ragflow

RAGFlow is a retrieval-augmented generation engine built around deep document understanding. It is one of the highest-velocity AI projects of 2026, and most of the attention is earned: it actually handles the messy parts of RAG that other open source frameworks hand-wave past. PDFs with tables, slides, Excel files, scanned images, mixed-language content. Apache 2.0.

Self-hosting is a real project, not a `docker run`. You need 4+ CPU cores, 16+ GB RAM, 50+ GB disk, and Docker 24. The compose file spins up the RAGFlow service plus ElasticSearch (or Infinity, their in-house vector DB) and MinIO for blob storage. Pick your LLM and embedding models per knowledge base. The template-based chunking is the killer feature: you can visualize and tweak how documents are segmented before they hit the index.

Solo developers and small teams that want full control of their RAG pipeline can self-host this and feel good about it. Larger teams or anyone without infra time should look at RAGFlow Cloud, which gives you the same engine without the ops.

The catch is the operational weight. RAG is not a side feature, and RAGFlow treats it like the production system it is. If you just need to chat with a few documents, LlamaIndex or LangChain over a simple vector store is faster to ship. Pick RAGFlow when document parsing quality and citation grounding are the thing you are buying.