Langgraph Supervisor Cookbook

This project contains cookbook for the langgraph-supervisor. It provides examples and documentation to help users effectively utilize the langgraph-supervisor in their projects.

Overview

This repository contains a collection of Jupyter notebooks that demonstrate the use of the Langgraph-Supervisor framework for building advanced multi-agent systems. The focus is on creating sophisticated, production-style systems that can handle complex tasks through the orchestration of specialized agents.

Basics

The basics.ipynb notebook introduces the fundamental concepts of the Langgraph-Supervisor framework. It includes:

• Math and Research Experts: These agents are designed to work together under the supervision of a central controller, ensuring that tasks are delegated to the most appropriate agent.

• Customer Support Swarm: A system that utilizes multiple agents to handle customer support tasks.

Customer Support System

The customer_support_system.ipynb notebook showcases a sophisticated multi-agent customer support system. Key features include:

• Specialized Agents: The system includes agents for intent classification, FAQ response, troubleshooting, billing, escalation, sentiment analysis, and quality assurance.
• Supervisor Orchestration: A central supervisor routes conversations to the appropriate agent, tracks hand-offs, and enforces quality gates.
• Custom Handover Tools: These tools carry context, reason, and priority when agents transfer a customer.
• Rich Toolset: Includes diagnostics, billing actions, escalation tickets, sentiment analysis, and quality review.
• Metrics Tracking: Monitors interactions, escalations, resolutions, and per-agent usage.
• End-to-End Demonstration: Features three realistic support scenarios to illustrate the system's capabilities.

Getting Started

To get started with the langgraph-supervisor, you will need to set up your environment and obtain the necessary API keys.

Setting Up LLM Providers

To use language models with the langgraph-supervisor, you can choose from various providers. Below are instructions for setting up OpenRouter and Google. Feel free to add more providers as needed.

Using OpenRouter

To use OpenRouter, follow these steps to obtain and set up your API key:

1. Create an API Key:

   • Visit the OpenRouter to create a new API key.

2. Set Up Your API Key:

   • Add to your .env file:

   OPENROUTER_API_KEY=<YOUR_API_KEY_HERE>
   

Using Google AI Studio

To use the Gemini API, you need an API key. Follow these steps to obtain your API key:

1. Create an API Key:

   • Go to Google AI Studio.
   • Click on "Get a Gemini API key" to create a new API key.

2. Set Up Your API Key:

   • Add to your .env file:

   GEMINI_API_KEY=<YOUR_API_KEY_HERE>
   

Adding More Providers

If you wish to use other LLM providers, please follow their respective documentation to obtain API keys and set them up in your environment.

Running with uv in VS Code

To run the project with uv in VS Code, follow these steps:

1. Create uv virtual environment:

   • Run the following command to create a virtual environment:

     uv create .venv
     

2. Activate the virtual environment:

   • Use the following command to activate the virtual environment:

     source .venv/bin/activate
     

3. Synchronize your environment:

   • Run the following command to add your dependencies:

     uv sync
     

4. Create and Select Kernel in VS Code:

   • Open the project directory in VS Code.
   • When prompted to select a kernel, choose "Python Environments" and select the virtual environment you created earlier (e.g., .venv/bin/python on macOS and Linux, or .venv\Scripts\python on Windows).

Usage

Once you have your API key set up, you can start using the langgraph-supervisor by following the examples provided in the Jupyter notebooks.

Contributing

Contributions are welcome! Please feel free to submit a pull request or open an issue.

License

This project is licensed under the MIT License.