wtf is tagpilot-lm?

TagPilot-LM is a single-file browser tool for captioning and tagging images, built around AI image datasets used to train custom image models. It is a fork of an existing project called TagPilot, and this version is aimed mainly at Windows users who want to run captioning locally on their own machine instead of paying for a hosted AI service. The core idea is that you run LM Studio, a program that lets you load and run vision capable AI models on your own computer, and TagPilot-LM talks to that local model over your own network connection to generate captions for a folder of images. Because everything runs locally by default, your images do not leave your PC unless you choose to use one of the other supported providers, which include Gemini, Grok, OpenAI, Claude, vLLM, DeepDanbooru, and WD1.4, carried over from the original project. To get started on Windows you start the LM Studio server, then start a small local web server from the TagPilot-LM folder, open the tool in a browser, connect to LM Studio, detect which model is loaded, test the setup on one image, and only then run captioning across the whole batch. There is also an optional mode for reaching the LM Studio server from a different device on the same private network, though the README is careful to warn that this opens the server up more broadly and should only be used on a trusted network, never over the open internet. Other features include automatic detection of the exact model being used, checks that catch bad or malformed captions before they are accepted, and a way to insert a fixed identity trigger word into exported files exactly once, so it cannot be duplicated or misspelled. The project has no stars yet and is released under the MIT license, inherited from the original TagPilot project it builds on.