AI Stack for Linux — runs on any hardware
مكدس الذكاء الاصطناعي للينكس — يعمل على أي جهاز
⚡ Quick Install • Download ISO • Features • Quick Start • Docs • Roadmap
Turn any Linux machine into an AI workstation in one command:
curl -fsSL https://raw.githubusercontent.com/yahyasaqban-lab/y7os/main/tools/y7-install | sudo bashOr download and run locally:
wget https://raw.githubusercontent.com/yahyasaqban-lab/y7os/main/tools/y7-install
chmod +x y7-install
sudo ./y7-install- ✅ Auto-detects your hardware (RAM, CPU, GPU, device type)
- ✅ Picks the best AI backend — Ollama (8GB+) or llama.cpp (4-8GB)
- ✅ Installs Python, Docker, Open WebUI
- ✅ Sets up Y7 CLI tools
- ✅ Enables ZRAM compression for low-resource devices
- ✅ All in English or Arabic — auto-detects your language
sudo ./y7-install --lang ar # Force Arabic
sudo ./y7-install --backend llama # Force llama.cpp
sudo ./y7-install --dry-run --verbose # Simulate first
sudo ./y7-install --skip-validation # Skip post-install checksCompatible with: Ubuntu 20.04+, Debian 11+, CentOS 8+, Raspberry Pi OS Minimum: 4GB RAM, 20GB storage
📖 Full installer docs: docs/INSTALLING.md 🧪 Test suite: tools/test-installer.sh 🔧 CI/CD: .github/workflows/test-installer.yml
Y7 OS is a Linux distribution built for local AI. It turns any computer — even old laptops with 4GB RAM — into a private AI workstation. No cloud. No subscriptions. No data leaving your device.
┌─────────────────────────────────────────────────────────────┐
│ Y7 OS v0.3.0 │
│ │
│ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │
│ │ Ollama │ │ Open WebUI │ │ Docker │ │
│ │ (LLMs) │ │ (Chat UI) │ │ (Containers)│ │
│ └─────────────┘ └─────────────┘ └─────────────┘ │
│ │
│ ┌─────────────────────────────────────────────────┐ │
│ │ y7-ai │ y7-models │ y7-status │ y7-gpu │ y7-persist │
│ └─────────────────────────────────────────────────┘ │
│ │
│ ┌─────────────────────────────────────────────────┐ │
│ │ Debian Bookworm (AMD64 + ARM64) │ │
│ └─────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────────┘
| Architecture | Download | Size | Use Case |
|---|---|---|---|
| AMD64 | y7os-0.3.0-amd64.iso | ~1.5 GB | Intel/AMD PCs, laptops |
| ARM64 | y7os-0.3.0-arm64.iso | ~1.5 GB | Raspberry Pi 4/5, Pine64 |
- ARM64 support — Raspberry Pi 4/5, Apple Silicon (via UTM)
- NVIDIA GPU detection —
y7-gpu detectand driver installation - Persistence mode — Save data across reboots with
y7-persist - TinyLlama bundled — Works offline immediately!
# Linux/macOS
sudo dd if=y7os-0.3.0-amd64.iso of=/dev/sdX bs=4M status=progress
# Or use Balena Etcher, Rufus, Ventoy- Ollama — Local LLM inference (auto-selected for 8GB+ RAM)
- llama.cpp — Lightweight alternative (auto-selected for 4-8GB)
- Open WebUI — ChatGPT-like interface at localhost:3000
- Docker — For Open WebUI and extensibility
- Auto-detects RAM, CPU, GPU; picks smallest working model
- ZRAM compression enabled on devices with <8GB RAM
- Light mode:
--lightflag installs only Ollama + CLI tools - Works on 4GB RAM machines, Raspberry Pi, old laptops
- Everything runs locally — no data leaves your machine
- No accounts, no subscriptions, no telemetry
- All models are open-source quantized variants
- Auto-detects system language
- All tools respond bilingually
- Full Arabic install:
sudo ./y7-install --lang ar
| Tool | Description |
|---|---|
y7-ai |
AI chat launcher, auto-picks best model for your RAM |
y7-code |
Code review, generation, translation across 6 languages |
y7-models |
Model manager (list, download, remove, info) |
y7-agents |
Autonomous AI agents for multi-step tasks |
y7-rag |
RAG pipeline (ingest docs, query with context) |
y7-voice |
Voice command interface for the AI |
y7-serve |
REST API server for all Y7 tools |
y7-web |
Web dashboard at localhost:9090 |
y7-backup |
Backup models, configs, and user data |
y7-status |
Real-time terminal dashboard |
y7-doctor |
Diagnose and fix common issues |
y7-update |
Self-update tools from GitHub |
Username: y7
Password: y7
y7-ai # Interactive chat
y7-ai "explain kubernetes" # Single questionhttp://localhost:3000
y7-models list # Show downloaded models
y7-models recommend # Best model for your RAM
y7-models download phi3 # Download a modely7-status # Dashboard
y7-bench # Benchmark your system| Spec | Minimum | Recommended |
|---|---|---|
| RAM | 4 GB | 8 GB+ |
| Storage | 20 GB | 50 GB+ |
| CPU | 64-bit x86 | 4+ cores |
| Internet | Required for first boot | Optional after setup |
| Model | RAM Required | Quality |
|---|---|---|
| TinyLlama | 2 GB | Basic |
| Phi-3 Mini | 4 GB | Good |
| Qwen 2.5 7B | 6 GB | Great (Arabic) |
| Llama 3.1 8B | 8 GB | Best |
- ARM64 support (Raspberry Pi, Apple Silicon)
- GPU acceleration (NVIDIA, AMD)
- Persistence mode for live USB
- Auto-update system
- Fine-tuning tools (Unsloth, Axolotl)
- RAG pipeline (ChromaDB, LlamaIndex)
- Voice interface (Whisper, TTS)
- Mobile companion app
- Stable release
- Enterprise support
- Certified hardware partners
See ROADMAP.md for details.
قم بتحميل ملف ISO من صفحة الإصدارات
# تسجيل الدخول
المستخدم: y7
كلمة المرور: y7
# محادثة بالعربية
y7-ai --lang ar
# لوحة الحالة
y7-status --lang ar
# أفضل نموذج لجهازك
y7-models recommend| النموذج | الذاكرة | جودة العربية |
|---|---|---|
| qwen2.5:7b | 6 GB | ممتازة |
| llama3.1:8b | 8 GB | جيدة جداً |
| phi3:mini | 4 GB | جيدة |
We welcome contributions! See CONTRIBUTING.md.
git clone https://github.com/yahyasaqban-lab/y7os.git
cd y7os
make build- Website: y7os.dev
- GitHub: github.com/yahyasaqban-lab/y7os
- Twitter/X: @y7os_ai
Found a vulnerability? Please report it responsibly. See SECURITY.md.
MIT License — free forever. See LICENSE.
Built by Yahya Saqban — Kuwait 🇰🇼
For everyone who deserves access to AI
🧠 AI Contributors: DeepSeek AI • Claude (Anthropic)