RAG: build a chatbot over your own documents

12 steps · about 17 h 25 min · Intermediate

  1. Retrieval-Augmented Generation (RAG) (article, Pinecone, 15 min)

    It explains in plain terms why models need external data and how ingestion, retrieval and generation fit together, so the code later makes sense.

  2. Building a Retrieval Augmented Generation Chatbot: A Practical Walkthrough (guide, Obot, 25 min)

    It gives the whole chatbot as a small set of functions: split, embed, store, retrieve and answer with citations. That makes a good mental map before the longer course.

  3. Learn RAG From Scratch – Python AI Tutorial from a LangChain Engineer (video, freeCodeCamp.org / Lance Martin, 2 h 30 min)

    A LangChain engineer builds indexing, retrieval and generation step by step, then covers query translation and advanced patterns. The pipeline stops being a black box.

  4. rag-from-scratch (repo, LangChain (GitHub), 45 min)

    Running the course notebooks yourself turns the video into hands-on practice. Swap in your own documents here.

  5. Build Your Own RAG Chatbot with JavaScript! (video, freeCodeCamp.org / Ania Kubow, 1 h 30 min)

    If you work in JavaScript, this shows the same pipeline built as a deployed web chatbot. It also shows the idea is not tied to Python.

  6. Best Practices in RAG Evaluation: A Comprehensive Guide (article, Qdrant, 25 min)

    Your first chatbot will sometimes answer badly. This guide covers the usual causes (chunk size, overlap, embedding choice) and fixes such as re-ranking.

  7. RAG That Actually Works: Choosing Chunking, Embeddings, and Retrieval with an Eval-First Playbook (blog, AI Engineer Playbook, 20 min)

    It teaches you to judge chunking and embeddings by whether they retrieve the right evidence, and to treat keyword and vector search as complementary.

  8. Build a custom RAG agent with LangGraph (doc, Docs by LangChain, 40 min)

    Once the basic pipeline works, this official tutorial shows the next step. The model decides when to retrieve and when to answer directly.

  9. Retrieval (doc, Docs by LangChain, 20 min)

    It sets out the official ways to give a model retrieval, as a fixed pipeline or as an agent tool. Use it to choose between them.

  10. Retrieval Augmented Generation (guide, Pinecone, 1 h 30 min)

    A free ebook-style series on rerankers, two-stage retrieval, choosing embedding models and metrics-driven development. It's the way to improve retrieval beyond a prototype.

  11. Production RAG with LangChain & Vector Databases (video, freeCodeCamp.org / Paulo Dichone, 8 h)

    This longer course covers what tutorials usually skip: debugging, hybrid search, observability, scaling and hosting. Do it once you have a working prototype.

  12. Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks (reference, Lewis et al., arXiv, 45 min)

    The 2020 paper that introduced the term RAG. Reading it last shows how the research idea became the chatbot you built.