Getting Started with EdgeAI-OS: Your First 10 Minutes

EdgeAI-OS is designed to get you from zero to AI in under 60 seconds. This guide walks you through your first 10 minutes with the system.

What You’ll Need

• A computer or VM with 4GB RAM (2GB minimum)
• The EdgeAI-OS ISO file (1.2 GB download)
• 10 minutes of your time

Step 1: Download the ISO

Grab the latest release from GitHub:

Download EdgeAI-OS v0.1.7

The ISO is about 1.2 GB and includes everything you need:

• Debian-based Linux OS
• TinyLlama 1.1B (Oracle - complex tasks)
• SmolLM 135M (Sentinel - fast tasks)
• ai-sh and all tools pre-configured

Step 2: Boot the ISO

Option A: Virtual Machine (Recommended for Testing)

VirtualBox:

1. Create new VM → Type: Linux, Version: Debian 64-bit
2. RAM: 4096 MB (or 2048 MB minimum)
3. Storage: Create virtual disk, 8GB is plenty
4. Settings → Storage → Add ISO to IDE Controller
5. Start the VM

VMware:

1. Create new VM → Installer disc image: select the ISO
2. Guest OS: Linux, Debian 10+ 64-bit
3. RAM: 4GB, Disk: 8GB
4. Finish and boot

Option B: USB Boot (For Real Hardware)

On Linux:

sudo dd if=edgeai-os-0.1.7-x86_64.iso of=/dev/sdX bs=4M status=progress

On Windows, use Rufus or Balena Etcher.

Step 3: First Boot

The system will boot and auto-login as edgeai (password: edgeai).

You’ll see the service status:

=== EdgeAI-OS Services ===

  [OK] Sentinel (SmolLM 135M)
  [OK] Oracle (TinyLlama 1.1B)
  [OK] AI Daemon

  [OK] Socket ready

Quick start: ai-sh

Note: If services show [--], wait about 60 seconds for the models to load.

Step 4: Launch ai-sh

Type ai-sh and press Enter:

ai-sh

You’ll see the AI shell prompt:

┌─────────────────────────────────────┐
│  ai-sh - AiOS AI Shell v0.3.0      │
│  Type naturally or use bash commands│
│  Type !help for commands            │
└─────────────────────────────────────┘

[ai-sh] 

Step 5: Try Your First Queries

Instant Templates (No AI Wait)

These resolve immediately without touching the LLM:

[ai-sh] what time is it?
  [template] date
  Sun Feb  9 10:30:00 UTC 2025

[ai-sh] disk space
  [template] df -h
  Filesystem      Size  Used Avail Use% Mounted on
  /dev/sda1       7.9G  2.1G  5.8G  27% /

[ai-sh] memory usage
  [template] free -h
                 total        used        free
  Mem:           3.9Gi       1.2Gi       2.7Gi

[ai-sh] files larger than 100mb
  [template] find /home/edgeai -type f -size +100M

File Search Examples

[ai-sh] find all python files
  [template] find / -type f -iname '*.py' 2>/dev/null | head -30

[ai-sh] files modified in the last week
  [template] find . -mtime -7

[ai-sh] files starting with test
  [template] find . -name 'test*'

AI-Generated Responses

For complex queries, the LLM kicks in (takes 1-2 seconds):

[ai-sh] explain how to set up a cron job
  [ai-generated]
  To set up a cron job:
  1. Open crontab: crontab -e
  2. Add a line with format: minute hour day month weekday command
  3. Example: 0 2 * * * /path/to/script.sh (runs at 2 AM daily)
  4. Save and exit

Step 6: Use AI from Scripts

EdgeAI-OS isn’t just a shell - any program can use AI through the Unix socket.

From Bash

echo "what is 2+2?" | socat - UNIX-CONNECT:/tmp/aios/ai0.sock

From Python

import socket

sock = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
sock.connect('/tmp/aios/ai0.sock')
sock.send(b'what is 2+2?')
sock.shutdown(socket.SHUT_WR)
print(sock.recv(4096).decode())
sock.close()

From Any Language

Anything that can connect to Unix sockets can use AI. No libraries, no dependencies, no API keys.

Step 7: Check Status

Use !status in ai-sh to verify services:

[ai-sh] !status
Checking AI services...
  [OK] AI Socket
  [OK] Oracle (TinyLlama 1.1B)
  [OK] Sentinel (SmolLM 135M)

What to Try Next

Now that you’re up and running:

1. Explore templates - Try natural language queries and see what resolves instantly
2. Analyze logs - cat /var/log/syslog | ai-sh "summarize any issues"
3. Script with AI - Build a bash script that uses the AI socket
4. Read the docs - Check the GitHub repo for architecture details

Common Issues

Services Not Starting

Wait 60 seconds after boot. The models take time to load into memory.

Out of Memory

If you see OOM errors, ensure your VM has at least 2GB RAM (4GB recommended).

Slow Responses

First query may be slow as the model warms up. Subsequent queries are faster.

Summary

In 10 minutes, you’ve:

1. ✅ Downloaded EdgeAI-OS
2. ✅ Booted the ISO
3. ✅ Launched ai-sh
4. ✅ Run instant template queries
5. ✅ Seen AI-generated responses
6. ✅ Learned how to use AI from scripts

That’s it. You’re now running a complete AI system locally, with no cloud, no API keys, and no dependencies.

Welcome to AI as a system primitive.

Have questions? Check our GitHub Discussions or open an issue.