RAGFlow — поиск и ответы по сложным документам

★ 84K

RAGFlow is an open-source engine for question answering over documents whose strength is how deeply it parses the documents themselves. Reach for it when the knowledge base is not plain text: scans, contracts full of tables, reports with complex layout, decks, multi-column PDFs. Ordinary pipelines fall apart on those, losing table structure and column order; RAGFlow reads the layout and splits a document into meaningful chunks rather than fixed character counts. From there it is the familiar shape: semantic retrieval, a model answer citing the exact place in the source so it can be verified. It ships a web interface, an API and agentic flows on top of the knowledge base, and you host it yourself. It is heavier on resources than simple RAG libraries, but on messy real-world documents the difference in quality is clear.

• #RAG