deepseek-ai/Janus ? reverse-engineered prompt

Build me a simple local app that lets me talk to a multimodal AI and also generate images from text.

I want it to support things like uploading a picture and asking questions about it, chatting back and forth, and entering a text prompt to create an image. It should feel polished and easy to use, with a clean demo style interface and a few example prompts so I can test it quickly. If you need to, look up the current docs online and wire it up the right way.

Please use the model files and inference scripts in this repo, and make sure the experience works smoothly on my machine with clear setup steps. If there are separate modes for chat, image understanding, and image generation, keep them easy to switch between. Add sensible error messages, loading states, and make the output easy to view and save.