Yuheng documentation
Yuheng is an open-source (MIT) enterprise knowledge platform: the knowledge layer for AI agents. It brings in documents, web pages and content synced from Feishu/Lark, Notion, Yuque, RSS, GitLab and Tencent ima, or written in its own optional collaborative documents; organizes them into knowledge bases, FAQ entries, an LLM-generated Wiki and an optional knowledge graph; answers questions with hybrid retrieval and cited, streamed answers; keeps the knowledge maintained through knowledge health (duplicates, divergent copies, documents due for review, answers people found unhelpful, each routed to a document owner); and exposes all of it through a REST API (/api/v1), an MCP server with 22 tools, a Go SDK, the yuheng CLI and a DeepSeek Harness plugin.
Yuheng is not an agent framework. It does not run agents; it is what agents call for knowledge they can trust.
Chinese is the primary language of this documentation. Only the two pages below are translated into English. Everything else, including the product introduction, the feature guides, the API reference and the architecture notes, is Chinese.
- Installation: build and run Yuheng with Docker Compose (no Docker images are published, so you build them locally), the one-command deploy script, development mode, Helm and running from source.
- Quick start: register the first account, configure models, upload documents and ask a question, in the UI and with curl.
Yuheng 0.1.0 is a preview: the /api/v1 REST API may change between 0.x releases, and MCP tool names are stable. It began as a fork of Tencent's WeKnora and now evolves independently with a substantially reworked architecture. The source is at github.com/magicyuan876/Yuheng; the README is in English.