RWiki

Self-hosted knowledge base Q&A powered by RAG. Single binary, zero external databases.

Upload Markdown, XLSX, or OpenAPI specs — RWiki chunks and vectorizes them, then serves streaming answers with source citations. Runs on SQLite, deploys with one command, works with any OpenAI-compatible LLM.

中文文档

Quick Start

docker run -d -p 8080:8080 \
  -v rwiki-data:/app/data \
  -e OPENROUTER_API_KEY=your-llm-key \
  -e OPENAI_API_KEY=your-embedding-key \
  ghcr.io/timzaak/rwiki

Open http://localhost:8080, upload a document, publish, start chatting.

Or try the demo:

cd scripts && pip install -r requirements.txt && python demo-start.py

Why RWiki

Most RAG setups need PostgreSQL + pgvector, Redis, a vector database, and a Docker Compose file with 5 services. For teams that just want "upload docs, ask questions," that's overkill.

RWiki does one thing — knowledge base Q&A — and keeps the infrastructure to a single binary with SQLite.

	RWiki	Typical RAG Stack
Database	SQLite (built-in)	PostgreSQL + pgvector
Dependencies	None	Redis, vector DB, message queue
Deployment	Single binary	Docker Compose, 3–5 services
Setup	docker pull and run	Hours of configuration

Features

• Streaming chat Q&A — Ask questions, get answers with source citations from your documents
• Embeddable chat widget — Single JS file, Shadow DOM, add to any site with two lines of HTML
• Multi-format ingestion — Markdown files, XLSX spreadsheets, OpenAPI specifications
• API documentation assistant — Upload OpenAPI specs, ask questions about your APIs
• Provider-agnostic — OpenAI, OpenRouter, BigModel, any OpenAI-compatible endpoint
• Self-hosted — Your data stays on your server, no cloud dependency

Deploy

Docker (recommended)

docker pull ghcr.io/timzaak/rwiki
docker run -d -p 8080:8080 \
  -v rwiki-data:/app/data \
  -e OPENROUTER_API_KEY=your-llm-key \
  -e OPENAI_API_KEY=your-embedding-key \
  ghcr.io/timzaak/rwiki

From Source

Prerequisites: Rust (latest stable), Node.js 20+, an OpenAI-compatible embedding API key.

git clone https://github.com/timzaak/rwiki
cd rwiki/backend
cp config/config.example.toml config/config.toml
# Edit config.toml — set your API keys
cargo run

E2E Tests

cd demo
npm install
npx playwright install chromium
npm test

Configuration

• Backend config: backend/config/config.toml (copy from config.example.toml)
• API keys (env vars override config):
  • OPENROUTER_API_KEY — LLM provider key
  • OPENAI_API_KEY — Embedding provider key
• Embedding providers: OpenAI (text-embedding-3-small), BigModel (embedding-3), or any OpenAI-compatible API
• OpenAPI docs: config.toml → server.enable_openapi
• Frontend proxy: frontend/vite.config.ts → server.proxy

AI-Built

This project is built entirely with Claude Code + web-dev-skills.

git clone https://github.com/timzaak/web-dev-skills
claude --plugin-dir /path/to/web-dev-skills

License

Apache License 2.0