Build a RAG Pipeline from Scratch

Prerequisites: Python 3.9+, comfort with the command line, an Anthropic API key. No GPU required.

Retrieval-Augmented Generation (RAG) gives a language model access to a knowledge base at query time so it can ground its answers in real documents rather than relying only on what it memorized during training. The pipeline has two phases: an indexing phase (load documents, split into chunks, embed each chunk, store the embeddings) that runs once or periodically, and a retrieval phase (embed the incoming question, search for the closest chunks, hand them to the LLM) that runs on every query.

You'll build this with three free, open-source components — Python's text splitting, the sentence-transformers library for embeddings (runs locally, no API key needed), and ChromaDB as the vector store — and use the Claude API only for the final generation step, which keeps the indexing phase completely free to experiment with.