google-ai-edge/LiteRT-LM ? reverse-engineered prompt

Build me LiteRT LM, a fast on device LLM runtime that can run language models locally on desktop, mobile, web, and small devices.

I want a simple command line app where I can download or point to a model, start a chat, and ask it questions from my terminal. It should support the main model formats this project is meant for, work well with GPU acceleration when available, and include a clean way to plug into Python, Kotlin, Swift, and JavaScript. If you can, add a small demo app or sample that shows local chat running end to end with a real model.

Make it feel production ready, with good defaults, clear errors, and a straightforward setup flow. If you need to check current docs or platform details online while building, do that.