Learn LangGraph

Monorepo for learning how to build stateful LLM workflows with LangGraph.

It contains two small, practical projects:

• calculator: tool-calling arithmetic agent
• support_email: customer support workflow with routing and human review

What Is in This Monorepo

1) calculator (Beginner project)

A compact project to understand the LangGraph fundamentals:

• graph state
• nodes
• conditional routing
• tool execution loop

How it works:

1. llm_call receives messages and decides whether to call a tool.
2. tool_node executes one of the arithmetic tools (add, subtract, multiply, divide).
3. should_continue loops back to the LLM or ends the graph.

Main files:

• calculator/tools.py: model setup + tools
• calculator/state.py: graph state schema
• calculator/nodes.py: node logic
• calculator/build_compile.py: build/compile graph, generate image
• calculator/invoke.py: runnable example

2) support_email (Intermediate project)

A larger workflow showing multi-step orchestration:

• read incoming email
• classify intent and urgency
• route to docs search or bug tracking
• draft response
• pause for human approval (interrupt)
• resume and send reply

Main files:

• support_email/state.py: typed workflow state
• support_email/compile.py: graph assembly + memory checkpointer
• support_email/nodes/*: business steps (classification, search, response)
• support_email/run_agent.py: end-to-end example with resume

Quick Start

Requirements

• Python 3.11+ (3.12 recommended)
• virtual environment
• provider API key in .env

Setup

python -m venv venv
.\venv\Scripts\Activate.ps1
pip install langgraph langchain langchain-deepseek langchain-openai python-dotenv ipython taskipy

Create .env at the project root:

DEEPSEEK_API_KEY=your_key_here

Run

With tasks (from pyproject.toml):

task calculator
task support_email

Or directly:

python -m calculator.invoke
python -m support_email.run_agent

To generate the calculator graph image:

python -m calculator.build_compile

Learning Path

Recommended order:

1. Start with calculator to learn LangGraph basics.
2. Move to support_email to learn branching, persistence, and human-in-the-loop review.