Project-MONAI/VLM-Radiology-Agent-Framework — reverse-engineered prompt

I want this repo turned into a working local demo for a radiology focused vision language assistant. It should let me start the app, open a simple web UI, upload medical images, and chat with the model about what it sees in a way that feels useful for radiology workflows. If the project already supports expert tools for things like chest X ray classification or 3D CT segmentation, wire those in so they can be used from the same experience when available.

Please use the recommended setup from the repo, Docker is fine if that is the easiest path, and make the default model download work out of the box. If there is an option to use a local checkpoint instead, keep that available too. I would also like a clear run flow so I can launch the demo without hunting around.

If anything is unclear, look up the current docs or linked model pages online and follow the repo’s intended path rather than inventing extra features.