default

go-light-rag Example Implementation

This example demonstrates how to use the go-light-rag library to build a document retrieval system. The implementation showcases the library's core functionality in a straightforward manner, using the default handler which closely mirrors the behavior of the official Python LightRAG implementation.

Prerequisites

Before running this example, you'll need:

• A graph database (Neo4J or MemGraph)
• An OpenAI API key
• A text document for processing

Setup Instructions

1. Clone & Navigate

After cloning the repository, change to the example directory:

cd example/default

2. Set Up Graph Database

Choose one of these options:

• Create a free account on Neo4J
• Self-host a compatible graph database like MemGraph

Make note of your graph database URI, username, and password.

3. Get OpenAI API Key

Create an account on OpenAI and generate an API key.

4. Prepare Your Document

Create a text file named book.txt containing the document you want to analyze. The default entity types in this example are configured for A Christmas Carol by Charles Dickens.

5. Configure the Application

Copy the example configuration file and update it with your credentials:

cp config.example.yaml config.yaml

Edit config.yaml to include your:

• Neo4J connection details
• OpenAI API key and model
• Preferred logging level

Running the Example

Execute the application:

go run main.go

First run: The system will process your document and build the RAG database (approximately 5 minutes depending on document size).

Subsequent runs: As long as the document hasn't changed and the database files (vec.db and kv.db) exist, the system will skip processing and directly open the query interface.

How It Works

The system:

1. Embeds document text using OpenAI
2. Stores vectors in ChromeM and metadata in BoltDB
3. Creates a knowledge graph in Neo4J
4. Provides an interactive query interface
5. Retrieves relevant information for each query
6. Uses a prompt template taken directly from the official Python implementation to ensure consistent output quality

Demo

Note: The demo video is played at 2x speed and uses the gpt-4o-mini model.