System Details

Base OS: Debian

Rock-solid stability, massive package repo, predictable release cycle. Your AI environments won't break from bleeding-edge updates.

Desktop: MATE

Traditional, lightweight, and fast. Minimal GUI resource footprint means maximum CPU/RAM for your AI models.

Display Server: X11

Proven compatibility with proprietary GPU drivers (NVIDIA CUDA), critical for hardware-accelerated AI inference.

Init System: systemd

Modern service management. AI daemons (Ollama, model servers) run securely and efficiently in the background.

Writing to USB

Flash the downloaded .iso to a USB drive (8GB+ recommended). Three methods:

Ventoy (Recommended)

Install Ventoy to your USB once, then simply copy the ISO file onto it. You can keep multiple ISOs and other files on the same drive. Boot from USB and select Logilocky from the Ventoy menu.

BalenaEtcher

Cross-platform GUI tool. Open BalenaEtcher → Select ISO → Select USB → Flash. Includes automatic verification.

dd (Linux CLI)

Warning: Double-check the of= parameter to avoid overwriting your system drive.

lsblk                    # identify your USB (e.g. /dev/sdb)
sudo dd if=logilocky.iso of=/dev/sdX bs=4M status=progress oflag=sync

Using the Debian Installer

Boot from USB → select Graphical Install or Install from GRUB.

  1. Select language, timezone, keyboard layout.
  2. Partitioning: "Guided – use entire disk" for simple installs, or "Manual" for dual-boot / custom partitions.
  3. For AI workloads, consider a separate /home or /opt on your fastest NVMe for storing large model files (GGUF/Safetensors).
  4. Create generous Swap (32GB+ on NVMe) to prevent OOM crashes when loading large LLMs.
  5. Configure network during install for latest security updates.

Kernel: Sysctl Tuning

Custom kernel tuning in /etc/sysctl.d/tuning.conf optimizes VM and networking for AI workloads:

vm.swappiness = 10              # keep model data in RAM longer
vm.dirty_ratio = 60             # aggressive memory buffering for model I/O
vm.dirty_background_ratio = 2   # early background writeback

Low swappiness prevents the kernel from paging model data to disk during inference. High dirty_ratio speeds up model saving/loading by buffering writes in memory.

Kernel: GRUB Settings

Boot parameters in /etc/default/grub.d/99-performance-tuning.cfg disable aggressive CPU sleep states and force the governor into a burst-ready mode — critical for instant response when you run AI inference.

LM Studio — Offline AI Running Guide

LM Studio is the recommended way to run local LLMs on Logilocky. It provides a polished GUI for downloading, managing, and chatting with GGUF models — with full GPU acceleration support.

Installation

  1. Open AI Center from the application menu.
  2. Find LM Studio in the Flatpak list and click Install.
  3. If this is your first Flatpak app, AI Center will install the Flatpak runtime automatically. A reboot will be required.

GPU Access (Critical Step)

Flatpak apps run in a sandbox and cannot access your GPU by default. You must grant GPU access manually:

  1. Open Flathub AI GPU Settings on application menuor use the terminal:
flatpak-ai-gpu-settings

This tool provides a simple CLI to manage GPU access for Flatpak apps. Just select LM Studio from the list to enable GPU acceleration.

Alternatively, install Flatseal (a Flatpak permissions manager) and toggle "GPU acceleration" / "All devices" for LM Studio.

flatpak install flathub com.github.tchx84.Flatseal

Downloading Models

Inside LM Studio, go to the Discover tab, search for models (e.g. llama-3.2-3b, mistral-7b, phi-3-mini), and click download. Models are stored locally and never leave your machine.

Performance Tips

  • Use Q4_K_M or Q5_K_M quantized models for the best speed/quality balance.
  • Set GPU layers to maximum in LM Studio settings for full VRAM offloading.
  • Monitor usage with AI Monitor while running models to track VRAM and CPU load.
  • For NVIDIA GPUs: ensure proprietary drivers + CUDA are installed via AI Center.

AI Center

The central hub for managing AI dependencies and applications. Install AI frameworks, Flatpak apps (LM Studio, Alpaca, GPT4All, Jan, Cherry Studio, Upscayl, etc.), Python ML libraries (PyTorch, scikit-learn, pandas, NumPy, XGBoost), and CLI tools (Gemini CLI, Codex CLI, GitHub Copilot CLI).

AI Center handles package backends (apt, flatpak, npm) automatically. First-time Flatpak installs trigger a reboot prompt to register the runtime. Status indicators show installed/not-installed for each app. You can launch AI Center from the application menu

Ollama Manager

A rich terminal interface for Ollama. Pull, delete, inspect, copy, and chat with local LLMs without memorizing CLI commands.

  • Pull Model: Download models by name (e.g. llama3.2, deepseek-r1:7b, phi3:mini).
  • Model Info: View quantization level, parameter count, context length, and the raw Modelfile.
  • Chat: Interactive multi-turn conversation with streaming output.
  • Quick Prompt: Single-shot generation for quick tasks.
  • Running Models: See which models are loaded in VRAM with their memory usage.
  • Service Control: Start, stop, restart the Ollama daemon without leaving the tool.
  • Install Ollama: One-click installation via the official install script.

You can launch Ollama Manager from the application menu

Alternatively, you can launch Ollama Manager from the terminal

# Launch
ollama-manager

AI Chat

Native desktop chat application with a PyQt6 UI. Supports multiple cloud AI providers — OpenAI, Anthropic (Claude), DeepSeek, xAI (Grok), and OpenRouter — with per-provider API key management and model selection.

  • Sidebar for provider/model/API key configuration.
  • Custom-painted chat bubbles with typing animation.
  • Conversation history with clear chat option.
  • Settings saved securely to ~/.config/logilocky_ai_settings.json (chmod 600).
  • Enter to send, Shift+Enter for newline.

You can launch AI Chat from the application menu

Alternatively, you can launch AI Chat from the terminal

# Launch
aichat

Terminal AI

Multi-provider AI chat right in your terminal. Supports OpenAI, Anthropic, Google Gemini, Groq, and OpenRouter with streaming responses.

  • Select provider and model from a menu, then chat interactively.
  • Session-only API key storage — keys are never written to disk.
  • Commands: !clear (reset history), !history (show messages), !tokens (estimate token count).
  • Supports custom/fine-tuned model names.

You can launch Terminal AI from the application menu

Alternatively, you can launch Terminal AI from the terminal

# Launch
terminal-ai

AI Monitor

Real-time system monitoring built for AI workloads. Tracks CPU usage, RAM allocation, GPU VRAM usage (NVIDIA via nvidia-smi), temperatures, and compute utilization. Designed to run alongside your AI inference to identify bottlenecks instantly.

You can launch AI Monitor from the application menu

Alternatively, you can launch AI Monitor from the terminal

# Launch
ai-monitor

AI Scraper

Web scraping tool for AI works. Extract structured data from websites, clean and process it with local models. Useful for building custom datasets, research, and data collection for fine-tuning.

You can launch AI Scraper from the application menu

Alternatively, you can launch AI Scraper from the terminal

# Launch
ai-scraper

OpenClaw Installer

Custom installer utility for deploying OpenClaw AI models seamlessly. Automates the environment setup and dependency handling for OpenClaw components.

You can launch OpenClaw Installer from the application menu

Alternatively, you can launch OpenClaw Installer from the terminal

# Launch
openclaw-installer

Huggingface CLI Store

A specialized tool to browse and download models/datasets from Huggingface directly via CLI. Helps manage your local model repository without needing a web browser.

You can launch Huggingface CLI Store from the application menu

Alternatively, you can launch Huggingface CLI Store from the terminal

# Launch
huggingface-cli-store

System Manager

Comprehensive GUI for system administration. Manage hostname, services (systemd), network diagnostics, hardware/disk monitoring, and core OS settings from a single unified interface. Built with PyQt6 and integrated with pkexec for privilege escalation.

You can launch System Manager from the application menu

Alternatively, you can launch System Manager from the terminal

system-manager

Package Manager

GUI frontend for apt package management. Search, install, remove, and update packages without touching the terminal. Shows package descriptions, installed status, and handles dependency resolution automatically.

You can launch Package Manager from the application menu

Alternatively, you can launch Package Manager from the terminal

packagemanager

Update Tool

One-click system update GUI. Runs apt update && apt upgrade with real-time progress output. Also handles Flatpak and Snap updates if those backends are installed. Uses pkexec for safe privilege escalation.

You can launch Update Tool from the application menu

Alternatively, you can launch Update Tool from the terminal

update

System Cleaner

Free up disk space safely. Clears apt cache, removes orphaned packages, cleans temporary files, and purges old kernels. Shows space saved after each operation.

You can launch System Cleaner from the application menu

Alternatively, you can launch System Cleaner from the terminal

system-cleaner

System Information

Hardware and system information dashboard. Displays CPU model, core count, RAM, GPU, disk usage, kernel version, uptime, and network interfaces in a clean GUI layout.

You can launch System Information from the application menu

Alternatively, you can launch System Information from the terminal

sysinfo

USB Toolkit

Format and prepare USB drives from the GUI. Supports ext4, FAT32, NTFS, and exFAT. Includes safe unmount, partition table creation, and drive identification to prevent accidental formatting of system drives.

You can launch USB Toolkit from the application menu

Alternatively, you can launch USB Toolkit from the terminal

usb-toolkit

Browser Manager

Install and manage web browsers. Supports Chromium, Firefox, Brave, and other browsers with one-click install/remove and default browser configuration.

You can launch Browser Manager from the application menu

Alternatively, you can launch Browser Manager from the terminal

browser_manager

Conky Settings

CLI configuration for your desktop conky widgets and performance monitors. Easily tweak appearance, update intervals, and displayed metrics like CPU, RAM, and Network.

You can launch Conky Settings from the application menu

Alternatively, you can launch Conky Settings from the terminal

conky-settings

Firewall Manager

GUI for managing nftables firewall rules. Toggle firewall on/off, configure inbound/outbound rules, block specific ports, and apply hardened presets. Rules are embedded directly in the tool — no external config files needed.

You can launch Firewall Manager from the application menu

Alternatively, you can launch Firewall Manager from the terminal

firewall

Privacy Tools

Privacy tools suite. Metadata removal, IP Checker, Cheksum Verification, Password Generator utilities.

You can launch Privacy Tools from the application menu

Alternatively, you can launch Privacy Tools from the terminal

privacy

Logilocky Crypt

File and directory encryption tool. Encrypt sensitive files, private datasets, or proprietary model weights with several encryption algorithms. Supports single file and batch directory encryption/decryption with password protection.

You can launch Logilocky Crypt from the application menu

Alternatively, you can launch Logilocky Crypt from the terminal

crypt

DNS Changer

Quickly switch your system DNS to privacy-focused providers (Cloudflare 1.1.1.1, Google 8.8.8.8, Quad9, etc.). Overrides NetworkManager persistently so changes survive reboots. GUI with one-click provider selection.

You can launch DNS Changer from the application menu

Alternatively, you can launch DNS Changer from the terminal

dns-changer

Notes

Lightweight note-taking app with persistent local storage. Create, edit, search, and delete notes. Data saved as JSON in ~/.config/. Designed for quick capture during development sessions.

You can launch Notes from the application menu

Alternatively, you can launch Notes from the terminal

notes

Todo

Task management app. Add tasks with priorities, mark as complete, filter by status. Persistent storage via JSON. Clean PyQt6 interface matching the Logilocky design language.

You can launch Todo from the application menu

Alternatively, you can launch Todo from the terminal

todo

Calendar

Desktop calendar with event management. Add events with dates and descriptions, view monthly/weekly layouts. Data persisted locally in JSON format.

You can launch Calendar from the application menu

Alternatively, you can launch Calendar from the terminal

calendar

Document Converter

Convert between document formats (PDF, HTML, DOCX, TXT, Markdown). Uses WeasyPrint for PDF generation. Batch conversion support with drag-and-drop interface.

You can launch Document Converter from the application menu

Alternatively, you can launch Document Converter from the terminal

document-converter

Frequently Asked Questions

Is Logilocky AI Linux free?

Yes, completely free and open-source. No subscriptions, no data harvesting.

Do I need internet to run AI models?

No. You only need internet to download the OS and pull model weights initially. Once downloaded, everything runs 100% locally. No data ever leaves your machine.

Minimum system requirements?

Base OS: 4GB RAM, 20GB disk. For AI models: 8GB+ RAM, NVIDIA GPU with 4GB+ VRAM strongly recommended. Use Q4/Q5 quantized models for best performance on limited hardware.

Can I install NVIDIA proprietary drivers?

Yes. System manager make it easy to install NVIDIA drivers.

What's the best way to run local LLMs?

Strongly recommend LM Studio. Install it via AI Center.

Also Ollama Manager is a good option.For terminal based local AI usage.

Contact & Support

Found a bug, have a feature request, or need help? Report all issues through official contact portal:

Bug reports · Feature requests · Security concerns

nixovena.org/contact

Disclaimer

Disclaimer of Liability

Logilocky AI Linux is provided "as is" without warranties of any kind. The developers are not liable for data loss, hardware damage, security breaches, or other issues arising from use. Running intensive AI workloads can stress hardware to thermal limits — ensure adequate cooling. Always back up critical data before installing a new OS or modifying kernel parameters.