physics-agent-cli

Automate physics property classification for USD assets by running the Physics Agent CLI to apply VLM-based predictions for component classification and material physics characteristics. This reduces manual scripting and error-prone workflows in 3D asset pipelines.

Yes, you can generate simulation-ready UsdPhysics schemas from existing USD files. The CLI predicts physics characteristics and writes schemas directly into the assets, preparing them for downstream simulation pipelines.

To run the Physics Agent CLI with a VLM backend, you need an activated Python environment, the physics-agent CLI installed, and valid credentials and endpoints for your chosen VLM backend and render/optimizer services.

The best way to build per-asset datasets from USD files is using the Physics Agent CLI, which classifies component parts, identifies asset types, and predicts material properties to generate structured datasets for automation pipelines.

VLM-based physics prediction for 3D asset typing works by applying vision-language models to USD assets, analyzing component geometries to identify asset types and predict physics characteristics without manual inspection.