RAG Crash Course

RAG Crash Course teaches retrieval-augmented generation by having the learner build every part of a question-answering system themselves: calling a language model, creating embeddings, chunking text, and searching a vector database, using a fictional company's documents as the working example. No prior machine-learning knowledge is assumed.

The course is organized as seven ordered Jupyter notebooks, moving from what RAG is and why it is needed, through embeddings, chunking and vector search, to assembling a complete pipeline by hand and then reproducing it with LangChain in far less code. A setup guide covers installing Python, getting an API key, and an alternative that uses local Ollama models at no cost. Running the whole course with the cheapest OpenAI models typically costs a few cents.