Level 1 · Entry Foundations¶
Goal: understand every moving part of a RAG pipeline — embeddings, chunking, vector storage, retrieval, and grounded generation — and ship a working docs Q&A bot built from those parts in plain Python.
Modules¶
- What Is RAG & Why It Exists
- Embeddings & Semantic Similarity
- Chunking Strategies
- Vector Stores (ChromaDB)
- Retrieval
- Prompt Assembly & Generation
- A Minimal End-to-End Pipeline
- Evaluating RAG
- Common Failure Modes
- Project — Docs Q&A Bot
By the end of this level you'll be able to take a folder of documents, index it into a vector store, answer questions over it with cited sources, measure how well your pipeline retrieves and answers, and recognize (and fix) the most common ways RAG systems fail.
Setup for this level
sentence-transformers and chromadb run locally and are free — no API
key, no network calls after the first model download. Only the generation
lessons (06, 07, 08, 10) additionally use the Anthropic API, which needs an
ANTHROPIC_API_KEY environment variable. Everything else runs without one.