dgx-spark-ops

PackCloud & infra

dgx-spark-ops is a plugin for AI agents that handles NVIDIA DGX Spark (GB10) environment operations. It helps set up, preflight, and debug machine learning workloads on DGX Spark or aarch64 Blackwell hardware, covering tasks like CUDA-13 wheel and ABI checks, unified-memory OOM ladders, and thermal monitoring.

Unavailable. Delivery for this kind is on the roadmap — not serving yet.

Have access to NVIDIA DGX Spark or aarch64 Blackwell hardware running ML workloads.

What your AI can do with it

  • Set up ML environments on NVIDIA DGX Spark (GB10) hardware
  • Run preflight checks before launching ML workloads
  • Debug workloads on DGX Spark or aarch64 Blackwell hardware
  • Check CUDA-13 wheel and ABI compatibility
  • Walk through unified-memory OOM ladders
  • Monitor thermals during workloads

Getting started

  1. Have access to NVIDIA DGX Spark or aarch64 Blackwell hardware running ML workloads.
  2. Add the dgx-spark-ops plugin to the agent's available plugins.
  3. Ask the agent to set up, preflight, or debug a workload, and it will use the plugin's operations.

Signals

GitHub stars
40k
Forks
4k
Last commit
Sep 2026

Questions

What is dgx-spark-ops?
It is a plugin that helps an AI agent set up, preflight, and debug machine learning environments on NVIDIA DGX Spark (GB10) and aarch64 Blackwell hardware.
When should it be used?
Use it when setting up, preflighting, or debugging ML workloads on DGX Spark or aarch64 Blackwell hardware, including CUDA-13 checks, memory issues, or thermal problems.
Does it work on non-NVIDIA hardware?
It is built for DGX Spark (GB10) and aarch64 Blackwell hardware; other environments are not covered.
Advanced
Item type
plugin
Key
wshobson-agents-dgx-spark-ops
Source
github.com/wshobson/agents