Simple LangChain Agent

Introduction

In this lesson, you’ll build your first LangChain agent that can interact with your Neo4j knowledge graph.

Key insight: The agent chooses which tool to use based on your question.

When you ask: "Summarize the schema of the graph database"

Agent receives your natural language question
Agent decides to use the Get-graph-database-schema tool
Tool executes and returns the Neo4j schema
Agent processes the schema and creates a summary
Agent responds in natural language

What You’ll Build

Open the notebook: 02_01_simple_agent.ipynb

This agent:

Uses LangChain and LangGraph for agent functionality
Connects to your Neo4j database
Has a single tool: Get-graph-database-schema
Can answer questions about your graph structure

The goal: Understand basic agent architecture before adding retriever capabilities.

== Agent Architecture

python

# 1. Initialize LLM
model = init_chat_model("gpt-4o", model_provider="openai")

# 2. Connect to Neo4j
graph = Neo4jGraph(url=NEO4J_URI, username=username, password=password)

# 3. Define tools
@tool("Get-graph-database-schema")
def get_schema():
    """Get the schema of the graph database."""
    return graph.schema

# 4. Create agent
agent = create_react_agent(model, tools)

Try These Questions

Schema exploration:

"Summarize the schema of the graph database"
"How are Products related to other entities?"
"What questions can I answer using this graph database?"

Perfect for understanding data structure before building queries!

Summary

In this lesson, you built your first agent with schema introspection capabilities:

Key Concepts:

Schema tool: Database structure exploration through conversation
Agent setup: LangChain ReAct agent with introspection tool
LangGraph: Modern agent framework for tool orchestration
Graph structure: Understanding nodes, relationships, and properties

Next lesson: Add Vector Retriever tool for semantic content search.

Conversation flow: Natural language → tool selection → retriever execution → formatted response

What You Built:

Agent that can answer questions using semantic search
Conversational interface to graph database schema
Foundation for adding more retriever tools

Limitations:

Only semantic search (no relationship traversal)
No precise queries or aggregations
Single tool agent

In the next lesson, you will enhance your agent by adding the Vector + Cypher Retriever tool for contextual relationships and richer answers.

Neo4j & Generative AI Hands-On Workshop

Generative AI

Retrievers

Agents

Simple LangChain Agent

Introduction

What You’ll Build

Try These Questions

Summary

Chatbot