Developer builds in-memory NSFW image filter in Node.js with 200ms classification

A developer created a Node.js endpoint that classifies explicit images entirely in RAM, ensuring uploaded content never touches a hard drive or cloud storage. The solution uses the open-source nsfwjs library alongside TensorFlow.js and multer's memory storage to process image buffers and discard them immediately after classification. The approach was motivated by an Apple App Store rejection under Guideline 1.2 for inadequate user-generated content moderation. Traditional moderation pipelines save images to disk before scanning them, creating a window where policy-violating content may temporarily reside on servers. The developer reports roughly 200ms latency per image on a standard VPS and has since packaged the logic into a public API called Tabu.