Setting Up the Graph RAG Environment: Core Libraries

We are entering the "Implementation Phase" of the course. Up until now, we have talked about the Theory of graphs and the Logic of retrieval. Now, we are going to build it. To build a modern Graph RAG system, we need an orchestrator that can handle LLM calls, chain them together, and communicate with our graph database.

In this lesson, we will set up our Python Environment. We will install the three pillars of our stack: Neo4j (The Storage), LangChain (The Orchestration), and LangGraph (The Agentic Logic). We will also configure our API keys and verify the connection between our code and our graph.

---

1. The Python Stack for Graph RAG

We will use Python 3.10+ for this implementation.

Essential Libraries:

1. langchain / langchain-community: The standard for building LLM applications.
2. neo4j: The official driver for communicating with Neo4j.
3. langgraph: For building "Cyclic" agents that can query, reflect, and re-query the graph.
4. openai / langchain-openai (or Google Gemini/Anthropic): The "Brains" for generating Cypher.

---

2. Installation and Dependency Management

We recommend using a Virtual Environment (venv or conda).

# 1. Create a virtual environment
python -m venv venv
source venv/bin/activate

# 2. Install the Graph RAG stack
pip install langchain langchain-openai langchain-community neo4j langgraph python-dotenv

---

3. Configuration: The .env File

Never hardcode your credentials. Use a .env file to store your API keys and database URI.

OPENAI_API_KEY=sk-....
NEO4J_URI=bolt://localhost:7687
NEO4J_USERNAME=neo4j
NEO4J_PASSWORD=your_password

---

4. Verifying the Connection

Before we build complex chains, we must ensure our "Bridge" works.

import os
from dotenv import load_dotenv
from langchain_community.graphs import Neo4jGraph

load_dotenv()

# 1. Initialize the Graph connection
graph = Neo4jGraph(
    url=os.getenv("NEO4J_URI"),
    username=os.getenv("NEO4J_USERNAME"),
    password=os.getenv("NEO4J_PASSWORD")
)

# 2. Run a simple sanity check
# This returns the current schema of your graph
print(graph.get_schema)

# SUCCESS: If you see your node labels (e.g., :Person, :Project)
# your environment is ready for Module 10!

---

5. Summary and Exercises

Setting up the environment is the first step toward automation.

• LangChain provides the pre-built connectors for graphs.
• Environment Variables keep your production keys secure.
• Graph Object Initialization verifies that your schema is "Visible" to the AI.

Exercises

1. Environment Check: Run the Python script above. Do you see your graph schema? If not, check if your Docker container (from Module 7) is still running.
2. Library Exploration: Look up the Neo4jGraph class in the LangChain documentation. What is the refresh_schema() method used for?
3. Security Drill: Why is it safer to use the bolt:// protocol instead of http:// for your Neo4j connection string?

In the next lesson, we will let the AI write its first query: Using the Neo4jCypherChain for Easy Queries.