Hello everyone,
I built an open source web control plane for spark-vllm-docker and I’d love DGX Spark owners to give it a try.
Spark Pulse brings recipe discovery, deployment management, live monitoring, cache cleanup, and configuration into one dashboard — so you don’t have to juggle the CLI, Docker, and separate monitoring tools.
What it shows and does:
• Recipe browsing — Explore deployment recipes, model variants, and mod combinations in a clean catalog view
• Deployment jobs — Launch deployments, watch live logs, inspect status, and stop running jobs from the UI
• Real-time monitoring — Track GPU, CPU, RAM, and disk usage with streaming updates
• Cache management — Review and clean Hugging Face, wheel, ccache, Triton, and related caches
Installation (under user):
python3 -m pip install --user spark-pulse --break-system-packages --upgrade
spark-pulse install --user
spark-pulse status --user
Open the UI in your browser (default port 8100).
Repo: https://github.com/kharkevich-engineering-lab/spark-pulse
Feedback welcome here or as GitHub issues. Thanks to anyone who gives it a try
Recipes
Monitoring




