RAG Guide

RAG (Retrieval-Augmented Generation) combines the power of LLMs with your own documents. The AI can answer questions about your data - without training.

How RAG Works

1. User asks question
2. Question → Embedding Model
3. Embedding → Vector Database
4. Find similar documents
5. Documents + Question → LLM
6. LLM generates answer

Components

• Document Loader: PDF, Markdown, HTML, Text
• Text Splitter: Split into chunks
• Embedding Model: Convert to vectors
• Vector Database: Store and search
• LLM: Generate answer from context

Popular Tools

Basic RAG Pipeline

# 1. Load and split documents
from langchain_community.document_loaders import TextLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter

loader = TextLoader("my-docs.txt")
docs = loader.load()
splitter = RecursiveCharacterTextSplitter()
chunks = splitter.split_documents(docs)

# 2. Create embeddings and store
from langchain_community.embeddings import OllamaEmbeddings
from langchain_community.vectorstores import Chroma

embeddings = OllamaEmbeddings(model="nomic-embed-text")
db = Chroma.from_documents(chunks, embeddings)

# 3. Query
query = "What is our return policy?"
docs = db.similarity_search(query)

# 4. Get answer from LLM
from langchain_community.chat_models import ChatOllama
llm = ChatOllama(model="llama3:8b")
result = llm.invoke(f"Answer based on: {docs}")