RAGFlow
Self-hosted RAG engine with document parsing and agent workflows
What it is
RAGFlow ingests Word, PDF, Excel, slides, scans, web pages and other sources, splits them with template-based chunking, and answers questions over them with traceable citations. It adds multi-step agentic retrieval, fused re-ranking, configurable LLMs and embedding models, and can compile datasets into wikis, graphs, mind maps and timelines. Developers and companies run it with Docker to build question-answering systems and agents over their own documents, using its web interface or API.