NerdearLA 2025 - LangChain & LangGraph Workshop

A comprehensive workshop demonstrating the evolution from simple AI responses to complex agent workflows using LangChain, LangGraph, and MCP (Model Context Protocol) servers.

🚀 Workshop Overview

This repository contains a progressive series of examples showcasing:

• Basic AI agent responses
• Tool integration with agents
• MCP server architecture
• Manual graph construction with LangGraph
• Interactive workflows with user permissions
• Multi-server deployments

📋 Prerequisites

• Python 3.12+
• uv package manager
• OpenAI API key

🛠️ Setup

1. Clone and navigate to the repository

cd 08_nerdearla

2. Set up environment variables

cp .env.example .env
# Edit .env and add your OpenAI API key

3. Install dependencies

uv sync

4. Initialize the database

cd 00_tables_creation
# Run the SQL commands from init_db.sql to create data.db in the parent folder
cd ..

📚 Workshop Modules

00 - Database Setup

Location: 00_tables_creation/

Sets up SQLite database with sample sales data for the workshop examples.

Files:

• init_db.sql - Database schema and sample data

01 - Simple Response

Location: 01_simple_response/

Basic implementation of an AI agent that can query a database and provide simple responses.

Run:

cd 01_simple_response
uv run main.py

Example queries:

• "Give me the sales grouped by genre"
• "Give me the total sales grouped by item category"

02 - React Agent with Tools

Location: 02_react_agent_with_tools/

Introduction to ReAct (Reasoning + Acting) agents with tool calling capabilities. Note: This implementation doesn't have memory between interactions.

Run:

cd 02_react_agent_with_tools
uv run gradio_interface.py

Features:

• Gradio web interface
• Synchronous and asynchronous response modes
• Database query tools

03 - MCP Tools

Location: 03_mcp_tools/

Demonstrates Model Context Protocol (MCP) server architecture for tool isolation and modularity.

Run:

cd 03_mcp_tools
# Start MCP server
uv run tool_mcp.py

# In another terminal, start the interface
uv run gradio_interface.py

Benefits:

• Tool isolation
• Modular architecture
• Same frontend interface

04 - LangGraph Manual Graph

Location: 04_langgraph_manual_graph/

Manual construction of ReAct agents using LangGraph instead of built-in functions, providing more control over the agent workflow.

Implementations:

1. 01_agent_simple_decoupled.py - Decoupled MCP connection
2. 01_agent_simple.py - Standard implementation
3. decoupled_yield.py - Yield-based approach

Run with LangGraph Studio:

cd 04_langgraph_manual_graph
uv run langgraph dev

Features:

• Visual graph execution monitoring
• Custom agent compilation
• Multiple MCP connection patterns

05 - Graph with Pause

Location: 05_graph_with_pause/

Enhanced workflow with user permission system - the agent asks for permission before executing tools.

Run:

cd 05_graph_with_pause
uv run gradio_interface.py

Features:

• Interactive tool execution approval
• User feedback integration
• Custom checkpoint management

LangGraph Studio:

# Comment the checkpointer line in agent.py first
uv run langgraph dev

06 - Multiple MCP Servers

Location: 06_langgraph_with_multiple_servers/

Demonstrates deployment of multiple specialized MCP servers for different capabilities.

Available Servers:

1. Image Analysis - JPG image processing
2. GitHub Integration - Repository interactions
3. Database Tools - SQL query capabilities

Use Cases:

• Analyze data from images
• Fix GitHub issues automatically
• Extract database insights

Exercise: Implement simultaneous connection to multiple MCP servers using AsyncExitStack from contextlib.

07 - Images

Location: 00_images/

Contains workshop screenshots and visual documentation.

🎯 Learning Path

1. Start with basics (01) - Understand simple AI responses
2. Add complexity (02) - Learn ReAct patterns and tools
3. Introduce MCP (03) - Understand server architecture
4. Manual control (04) - Build custom LangGraph workflows
5. Interactive flows (05) - Add user decision points
6. Scale up (06) - Deploy multiple specialized servers

🔧 Key Technologies

• LangChain: Framework for building applications with LLMs
• LangGraph: Library for building stateful, multi-actor applications
• MCP (Model Context Protocol): Standard for connecting AI models to external tools
• Gradio: Web interface for machine learning applications
• SQLite: Lightweight database for examples

📝 Workshop Session Details

Date: March 23, 2025 Time: 11:50 AM - 1:00 PM (1 hour 10 minutes) Format: Online workshop with live coding demonstrations

🎥 Watch the Workshop: YouTube Live Stream

🤝 Contributing

This workshop is designed for educational purposes. Feel free to experiment with the examples and extend them for your own learning.

📄 License

Educational content for NerdearLA 2025 workshop.