Skip to main content

COMPEL Glossary / agentic-rag

Agentic RAG

A retrieval pattern in which the agent decides — within its loop — when and what to retrieve, rather than a retrieval step running before generation as in naive RAG.

What this means in practice

Yields higher-quality answers on complex queries but trades away predictable cost and latency; multi-hop query governance becomes harder.

Synonyms

agent-driven RAG , dynamic RAG

See also

  • Naive RAG — The baseline retrieval-augmented-generation pattern: a single retrieval step returns the top-k chunks by vector similarity, and those chunks are concatenated into the prompt for generation.
  • Agent loop — The iterative pattern in which a model plans, calls tools, observes the results, and replans — continuing until a goal is met, an iteration budget is exhausted, or a human intervenes.
  • Tool use / function calling — A prompt pattern authorizing the model to request named functions with structured arguments — searching the web, reading a database, calling a calculator, triggering an API — rather than generating all answers from its weights.

Related articles in the Body of Knowledge