xxWikiJS: Transform docs into AI-powered Q&A with verified citations—100% client-side

xxWikiJS transforms static documentation into an AI-powered knowledge base that answers questions with verified citations. Upload PDF, TXT, or DOCX files; the system indexes them and responds to queries only with information actually in your docs—eliminating hallucinations through source grounding. Running 100% in the browser and requiring zero backend, xxWikiJS is the missing link between dumb search and unreliable AI, deployed in seconds to GitHub Pages or any static host.

How it works

Three-step workflow:

1. Upload documents – Add PDFs, plain text, or Word docs to build a knowledge base.
2. Compile wiki – Index pages with chunking (8KB–40KB segments) and BM25 search.
3. Query with citations – Ask questions; get AI-generated answers with line-specific references back to source material.

The innovation: every answer is restricted to information in the uploaded documents. If the knowledge base doesn't contain an answer, the system says so—no speculation.

Technical approach

xxWikiJS implements Retrieval-Augmented Generation (RAG) entirely in the browser using ES modules. BM25 indexing retrieves relevant document segments; an LLM generates answers grounded in those segments. Data persists in IndexedDB (browser storage), so no server sees your documents. API keys for LLM providers (stored locally only) keep sessions private.

Supported LLM providers (6 auto-updating model catalogs):

• Google Gemini
• Mistral AI
• Groq
• OpenRouter
• HuggingFace
• Additional providers via extensible architecture

Trial API keys pre-loaded for testing; users substitute their own keys (never leave device).

Deployment is trivial

No database, no backend, no Docker. Start with: python -m http.server 8123 locally, or deploy to GitHub Pages, Netlify, Vercel. That's it. Assets stay on the user's device; code runs client-side.

Use cases

• Enterprise knowledge base: Searchable docs with verified AI Q&A
• Customer support: Self-service knowledge base reducing support tickets
• Internal runbooks: Private procedures with instant answers
• Developer documentation: API docs with cited sources
• Legal/compliance: Sensitive document repositories (100% private)
• Research: Extract insights from papers with citations
• Education: Course materials and knowledge repositories

Availability and community

xxWikiJS is actively maintained open source by Giuseppe Materni. Recently featured on Show HN (October 2026) with positive reception. Demo site: https://gmaterni.github.io/wikijs/wikijs.html. GitHub: https://github.com/gmaterni/wikijs.

Why it matters

RAG (Retrieval-Augmented Generation) is the industry solution to LLM hallucinations, but most RAG platforms require backends, vector databases, and API keys stored server-side. xxWikiJS proves RAG can be simpler: client-side indexing, standard algorithms (BM25), and local storage. For teams handling sensitive docs, privacy concerns, or wanting to avoid vendor lock-in, this matters enormously.

The broader trend: as organizations realize RAG is essential for trustworthy AI, tools that make RAG accessible without infrastructure complexity win. xxWikiJS is the first to make it painless—and private.