HuixiangDou is an open-source, LLM-powered technical assistant built by the InternLM team specifically for group chat scenarios. Instead of answering every message like a generic chatbot, it watches group conversations in apps such as WeChat and Lark, decides whether a question falls inside its knowledge base, and only replies when it can genuinely help. The project has earned more than 2,500 stars on GitHub and was last actively updated in November 2025.
Busy technical groups suffer from the same questions being asked again and again, while a naive bot that answers everything quickly becomes spam. HuixiangDou takes a different approach with a retrieval-augmented generation pipeline: incoming messages are checked against a document knowledge base, and the assistant stays silent when the topic is out of scope or the confidence is low. When it does answer, it grounds its response in the retrieved material, which makes it well suited to product support groups, open-source communities, and internal engineering teams. It also supports multimodal input, including image retrieval, so questions that reference screenshots can be handled in context.
HuixiangDou is free and open source under the BSD-3-Clause license. You can self-host it with your own LLM endpoints; costs are limited to whatever model API or compute you choose to run it with.
HuixiangDou is for teams and communities that run support or discussion groups and want an assistant that answers repetitive technical questions without flooding the chat. It is a strong fit for open-source maintainers, developer relations teams, and internal support channels.