Mini RAG lab

A laptop-scale lab: chunk a few docs, embed, retrieve, generate with citations—wiring Course 09 ideas into code.

RAG clicks when you see a wrong chunk cause a wrong answer.

1. Put 3–10 markdown files in a folder
2. Chunk + embed (local or API)
3. Ask a question → top-k chunks
4. Prompt model with chunks + require citations
5. Break it on purpose (bad chunking) and observe

Include one question whose answer is only in doc B; confirm retrieval returns B.

⚠️ Myth: You need a huge vector DB to learn RAG.

✓ Reality: A folder + cosine ranking teaches the loop.

• Course 09 RAG chapters; Course 21 setup + safe API calls
• https://www.ainerdnetwork.com/learn/rag-retrieval-augmented-generation ↗