MetaChar

Scraper of popular Character sites for use with large language models.

Currently, supports Chub and Janitor.

Installation

Below is a list of steps to get this running in docker environment, but you can adapt these steps if you're planning on running this without docker.

0. Check out the repo

git clone https://github.com/kubernetes-bad/metachar

1. Make the env file

Copy the backend/example.env file to backend/.env - note the comma at the beginning of the file name.
This file contains some sensitive settings that you may want to change like your S3/Minio credentials.

cp backend/example.env backend/.env

Build containers

docker compose build

Run

docker compose up -d

The app will be available on port 8006.

MetaChar uses MySQL as a database, Minio as image storage, FlareSolverr as request proxy.
All these components are pluggable, meaning that instead of using bundled MySQL, or S3 you can use your own - just edit the config in backend/config/default.yml

Configuration - backend

You can create a file called backend/local.yml that will serve as an override to the default config. That file contains paths to flaresolverr instance and various credential info.

Optional: Serve backend and frontend from separate URLs

Copy file frontend/example.env to frontend/.env, edit your path to backend API and re-build the docker image or run npm run build.

Warning

Both Chub and Janitor have very little moderation of contents. Do not expose your MetaChar installation to the internet.