J.A.R.V.I.S.

Just A Rather Very Intelligent System

A fully local AI assistant for KDE Plasma 6, inspired by Tony Stark's iconic AI.
Powered by open-source LLM, speech recognition, and text-to-speech — 100% offline, 100% private.

---

Screenshots

COMM — Chat Interface	SYSTEMS — Diagnostics	CONFIG — Quick Controls
Settings — LLM & Models	Settings — Voice Commands	Settings — J.A.R.V.I.S.

---

Features

Feature	Description
Local LLM Chat	Conversational AI via llama.cpp — no cloud, no API keys
Self-Contained	Bundled llama.cpp + whisper.cpp — no separate installs needed
Bundled LLM Server	llama-server included, start/stop from settings UI
Wake Word	Say "Jarvis" to activate voice commands (bundled whisper.cpp, CPU-only)
Voice Commands	14 built-in commands + fully customizable mappings
System Interaction	LLM can open apps, run commands, write files, type text
Text-to-Speech	Piper TTS with downloadable voices (espeak-ng fallback)
System Monitor	Real-time CPU, RAM, temperature, uptime display
Timers & Reminders	Quick presets and custom timers with audio alerts
Iron Man HUD	Arc reactor animation, waveform visualizer, holographic UI
Privacy First	Everything runs locally — no data ever leaves your machine

Requirements

System Dependencies

# Arch Linux / Manjaro
sudo pacman -U jarvis-plasmoid-0.2.0-1-x86_64.pkg.tar.zst
# Or build from source:
sudo pacman -S qt6-base qt6-declarative qt6-multimedia qt6-speech \
  plasma-framework extra-cmake-modules ki18n pipewire xdotool

# Fedora
sudo dnf install qt6-qtbase-devel qt6-qtdeclarative-devel qt6-qtmultimedia-devel \
  qt6-qtspeech-devel plasma-framework-devel extra-cmake-modules kf6-ki18n-devel \
  piper pipewire xdotool

# Ubuntu / Kubuntu (24.04+)
sudo apt install qt6-base-dev qt6-declarative-dev qt6-multimedia-dev qt6-speech-dev \
  libplasma-dev extra-cmake-modules libkf6i18n-dev piper pipewire xdotool

Bundled Components (No Separate Install Required)

• llama-server — Automatically built from llama.cpp source during installation
• whisper.cpp — Bundled and built from source (v1.7.3)
• Piper TTS — Optional download via settings UI

Whisper Model

Download the tiny English model for wake word detection:

mkdir -p ~/.local/share/jarvis
wget -O ~/.local/share/jarvis/ggml-tiny.en.bin \
  https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-tiny.en.bin

LLM Server

J.A.R.V.I.S. now includes a bundled llama-server. You can start/stop it from the plasmoid settings, or use your own external server.

To use the bundled server:

1. Open J.A.R.V.I.S. settings → General tab
2. Click "Start Bundled LLM Server"
3. Download a model (e.g., Qwen 2.5 0.5B) using the built-in downloader
4. The server will load the model automatically

Or continue using an external llama-server:

llama-server -m your-model.gguf --port 8080

Installation

Universal Installation (Recommended)

Works on any Linux distribution with KDE Plasma:

# Clone the repository
git clone https://github.com/novik133/jarvis.git
cd jarvis

# Run the universal installer (handles all dependencies automatically)
chmod +x install.sh
./install.sh

The universal installer automatically:

• Detects your Linux distribution (Ubuntu, Debian, Fedora, openSUSE, Arch, etc.)
• Installs all required dependencies using the system package manager
• Builds llama.cpp and whisper.cpp from source
• Installs J.A.R.V.I.S. to your system

Distribution-Specific Packages

Arch Linux / Manjaro

# Install from AUR or build package
makepkg -f
sudo pacman -U jarvis-plasmoid-*.pkg.tar.zst

Manual Dependency Installation

# Fedora
sudo dnf install qt6-qtbase-devel qt6-qtdeclarative-devel qt6-qtmultimedia-devel \
  qt6-qtspeech-devel plasma-framework-devel extra-cmake-modules kf6-ki18n-devel \
  piper pipewire xdotool

# Ubuntu / Kubuntu (24.04+)
sudo apt install qt6-base-dev qt6-declarative-dev qt6-multimedia-dev qt6-speech-dev \
  libplasma-dev extra-cmake-modules libkf6i18n-dev piper pipewire xdotool

Build from Source (if dependencies already installed)

git clone https://github.com/novik133/jarvis.git
cd jarvis

mkdir build && cd build
cmake .. -DCMAKE_INSTALL_PREFIX=/usr
make -j$(nproc)
sudo make install

# Restart Plasma
plasmashell --replace &

Add to Panel

1. Right-click your Plasma panel
2. Add Widgets... > Search for "Jarvis"
3. Drag it to your panel or desktop
4. Right-click the widget > Configure to set up LLM server and download voices

Configuration

Right-click the J.A.R.V.I.S. widget > Configure to access:

General

• LLM server URL and model selection
• TTS voice selection and download
• Speech rate, pitch, and volume
• Wake word and audio settings
• Conversation memory length
• AI personality customization

Voice Commands

• View and edit all voice command mappings
• Add custom commands (app launchers or shell commands)
• Reset to defaults

Advanced System Interaction

J.A.R.V.I.S. can interact with your system through natural language:

"Jarvis, open the terminal and run htop"
"Jarvis, create a file on my desktop called notes.md with today's meeting notes"
"Jarvis, open Kate and write a Python hello world script"
"Jarvis, what's my CPU temperature?"

The LLM generates structured action blocks that the backend executes:

• run_command — execute shell commands silently
• open_terminal — open a terminal with a command
• write_file — create files with content
• open_app — launch GUI applications
• open_url — open URLs in browser
• type_text — type text into the focused window

Architecture

src/plugin/
  jarvisbackend.h/.cpp        Coordinator — composes all modules
  jarvisplugin.h/.cpp         QML plugin registration
  settings/
    jarvissettings.h/.cpp     Persistence, model/voice lists, downloads
  tts/
    jarvisTts.h/.cpp          Piper TTS + espeak-ng fallback
  audio/
    jarvisaudio.h/.cpp        Audio capture, whisper.cpp, transcription
  system/
    jarvissystem.h/.cpp       CPU/RAM/temp monitoring, clock
  commands/
    jarviscommands.h/.cpp     Voice command mappings
package/
  metadata.json               Plasma applet metadata
  contents/
    ui/main.qml               Main widget UI
    ui/configGeneral.qml      General settings page
    ui/configCommands.qml     Voice commands settings page
    ui/configAbout.qml        About page
    config/config.qml         Configuration tab model

Resource Usage

J.A.R.V.I.S. is designed to be lightweight:

Component	Memory	CPU (idle)	Notes
Plugin (C++ backend)	~15 MB	<1%	Modular, lazy initialization
Whisper (wake word)	~40 MB	~2%	tiny.en model, 2 threads, 512 audio context
Piper TTS	~0 MB idle	0% idle	Spawns only when speaking, exits after
LLM Server	~500 MB+	Variable	Bundled llama-server — user starts via UI

• Whisper runs in a background thread, never blocks the UI
• Piper TTS is a one-shot process per utterance (no persistent daemon)
• Audio buffer has amplitude threshold — whisper skips silent frames
• System monitor polls every 3 seconds (minimal /proc reads)

Contributing

Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

License

This project is licensed under the GNU General Public License v3.0 — see the LICENSE file for details.

Author

Kamil 'Novik' Nowicki

• Website: kamilnowicki.com
• GitHub: github.com/novik133
• Email: kamil@kamilnowicki.com

Support

If you find J.A.R.V.I.S. useful, consider supporting its development:

---

"At your service, Sir. All systems nominal."

Copyright © 2026 Kamil Nowicki. All rights reserved.