LunaAI

LunaAI is a multi-agent personality system designed to produce the most human-like simulation of consciousness possible in conversation with modern LLMs.

Features

• Multi-Agent Architecture: Specialized agents handling different aspects of cognition and conversation
• Emergent Behaviors: Carefully structured system prompts, agent design, and tools allowing for emergent consciousness-like behaviors
• Emotion Processing: Simulation of emotional responses and relationship development
• Memory Systems: Short and long-term memory mechanisms for contextual awareness
• Self-Reflection: Internal thought processes for deeper reasoning capabilities
• Persona Evolution: Dynamic personality development based on interactions
• Multi-Provider Support: Flexible adapter interface supporting multiple LLM providers

LLM Provider Support

Luna now supports multiple LLM providers through a unified adapter interface:

• Anthropic: Claude models (default)
• OpenAI: GPT-4 models
• Google: Gemini models

You can easily switch between providers by setting the PROVIDER environment variable:

# Use Anthropic Claude (default)
export PROVIDER=anthropic

# Use OpenAI GPT
export PROVIDER=openai

# Use Google Gemini
export PROVIDER=gemini

Provider Feature Comparison

Feature	Anthropic	OpenAI	Gemini
Text Generation	✅	✅	✅
Tool Calling	✅	✅	✅
System Instructions	✅	✅	✅
Content Blocks	✅	❌	✅
Thinking Blocks	✅	❌	❌
Image Input	✅	✅	✅
Streaming	✅	✅	✅

Installation

# Clone the repository
git clone https://github.com/yourusername/luna-ai.git
cd luna-ai

# Set up development environment
./scripts/setup_dev.sh

# Set up your API keys (for the providers you want to use)
export ANTHROPIC_API_KEY=your_key_here
export OPENAI_API_KEY=your_key_here
export GEMINI_API_KEY=your_key_here

# Set your preferred provider (optional, defaults to anthropic)
export PROVIDER=anthropic

Development

Code Quality Tools

The project uses several tools to maintain code quality:

• Black for code formatting
• isort for import sorting
• flake8 for linting
• mypy for type checking
• pytest for testing

Development Commands

# Format code
./scripts/format.sh

# Run linting checks
./scripts/lint.sh

# Run tests
./scripts/test.sh

# Run specific test sets
./scripts/test.sh --unit
./scripts/test.sh --integration

# Run tests with coverage
./scripts/test.sh --coverage

# Compare responses from different LLM providers
./scripts/compare_adapters.py "Your question here?"

# Compare providers with tools enabled
./scripts/compare_adapters.py "Calculate 25 * 16" --with-tools

Project Structure

• adapters/: External service interfaces
• core/: Core application logic
• domain/: Business models and tools
  • models/: Domain model classes
  • tools/: Specialized tools for different capabilities
• services/: Business logic services
• system_prompts/: Agent prompts and configuration
• config/: Application configuration
• tests/: Unit and integration tests

License

MIT License