AAAI (All Autonomous AI Assistant)

All Autonomous AI Assistant (AAAI) is an AI project that aims to create an AI system capable of assisting with various aspects of daily life, ranging from the simple to the complex.

The project explores how a self-hosted, local-first AI system can continuously observe, organize, and assist with daily schedules, multi-platform communications, personal tasks, and local hardware automation without depending on proprietary cloud AI services.

Status: Early development / Experimental

Quick Navigation: Project Goals • Roadmap • Design Principles • Hardware Requirements • Installation Guide • Telegram Commands • Project Structure • Development Status • Development • Disclaimer • License

Project Goals

The long-term goal of AAAI is to build an autonomous, always-on personal assistant capable of integrating across multiple data sources and hardware interfaces:

Daily schedules, reminders, and personal task management
Messaging platforms (WhatsApp, Telegram, Discord)
Educational and productivity platforms (School LMS, calendars, documents)
Local room hardware (Ambient wall display, audio alerts, power management)
Potentially more complex autonomous workflows in the future

The core reasoning engine and personal database should remain local, private, and independent from the underlying data sources or control interfaces.

Conceptually:

                       AAAI
                         │
            Autonomous Reasoning Layer
           (Local LLM & Vision Engine)
                         │
              Context & Memory Layer
          (SQLite DB / Schedule / Logs)
                         │
          ┌──────────────┼──────────────┐
          │              │              │
    Data Sources   Control Interfaces  Hardware Layer
          │              │              │
     ┌────┼────┐    ┌────┴────┐    ┌────┼────┐
     │    │    │    │         │    │    │    │
    WA   TG   LMS  Web UI    Bot  X11 Audio Power

This architecture is intended to allow AAAI to gradually expand from simple daily reminders into complex multi-platform life assistance without rewriting the core local AI foundation.

Roadmap

The project is developed incrementally. Early versions prioritize stability, 24/7 local operation on modest hardware, privacy, and practical daily utility over feature count.

v0.0 — Foundation

Establish the core local AI bridge, local database, schedule structure, hardware display/audio control, and power safety foundations.

Completed

Remaining

Implement structured rotating application logs
Establish automated backup for SQLite database and schedule JSON
Implement configurable multi-sound alarm profiles
Implement system temperature and resource monitoring on the dashboard

Current goal:

AAAI runs reliably 24/7 on a wall-mounted Linux system, managing daily schedules, tasks, and hardware wake-up alerts autonomously.

v0.1 — Multi-Platform Ingestion & Remote Control

Focus on connecting AAAI to daily messaging platforms and remote control channels.

v0.1.0 — WhatsApp & Telegram Foundation

v0.1.1 — Personal Telegram & Discord Ingestion

Read-only ingestion for personal Telegram channels and school groups
Discord bot/listener for school and community announcement channels
Unified message normalization across WhatsApp, Telegram, and Discord
Keyword and urgency detection (e.g., holiday announcements, schedule changes, deadlines)

v0.1.2 — School LMS Integration

School LMS connector / scraper foundation
Automatic extraction of new assignments and due dates
Synchronization between LMS deadlines and the local AAAI To-Do list
Proactive morning briefing alerts for upcoming LMS deadlines

v0.1 goal:

AAAI aggregates announcements across WhatsApp, Telegram, Discord, and School LMS to provide accurate daily briefings and deadline tracking.

v0.2 — Context Memory & Proactive Assistance

Focus on moving from reactive commands to proactive daily assistance.

Long-term conversational memory and user preference storage
Automatic conflict detection between school schedule changes and personal reminders
Offline progress and habit tracking
Context-aware study and task recommendations
Weekly summary reports via Telegram

v0.2 goal:

AAAI proactively identifies schedule changes, cancellations, and pending tasks without requiring manual checks.

v0.3 — Voice & Local Interaction

Explore hands-free room interaction in addition to keyboard and Telegram triggers.

Local wake-word detection research
Local Speech-to-Text (Whisper.cpp) integration
Natural local Text-to-Speech (Piper TTS) integration
Offline voice command execution

v0.3 goal:

Hands-free voice interaction inside the room while maintaining strict privacy modes when away.

v0.4 — Complex Autonomous Workflows

Expand AAAI toward complex multi-step digital tasks.

Modular tool-use / agent plugin framework
Local file and document summarization
Automated study material and flashcard generation
Cross-device synchronization between desktop PC, Linux server, and Android

v0.4 goal:

AAAI assists with complex multi-step workflows across devices.

v1.0 — Stable Release

The first major stable release.

Potential requirements:

Stable multi-platform ingestion (WhatsApp, Telegram, Discord, LMS)
Reliable 24/7 daemon operation and crash recovery
Complete hardware and power management integration
Comprehensive documentation and setup scripts
Modular configuration system
Automated test suite

Design Principles

AAAI is developed around several core principles:

1. Local-first and self-hosted

Core AI reasoning, personal schedules, and message logs should run locally on user-owned hardware without relying on paid third-party AI APIs.

2. Passive and safe platform observation

When reading external messaging platforms such as WhatsApp, AAAI operates strictly in read-only mode with explicit group whitelisting to prevent spam and protect account safety.

3. Hardware and environment awareness

An always-on room assistant must respect physical privacy—staying dark and silent when idle or when the user is away (/away), and waking the display or speaker only when triggered or scheduled.

4. Resilience on modest hardware

The system is optimized to run 24/7 on low-power dual-core laptops (such as an Intel Core i7-6600U) with automatic safe shutdown during power outages.

5. Bilingual comprehension, English immersion

To support daily language learning, AAAI accepts natural input in both Indonesian and English while consistently delivering interfaces and responses in clear English.

6. Incremental development

Each stage produces a practical, working improvement used in real daily life before expanding into more complex autonomous capabilities.

Hardware & System Requirements

Minimum Specifications

Component	Requirement
CPU	Dual-Core 64-bit x86_64 with AVX2 support (e.g., Intel Core i5/i7 6th Gen Skylake or newer)
RAM	8 GB RAM (for `qwen2.5:1.5b` + headless Chromium + Debian XFCE)
Storage	64 GB SSD (~15 GB free space for OS, Node.js, and Ollama models)
OS	Debian 12 (Bookworm) with XFCE Desktop Environment (X11/Xorg)
Power	AC Adapter + working internal battery (for 60-second blackout buffer)

Note: An X11 session (such as XFCE on Xorg) is required because AAAI uses xset dpms and xdotool to control physical screen power states.

Reference Hardware (Tested 24/7 Build)

Device: Toshiba Portege Z30-C
CPU: Intel Core i7-6600U (2 Cores / 4 Threads)
RAM: 16 GB DDR3L
Storage: 256 GB SATA SSD

Step-by-Step Installation

1. BIOS Configuration (Auto Power-On)

To allow the laptop to boot automatically when electricity returns after a power outage:

Enter BIOS setup on boot (press F2 or Del).
Navigate to Power Management or Advanced.
Set Wake on AC / Power On by AC to Enabled.
Save and exit (F10).

2. System Packages & Display Setup (Debian 12 XFCE)

Install the required system packages and remove default screen lockers so remote screen wake commands never get trapped on a login prompt:

sudo apt update && sudo apt upgrade -y
sudo apt install -y xfce4 xfce4-goodies lightdm x11-xserver-utils xdotool \
acpi alsa-utils mpg123 espeak-ng chromium curl git sqlite3 \
build-essential python3 tlp lm-sensors fonts-noto-color-emoji

sudo apt purge -y light-locker xfce4-screensaver
sudo systemctl enable --now tlp

Enable auto-login in /etc/lightdm/lightdm.conf under [Seat:*]:

autologin-user=YOUR_LINUX_USERNAME
autologin-user-timeout=0

Configure /etc/systemd/logind.conf to prevent suspend on lid close or idle:

HandleLidSwitch=ignore
HandleLidSwitchExternalPower=ignore
IdleAction=ignore

Restart systemd-logind:

sudo systemctl restart systemd-logind

3. Node.js 22 LTS & PM2

Install official Node.js 22 LTS from NodeSource to ensure native compatibility with better-sqlite3:

curl -fsSL [https://deb.nodesource.com/setup_22.x](https://deb.nodesource.com/setup_22.x) | sudo -E bash -
sudo apt install -y nodejs
sudo npm install -g pm2

4. Ollama & Local AI Models

Install Ollama and pull the text and vision models:

curl -fsSL [https://ollama.com/install.sh](https://ollama.com/install.sh) | sh
ollama pull qwen2.5:1.5b
ollama pull huihui_ai/qwen2.5-vl-abliterated:3b

Lock the active model in RAM and limit CPU threads to keep temperatures low during 24/7 operation:

sudo mkdir -p /etc/systemd/system/ollama.service.d
cat << 'EOF' | sudo tee /etc/systemd/system/ollama.service.d/override.conf
[Service]
Environment="OLLAMA_NUM_THREADS=3"
Environment="OLLAMA_KEEP_ALIVE=-1"
Environment="OLLAMA_NUM_PARALLEL=1"
Environment="OLLAMA_MAX_LOADED_MODELS=1"
EOF

sudo systemctl daemon-reload
sudo systemctl restart ollama

5. Clone & Configure AAAI

git clone [https://github.com/CloroLab/AAAI.git](https://github.com/CloroLab/AAAI.git) ~/aaai
cd ~/aaai
npm install
cp .env.example .env
nano .env

Fill in your TELEGRAM_TOKEN (from @BotFather) and MY_CHAT_ID inside .env (or leave MY_CHAT_ID blank and send /start to your bot on Telegram to auto-detect it). Optionally place a custom alarm.mp3 file in ~/aaai/alarm.mp3.

6. Link WhatsApp & Start 24/7 Service

Run node server.js once in the foreground to scan the WhatsApp Web QR code using WhatsApp -> Linked Devices on your phone:

node server.js

Once ✅ WhatsApp Listener is Ready! appears, press Ctrl+C and start the persistent PM2 service:

pm2 start server.js --name "aaai"
pm2 save
pm2 startup

(Run the sudo env PATH=... command output by pm2 startup).

7. Kiosk Autostart & 60s Power Watchdog

Copy the autostart desktop entry (launches the fullscreen dashboard on boot and turns off the display after 60 seconds of inactivity):

mkdir -p ~/.config/autostart
cp ~/aaai/scripts/dashboard-kiosk.desktop ~/.config/autostart/dashboard-kiosk.desktop

Setup the 60-second AC power outage auto-shutdown daemon:

sudo cp ~/aaai/scripts/power-watchdog.sh /usr/local/bin/power-watchdog.sh
sudo nano /usr/local/bin/power-watchdog.sh   # Insert your TELEGRAM_TOKEN and CHAT_ID
sudo chmod +x /usr/local/bin/power-watchdog.sh
sudo touch /var/log/power-history.log && sudo chmod 666 /var/log/power-history.log

cat << 'EOF' | sudo tee /etc/systemd/system/power-watchdog.service
[Unit]
Description=Auto Safe Shutdown on AC Power Loss (60s)
After=network-online.target

[Service]
ExecStart=/usr/local/bin/power-watchdog.sh
Restart=always
User=root

[Install]
WantedBy=multi-user.target
EOF

sudo systemctl daemon-reload
sudo systemctl enable --now power-watchdog.service

Telegram Commands

Command	Action
`/help`	Display all available bot commands.
`/morning`	Generate today's briefing (Schedule + To-Do List + WhatsApp updates).
`/groups`	List allowed WhatsApp groups and view recently detected chats.
`/allow_group <name>`	Add a WhatsApp group or contact name to the whitelist.
`/remove_group <name>`	Remove a WhatsApp group from the whitelist.
`/clear_wa`	Clear all saved WhatsApp messages from the dashboard feed.
`/add <task>`	Add a new task to the To-Do list.
`/done <id>`	Mark a To-Do item as completed by ID.
`/away`	Stealth Away Mode: Turn screen OFF & mute speaker; incoming alarms stay silent.
`/home`	Home Mode: Restore normal screen wake and audio alerts.
`/mute` / `/unmute`	Mute or unmute the laptop speaker remotely.
`/screen_off` / `/screen_on`	Force the laptop display OFF or ON remotely.
`/stop_alarm`	Stop a currently ringing alarm.
`/power_log`	View the last 15 AC power outage and system boot logs.
Send Photo	Parse a timetable image into `jadwal.json` using AI Vision.
Free Chat (EN / ID)	Set reminders (`"Panggil saya jam 13.00 untuk belajar"`) or ask questions.

Project Structure

AAAI/
├── README.md
├── LICENSE
├── package.json
├── .env.example
├── .gitignore
├── jadwal.example.json
├── server.js
├── public/
│   └── index.html
└── scripts/
    ├── dashboard-kiosk.desktop
    └── power-watchdog.sh

The repository structure will evolve as new data connectors and autonomous modules are separated into internal packages.

Development Status

Component	Status
Project foundation & Express API	🟢 Complete
SQLite database (`wa_messages`, `todos`)	🟢 Complete
Local LLM integration (Ollama `Qwen2.5`)	🟢 Complete
Vision schedule parser (`Qwen2.5-VL`)	🟢 Complete
Bilingual prompt & regex reminder parser	🟢 Complete
X11 DPMS display & ALSA audio control	🟢 Complete
Fullscreen Kiosk Web UI (`public/`)	🟢 Complete
WhatsApp read-only listener & whitelist	🟢 Complete
Telegram remote control & Away Mode	🟢 Complete
60s AC power loss auto-shutdown watchdog	🟢 Complete
Personal Telegram group ingestion	🟡 Planned
Discord announcement ingestion	⚪ Planned
School LMS assignment sync	⚪ Planned
Proactive conflict & holiday detection	🟡 In progress
Local voice interaction (STT/TTS)	⚪ Planned

Status Legend

🟢 Complete — implemented and tested
🟡 In progress — actively being developed
⚪ Planned — not yet implemented

Development

AAAI is currently developed in Node.js (v22 LTS) using Express, better-sqlite3, whatsapp-web.js, node-telegram-bot-api, and Ollama.

Basic commands used during development and operation:

# Run manually in foreground
node server.js

# Monitor 24/7 logs via PM2
pm2 logs aaai

# Check loaded Ollama models
ollama list

Disclaimer

AAAI is an experimental personal AI and hardware automation project created for educational, research, and daily productivity purposes.

WhatsApp Web integration relies on headless browser automation (whatsapp-web.js) and is strictly intended for passive, read-only personal notification monitoring. Hardware power and display controls (xset, xdotool, acpi) are designed for Debian/XFCE (X11) systems and may behave differently on other Linux distributions or laptop models.

License

This project is licensed under the GNU General Public License v3.0 (GPL-3.0). See LICENSE for details.

Name		Name	Last commit message	Last commit date
public		public
scripts		scripts
.env.example		.env.example
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
jadwal.example.json		jadwal.example.json
package.json		package.json
server.js		server.js

Latest commit

History

Folders and files

Repository files navigation