LangGraph Learning Repository

This repository contains a collection of Jupyter notebooks exploring various concepts and patterns in LangGraph, a library for building stateful, multi-actor applications with LLMs (Large Language Models).

Overview

LangGraph enables the creation of complex AI workflows with multiple agents, state management, and flexible graph-based architectures. This repository serves as a learning resource covering fundamental concepts to advanced patterns.

Prerequisites

• Python 3.11+
• Virtual environment (recommended)
• API keys for LLM providers (Anthropic, OpenAI, etc.)
• Environment variables configured in .env file

Installation

1. Create and activate a virtual environment:

python -m venv langgraph
source langgraph/bin/activate  # On Windows: langgraph\Scripts\activate

2. Install dependencies:

pip install langgraph langchain langchain-openai langchain-anthropic langchain-community langchain-tavily langchain-text-splitters

3. Set up environment variables: Create a .env file in the root directory with your API keys:

ANTHROPIC_API_KEY=your_key_here
OPENAI_API_KEY=your_key_here
TAVILY_API_KEY=your_key_here

Notebooks

1. Simple Chatbot (simple_chatbot.ipynb & simple_chatbot.py)

• Description: Basic introduction to LangGraph with a simple chatbot implementation
• Concepts:
  • StateGraph creation
  • Node and edge definition
  • Basic message handling
  • Streaming responses
• Key Learning: Understanding the fundamental structure of a LangGraph application

2. Chatbot with Conversation Memory (chatbot_with_conversation_memory.ipynb)

• Description: Extends the simple chatbot with persistent conversation memory
• Concepts:
  • MemorySaver for checkpointing
  • State persistence across conversations
  • Production-ready memory solutions (SqliteSaver, PostgresSaver)
• Key Learning: How to maintain conversation context and state

3. Chatbot with Search (chatbot_with_search.ipynb)

• Description: Adds web search capabilities to the chatbot using Tavily
• Concepts:
  • Tool binding to LLMs
  • Tool invocation patterns
  • Integrating external APIs
• Key Learning: Extending agent capabilities with tools

4. Exploring Agents (exploring_agents.ipynb)

• Description: Exploration of prebuilt ReAct agents and their capabilities
• Concepts:
  • create_react_agent from langgraph.prebuilt
  • Structured outputs with agents
  • Custom prompts for agents
  • Model initialization patterns
• Key Learning: Using prebuilt agent components vs building from scratch

5. Agent Supervisor (agent_supervisor.ipynb)

• Description: Multi-agent system with a supervisor agent coordinating specialized agents
• Concepts:
  • Supervisor pattern
  • Specialized agent creation (research agent)
  • Multi-agent coordination
  • Agent routing and delegation
• Key Learning: Building complex multi-agent systems

6. Async LangGraph (async_langgraph.ipynb)

• Description: Asynchronous patterns in LangGraph
• Concepts:
  • Async/await with LangGraph
  • ainvoke for async execution
  • Async node definitions
• Key Learning: Performance optimization through async operations

7. Agentic RAG (agentic_rag.ipynb)

• Description: Retrieval-Augmented Generation implementation using agents
• Concepts:
  • Document loading and splitting
  • Vector store creation (InMemoryVectorStore)
  • Embeddings and retrieval
  • RAG patterns with LangGraph
• Key Learning: Combining retrieval systems with agent workflows

8. Corrective RAG (corrective_rag.ipynb)

• Description: Advanced RAG pattern with retrieval grading and correction
• Concepts:
  • Retrieval quality assessment
  • Self-correction mechanisms
  • Advanced RAG workflows
• Key Learning: Improving RAG accuracy through feedback loops

9. LangGraph Agents (langgraph_agents.ipynb)

• Description: Advanced agent patterns including recursion limits
• Concepts:
  • Recursion limits and error handling
  • GraphRecursionError management
  • Configuring agent iteration limits
  • Tool integration patterns
• Key Learning: Controlling agent behavior and preventing infinite loops

Project Structure

LangGraph/
├── README.md                          # This file
├── .gitignore                         # Git ignore rules
├── documentation.txt                  # Reference links and package ecosystem
├── simple_chatbot.py                  # Simple chatbot script
├── simple_chatbot.ipynb               # Simple chatbot notebook
├── chatbot_with_conversation_memory.ipynb
├── chatbot_with_search.ipynb
├── exploring_agents.ipynb
├── agent_supervisor.ipynb
├── async_langgraph.ipynb
├── agentic_rag.ipynb
├── corrective_rag.ipynb
├── langgraph_agents.ipynb
└── langgraph/                         # Virtual environment directory

Key Concepts Covered

State Management

• TypedDict for state definition
• Annotated types with reducer functions (add_messages)
• State updates and persistence

Graph Construction

• StateGraph creation
• Node addition and configuration
• Edge definition (START, END, conditional edges)
• Graph compilation

Agents

• ReAct agent pattern
• Tool integration
• Structured outputs
• Custom prompts and configurations

Advanced Patterns

• Multi-agent systems
• Supervisor patterns
• RAG implementations
• Async execution
• Error handling and recursion limits

LangGraph Package Ecosystem

As documented in documentation.txt, the LangGraph ecosystem includes:

• langgraph (base): Core LangGraph functionality
• langgraph-prebuilt: Prebuilt components for creating agents
• langgraph-supervisor: Tools for building supervisor agents
• langgraph-swarm: Tools for building swarm multi-agent systems
• langchain-mcp-adapters: Interfaces to MCP servers
• langmem: Agent memory management (short-term and long-term)
• agentevals: Utilities to evaluate agent performance

Resources

• LangGraph Documentation
• LangGraph Tutorials
• ReAct Agent Guide
• Agent Overview
• Tool Usage Documentation

Notes

• All notebooks use environment variables from .env file
• The virtual environment directory (langgraph/) is excluded from git
• Some notebooks may require additional dependencies (check individual notebook cells)
• API keys should never be committed to version control

Contributing

This is a personal learning repository. Feel free to explore and adapt the code for your own learning purposes.

License

This repository is for educational purposes.