LLMOS

Natural language Linux orchestrator: describe what you want → LLM generates commands → validated for safety → executed in Docker containers.

Architecture

Next.js Frontend → FastAPI Backend → Rust CLI (llmos-exec) → Docker

• Frontend (frontend/): Next.js app with xterm.js terminal, WebSocket live streaming
• Backend (backend/): FastAPI with SQLite, job queue, LLM integration (OpenAI)
• Rust CLI (llmos-exec/): Streams NDJSON from Docker containers (stdout/stderr + exit code)

Supported Distros

Ubuntu, Debian, Fedora, Arch, Alpine, CentOS/Rocky, NixOS

Setup

Prerequisites

• Docker, Node.js 18+, Python 3.11+, Rust toolchain

Backend

cd backend
cp .env.example .env  # Add your OPENAI_API_KEY
pip install -r requirements.txt
python start_uvicorn.py

Frontend

cd frontend
npm install
npm run dev

Rust CLI

cd llmos-exec
cargo build --release

Environment Variables

Variable	Description	Default
OPENAI_API_KEY	OpenAI API key	(required)
OPENAI_MODEL	Model to use	gpt-4o-mini
LLM_PROVIDER	LLM provider	openai
DATABASE_URL	SQLite path	sqlite:///./llmos.db
NEXT_PUBLIC_API_URL	Backend URL for frontend	http://localhost:8000

API Endpoints

• POST /generate - Generate commands from natural language
• POST /validate - Validate commands for safety
• POST /execute - Execute commands (SSE streaming)
• POST /enqueue - Queue execution for async processing
• GET /history - Execution history
• GET /executions/{id}/status - Job status
• WS /ws/execute-live - WebSocket live terminal
• WS /ws/execute/{id} - WebSocket stream for existing job

Execution Options

The Rust CLI accepts resource limits via the JSON payload:

{
  "distro": "ubuntu",
  "commands": ["echo hello"],
  "memory_limit": "512m",
  "cpu_limit": "1.0",
  "container_name": "my-session",
  "keep_alive": true
}