A self-hosted wiki that makes the LLM read the manual

What it does

PandaWiki is a knowledge-base builder that wraps your product docs, technical wikis, FAQs, or blog posts in a web interface and hands the content to a large language model. It offers a rich text editor that handles Markdown and HTML, plus AI-assisted creation, Q&A, and search. You can also turn the whole thing into a chatbot for DingTalk, Feishu, or WeChat Work, or drop it onto another site as an embedded widget.

The interesting bit

Most AI knowledge-base tools ship under permissive licenses to encourage hosted forks; PandaWiki uses AGPL-3.0 and explicitly warns that network use requires open-sourcing your modifications. The README also treats the LLM as the engine, not the chrome—without a configured model, the AI creation, Q&A, and search simply do not run.

Key highlights

• AI features are first-class, not bolted-on: the editor, search, and Q&A all lean on a configured LLM, and the app recommends Baizhi Cloud for quick setup.
• Content can be ingested from URLs, Sitemaps, RSS feeds, or offline files, so you don’t have to start from a blank page.
• Exports go both ways: write in Markdown or HTML, then spit pages back out as Word, PDF, or Markdown.
• Chatbot and widget integrations mean the knowledge base can live inside your existing team tools instead of siloed on a subdomain.

Caveats

• The AI features are disabled until you configure an LLM; the README explicitly states they cannot work without one.
• AGPL-3.0 is a strict copyleft license: the README notes that providing the software over a network requires you to open-source your modifications.
• Installation targets Linux with Docker 20.x or newer; there is no mention of Windows or macOS support.

Verdict

A solid choice for technical teams or open-source projects that want a self-hosted, AI-augmented docs site and can live with AGPL obligations. Look elsewhere if you need a permissive license or a wiki that works without an external model API.