RAG cheat sheet

1. 1RAG basicsWhat retrieval augmented generation is, the two halves of every pipeline and when it is the right tool.6 concepts

   1. 1Retrieval augmented generationA pattern where your code first searches your own documents for passages relevant to a question, then gives those passages to an LLM to answer from.
   2. 2Indexing and query pipelinesIndexing runs ahead of time: load, clean, chunk, embed and store documents.
   3. 3RAG vs fine-tuningRAG adds knowledge at question time; fine-tuning changes model weights to learn a style, format or narrow skill.
   4. 4RAG vs long contextIf all your content fits comfortably in the model context window, you can sometimes paste it in instead of retrieving.
   5. 5Start with a naive baselineFixed-size chunks, one embedding model, top 5 by vector similarity and a simple grounded prompt.
   6. 6Knowledge baseThe collection of documents the system can retrieve from: docs, tickets, PDFs, wiki pages or database rows.

   6 in this chapterOpen this chapter
2. 2Loading and chunkingTurning messy documents into clean, well-sized chunks with the metadata you will need later for filters and citations.8 concepts

   1. 1Document parsingExtracting clean text from PDFs, HTML, slides and docs while keeping structure such as headings, tables and lists.
   2. 2ChunkingSplitting documents into passages small enough to embed and retrieve precisely, but large enough to make sense alone.
   3. 3Fixed-size chunks with overlapCutting text every N tokens or characters, with each chunk repeating the last part of the previous one.
   4. 4Structure-aware chunkingSplitting on natural boundaries first, such as headings, then paragraphs, then sentences, and only cutting mid-sentence as a last resort.
   5. 5Semantic chunkingPlacing chunk boundaries where the meaning shifts, detected by comparing embeddings of neighbouring sentences.
   6. 6Choosing chunk sizeSmall chunks match questions precisely but lose surrounding context; large chunks keep context but dilute the embedding and waste prompt space.
   7. 7Chunk metadataFields stored with each chunk such as source URL, title, section heading, date, language, product version and access group.
   8. 8Small-to-big retrievalEmbed small chunks for precise matching, but hand the model the larger parent section they came from.

   8 in this chapterOpen this chapter
3. Embeddings and the indexTurning chunks into vectors and storing them. The Vector databases cheat sheet covers index types and stores in depth; this is what RAG needs from them.6 conceptsOpens with credits
4. RetrievalFinding the right chunks: top k, keyword search, hybrid search, filters and rewriting the question before searching.8 conceptsOpens with credits
5. Reranking and context buildingChoosing the best few chunks from the candidates and assembling them into a context the model can use.6 conceptsOpens with credits
6. Grounded generationPrompting the model to answer only from the retrieved text, cite it and admit when the answer is not there.6 conceptsOpens with credits
7. Evaluating RAGMeasuring retrieval and generation separately, so you know which half to fix.7 conceptsOpens with credits
8. Failure modes and advanced patternsWhy RAG answers go wrong in production, and the patterns people reach for once the basics work.7 conceptsOpens with credits
9. 9Glossary, A to ZEvery concept in the sheet and the smaller terms it uses, in one list.61 terms

Retrieval augmented generation end to end: chunking, embeddings, hybrid retrieval, reranking, grounded prompts, evaluation and failure modes.