t2vee / nsfwjs-avatar-service

a simple nodejs express app that provides a endpoint for nsfwjs image checking

NSFWJS Avatar Service

A super simple service that checks the avatar for explicit images using the awesome nsfwjs library
It is designed to be used hand-in-hand with Logto Account Dashboard

minimum requirments

Unfortunately due to the sheer size of tensorflow and the model itself,
This applciation wont run on mini systems, so anything LESS than 1vcpu and 2gb of ram.
The docker package itself is also rougly ~2.25GB which is quite large, make sure that you also have a decent amount of avaliable space

install

You can isntall this app via docker and (unfortunately) only by docker as the moment

1. start grabbing the docker-compose.yml:

$ curl -O -L https://github.com/t2vee/nsfwjs-avatar-service/raw/main/docker-compose.yml

2. configure variables under environment in the docker-compose.yml
3. start the service with docker compose:

$ docker compose up -d

4. expose container with a reverse proxy

here is a nginx example:

server {
    listen 80;
    server_name check.yourdoma.in;
    return 301 https://$server_name$request_uri;
}

server {
    listen 443 ssl;
    server_name check.yourdoma.in;

    ssl_certificate     /path/to/your/ssl.pem;
    ssl_certificate_key /path/to/your/ssl.key;

    location / {
        proxy_set_header Host $host;
        proxy_set_header X-Real-IP $remote_addr;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
        proxy_set_header X-Forwarded-Proto $scheme;

        proxy_pass http://127.0.0.1:3005;
        proxy_redirect off;
    }
}