The Sparky Command Center: Open-Source Dashboard for Local AI Hardware and GPU Fleets

X postTech2WildTool

Tech2Wild released The Sparky Command Center v2, an open-source monitoring command center for self-hosted AI labs. It tracks GPU nodes, network switches, and inference servers without server-side dependencies, featuring a 3D fleet visualization and an integrated telemetry-grounded chat assistant.

Key points

  1. Agentless monitoring: Collects GPU metrics over SSH using nvidia-smi without requiring software installed on the target nodes.
  2. Network and inference tracking: Integrates MikroTik RouterOS switch stats and Prometheus metrics from vLLM, SGLang, and llama.cpp (including TTFT, tokens/sec, and KV-cache).
  3. ComfyUI lane monitoring: Measures rendering status and memory usage directly from the driver rather than ComfyUI's internal estimates.
  4. Grounded chat assistant: Connects to OpenAI-compatible endpoints (such as Ollama or LM Studio) and injects live dashboard telemetry as system context.
  5. Lightweight stack: Runs on Python 3.9+ standard library with zero pip dependencies, serving a pre-built React, Three.js, and Vite interface.

Why it matters

Self-hosted AI labs often require setting up complex monitoring stacks with various exporters. This tool offers a zero-dependency alternative tailored to LLM inference and generative workflows, pairing live metrics with an LLM interface capable of diagnosing fleet status in natural language.

Analysis

By relying solely on the Python standard library and pre-packaged frontend assets, the project removes deployment hurdles. Its specialized focus on model-serving metrics—such as prefill/decode speeds and KV cache consumption—makes it considerably more relevant for local AI operations than general-purpose infrastructure monitoring dashboards.

Original post

Tech2Wild

@Tech2Wild · 17 October 2026

I built a command center for my whole home AI lab 🖥️⚡ 📊 Every Spark + the 3090 box: temps, watts, memory, live throughput 🌌 3D constellation of the fleet 🤖 Jarvis chat that reads the live board Open source, just python3 t.co/sGJrsEGbFp 👇 t.co/4CNrWbA0fJ t.co/AptN8qpLA7

25 likes5 replies16 bookmarks1.6k views

View on X
More posts in the thread (1)
  1. @RyanJ_KOL Yes, it also marks when a fleet is offline due to that
Selected replies (5)
  • @RyanJ_KOL @Tech2Wild The chat reading the live board is a useful touch. Does it include each node’s last telemetry timestamp and mark disconnected machines clearly? That would keep an old temperature or throughput sample from being presented as the current state.
  • @GunFister @Tech2Wild 🔥🔥🔥
  • @Franzferdinan57 @Tech2Wild starred and forked
  • @AgentSparko @Tech2Wild Nice!
  • @peterislion @Tech2Wild RAIDFLASH-PD t.co/33lsAZS3cz new engine using all concurrencies for maximun prefill and decode!! FOR SINGLE USER!!! use it for better performance on tensorfold and exl3 on GLM5.3 flash!!!!

Mentioned links

Links and tools from this entry

Summary generated by AI (Gemini 3.8 Flash, 18 October 2026). It may contain errors – the original source is authoritative.

Related entries

Similar in content, found by the AI search.

  • Repository:tonyd2wild/The-Sparky-Command-Center

    Sparky Command Center: Dependency-Free Monitoring for GPU Fleets and LLM Inference

    Sparky Command Center is a lightweight, zero-dependency web dashboard designed to monitor DGX Spark fleets and custom Linux GPU rigs. It collects hardware metrics over SSH and scrapes live inference performance data from vLLM and llama.cpp without external databases.

    40starsPython

    AI & LLMs· Tool

  • Repository:eugr/spark-vllm-docker

    spark-vllm-docker: Optimised vLLM Docker Setups for DGX Spark

    spark-vllm-docker is an open-source tool providing Docker configurations and startup scripts to run vLLM on DGX Spark systems, supporting both single-node setups and multi-node clusters via Ray or PyTorch distributed mode.

    2.2kstarsShell

    AI & LLMs· Tool

  • Link:James & Milo

    GhettoRak: Building an 8-Node DGX Spark Cluster

    A detailed build log documents how eight NVIDIA DGX Sparks were assembled on a standard wire rack using MikroTik 400G switches to form a compact AI cluster.

    AI & LLMs· Resource

  • Link:spark-arena.com

    Spark Arena: LLM Performance Leaderboard on NVIDIA DGX Spark

    Spark Arena is a public leaderboard tracking Large Language Model performance on NVIDIA DGX Spark hardware. Powered by open-source tooling, it provides transparent benchmark comparisons across inference engines including vLLM, SGLang, TensorRT-LLM, and llama.cpp.

    AI & LLMs· Tool

Back to the knowledge base

Let's talk about your project

Get in touch

Locations

  • Mattersburg
    Johann Nepomuk Bergerstraße 7/2/14
    7210 Mattersburg, Austria
  • Vienna
    Ungargasse 64-66/3/404
    1030 Wien, Austria

Parts of this content were created with the assistance of AI.