Skip to main content

Compatibility

Support matrix for choosing a Toposync installation path.

Systems

SystemStatusRecommended path
LinuxSupportedPython, Docker CPU, Docker CUDA, or processing server
macOSSupported for CPUPython or processing server
WindowsSupportedPython or processing server as a Windows service
Home Assistant OSSupported through the add-onHome Assistant add-on

Architectures

ArchitectureStatusNotes
amd64 / x86_64SupportedMain path for Linux, Windows, Docker, and HAOS
aarch64 / arm64Supported for CPUSupported path for Docker CPU and Home Assistant OS
Apple SiliconSupported for CPUUse the Python installation on macOS
Intel macOSSupported for CPUUse the Python installation on macOS
armv7, armhf, i386Outside the support targetUse a 64-bit system

Installation scenarios

ScenarioStatusNotes
Python on Linux and macOSSupportedRecommended with Python 3.12
Python on WindowsSupportedRecommended with Python 3.12
Docker CPUSupportedValidated for amd64 and arm64
Docker CUDALinux + NVIDIARequires NVIDIA driver and NVIDIA Container Toolkit
Home Assistant add-onSupportedCPU-only on amd64 and aarch64
Processing server on Linux and macOSSupportedCPU on macOS; CPU or CUDA on Linux
Processing server on WindowsSupportedCPU, DirectML, or native CUDA
Processing server on DockerSupportedCPU; CUDA when the local image was built with the CUDA target

GPU

AccelerationStatusUse when
CPUDefaultFirst install, light usage, and broad compatibility
CUDA on LinuxSupportedLinux host with an NVIDIA GPU
CUDA in DockerSupported on LinuxLinux host with an NVIDIA GPU and NVIDIA Container Toolkit
Native CUDA on WindowsSupported as a Python bundleWindows machine with compatible NVIDIA driver and runtime
DirectML on WindowsSupportedWindows GPU compatible with DirectML
CUDA in the Home Assistant add-onOutside the current scopeUse an external processing server if you need GPU acceleration

Raspberry Pi and HAOS

EnvironmentStatusRecommendation
Raspberry Pi with 64-bit HAOSSupported through the add-onUse aarch64
Raspberry Pi 5 8 GB + NVMePractical referenceBetter baseline for modern usage
Raspberry Pi 4Best-effortSuitable for light usage and compatibility testing
SD card storageBest-effortAvoid it for many cameras or write-heavy workloads
Heavy vision/OpenCV on ARM CPULimitedDelegate to a remote processing server

Practical rule

Start with CPU. Add GPU acceleration or a processing server only when you hit a real bottleneck in vision, OpenCV, multiple cameras, or ONNX inference.