Retrieval-augmented generation

As explained in Data Pipelines with Apache Airflow, Second Edition

A RAG workflow is a retrieval-augmented generation workflow that retrieves relevant information and supplies it to a generative model for producing an output.

A recipe assistant needs to answer questions from a growing collection of family recipes. Retrieval-augmented generation (RAG) lets a language model use that external collection without retraining whenever a recipe is added: it retrieves relevant recipes, then supplies them with the question to the model for an answer. section 14.3

The workflow has two connected parts:

• Prepare the collection. Split each recipe into an ingredients chunk and an instructions chunk, retaining the recipe name with both. Embed the content and store its vectors in Weaviate. Airflow can automate ingestion and updates; existing versions are identified before replacement. section 14.5 section 14.10
• Retrieve context. Embed the user’s question with the same model used for the recipes, then search the vector database for relevant documents. Cosine similarity compares vector direction, emphasizing meaning rather than identical wording. section 14.10.1
• Generate an answer. Send the question and retrieved recipes to a language model so its response can draw on the collection. section 14.3

A recipe question

For “How can I make lasagna?”, the assistant searches the stored recipes before answering with ingredients and steps. If the requested recipe is absent, the chat assistant says it does not know that recipe. It can also suggest stored recipes from a list of ingredients or share recipes in another language. section 14.10.1 section 14.10.3

Figure 14.3 Simplified representation of a RAG workflow: (1) A user asks a question about a recipe. (2) Vector search retrieves (3) the most relevant recipes. (4) The question and retrieved recipes are combined and sent to an LLM, which (5) generates an answer based on the relevant context.

The diagram follows a question through vector search, retrieval, and generation with the retrieved recipes as context. section 14.3

Chunk size is an operational choice. One enormous chunk makes a source recipe hard to pinpoint; sentence-sized chunks can return many redundant matches from the same recipe. Keeping ingredients and instructions as separate chunks provides a practical middle ground while preserving a link to the original recipe. section 14.5

How Data Pipelines with Apache Airflow, Second Edition covers it

How deep it goes

Works through ingestion, two-part chunking, embedding, similarity search, and answer generation; leaves embedding-model training details aside.

Examples

Uses family recipes, a lasagna question, a two-dimensional flavor analogy, and a recipe chat assistant.

What it assumes

Introduces vectors and embedding models through a flavor-space analogy; implements the workflow with Python and Airflow tasks.

What you can do after

Build it from scratch with Airflow-managed recipe ingestion, Weaviate retrieval, and a language-model response to a user question.

How it connects

Connects embeddings, vector databases, cosine similarity, text chunking, Airflow ingestion, and language models.

liveBookmanning.com

Read Data Pipelines with Apache Airflow, Second Edition on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Grokking AI ApplicationsA generative AI technique that retrieves relevant external information and supplies it to a model as context for generation.
• 

  liveBookmanning.com
  Effective Conversational AIRetrieval-augmented generation is a technique that conditions generated responses on information retrieved from external documents or data sources.
• 

  liveBookmanning.com
  Domain-Specific Small Language ModelsAn application system that combines retrieval from external sources with text generation to ground outputs in outside knowledge.
• 

  liveBookmanning.com
  Software Testing with Generative AIRetrieval-augmented generation is a generation technique that retrieves relevant external information and supplies it to a language model as context.
• 

  liveBookmanning.com
  LLMs in ProductionRetrieval-augmented generation is a technique that supplies a generative model with information retrieved from external sources at response time.
• 

  liveBookmanning.com
  AI-Powered SearchRetrieval augmented generation is a technique that supplies retrieved external information to a generative model as context for producing a response.
• 

  liveBookmanning.com
  Natural Language Processing in Action, Second EditionRetrieval-augmented generation is a language-generation technique that retrieves external information and conditions generation on the retrieved information.
• 

  liveBookmanning.com
  Machine Learning Platform EngineeringRetrieval-augmented generation is a method that supplies retrieved external information to a generative model when producing an output.
• 

  liveBookmanning.com
  Building Reliable AI SystemsA retrieval-augmented generation approach that retrieves and synthesizes stored information without taking actions in the real world.
• 

  liveBookmanning.com
  Build AI-Enhanced Web AppsRetrieval-augmented grounding is a grounding technique that retrieves relevant external information and incorporates it into response generation.
• 

  liveBookmanning.com
  The Art of AI Product DevelopmentAn architecture that retrieves relevant external information and supplies it to a generative model as context for producing an output.
• 

  liveBookmanning.com
  Generative AI in ActionRetrieval-augmented generation is a generation technique that retrieves relevant external information and supplies it to a generative model as context for its response.
• 

  liveBookmanning.com
  Financial AI in PracticeA method that retrieves relevant external information and supplies it to a generative model when producing an output.
• 

  liveBookmanning.com
  Introduction to Generative AI, Second EditionRetrieval-augmented generation is a generative technique that retrieves external information and supplies it to a model as context before generation.
• 

  liveBookmanning.com
  AI Agents and ApplicationsRetrieval-augmented generation is a technique that retrieves relevant external information and supplies it to a language model as context for response generation.
• 

  liveBookmanning.com
  Investing for ProgrammersRetrieval-augmented generation is a technique that retrieves relevant external information and supplies it to a generative model when producing a response.
• 

  liveBookmanning.com
  Knowledge Graphs and LLMs in ActionRetrieval-augmented generation is a technique that supplies a generative language model with relevant information retrieved from an external source before answer generation.
• 

  liveBookmanning.com
  Build Python Web Apps with StreamlitRetrieval-augmented generation is a technique that retrieves relevant external information and supplies it to a language model when generating a response.
• 

  liveBookmanning.com
  Data Storytelling with Altair and AIRetrieval-augmented generation is a technique that supplies a language model with relevant information retrieved from an external knowledge source when generating output.
• 

  liveBookmanning.com
  AI Agents in Action, Second EditionRetrieval-augmented generation is a technique that retrieves pertinent information at generation or decision time and adds it to a model's context.
• 

  liveBookmanning.com
  Spring AI in ActionRetrieval-augmented generation is a technique that supplies retrieved external information to a generative model when producing a response.
• 

  liveBookmanning.com
  Just Use Postgres!Retrieval-augmented generation is a technique in which a generative model incorporates information retrieved from an external knowledge source when generating an answer.
• 

  liveBookmanning.com
  Transformers in ActionA retrieval-augmented generation system generates outputs using information retrieved from an external collection.
• 

  liveBookmanning.com
  AI Agents in ActionRetrieval augmented generation is a language-model pattern that retrieves relevant external information and incorporates it into generated responses.
• 

  liveBookmanning.com
  MongoDB 8.0 in Action, Third EditionRetrieval-augmented generation is an AI technique that retrieves external information and supplies it to a generative model when producing a response.
• 

  liveBookmanning.com
  Essential GraphRAGA retrieval-augmented generation pipeline combines information retrieval with language-model generation to produce responses grounded in retrieved information.
• 

  liveBookmanning.com
  Machine Learning System DesignA generation approach that retrieves relevant external information and supplies it to a generative model as input.
• 

  liveBookmanning.com
  Build Applications with Local AI Models on a MacRetrieval-augmented generation is a technique that retrieves relevant information from an external knowledge source and supplies it to a language model alongside a question or instruction.
• 

  liveBookmanning.com
  The AI Pocket BookA method of augmenting an LLM prompt with retrieved external documents so the model can answer using information not in its training data.
• 

  liveBookmanning.com
  Building LLM Applications with DSPyA language-model application pattern that retrieves external information and supplies it as context for response generation.
• 

  liveBookmanning.com
  Authorization in ActionRetrieval-augmented generation is a technique that supplies a generative model with externally retrieved information when producing an output.

Compare how 32 books explain retrieval-augmented generation →

Related concepts

• language modeluses · A language model is a computational model that represents patterns in linguistic tokens to estimate sequence probabilities and predict or generate language from context.
• similarity searchuses · Similarity search is a retrieval technique that finds or ranks items by comparing their representations with a query according to a similarity or distance measure.
• hallucinationaddresses · A hallucination is model-generated content that is false, fabricated, or unsupported by relevant evidence but presented as plausible or factual.
• knowledge baseuses · A knowledge base is an organized collection of information that can be accessed or queried to support reference, retrieval, or reasoning.
• information retrievaluses · Information retrieval is the process of locating and selecting relevant information from an available collection in response to a query or information need.
• fine-tuningcontrasts with · Fine-tuning is adapting a pretrained machine-learning model by training it further on targeted data or objectives for a task, domain, or behavior.

This explainer was generated by Manning's knowledge graph from the text of Data Pipelines with Apache Airflow, Second Edition. The section links point to the parts of the book it draws on.