Centralized architecture documentation for a distributed, cloud-agnostic AI and security laboratory engineered for resilience, privacy, and bare-metal performance.
This repository is the single source of truth for the architecture and operational design of the sovereign computing lab. It documents the multi-node topology, hardware specifications, software stack, and operational procedures.
The lab converts repurposed enterprise-grade hardware into a distributed, multi-node computing cluster capable of high-performance local AI inference with zero cloud dependencies.
| Node | Hostname | OS | Role | Hardware |
|---|---|---|---|---|
| 🧠 | Arch-GPU | Arch Linux | GPU Compute / LLM Inference | 4x NVIDIA P106-100 (24 GB VRAM) |
| 🔴 | Kali-Master | Kali Linux | Master Orchestrator / SecOps | x86 workstation |
| 🥧 | Raspberry-Pi | Raspbian | IDS / DNS / Network Services | ARM SBC |
| 🖧 | Dell-Gateway | Debian | Gateway / Monitoring | Enterprise rack unit |
| 💾 | AMD-Storage | Ubuntu Server | Storage Worker / CPU Compute | AMD Canwork189 |
🐧 OS Layer :: Arch Linux, Kali Linux, Debian, Raspbian
🐳 Containers :: Docker, Docker Compose
🤖 AI Runtime :: Ollama (local LLM), CUDA, Multi-GPU VRAM pooling
📊 Monitoring :: Prometheus, Grafana, custom Python agents
🛡️ Security :: Suricata IDS, custom log analytics, NVIDIA Morpheus (AI SecOps)
🌐 Networking :: VLAN segmentation, custom DNS, SSH hardening, Tailscale (remote access)
[Internet]
│
[Dell Gateway / Monitoring Node]
│
[Internal LAN (192.168.x.0/24)]
├── [Kali Master Orchestrator & SecOps Hub]
├── [Arch GPU Compute Cluster] ← 4x NVIDIA P106-100 / 24 GB VRAM
├── [Raspberry Pi IDS / DNS / Network Services]
└── [AMD Canwork189 Storage & CPU Worker]
| Component | Details |
|---|---|
| Runtime | Ollama (local, no cloud) |
| Model | Qwen 3.5:27B |
| GPU Setup | 4x NVIDIA P106-100 (6 GB each) = 24 GB VRAM total |
| Framework | CUDA multi-GPU with layer offloading |
| Host OS | Arch Linux (optimized kernel + DKMS) |
- Grafana → Real-time dashboards for GPU metrics, system health, network traffic
- Prometheus → Metrics collection across all nodes
- Suricata IDS → Network intrusion detection on Raspberry Pi node
- Custom Python Agents → Host-level log analytics and anomaly detection
- SSH key-based authentication across all nodes
- VLAN segmentation for service isolation
- MCP Bridge on Kali for direct security tool execution via AI agent
- AI-driven threat detection via NVIDIA Morpheus pipeline
- Zero cloud exposure — all data and inference remains on-premises
sovereign-ai-infrastructure/
├── README.md (this file)
├── docker-compose/
│ └── README.md Docker hosting layer documentation
├── hardware-bom/
│ └── README.md Hardware specifications & inventory
├── orchestration/
│ └── README.md Orchestration methodology & SOPs
├── network-topology/ Network diagrams & VLAN config (planned)
└── recovery-procedures/ Node recovery & disaster recovery (planned)
| Repository | Purpose | Status |
|---|---|---|
local-ai-sovereign-stack |
Docker AI stack deployment (Ollama + Grafana + n8n) | ✅ Active |
arch-linux-multi-gpu-llm |
GPU cluster optimization & CUDA tuning | ✅ Active |
cybersecurity-lab-automation |
Security automation & log monitoring agents | ✅ Active |
sovereign-ai-security |
NVIDIA Morpheus & AI-driven security tooling | ✅ Active |
sovereign-ai-skills |
Custom AI skills & Gemini CLI integration | ✅ Active |
n8n-automation-hub |
Workflow automation definitions | ✅ Active |
infrastructure-as-code-lab |
IaC experiments & deployment frameworks | 🔄 In Development |
proxmox-homelab-setup |
Proxmox hypervisor & LXC containerization | ✅ Active |
| Service | Status | Last Verified |
|---|---|---|
| Multi-GPU LLM Inference (Qwen 3.5:27B) | ✅ Stable | 2026-07-05 |
| Ollama Distributed Inference | ✅ Active | 2026-07-05 |
| Grafana / Prometheus Stack | ✅ Active | 2026-07-05 |
| Suricata IDS (Raspberry Pi) | ✅ Active | 2026-07-05 |
| MCP Bridge (Kali) | ✅ Active | 2026-07-05 |
| n8n Workflow Automation | ✅ Operational | 2026-07-05 |
| Docker Containerized Services | ✅ Operational | 2026-07-05 |
git clone https://github.com/Dinaverse/local-ai-sovereign-stack
cd local-ai-sovereign-stack
docker compose up -ddocker compose ps- Grafana: http://localhost:3000
- Prometheus: http://localhost:9090
- Ollama API: http://localhost:11434
Start here based on your goal:
| Goal | Start With |
|---|---|
| Understand the overall architecture | This README |
| Deploy Docker services | docker-compose/README.md |
| Check hardware specs | hardware-bom/README.md |
| Manage & orchestrate services | orchestration/README.md |
| Set up GPU inference | arch-linux-multi-gpu-llm |
| Deploy Proxmox hypervisor | proxmox-homelab-setup |
| Configure security automation | cybersecurity-lab-automation |
Sovereign by design — every component runs locally, owned entirely, zero cloud dependencies.