QwenLM/qwen ? reverse-engineered prompt

Build me a clean, easy to run local demo for Qwen so I can chat with the model in a browser and from the command line, and also expose an OpenAI style API endpoint for simple testing.

I want one project that helps me download or point to a Qwen model, run inference, and switch between a basic chat demo and a web UI. Please make the setup as straightforward as possible, with clear startup commands, sensible defaults, and helpful error messages if the model or GPU settings are wrong. If it makes sense, include support for smaller or quantized models so it can run on modest hardware too.

Also add a simple finetuning path or example config for LoRA or Q LoRA, since I may want to adapt the model later. Keep the README style instructions practical and beginner friendly, and look up current docs online if you need to.