LangGraph Multi-Agent Orchestration

This repository contains a series of notebooks and scripts exploring the capabilities of LangGraph for building agentic workflows, from simple ReAct Agents to complex Multi-Agent Orchestration with memory and human-in-the-loop features.

Table of Contents

1. Introduction
2. Key Concepts
3. Project Structure
4. Configuration
5. References

Introduction

The project is focused on learning and implementing various patterns of AI Agents using the LangGraph framework. It covers the transition from basic reasoning-acting loops to production-ready assistants.

This project is entirely based on the course "LangGraph: Orquestrando agentes e multiagentes" by Alura. The course, in Brazilian Portuguese, can be found here

Key Concepts

• ReAct Agents: Implementing the Reasoning + Acting pattern for autonomous task execution.
• Nodes and Edges: Building graph-based workflows where nodes represent computation and edges represent the flow of control.
• State Management: Using global state to track information across agent turns.
• Agentic Search: Integrating external search capabilities (Tavily) for real-time information retrieval.
• Persistence (Memory): Implementing checkpointers to allow agents to maintain state across sessions.
• Human-in-the-Loop (HITL): Integrating human approval or correction steps in automated workflows.
• Multi-Agent Orchestration: Coordinating multiple specialized agents to solve complex tasks.

Project Structure

• 01_react_agent.ipynb: Introduction to the ReAct pattern.
• 02_langgraph_components.ipynb: Deep dive into LangGraph nodes, edges, and state.
• 03_agentic_search.ipynb: Implementing agents with search tools.
• 04_persistence_and_streaming.ipynb: Adding memory using SQLite checkpointers.
• 05_human_in_the_loop.ipynb: Adding breakpoints for human intervention.
• 06_multiagent_orchestration.ipynb: Orchestrating a multi-agent team for content creation.
• 07_mail_assistant.ipynb: A practical email triage assistant implementation.
• 08_mail_assistant_with_memory_and_hitl.ipynb: Enhanced email assistant with full features.
• prompts.py: Centralized management of system and user prompts.

Configuration

Create a .env file in the root directory and add your API keys:

GEMINI_API_KEY=your_gemini_api_key
TAVILY_API_KEY=your_tavily_api_key

References

• LangGraph: Orquestrando agentes e multiagentes

Author

Luis Fernando Torres

---