RAG (Retrieval-Augmented Generation)
In one sentence
A method where AI retrieves relevant information from a knowledge base first, then generates an answer based on those materials.
RAG solves two problems: models do not know proprietary company data, and documents are too large to fit into context windows. Retrieving only the most relevant passages for the model ensures accuracy while saving tokens.
RAG performance depends on retrieval accuracy: how documents are chunked, the retrieval methods used, and how many passages are retrieved all affect the outcome.
Example
When a website sales advisor receives the inquiry 'Do you support overseas GEO?', it first retrieves the GEO Global edition documentation, then answers based on it.
Related terms
- Knowledge Base
- A repository of company documents and institutional knowledge that AI can search and reference, enabling it to answer questions and produce content grounded in internal facts.
- Vector (Embedding)
- Converting a block of text or an image into a series of numbers. Content with similar meanings gets similar numbers, making it possible to retrieve information by meaning.
- Vector Database
- A database that stores vectors and quickly finds semantically similar content, commonly used as a core component for RAG and knowledge base retrieval.
- Text Chunking
- Splitting long documents into appropriately sized chunks before saving them to a knowledge base, making it easy to retrieve only the relevant sections.
