Sifter is a self-hostable, open-source AI text classifier that processes text in parallel chunks through a local model to determine the probability it was written by an LLM. It features dynamic text highlighting on the frontend to pinpoint AI-generated content and keeps all data entirely local.