RAGFlow
An open-source RAG engine for document search and AI agents, with self-hosted deployment, hybrid retrieval, citations and a separate cloud service.
RAGFlow is an Apache 2.0 licensed RAG engine for teams building AI agents that need to answer questions from their own documents. It can run on a self-hosted server through Docker on Windows, macOS or Linux. A separate hosted cloud service is available.
Its ingestion pipeline handles Word files, slides, spreadsheets, images, scanned pages, structured data and web pages. It extracts content from complex layouts, including tables, then organizes it for retrieval. Teams can inspect and adjust how documents are split into searchable chunks, or compile document and dataset content into formats such as wikis, graphs and timelines.
Search combines vector retrieval, BM25 keyword matching, custom scoring and reranking. For complex questions, an agent can break a query into steps and retrieve more evidence before answering. Citations point back to the material used. Visual workflows bring retrieval, tools and MCP connections into the same agent platform, while configurable LLM, embedding and reranking models let teams choose their model providers.
The self-hosted deployment uses a Go service for its API, administration and document ingestion. Document layout analysis, OCR and table recognition run on the CPU. Resource needs depend on the document workload and the models you run.
