RAGFlow is an end-to-end open-source RAG engine: document parsing (DeepDoc), chunking, indexing, retrieval, plus the Q&A and assistant orchestration aimed at end users — all in one platform. For someone who wants a RAG assistant out of the box, that is the shortest path.
Langhuan does no chat and no agent orchestration. It handles only the "document → traceable retrieval" stage and exposes it as a knowledge foundation over REST and MCP over HTTP — including to agents you already run or plan to build.
If you want a RAG assistant out of the box, RAGFlow is the more direct path. If you want a stable, traceable, MCP-callable knowledge foundation to serve your own agents, Langhuan is that layer. They can also compose: Langhuan grounds the knowledge processing, and RAGFlow or another orchestrator sits on top.
Can Langhuan replace RAGFlow?
Different positioning, so not a drop-in replacement. RAGFlow is an end-to-end platform with Q&A and assistant orchestration; Langhuan is the knowledge layer beneath it. If you need a reusable retrieval foundation rather than another platform, Langhuan is the better fit.
How does Chinese retrieval differ between Langhuan and RAGFlow?
Both support Chinese. Langhuan runs a two-path recall with pgvector embeddings and PostgreSQL full-text search (zhparser tokenization), fused with deterministic RRF. RAGFlow uses DeepDoc for document parsing and Infinity for vector retrieval. The technical route and the positioning both differ.
Does Langhuan support MCP?
Yes, natively. Langhuan serves MCP over HTTP at /mcp, exposing tools like knowledge_search and document_ingest directly to MCP clients with no extra bridge.