                    # This overview
├── CONTRIBUTING.md              # How to contribute
├── LICENSE                      # MIT license
├── docker-compose.yml           # Docker orchestration for all services
├── docs/                        # High-level documentation
│   └── getting_started.md      # Step-by-step setup guide
├── postman/                     # API testing collections
│   ├── README.md               # Testing guide and workflows
│   ├── environments/           # Pre-configured endpoints
│   └── collections/            # Comprehensive API test collections
├── playground/                  # Standalone AI playground
│   ├── README.md               # Playground setup and usage
│   └── Dockerfile              # Container configuration
└── samples/                     # Self-contained example projects
    ├── README.md               # Detailed sample comparison table
    ├── 01-saas-sample-kit/     # Complete SaaS + MCP implementation
    ├── 02-weather-forecast-agent/ # Weather system with UI components
    ├── 03-rag-kb-strands-basic/   # RAG system with AWS Bedrock KB
    ├── 04-web-scraper/         # Web scraping with ethical guardrails
    ├── 05-chat-widget/         # Embeddable AI chat widget
    └── 99-scratchpad/          # Experimental area (not in CI)

Architecture Patterns

Model Context Protocol (MCP) Servers

• Dual Transport Support: stdio for Claude Desktop, HTTP for web clients
• Tool Organization: System management, domain logic, analytics categories
• Error Handling: Consistent error formats with graceful degradation
• TypeScript Quality: Strict typing, ES2022 modules, comprehensive tooling

Conversational Agents

• Strands Framework: Production-ready agent development with Python
• Structured Components: Chart.js charts, metrics, tables, markdown output
• Session Management: Conversation memory with efficient persistence
• Multi-Model Support: Claude, AWS Bedrock Nova, knowledge base integration
• RAG Integration: Official Strands patterns for S3 Knowledge Base retrieval
• Web Data Access: Ethical scraping with question-directed extraction
• Full Observability: OpenTelemetry traces to LangSmith, metrics collection, rotating file logs

API Integration

• RESTful Design: Consistent response structures, proper status codes
• Development-First: Memory persistence with deterministic seeding
• Analytics Ready: Time series, aggregations, business intelligence patterns
• Security Patterns: Input validation, CORS, authentication templates

Learning Objectives

For Development Teams

• MCP Protocol: Build tool servers that work with any MCP-compatible client
• Agent Architecture: Create intelligent conversational systems with external capabilities
• API Design: Modern TypeScript/Express patterns with comprehensive tooling
• UI Components: Structured output that integrates with modern frontend frameworks
• RAG Systems: Implement knowledge retrieval with AWS Bedrock and S3
• Ethical Scraping: Build web data collection with compliance and guardrails
• Embeddable Widgets: Self-contained frontend components for website integration

For Sister Companies

• Integration Patterns: How to MCP-enable existing SaaS platforms
• Knowledge Systems: Build RAG solutions for document retrieval and Q&A
• Data Collection: Implement ethical web scraping for business intelligence
• Widget Development: Create embeddable chat widgets for customer engagement
• Deployment Strategies: From development to production with Docker and cloud providers
• Customization Guide: Adapt samples to your specific business domain
• Best Practices: Security, performance, and maintainability patterns

Development

Getting Started

Common Patterns

• Docker-First Development: Single command deployment with docker-compose up -d
• Memory-Only Development: Fast iteration with deterministic seed data
• Tool-First Design: Build MCP tools before agent integration
• Structured Output: Component-based responses for rich UI integration
• Multi-Transport: Support both CLI and web clients from the same codebase
• API Testing: Comprehensive Postman collections for all endpoints and workflows

Extension Points

• New Tools: Add business-specific MCP tools following existing patterns
• UI Components: Extend structured output schemas for your data types
• Agent Capabilities: Customize system prompts and conversation flows
• API Integration: Connect to your existing services and databases

Production Deployment

Scaling Considerations

• Database Integration: Replace memory stores with persistent databases
• Authentication: Add proper auth/authz layers
• Monitoring: Implement logging, metrics, and alerting
• Load Balancing: Horizontal scaling patterns for stateless services
• Container Orchestration: Scale from Docker Compose to Kubernetes or ECS

AWS Permissions

Required AWS Services

• AWS Bedrock: Model inference and Knowledge Base operations
• Amazon S3: Document storage and artifact management
• OpenSearch Serverless: Vector embeddings for Knowledge Bases

Service-Specific Setup

Contributing

We welcome contributions from sister companies and the broader community!

See CONTRIBUTING.md for detailed guidelines, code standards, and the review process.

License

See the LICENSE file for details.

Fork Workflow

For detailed fork setup and development workflow, see our Getting Started Guide.

Quick Summary:

1. Fork the repository on GitHub
2. Clone your fork locally
3. Add upstream remote
4. Create feature branches for your work
5. Keep your fork synced with upstream