System Architecture
Full pipeline from natural language to bootable ISO
User Natural Language Input
Describe your ideal OS in plain language. The system understands architectures, platforms, hardware requirements, and software preferences.
React Web UI / CLI TUI
IA Parser
LLM processes the natural language input and generates a structured YAML blueprint with all configurations, packages, and build parameters.
Ollama / LangChain → YAML Blueprint
Nix Flake Generation
YAML blueprint is transformed into a valid Nix flake configuration with all required packages, kernel modules, and system settings.
Nix Flakes + nixpkgs-unstable
Build Engine
Cross-compilation engine builds the OS for the target architecture using Nix reproducible builds, Docker containers, and QEMU emulation.
Nix build + Docker + QEMU Cross-Arch
NVIDIA CUDA Layer
Integrates NVIDIA proprietary/open drivers, CUDA toolkit, and TensorRT for AI acceleration. Supports Jetson, RTX, and server GPUs.
CUDA 12.x + TensorRT + DXVK
UniOS Modules
Universal FUSE filesystem layer mounts any FS natively. AI translators convert executables across platforms (GhidraLLM, Darling, Waydroid).
FUSE-Universal + AI-Wine/Darling/Waydroid
Linux Kernel 6.9-LTS
Custom kernel with eBPF hooks for monitoring, Rust FUSE modules for performance, and Ollama Edge for on-device AI inference.
eBPF + Rust FUSE + Ollama Edge
ISO/IMG Output
Final bootable ISO/IMG ready for bare metal, VM, or USB. Includes demo scripts for mounting Windows USBs and running cross-platform apps.
Boot + Demo 'mount-win-usb'
Tech Stack
Backend
FastAPI/Python
LangChain
Ollama
Frontend
React
CLI TUI
Builds
Nix Flakes
Docker/Podman
Yocto
IA Accel
NVIDIA CUDA
TensorRT
Docker NVIDIA
Cross-Arq
QEMU
LLVM
Kernel
Rust FUSE
eBPF
Linux 6.9
Security
Sandbox
Trivy
AppArmor