Skip to content

10.08 Setting Up a ReAct Agent Project

Before writing any LangGraph code, we need to set up a clean, secure project environment. Setting up your workspace correctly from Minute 1 is the difference between a project that works smoothly and one that fails with endless "Module Not Found" or "Invalid API Key" errors.

[!TIP] Beginner's Checklist: The Anatomy of an AI Project Every AI agent project requires three things: 1. Packages: The necessary libraries (like langgraph or openai). 2. Environment Variables: Secret keys (like passwords) to talk to OpenAI or Google search. 3. Structure: The Python files organizing your graph.

1. Project Initialization & Packages

We will use uv (the fast, modern Python package manager) to install our dependencies.

  1. Create the Project: Navigate to your project folder using your terminal or command prompt.
  2. Initialize: Run uv init (this sets up a basic pyproject.toml file).
  3. Install Packages: Run the following command:
uv add langgraph langchain-openai langchain-community python-dotenv

What are these packages doing?

  • langgraph: The orchestration framework to build our nodes and edges.
  • langchain-openai: The unified connection module to talk to OpenAI's models (like GPT-4).
  • langchain-community: A massive library of community-built tools (like Wikipedia search, calculcators, or Tavily web search).
  • python-dotenv: A small utility that securely loads our secret API keys into our Python script.

(Optional but recommended formatting tools: uv add --dev black isort)

2. Managing API Keys (The .env File)

To use tools like OpenAI or web search, you need API keys. Never paste API keys directly into your Python files! If you commit them to GitHub by mistake, bots will find them and spend your money.

Step-by-Step Security

  1. Create a plain text file in the root of your project called .env (just .env, nothing before the dot).
  2. Inside .env, add your keys like this:
# .env file
OPENAI_API_KEY=sk-proj-your-actual-key-here...
TAVILY_API_KEY=tvly-your-actual-key-here...

# Enable the LangSmith Debugging Dashboard!
LANGSMITH_TRACING=true
LANGSMITH_API_KEY=lsv2_your-actual-key-here...
LANGSMITH_PROJECT=MyFirstLangGraphAgent
  1. Create exactly one more file called .gitignore. Add .env to it. This guarantees your secret keys will absolutely never be uploaded to GitHub.

3. Project Structure: Organizing the Code

For a LangGraph project, it is considered best practice to separate your agent's reasoning from its execution architecture. Don't put everything in one massive main.py file!

A beginner-friendly project structure looks like this:

my-react-agent/
├── .env                  <-- Your secret keys
├── .gitignore            <-- Tells GitHub to ignore .env
├── pyproject.toml        <-- Installed packages list (built by 'uv')
│
├── react.py              <-- The Agent's "Brain" (The LLM + The Tools)
├── nodes.py              <-- The Graph's "Workers" (The Reasoning Node + Executer Node)
├── graph.py              <-- The Graph's "Map" (Connecting the nodes and edges)
└── main.py               <-- The "Ignition" (The entry point to run the app)

Visualizing the File Architecture: 

graph TD
    classDef file fill:#f9f9f9,stroke:#333,stroke-width:2px;

    A[main.py: Ignition & Entry]:::file --> B(graph.py: The Map):::file
    B --> C(nodes.py: The Workers):::file
    C --> D(react.py: The Brain & Toolbox):::file

    style A stroke:#4CAF50,stroke-width:4px

4. The Entry Point (main.py)

To ensure everything is wired correctly, we usually start by writing a tiny script in main.py that simply loads our environment variables. 

# main.py
from dotenv import load_dotenv
import os

# 1. Look for the .env file and securely load the keys!
load_dotenv()

# 2. Check to see if it worked
if not os.getenv("OPENAI_API_KEY"):
    print("WARNING: OpenAI API Key is missing. The agent will crash!")

# (Later, we will import our compiled 'graph' here and run it.)

Summary

You have now properly initialized your environment. Your packages are installed, your API keys are secured in your .env file, and your workspace is structured logically. You are now ready to start coding the "brain" of the agent in react.py.