Installation¶
KangEngine currently ships as platform-specific wheels built from a source checkout. Use Python 3.12 and build the wheel on the same OS and architecture family where it will run.
Requirements¶
macOS on Apple Silicon, or Ubuntu 24.04
Python 3.12
CMake, a C++17 compiler, and vcpkg
PhysX under
$HOME/Physics/PhysX
Linux/NVIDIA GPU simulation additionally requires the supported PhysX 5.8 GPU build and CUDA toolkit. Start with the CPU build unless GPU simulation is the reason you are installing KangEngine.
Create the development environment¶
From the repository root:
uv venv python/.venv --python 3.12
source python/.venv/bin/activate
uv sync --project python
Build the Python extension¶
make build_python
The normal Python build intentionally disables OpenUSD. To build an optional USD-enabled development extension instead:
make build_usd_python
Install for development¶
uv pip install -e ./python
Install a Built Wheel¶
If you received a wheel built for your operating system, architecture, and Python 3.12 ABI, install it into an isolated environment:
python3.12 -m venv .venv
source .venv/bin/activate
python -m pip install /path/to/kangengine-0.1.0-cp312-cp312-platform.whl
python -c "import kangengine as ke; print(ke.__file__)"
KangEngine wheels are platform-specific. The Linux CUDA wheel additionally requires a compatible NVIDIA driver and the GPU runtime described in PhysX GPU Simulation. Distributed wheels do not include OpenUSD support.
Continue with Verify Installation.
For platform-specific PhysX setup, source builds, CUDA 13 notes, and wheel production, see Build from Source.