Skip to content
Back to search

sparkrun

MarketplaceWorks with Claude CodeGitHub

AI-assisted inference on NVIDIA DGX Spark - run, manage, and stop LLM workloads

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today

Check it deeper on your machine

It has changed since its scan on 11 Nov 2026.

As a Claude Code plugin, install it and then ask Claude Code:

/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/spark-arena/sparkrun with scanskillsafety

Or from a terminal, clone it and run it:

git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/spark-arena/sparkrun'

It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/spark-arena--sparkrun/badge.svg)](/listings/spark-arena--sparkrun)

HTML

<a href="/listings/spark-arena--sparkrun"><img src="/api/listings/spark-arena--sparkrun/badge.svg" alt="findsafeskills safety rating" /></a>

Install

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

uvx sparkrun setup

One command — installs sparkrun, then launches the guided setup wizard to create a cluster, configure SSH mesh, detect ConnectX-7 NICs, set up sudoers, and enable earlyoom.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

.claude-plugin/marketplace.json · Manifest · 33 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. {
  2. "$schema": "https://anthropic.com/claude-code/marketplace.schema.json",
  3. "name": "sparkrun",
  4. "description": "AI-assisted inference on NVIDIA DGX Spark - run, manage, and stop LLM workloads",
  5. "owner": {
  6. "name": "scitrera.ai",
  7. "email": "open-source-team@scitrera.com"
  8. },
  9. "plugins": [
  10. {
  11. "name": "sparkrun",
  12. "description": "Claude Code plugin for sparkrun - launch, manage, and stop LLM inference workloads on NVIDIA DGX Spark systems. Provides slash commands and skills for recipe browsing, cluster management, and multi-node orchestration.",
  13. "version": "0.0.5",
  14. "author": {
  15. "name": "scitrera.ai",
  16. "email": "open-source-team@scitrera.com"
  17. },
  18. "source": "./sparkrun-cc-plugin",
  19. "category": "development",
  20. "homepage": "https://github.com/spark-arena/sparkrun",
  21. "tags": [
  22. "inference",
  23. "llm",
  24. "dgx-spark",
  25. "nvidia",
  26. "vllm",
  27. "sglang",
  28. "gpu",
  29. "cluster"
  30. ]
  31. }
  32. ]
  33. }
README.md · README · 143 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. <p align="center">
  2. <img src="assets/sparkrun-banner.svg" alt="sparkrun — Part of the Spark Arena ecosystem" width="480" />
  3. </p>
  4. <p align="center">
  5. <a href="https://pypi.org/project/sparkrun/"><img src="https://img.shields.io/pypi/v/sparkrun?color=76b900" alt="PyPI version" /></a>
  6. <a href="https://github.com/spark-arena/sparkrun/blob/main/LICENSE"><img src="https://img.shields.io/github/license/spark-arena/sparkrun" alt="License" /></a>
  7. <a href="https://sparkrun.dev"><img src="https://img.shields.io/badge/docs-sparkrun.dev-1e40af" alt="Documentation" /></a>
  8. <a href="https://spark-arena.com"><img src="https://img.shields.io/badge/Spark_Arena-community-76b900" alt="Spark Arena" /></a>
  9. </p>
  10. <h3 align="center">One command to rule them all</h3>
  11. <p align="center">
  12. Launch, manage, and stop LLM inference workloads on one or more NVIDIA DGX Spark systems — no Slurm, no Kubernetes, no fuss.
  13. </p>
  14. <p align="center">
  15. <a href="https://sparkrun.dev">Documentation</a> &middot;
  16. <a href="https://sparkrun.dev/getting-started/quick-start/">Quick Start</a> &middot;
  17. <a href="https://sparkrun.dev/recipes/overview/">Recipes</a> &middot;
  18. <a href="https://spark-arena.com">Spark Arena</a>
  19. </p>
  20. ---
  21. ## Install
  22. ```bash
  23. uvx sparkrun setup
  24. ```
  25. One command — installs sparkrun, then launches the guided setup wizard to create a cluster, configure SSH mesh, detect ConnectX-7 NICs, set up sudoers, and enable earlyoom.
  26. ## Quick Start
  27. ```bash
  28. # Run an inference workload
  29. sparkrun run qwen3-1.7b-vllm
  30. # Multi-node tensor parallelism (TP maps to node count on DGX Spark)
  31. sparkrun run qwen3-1.7b-vllm --tp 2
  32. # Re-attach to logs, stop a workload, check status
  33. sparkrun logs qwen3-1.7b-vllm
  34. sparkrun stop qwen3-1.7b-vllm
  35. sparkrun status
  36. ```
  37. Ctrl+C detaches from logs — it never kills your inference job. Your model keeps serving.
  38. Watched launches also report Docker-start TTR and TTFT using a rank-local
  39. streaming readiness check for Docker vLLM/SGLang launches. By default, `run`
  40. follows startup logs and exits when ready; use `--follow` to stay attached,
  41. `--no-follow` to wait without model logs, or `--no-ready-wait` to return after launch. See
  42. [startup readiness](docs/STARTUP_READINESS.md) for configuration, timing
  43. boundaries, per-recipe overrides, and execution-strategy integration.
  44. See the [full CLI reference](https://sparkrun.dev/cli/overview/) for all commands and options.
  45. ## Updating
  46. ```bash
  47. sparkrun update
  48. ```
  49. Upgrades sparkrun (when installed via `uv tool`), applies compatible updates to trusted managed plugins, and refreshes recipe registries. Use `--no-update-plugins` for a local plugin compatibility audit without plugin updates. See [plugin management](docs/PLUGINS.md#install-and-manage-external-plugins).
  50. Moving from 0.3.x or a preview build? See the [0.4.0 upgrade guide](docs/UPGRADING_0_4.md)
  51. for plugin installation, changed defaults, and Python API migration.
  52. ### Update channels (advanced)
  53. Opt into preview builds installed from git instead of PyPI:
  54. ```bash
  55. sparkrun update --stable # PyPI stable release (default)
  56. sparkrun update --beta # develop branch preview
  57. sparkrun update --alpha # develop-next branch (bleeding edge)
  58. sparkrun update --yolo # alias for --alpha
  59. ```
  60. `sparkrun update` with no flag stays on your current channel; a channel flag switches and is remembered for future updates. The same flags work with `sparkrun setup install` and `sparkrun setup update`. Stable prints a plain version; beta/alpha add a channel suffix and commit (for example, `0.4.0-alpha+g1a2b3c4`). Switching from a preview channel back to `--stable` may downgrade.
  61. ## Highlights
  62. - **Multi-runtime** — vLLM, SGLang, llama.cpp out of the box
  63. - **Multi-node tensor parallelism** — `--tp 2` = 2 hosts, automatic InfiniBand/RDMA detection
  64. - **VRAM estimation** — know if your model fits before you launch (`sparkrun show <recipe>`)
  65. - **Git-based recipe registries** — we publish official recipes, community recipes, and benchmarked recipes via [Spark Arena](https://spark-arena.com), plus you can add your own registries.
  66. - **Guided setup wizard** — cluster creation, SSH mesh, CX7 auto-detection, sudoers, earlyoom
  67. - **Model & container distribution** — syncs models and images to cluster nodes over SSH automatically
  68. - **Manifest model caches** — automatic [expected-file validation, selective weight downloads and pluggable model transfer](docs/MODEL_CACHE.md)
  69. ## Python API and applications (0.4.0)
  70. The 0.4.0 branch introduces breaking Python API changes. Start with the
  71. [migration guide](docs/DISTRIBUTION_API_MIGRATION.md), including the supported
  72. imports and option/result contracts. [Application profiles](docs/APPLICATION_PROFILES.md)
  73. let CLIs, daemons, and desktop applications share the core with their own identity,
  74. paths, defaults, and plugins. Use `sparkrun.application.initialize()` for Python
  75. API use without the Sparkrun CLI. The [catalog API](docs/CATALOG_API.md) supports
  76. headless browsing, previews, and persistent recipe selection.
  77. [Preparation-only builds](docs/BUILD.md) build images or native Python environments
  78. and stage model assets before launch. Set `SPARKRUN_ADVANCED=1` to show
  79. `sparkrun build` in CLI help.
  80. [Benchmarking](docs/BENCHMARK_API.md) separates measurement frameworks from
  81. publication integrations. [Plugin authors](docs/PLUGINS.md) register installed
  82. integrations through `sparkrun.plugins`.
  83. The [proxy CLI](docs/PROXY.md) uses LiteLLM by default on every channel.
  84. [SparkRoute](docs/SPARKROUTE.md) is available as a separately installed plugin.
  85. Docker now stages a custom seccomp profile allowing io_uring on each launch node;
  86. see [executor configuration](docs/EXECUTORS.md#docker-seccomp-profiles-04).
  87. vLLM no longer sets `OMP_NUM_THREADS` by default; recipes can set it explicitly.
  88. The [daemon guide](docs/DAEMON.md) covers the authenticated local API, durable
  89. jobs, desktop integration, and systemd deployment. Daemon
  90. serving supports local TCP on Linux, macOS, and Windows behind the `daemon`
  91. feature flag; Linux also supports peer-authenticated Unix sockets.
  92. ## Spark Arena
  93. [Spark Arena](https://spark-arena.com) is the community hub for DGX Spark recipe benchmarks — browse benchmark results, then run them directly with sparkrun.
  94. ## Official Recipes
  95. [Official Recipes](https://github.com/spark-arena/recipe-registry) are maintained by the Spark Arena team and hosted on GitHub. They are tested and optimized for NVIDIA DGX Spark systems.
  96. ## Community Recipes
  97. [Community Recipes](https://github.com/spark-arena/community-recipe-registry) are contributed by the community and hosted on GitHub.
  98. ## License
  99. sparkrun is licensed under the Apache License 2.0 — see [LICENSE](LICENSE).
  100. ## Anonymous Telemetry
  101. sparkrun sends basic anonymous usage telemetry to `https://telemetry.sparkrun.dev` by default. Events include a random installation id stored in `~/.config/sparkrun/config.yaml`, sparkrun version, OS/version, system architecture, and command-specific metadata such as run runtime/model/parallelism/source/hardware counts, benchmark category/framework/profile/result keys, update version and registry counts, and setup-wizard step choices.
  102. The data allows us to make informed decisions about new features for sparkrun or the greater DGX Spark ecosystem.
  103. Telemetry very specifically does not include personally identifiable information or information that may reveal trade secrets. Telemetry does not include hostnames, usernames, local file paths, tokens, secrets, logs, private HF or local models, or full command arguments. Disable it persistently with `sparkrun setup telemetry --disable`, re-enable with `sparkrun setup telemetry --enable`, or opt out for one process with `SPARKRUN_NO_TELEMETRY=1`.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.