nsfw-server

A microservice to run nsfw.js for images on IPFS.

Requirements

• Node 16.x (LTS)
• Python (Windows / Mac OS X Requires Python 2.7)
• Running IPFS daemon or public IPFS gateway (not recommended, can be configured with environment variable)

Getting started

Install Node dependencies

npm install

Start IPFS node

In a separate terminal, after installing a local IPFS daemon (see requirements): ipfs daemon

Start server

Run (auto-reloading) dev server with full debug enabled: env DEBUG=* npm run-script dev

Example usage

$ curl -s http://localhost:3423/classify/bafkreiam7vh2raw2chmx6gllup6sd32xsicwus2riglpjtzlygz4lzmxbm | jq

(Wait a while... IPFS is slow. :/)

{
  "classification": {
    "neutral": 0.9980410933494568,
    "drawing": 0.001135041005909443,
    "porn": 0.00050011818530038,
    "hentai": 0.00016194644558709115,
    "sexy": 0.00016178081568796188
  },
  "modelCid": "QmfBNCmYLxwTr3CHaknd5HdzA6uXcTZqn1hsuLf8mRc3xS",
  "nsfwServerVersion": "0.9.0"
}

Production server

Run in production mode:

env NODE_ENV=production NSFW_SERVER_PORT=3342 npm start

Run tests

npm test

Docker support

Build docker image (tagged as nsfw-server)

docker build -t nsfw-server .

Run docker image in production on port 9000

docker run --env NODE_ENV=production -p 9000:3423 -t nsfw-server

API

To classify an IPLD CID:

/classify/<cid>

Supported formats

We're using TensorFlow's decodeImage, which detects and supports BMP, GIF, JPEG and PNG formats.

Configuration

nsfw-server is configured through the following environment variables:

• IPFS_GATEWAY_URL: Gateway to use for loading images. Defaults to http://localhost:8080 (local node).
• NSFW_SERVER_HOST: Host to run server on. Defaults to localhost.
• NSFW_SERVER_PORT: Port to run server on. Defaults to 3423.
• DEBUG: Debug verbosity level.
• NODE_ENV: Node environment. Set to production in ... production. Defaults to development.
• NSFW_MODEL: NSFW model to use.