Agents

Retrieval-Augmented Generation

1 min read

In Short

An architecture that enhances LLM responses by first retrieving relevant information from external knowledge sources.

RAG addresses LLM limitations around knowledge cutoffs and hallucination by grounding responses in retrieved documents.

Components

  • Retriever: Finds relevant documents
  • Knowledge base: Source documents
  • Generator: LLM that synthesizes answer

Benefits

  • Current information access
  • Reduced hallucination
  • Verifiable sources
  • Domain specialization
agentsarchitecturerag