A RAG (Retrieval-Augmented Generation) chatbot answers questions based on your own documents — not just its training data. This guide builds one from scratch using Python, ChromaDB, and Claude.

Originally published at kalyna.pro

What Is RAG?

RAG combines two things:

Retrieval : search your documents for relevant chunks

Generation : use an LLM to write an answer based on those chunks

Without RAG, Claude can only answer questions based on its training data. With RAG, you inject relevant context directly into the prompt.

Architecture

Indexing : load docs → split into chunks → embed → store in vector DB

Querying : embed question → find similar chunks → send to Claude → return answer

Step 1: Install Dependencies

pip install anthropic chromadb sentence-transformers pypdf2

Enter fullscreen mode

Exit fullscreen mode

Step 2: Load and Chunk Documents

import PyPDF2 from pathlib import Path

def load_pdf ( path : str ) -> str : with open ( path , " rb " ) as f : reader = PyPDF2 .…