MindSearch
A self-hosted AI search framework that answers questions using the web, with local InternLM or cloud GPT-4 support and an Apache 2.0 license.
MindSearch is a self-hosted AI search framework for people who want to build their own Perplexity-style answer engine. Multiple LLM agents search and read web pages to produce answers with a visible research process. It's open source under Apache 2.0.
The focus is on questions that need broader research than a single search result can provide. MindSearch draws on multiple pages and exposes the path behind its answer, so readers can inspect how it reached a conclusion. That visibility matters when you're comparing information across sources or checking the basis for a response.
You can run the model locally through an InternLM server using InternLM2.5-7b-chat, or use GPT-4 through a cloud service. The local model has optimizations for Chinese, and the framework supports English and Chinese. Self-hosting doesn't make web research offline: searches still contact external services, and choosing GPT-4 sends model requests to a cloud provider.
MindSearch works with GoogleSearch, DuckDuckGoSearch, BraveSearch, BingSearch and TencentSearch. These connections let you choose the search service behind your answer engine rather than depend on one fixed provider.
Its FastAPI backend supports a choice of interfaces, including React, Gradio and Streamlit, as well as terminal access. Applications can also query the backend directly without using a frontend.
