Glossary

Term	Definition
Chunk	Document segment. During ingestion, each document is split into chunks for more precise search.
Favorite	An answer deemed particularly useful by a user, which can be shared with all users after administrator validation.
Ingestion	The process of importing and processing a document: text extraction, splitting into chunks, embedding creation and indexing.
OCR	Optical Character Recognition. Extracts text from images or scanned PDFs.
RAG	Retrieval Augmented Generation. A technique combining document search and AI text generation to produce reliable, sourced answers.
Deep search	Search mode using reranking to reorder results by relevance, at the cost of slightly longer processing time.
Hybrid search	Search mode combining semantic search (by meaning) and keyword search for complementary results.
Re-ingestion	Reprocessing a document already in the base. Useful after a document update or to improve chunking.
Reranking	Technique of reordering search results using an AI model to improve relevance.
Source	Reference to the specific document and passage used to generate an answer. Allows information verification.
Product Universe	Grouping of documents by product range or theme. Allows targeting searches to a specific scope.
VLM	Vision Language Model. An AI model capable of analyzing and describing image content in documents.
Widget	An embeddable version of the RAGFab chat, designed to be inserted into external applications (intranet, portals, etc.).