Prerequisites¶

Before installing Home Security Intelligence, ensure your system meets the following requirements.

Hardware Requirements¶

GPU (Required)¶

The system runs two AI models simultaneously:

• YOLO26 (object detection): ~4GB VRAM (ai/start_detector.sh:6-7)
• Nemotron (risk analysis): ~14.7GB VRAM production (Nemotron-3-Nano-30B) or ~3GB development (Mini 4B)

Supported GPUs:

• NVIDIA RTX 30-series (3060 and above)
• NVIDIA RTX 40-series (any)
• NVIDIA RTX A-series (A2000 and above)
• NVIDIA Tesla/Quadro with 8GB+ VRAM

CPU & Memory¶

Note: Storage requirements increase with camera count and retention period. Plan for ~1GB/day per active camera.

Network¶

• Cameras must be able to FTP upload to the server
• Local network access (no internet required after setup)
• Default ports: 80 (web), 8000 (API), 8095 (detection), 8091 (LLM)

Software Requirements¶

Operating System¶

NVIDIA Drivers & CUDA¶

# Verify NVIDIA driver
nvidia-smi

# Required output should show:
# - Driver Version: 535+
# - CUDA Version: 12.0+

Installation guides:

• Ubuntu: NVIDIA CUDA Installation Guide
• macOS: CUDA not available; use MPS backend

Python¶

# Verify Python version
python3 --version
# Python 3.10.x or higher

Installation:

# Ubuntu/Debian
sudo add-apt-repository ppa:deadsnakes/ppa
sudo apt update
sudo apt install python3.10 python3.10-venv python3.10-dev

# macOS (via Homebrew)
brew install python@3.10

Node.js¶

Note: Vite 7 requires Node.js 20.19+ or 22.12+ for native ESM support. Node.js 18 is NOT supported.

# Verify Node version
node --version
# v20.19.x or v22.12.x or higher

Installation:

# Ubuntu/Debian (via NodeSource)
curl -fsSL https://deb.nodesource.com/setup_20.x | sudo -E bash -
sudo apt install nodejs

# macOS (via Homebrew)
brew install node@20

Container Runtime¶

This project supports both Docker and Podman. Choose whichever is available on your system.

# Verify Docker
docker --version
docker compose version

# OR verify Podman
podman --version
podman-compose --version

Installation:

# Ubuntu/Debian
sudo apt update
sudo apt install docker.io docker-compose-v2
sudo usermod -aG docker $USER  # Log out and back in

# macOS (Docker Desktop)
# Download from https://www.docker.com/products/docker-desktop

# Ubuntu/Debian
sudo apt install podman podman-compose

# Fedora/RHEL
sudo dnf install podman podman-compose

# macOS (via Homebrew)
brew install podman podman-compose
podman machine init
podman machine start

macOS Note: If using Podman on macOS, set AI_HOST=host.containers.internal before starting containers. Docker Desktop uses host.docker.internal by default.

llama.cpp¶

Required for running the Nemotron LLM server.

# Verify llama-server is available
llama-server --version

Installation:

# Build from source (recommended for GPU support)
git clone https://github.com/ggerganov/llama.cpp
cd llama.cpp
make LLAMA_CUDA=1  # For NVIDIA GPU support

# Add to PATH
export PATH="$PATH:$(pwd)"

Alternative: Pre-built binaries available at llama.cpp releases.

Verification Checklist¶

Run these commands to verify all prerequisites:

# GPU
nvidia-smi | head -5

# Python
python3 --version

# Node.js
node --version && npm --version

# Container runtime (choose one)
docker --version && docker compose version   # Docker
# OR
podman --version && podman-compose --version  # Podman

# llama.cpp
which llama-server

Expected output (Docker example):

NVIDIA-SMI 535.xxx  Driver Version: 535.xxx  CUDA Version: 12.x
Python 3.10.x or higher
v20.19.x (or v22.12.x+)
10.x.x
Docker version 24.x.x
Docker Compose version v2.x.x
/usr/local/bin/llama-server

Expected output (Podman example):

NVIDIA-SMI 535.xxx  Driver Version: 535.xxx  CUDA Version: 12.x
Python 3.10.x or higher
v20.19.x (or v22.12.x+)
10.x.x
podman version 4.x.x
podman-compose version 1.x.x
/usr/local/bin/llama-server

Next Steps¶

Once all prerequisites are met, proceed to:

Installation - Set up the environment and download models.