langchain-ai/rag-from-scratch — reverse-engineered prompt

Build me a beginner friendly notebook series that teaches retrieval augmented generation from scratch. I want to understand why normal language models struggle with private or recent information, then learn how to connect them to outside documents so answers are grounded in retrieved context.

Start simple and build up step by step. Show how to load documents, split them, index them, search for the most relevant pieces, and pass that context into a model to generate an answer. Then expand into more advanced RAG ideas across later notebooks, while keeping the explanations clear enough for someone who is learning.

Please make it practical, with runnable Jupyter notebooks, short explanations beside the code, and small examples I can follow without already being an expert. Use LangChain style patterns where appropriate, and look up current docs online if you need to. The end result should feel like a hands on course that could go along with a video playlist.