deepseek-ai/DeepSeek-VL2 ? reverse-engineered prompt

Build me a simple local web app that lets me chat with DeepSeek VL2 using one or more images.

I want to be able to upload a photo, ask questions about it, and get answers about what’s in the image, text in the image, charts or tables, and general visual reasoning. It should also support pointing out or describing a specific object in the picture when I ask for that.

Please make it easy to run from the command line, with one script for quick inference and another for a small web demo. Use the model from Hugging Face, and if you need to, look up the current docs online so the setup actually works. Keep the interface clean and simple, with a chat style layout, image upload, and a place to type prompts.