What it does

PandaWiki is a self-hosted knowledge base system that wraps a wiki engine around an LLM. You create document collections, and it spins up public-facing wiki sites with AI-assisted writing, Q&A, and search. It also exports to Word, PDF, and Markdown, and can ingest content via URLs, sitemaps, RSS, or file uploads.

The interesting bit

The project treats the LLM as infrastructure, not garnish: without a configured model, the AI features simply don’t work. The README is unusually honest about this dependency, and even nudges you toward a specific Chinese model provider (Baizhi Cloud) with a ¥5 signup credit. The chatbot integrations—DingTalk, Feishu, WeChat Work—suggest the authors actually ship to enterprise users who live inside those apps.

Key highlights

• Dual interface: an admin console for managing knowledge bases, plus generated public wiki sites
• Embeddable widgets and chatbot integrations for third-party sites
• Rich text editor with Markdown/HTML support and multi-format export
• Content import via URL, sitemap, RSS, or offline files
• AGPL-3.0 license with explicit network-use clause (self-hosting is the only practical path)

Caveats

• Requires Docker 20.x+ on Linux; the install script demands root and curls a remote bash script
• AI model configuration is mandatory for core features; the “one-click” setup ties to an external commercial provider
• Documentation and community support are primarily in Chinese

Verdict

Worth a look if you need a self-hosted, LLM-augmented docs/FAQ system for a Chinese-speaking team or customer base. Skip it if you want vendor-neutral AI, non-Linux deployment, or a wiki without runtime dependencies on external model APIs.