Linux and Dell GB10-Class Workstations

Use this path for Linux workstations, Dell GB10-class systems, and similar local AI machines.

The Wright control plane does not require GPU access by default. Run heavy model servers or selected engineering backends beside Wright, then point Wright at their OpenAI-compatible endpoint.

Recommended First Run

Use the Docker appliance and configure LLM_API_URL to reach your local or hosted model endpoint. On Linux, add host.docker.internal:host-gateway in a local compose override if needed.

Validation Boundary

linux/amd64 is the smoked alpha target. linux/arm64, NVIDIA Container Toolkit, and --gpus all are follow-up validation items until the release workflow proves them.

For deeper workstation notes, see GB10 and DGX Workstations.