Docker

Use Docker when you want Toposync isolated from the host system and easier to rebuild or move between machines.

Toposync publishes official images on GitHub Container Registry:

ghcr.io/toposync/toposync

Public GHCR images can be pulled without signing in to Docker or GitHub.

Choose a Docker path

Scenario	Recommended guide
Regular server, no GPU	Docker CPU
Linux host with NVIDIA GPU	Docker CUDA
Remote processing server in a container	Processing server on Docker

Default recommendation

Start with Docker CPU.

Move to Docker CUDA only when you have a Linux host with an NVIDIA GPU and a real vision workload that benefits from acceleration.

What Docker changes

• Toposync runs in a container.
• Runtime data is mounted from the host into /data.
• The UI and API are still served from the same public port.
• GPU support is a separate CUDA path, not part of the default CPU container.
• Streaming support is included in the public CPU and CUDA images.

For system and architecture support, see Compatibility.