HimeWiki is a simple wiki engine built with Go + PostgreSQL.
It features a minimal markup language called Nomark and optional AI-based moderation.

Released under the MIT License.

Features

Lightweight - runs as a single binary
Nomark Markup - a simple custom markup language (Markdown / Creole planned)
PostgreSQL Storage - stores both page content and images in the same DB
Optional AI Filter - integrates with OpenAI API for spam filtering and style unification

Requirements

Go 1.24 or later
PostgreSQL 15 or later
Tested on Linux and OpenBSD
Not tested on macOS, but expected to work anywhere Go and PostgreSQL are available

Installation

git clone https://tea.kareha.org/pot/himewiki.git
cd himewiki
make

This will build the himewiki binary.

Database Setup

Create a PostgreSQL database named himewiki:

createdb himewiki

Tables and indexes will be created automatically on the first run.

Configuration

Copy the example config and edit it:

cp config.yaml.example config.yaml

(!) config.yaml must exist in the current working directory.

Basic Example

app:
  mode: "devel"
  addr: ":4444"

database:
  host: "localhost"
  port: 5432
  user: "hime"
  password: "SuperStrongPassw0rd"
  name: "himewiki"
  sslmode: "disable"

site:
  base: "https://wiki.example.org/"
  name: "HimeWiki"
  card: "https://icon.example.org/hime/card.png"

AI Filter (Optional)

Enable AI filtering for posts and images by setting the OpenAI API Key.
To disable a filter, set agent: "nil".

filter:
  agent: "ChatGPT"   # use "nil" to disable
  key: "(Your OpenAI Key Here)"
  system: "You are a wiki content filter..."
  prompt: "Please rewrite in mild style..."

image-filter:
  agent: "ChatGPT"   # use "nil" to disable
  key: "(Your OpenAI Key Here)"
  max-length: 4194304
  max-size: 512

Run

./himewiki

Then open your browser at http://localhost:4444/.

License

MIT License

Author: Aki Kareha aki@kareha.org

For Researchers

Although HimeWiki is a hobby project and intentionally kept simple,
its design may be of interest for academic research in fields such as CSCW, HCI, and NLP.

CSCW: collaborative editing systems and AI-assisted cooperation
HCI: user experience and interaction with AI-mediated content
NLP: style transfer, politeness adjustment, and malicious-to-benevolent text transformation

If you are a researcher looking for an experimental platform, feel free to fork this project
and extend it for your study.

HimeWiki itself will remain minimal, but we encourage forks to explore research-oriented features such as:

logging raw user input (before AI filtering)
tracking user behavior for analysis
comparing AI-filtered vs. original edits
experimenting with different filtering prompts and styles

We would be delighted if HimeWiki could serve as a starting point for future studies.