DGX Spark Control Center

A professional Windows desktop application for managing AI workloads on NVIDIA DGX Spark systems. Built with .NET 8 WPF.

Features

Current (v6.0)

• SSH Connection - Connect to DGX Spark via LAN or Tailscale VPN
• Process Management - View running Python processes with PID, CPU%, Memory%
• Service Control - Start/Stop/Restart services via configurable control scripts
• GPU Monitoring - Real-time GPU utilization, memory, temperature, and power (Multi-GPU support)
• System Metrics - CPU, RAM, Disk, Network I/O, and Uptime monitoring
• Log Viewing - Dynamic log discovery and viewing in popup window
• Auto-Refresh - Configurable refresh interval (default: 30 seconds)
• Multi-Profile Support - Save multiple connection profiles
• Configurable Services - Define services with process patterns, control scripts, and log paths
• Settings UI - In-app settings window for profile and service management

NVIDIA Playbooks (v3.0+)

One-click deployment for AI workloads based on build.nvidia.com/spark:

Training Frameworks:

• LLaMA Factory, Unsloth, NeMo Fine-tune

Inference Engines:

• vLLM, SGLang, TensorRT-LLM, Ollama

Applications:

• Open WebUI, ComfyUI, JupyterLab

Quantization:

• llama.cpp (GGUF inference and quantization)

NVIDIA NIM Microservices (v4.0)

Deploy containerized AI microservices from docs.nvidia.com/nim:

LLM NIMs:

• Llama 3.1 8B, Llama 3.1 70B, Mistral 7B

Retrieval NIMs:

• NV-EmbedQA (embeddings), NV-RerankQA (reranking)

Speech NIMs:

• Riva ASR (Speech-to-Text), Riva TTS (Text-to-Speech)

Safety NIMs:

• NeMoGuard (content safety & jailbreak detection)

Training Job Management (v5.0)

• Job Queue - View running, queued, and completed training jobs
• Progress Tracking - Real-time progress bars, epoch/step tracking, loss display
• Job Templates - LLaMA Factory, Unsloth, PyTorch SFT, NeMo, FLUX.1, DPO
• Job Control - Start, pause, resume, and cancel training jobs

Model Management (v5.0)

• Model Registry - Scan and list all models (base, fine-tuned, LoRA, GGUF)
• Model Operations - GGUF conversion, push to HuggingFace
• Filter & Search - Filter by model type and format

Multi-Spark Cluster (v5.0)

• NCCL Setup - Configure distributed training across multiple DGX Sparks
• DeepSpeed - Efficient distributed training support

AI Blueprints (v5.0)

Deploy NVIDIA AI Blueprints with one click:

• Multimodal RAG, Digital Human, Video Search & Summary
• Multi-Agent Chatbot, AI-Q Research Assistant, Data Flywheel

Robotics & Simulation (v6.0)

• Isaac Sim - Docker container launcher for robotics simulation
• Isaac Lab - Robot learning and reinforcement learning
• Cosmos NIMs - Text2World, Video2World, Transfer models

Advanced Features (v6.0)

• SSH Key Authentication - Connect with private key files
• Scheduled Jobs - Cron-based training job scheduling
• Desktop Notifications - Windows toast notifications for job events
• DevTools - VS Code Remote, Ollama+Continue, Code Server

Features:

• Auto-detect installed playbooks and NIMs
• One-click Setup/Start/Stop
• Status indicators (INSTALLED/RUNNING)
• Direct links to documentation

Screenshots

Dashboard

The main dashboard provides a system overview with GPU status, system metrics, and quick actions.

Playbooks

40+ one-click playbooks for training, inference, NIMs, robotics, and developer tools.

Training Management

Monitor and control training jobs with real-time progress tracking.

Requirements

• Windows 10/11 (64-bit)
• SSH access to your DGX Spark system
• No .NET runtime required for standalone exe

Installation

Quick Start (Recommended)

1. Download DGX-Spark-Control-Center-v6.0.zip from the Releases page
2. Extract the zip file
3. Run ProjectControlCenter.exe - no installation required!
4. Enter your DGX Spark connection details and click Connect

From Source

1. Clone the repository:

   git clone git@github.com:RealShocky/DGX-Spark-Control-Center-.git
   cd DGX-Spark-Control-Center-
   

2. Build and run:

   dotnet build
   dotnet run
   

3. Or build a standalone exe:

   dotnet publish -c Release -r win-x64 --self-contained true -p:PublishSingleFile=true -o ./publish
   

Configuration

Settings are stored in %APPDATA%\ProjectControlCenter\settings.json

Connection Profiles

Configure your DGX Spark connection profiles:

{
  "Profiles": [
    {
      "Name": "DGX Spark - LAN",
      "Host": "192.168.x.x",
      "Username": "your-username",
      "Description": "Local network connection",
      "IsDefault": true
    },
    {
      "Name": "DGX Spark - Tailscale",
      "Host": "your-spark.tailnet.ts.net",
      "Username": "your-username",
      "Description": "Tailscale VPN connection"
    }
  ]
}

Service Configuration

Define services to monitor and control:

{
  "Services": [
    {
      "Name": "My API Server",
      "Description": "FastAPI server for inference",
      "ProcessPattern": "uvicorn main:app",
      "ControlScript": "~/scripts/api-ctl.sh",
      "LogPath": "~/logs/api.log",
      "Port": 8000,
      "UrlPath": "/docs",
      "Enabled": true
    },
    {
      "Name": "Training Job",
      "Description": "Model fine-tuning process",
      "ProcessPattern": "train.py",
      "ControlScript": "",
      "LogPath": "~/logs/training.log",
      "Port": 0,
      "UrlPath": "",
      "Enabled": true
    }
  ]
}

Settings Options

Control Scripts

Create control scripts on your DGX Spark for service management:

#!/bin/bash
# ~/scripts/api-ctl.sh

case "$1" in
  start)
    cd ~/myproject && nohup python -m uvicorn main:app --host 0.0.0.0 --port 8000 > ~/logs/api.log 2>&1 &
    ;;
  stop)
    pkill -f "uvicorn main:app"
    ;;
  restart)
    $0 stop
    sleep 2
    $0 start
    ;;
esac

Roadmap

See ROADMAP.md for the full feature roadmap.

Completed Phases ✅

Upcoming Features

• Phase 11: Mobile & Web dashboard (planned)

Tech Stack

Playbook Categories

Contributing

This is an IAMVC Holdings project. Contact the team for contribution guidelines.

License

Proprietary - IAMVC Holdings

---

Built by IAMVC Holdings | Powered by NVIDIA DGX