joeynyc

joeynyc / spark-doctor

Public

Local diagnostic CLI for NVIDIA DGX Spark (GB10). Detects power caps, unified memory pressure, thermal risk, Docker/runtime issues, and validates vLLM/Ollama/llama.cpp/SGLang recipes.

cli dgx dgx-spark diagnostics gb10
14
0
100% credibility
Found Apr 26, 2026 at 14 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

Spark Doctor is a diagnostic tool for NVIDIA DGX Spark systems that collects hardware metrics like GPU usage, memory pressure, and Docker status, then applies rules to deliver plain-English explanations of issues and recommended fixes.

How It Works

1
🔍 Hear about the helper tool

You're frustrated because your AI computer is running slow or crashing, and you find Spark Doctor recommended in online forums to figure out why.

2
📥 Get the tool ready

You bring Spark Doctor to your computer in a simple way so it's all set up and waiting to help.

3
🩺 Run a quick check-up

With one easy command, it looks at your computer's health, GPU, memory, and setup to spot common problems automatically.

4
📊 Get clear advice

It shows you in simple words what's wrong, why it's happening, and exactly what to try next, like improving cooling or freeing memory.

5
Share or fix yourself
✏️
Follow the suggestions

Try the easy fixes like stopping extra programs, and check again to see it working better.

📤
Make a safe report

Generate a cleaned-up version of your check-up results to post in forums without sharing personal details.

🎉 Everything runs great

Your AI computer is back to full speed, handling big tasks smoothly without surprises.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 14 to 14 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is spark-doctor?

Spark-doctor is a Python CLI serving as a local diagnostic app for NVIDIA DGX Spark (GB10) systems. It runs a quick scan to detect power caps, unified memory pressure, thermal risks, Docker runtime issues, and validates recipes for vLLM, Ollama, llama.cpp, or SGLang. You get a concise report: what's wrong, evidence from nvidia-smi, logs, and PSI metrics, plus next steps—no dashboards or telemetry, just read-only facts.

Why is it gaining traction?

Unlike scattered manual checks across nvidia-smi, dmesg, and docker info, it aggregates signals into one command with DGX Spark-specific rules, outputting anonymized reports ready for GitHub issues or forums. Recipe validation catches tensor-parallel mismatches or aggressive memory settings before deployment, and exit codes (0-3) make it CI-friendly. Privacy redaction and formats like markdown/forum keep diagnostics shareable without leaks.

Who should use this?

DGX Spark owners troubleshooting stalled inference under load, AI devs deploying local LLMs via Docker containers, or forum users needing evidence for power/thermal bugs. Ideal for ops engineers validating vLLM recipes on arm64 or spotting multi-backend memory fights.

Verdict

Grab it if you run DGX Spark—installs easily via pip, solid README with commands, and MIT license; early ruleset covers top pain points. At 10 stars and 1.0% credibility, it's raw but reliable for targeted diagnostics; contribute rules as issues emerge.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.