rag_basic

Basic RAG Example

This example demonstrates a basic Retrieval-Augmented Generation (RAG) pipeline using LangGraphGo.

Overview

The basic RAG pattern follows these steps:

1. Retrieve: Find relevant documents from a vector store based on semantic similarity
2. Generate: Use an LLM to generate an answer based on the retrieved context

This is the simplest and most straightforward RAG implementation, suitable for:

• Quick prototyping
• Simple Q&A systems
• Applications with high-quality document collections

Features

• Vector-based document retrieval using embeddings
• In-memory vector store for fast similarity search
• LLM-based answer generation with context
• Visualization of the RAG pipeline

Running the Example

cd examples/rag_basic
go run main.go

Key Components

• Document Store: In-memory vector store with mock embeddings
• Retriever: Vector store retriever that finds top-k similar documents
• LLM: DeepSeek-v3 for answer generation
• Pipeline: Basic RAG pipeline (Retrieve → Generate)

Example Output

The example runs several queries and shows:

• Retrieved documents with sources
• Generated answers based on context
• Pipeline visualization in Mermaid format

Customization

You can customize:

• Number of documents to retrieve (TopK)
• System prompt for the LLM
• Document corpus
• Embedding dimension