Vogadero/AIProxyBridge ? reverse-engineered prompt

Build me a desktop app that lets me use a local OpenAI compatible API on my computer, while it connects through a browser session to LMArena in the background. I want it to work on Windows, macOS, and Linux, with a simple modern interface where I can start and stop the service, see connection status, refresh the available models, and test a model with streamed responses.

It should expose endpoints like a normal OpenAI style API, including a model list and chat completions, so I can point apps like Cherry Studio or Continue at localhost and use them without changing much. Please include a way to connect the browser either with a Tampermonkey script or with an automated browser window, and make the app handle multiple browser instances if possible. I’d also like real time logs and a dark and light theme. If you need to, look up current docs online for Electron and browser automation.