RAGFlow overview and key facts (October 2026)

RAGFlow is an open-source retrieval augmented generation engine that turns complex documents into a searchable knowledge base and answers questions with citations. It fits teams that want a ready application with an interface instead of assembling a framework themselves.

RAGFlow is developed by InfiniFlow and published under the Apache 2.0 license. It ingests Word, slides, Excel, text, images, scanned copies, structured data and web pages, splits them with configurable chunking templates and shows the chunks so people can correct them. Data can be synced from Confluence, S3, Notion, Discord and Google Drive, and MinerU or Docling can serve as parsers. Self-hosting needs Docker and at least 4 CPU cores, 16 GB RAM and 50 GB disk. As of September 28, 2026 the repository had about 91,400 GitHub stars. A hosted version runs at cloud.ragflow.io.

• Reads complex documentsThe DeepDoc component extracts text, tables and layout from difficult formats, and a multimodal model can interpret images inside PDF or DOCX files. Teams get usable knowledge out of scans and slide decks that plain text extraction misses.
• Finds the right passagesRetrieval combines several recall methods with fused re-ranking, and the embedding model and language model are configurable. Answers draw on the most relevant passages, even in large document collections.
• Shows where answers come fromEach answer links to the passages it used, and the chunking view lets people inspect and adjust how documents were split. Reviewers can check answers quickly and fix poor chunking by hand.
• Builds agent workflowsA visual editor combines RAG, tools, code execution and MCP servers into agent workflows, with a memory feature for agents. Teams can automate multi-step tasks on top of their knowledge base.

• Ships as a complete application with web interface, not only as a library.
• Chunking results are visible and editable, which supports human review of the knowledge base.
• The cloud version publishes monthly prices and a public data processing agreement with Standard Contractual Clauses for EU transfers.

Open-source project without a stated country focus; the readme is available in English, Chinese, Japanese and Korean among other languages.

For the cloud version the privacy policy does not name the legal entity's address or hosting regions, and the subprocessor list is available on request, so ask before you upload personal data. Self-hosting avoids this but needs a server with Docker, at least 16 GB RAM and someone to maintain it. Credits in the cloud plans limit usage.

No reviews yet. Be the first to share how your team uses it.

Every review shows its verification level. We read each one before it goes live. How we check reviews