RAGFlow

An open-source retrieval augmented generation engine that turns documents into answerable knowledge through deep document understanding, so tables, scans and complex layouts survive into the answer. Its pipeline supports configurable chunking, a visual chunk editor for inspecting how a document was split, retrieval tuning, and pluggable language and embedding models.

Teams use it to ground an assistant in internal manuals, contracts, tickets or research papers, and to trace each claim back to the source passage.

Hosting it in-house means documents and derived indexes never leave your servers, there is no per-query billing, and the database and object store are backed up on your own schedule.