> For the complete documentation index, see [llms.txt](https://www.mlcompendium.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://www.mlcompendium.com/generative-ai/rag.md). # RAG A language model only knows what it was trained on, so answering from your own documents means retrieving the right text first and generating from it. This page starts with practical retrieval-augmented generation (RAG) tutorials and ways to improve the pipeline, then moves to Graph RAG, where the retrieved context comes from a knowledge graph. The same notes are in [GenAI Applications](/generative-ai/genai-applications.md), [Search](/natural-language-processing-nlp/search.md), [Tools](/generative-ai/large-language-models-llms.md#tools), and [Vector databases](/data/engineering-intro/lakes-and-warehouses.md#vector-databases). ## Tutorials The first step is to build one pipeline end to end on real, messy documents. [RAG on Complex PDF using LlamaParse, Langchain and Groq](https://medium.com/the-ai-forum/rag-on-complex-pdf-using-llamaparse-langchain-and-groq-5b132bd1f9f3) is Plaban Nayak's walkthrough; it describes RAG as an approach that uses LLMs to automate knowledge search, synthesis, extraction, and planning from unstructured data, adding contextual information to LLM applications, and it lays out the components of the RAG data stack. Once a pipeline works, [Optimization methods for RAG](https://pub.towardsai.net/so-you-want-to-improve-your-rag-pipeline-28b0cfadbfd7) is "So, You Want To Improve Your RAG Pipeline", which points to an implementation of improving the pipeline with different indexing. Both tutorials end in a working bot, and [Q\&A Bot Using Gen-AI](https://pub.towardsai.net/q-a-bot-using-gen-ai-2ab934b180af) (October 2024) is a retrieval bot built the way this page describes. ## Graph RAG Plain retrieval returns chunks of text; Graph RAG builds a knowledge graph first so the model can reason over connections in the data. The same notes are in [Knowledge Graphs](/natural-language-processing-nlp/knowledge-graphs.md). [Microsoft on GraphRAG: Unlocking LLM discovery on narrative private data](https://www.microsoft.com/en-us/research/blog/graphrag-unlocking-llm-discovery-on-narrative-private-data/) is the Microsoft Research post, by Brenda Potts, on using LLM-generated knowledge graphs to improve question answering over complex information, where it consistently outperforms baseline RAG. An easier explainer of how GraphRAG works used to sit here; that address no longer opens and is kept at the end of the page. ## Deprecated links {% hint style="warning" %} These links and images no longer work. The original wording is kept here. A same-resource copy, when one was checked, is used above. {% endhint %} * Understanding Graph RAG. This address no longer opens: --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://www.mlcompendium.com/generative-ai/rag.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.