🦋 SL5 Auri (Lite)

The lightweight, offline Voice Control for Embedded Systems & IoT.

Run powerful voice commands on < 500 MB RAM. No Cloud. No AI Hallucinations.

Based on SL5 Aura Core • Installation • Features

---

🧐 What is Auri?

Auri is the "little sister" of SL5 Aura. While Aura is designed for high-end desktops with large AI models (Llama 3), Auri is optimized for efficiency.

It uses the same deterministic RegEx engine but strips away all heavy AI dependencies. This makes it perfect for:

• 🍓 Raspberry Pi (3, 4, 5, Zero 2W)
• 🤖 Robotics & IoT Projects
• 💻 Older Laptops / Netbooks
• 🔒 Privacy Purists (Strictly deterministic, no "fuzzy" AI logic)

Feature	🦁 SL5 Aura (Ultimate)	🦋 SL5 Auri (Lite)
Logic Engine	RegEx + Llama 3.2 (LLM)	RegEx Only (100% Deterministic)
RAM Requirement	4 GB - 8 GB	< 500 MB
Speech Model	Vosk Big (1.9 GB)	Vosk Small (45 MB)
Install Size	~ 8 GB	**~ 300 MB**

---

⚠️ Status: Early Development Installation scripts and service wrappers are currently in active development. The architecture and configuration format are stable — contributions welcome!

🚀 Installation

Auri uses a smart wrapper to install the SL5 Core in "Diet Mode".

1. One-Line Installer

# Clone this lightweight repo
git clone https://github.com/sl5net/auri.git
cd auri

# Run the installer wrapper
chmod +x install.sh
./install.sh

2. What happens now?

1. The installer fetches the latest stable SL5 Aura Core.
2. It installs system dependencies (Python, Vosk, Sound-Utils).
3. It EXCLUDES all heavy AI tools (Ollama, Training Data, Large Models).
4. It automatically downloads the lightweight German model (45 MB).

---

🎮 Usage

After installation, Auri behaves exactly like Aura, just faster and without the "Chat with AI" feature.

Start the Service

# Start the background service
./scripts/start_aura_service.sh

Add your own Commands

Edit the map files in config/maps/. Auri uses pure Python RegEx for matching.

# config/maps/my_commands.py
from plugins import BaseMap

# Simple Rule: When you say "Light on", it runs the command
# Auri is deterministic: precise input = precise output.
my_map = [
    ("Licht an", r"(licht|lampe)\s+an", "kasa --plug 1 on"),
    ("System Status", r"status", "htop"),
]

---

🛠️ Hardware Recommendations

• CPU: Any 64-bit CPU (ARM64 or x86_64). Single Core is enough.
• RAM: 512 MB minimum (1 GB recommended for comfort).
• Microphone: Any USB microphone or ReSpeaker HAT.

---

🗺️ Roadmap

Status	Feature
✅ Done	Deterministic RegEx engine (shared with Aura Core)
✅ Done	Vosk Small model support (45 MB)
🔄 In Progress	Installation wrapper (install.sh)
🔄 In Progress	Optional connect to SL5 Aura Service (full LLM mode)
🔜 Planned	Multi-user session support (shared Vosk server, isolated user contexts)
🔜 Planned	JWT-based authentication for multi-client setups
🔜 Planned	REST API for IoT device integration
🔜 Planned	k3s / Docker deployment for edge computing

❤️ Credits

Auri is maintained by the SL5NET team. If you need the full power of LLMs (Ollama), check out the main project: SL5 Aura.