Janus.Guru

A glossary of RAG and document assistants

These terms describe a managed retrieval and evidence workflow, not an abstract chat experience.

Launch a pilot Open demo

Core concepts

RAG — retrieval-augmented generation: a model answers after searching a collection
Corpus — documents, fragments, and metadata used by an assistant
Chunking — preparing documents for precise retrieval
Embeddings — numerical representations used for semantic search

Answer controls

Source references — retrieved fragments behind an answer
Source-backed answer — wording grounded in collection material
Unsupported-claim control — rules that limit free conclusions without evidence