NSFW API Policy
Downstream MR: https://gitlab.com/soapbox-pub/soapbox/-/merge_requests/35/
This adds a new MRF Policy: NsfwApiPolicy. It relies on an external FOSS service that can be easily self-hosted. During MRF filtering, it submits the attachments to the API service which uses artificial intelligence to rate it on a scale of 0-1, with 1 being most certainly NSFW.
Once you have the score, you can configure a threshold (default 0.7) for which to take action. Then you can either reject, unlist, or mark sensitive any content above the threshold.
Under the hood this uses Yahoo's open_nsfw and it's pretty accurate.
If the API server fails or any reason, it will treat the attachment as SFW and pass it through. It only works on images.