LangGraph Multi-Agent Orchestrator

A configurable multi-agent orchestration system built with LangGraph Studio, featuring dynamic LLM-driven routing and comprehensive agent management.

Features

• LLM-Driven Orchestration: Intelligent routing based on query understanding
• Multiple Routing Strategies: LLM dynamic, rule-based, and hybrid approaches
• Configurable Agents: Specialized agents for different tasks (data analysis, research, code assistance, etc.)
• LangGraph Studio Integration: Visual workflow design and debugging
• Extensible Architecture: Easy to add custom agents and tools
• Mock Tools: Built-in mock implementations for testing

Quick Start

Prerequisites

• Python 3.10+
• Make (for using Makefile commands)

Installation

1. Clone the repository

2. Initialize the project:

   make init

3. Copy .env.example to .env and add your API keys:

   cp .env.example .env
   # Edit .env with your favorite editor

4. Start LangGraph Studio:

   make start

5. Open http://localhost:8123 in your browser

Project Structure

.
├── langgraph.json              # Main LangGraph configuration
├── orchestrator/               # Orchestrator configuration
│   ├── config.json            # Core orchestrator settings
│   ├── routing_strategies/    # Routing strategy definitions
│   └── prompts/              # Orchestrator prompts
├── agents/                    # Agent definitions
│   ├── registry.json         # Agent registry
│   └── specialized/          # Specialized agent configs
├── tools/                     # Tool implementations
│   ├── core/                 # Core tools (web search, data processing)
│   └── custom/               # Custom tools
├── graphs/                    # Graph templates
└── examples/                  # Example workflows

Available Commands

make help        # Show all available commands
make init        # Initialize the project
make start       # Start LangGraph Studio
make dev         # Start in development mode with auto-reload
make validate    # Validate all configuration files
make test        # Run tests
make clean       # Clean up generated files

Configuration

Main Configuration (langgraph.json)

The main configuration file defines:

• Python version requirements
• Orchestrator settings
• Agent registry location
• LangGraph Studio settings

Orchestrator Configuration

Located in orchestrator/config.json, defines:

• Model settings (provider, temperature, etc.)
• Routing strategy
• Graph workflow definition
• Error handling policies

Agent Registry

The agents/registry.json file maintains a list of all available agents with:

• Capabilities
• Requirements
• Cost and latency estimates
• Routing hints

Adding Custom Agents

1. Create a new agent directory:

   make new-agent

2. Update the agent configuration in the created directory

3. Register the agent in agents/registry.json

4. Restart LangGraph Studio

Routing Strategies

LLM Dynamic

Fully AI-driven routing based on query analysis and agent capabilities.

Rule-Based

Deterministic routing using keyword and pattern matching.

Hybrid

Combines rule-based and LLM approaches for optimal performance.

Examples

Run examples with:

make run-example EXAMPLE=simple_query

Available examples:

• simple_query: Basic single-agent query
• complex_workflow: Multi-agent workflow
• custom_agent: Custom agent implementation

Development

Running Tests

make test

Code Formatting

make format

Linting

make lint

Environment Variables

Key environment variables (set in .env):

• OPENAI_API_KEY: OpenAI API key
• ANTHROPIC_API_KEY: Anthropic API key
• LANGGRAPH_STUDIO_PORT: Studio port (default: 8123)
• ORCHESTRATOR_MAX_ITERATIONS: Max orchestration iterations
• AGENT_VALIDATION: Validation strictness level

Troubleshooting

Studio Won't Start

• Check that all dependencies are installed: make install
• Verify .env file exists and has valid API keys
• Check port 8123 is not in use

Configuration Validation Errors

Run make validate to check all configurations

Agent Not Found

• Ensure agent is registered in agents/registry.json
• Verify agent configuration file exists at the specified path

License

[Your License Here]

Contributing

[Contributing Guidelines]