RAG and Retrieval
Retrieval-augmented generation helps AI find relevant information. These terms explain how retrieval works, where it succeeds, and what governance must wrap around it so answers stay grounded and permission-aware.
RAG
RAG
Retrieval-Augmented Generation: a pattern where AI retrieves relevant documents before generating an answer.
14 min read
RAG
Retrieval-Augmented Generation
An AI architecture that searches a knowledge store for relevant context, then uses that context to generate a response.
13 min read
RAG
RAG Governance
The controls around retrieval that decide which sources are eligible, current, permissioned, and cite-worthy for each AI app.
12 min read
RAG
RAG Pipeline
The end-to-end flow from documents through chunking, embedding, retrieval, and answer generation.
8 min read
RAG
RAG Architecture
How indexing, retrieval, context assembly, and generation connect in a production RAG system.
8 min read
RAG
RAG Evaluation
Measuring whether retrieval finds the right sources and whether answers grounded in those sources are correct.
7 min read
RAG
RAG Observability
Visibility into what documents were retrieved, why they ranked, and how they influenced the final answer.
7 min read
RAG
RAG Retrieval
The search step in RAG that finds candidate document chunks matching a user query.
7 min read
RAG
Retrieval Quality
How often retrieval returns the right documents for a given question.
7 min read
RAG
Retrieval Relevance
Whether retrieved chunks actually relate to the user's question and intent.
6 min read
RAG
Context Window
The maximum amount of text a model can consider at once, including retrieved knowledge and conversation history.
7 min read
RAG
Context Engineering
Designing what information enters the model context, in what order, and with what constraints.
7 min read
RAG
Semantic Search
Search that matches meaning and intent, not just exact keywords, often using embeddings.
7 min read
RAG
Vector Search
Finding similar content by comparing embedding vectors in a vector database.
7 min read
RAG
Embeddings
Numerical representations of text that capture semantic meaning for similarity search.
7 min read
RAG
Chunking
Splitting documents into smaller segments for indexing and retrieval.
7 min read
RAG
Hybrid Search
Combining keyword and semantic search to improve retrieval recall and precision.
7 min read
RAG
Metadata Filtering
Narrowing retrieval results using document metadata like collection, audience, status, or owner.
6 min read
RAG
Grounded AI
AI responses constrained to information from approved sources rather than unconstrained generation.
8 min read
RAG
AI Grounding
Connecting model outputs to specific source documents or facts that support each claim.
7 min read
Related reading paths
- I am building an AI product
Put this into practice
Retrieval, context, embeddings, and the systems that connect documents to model answers.
See governed retrieval