// Open-Source AI Assistant

N.O.V.A.

Explore Features ★ View on GitHub
v0.5 Active
MIT Licensed
Windows · Linux · macOS (untested)
Groq · Gemini · Local
01Capabilities

Built different.
By design.

Not another chatbot. NOVA is a fully local, voice-activated AI system with a cinematic HUD — running on your machine, on your terms.

🎙️

Wake Word Activation

Just say "Nova" and start talking. No clicks, no typing — it listens the moment you speak.

🌐

Holographic HUD

A stunning glassmorphism overlay rendered by Electron — looks exactly like the movies, runs on your desktop.

Dual AI Engine

Switch between Groq's ultra-fast inference or Google Gemini's massive context window — or run both simultaneously.

🛑

Instant Interrupt

Click the NOVA core at any moment to instantly stop a response mid-sentence. Full control, always.

🖥️

Cross-Platform

Windows and Linux fully supported. macOS is included but untested (I don't own a Mac) — Mac contributors welcome, see below.

🔓

Fully Open Source

MIT licensed. Fork it, mod it, break it, rebuild it. The source is yours — no black boxes.

02Step 1 — AI Engine

Pick your AI engine.

NOVA can run on free cloud API keys or completely offline with local models via llama.cpp.

"Give me a few hours. I'll have something better figured out." — Tony Stark

Groq API

Blazing fast inference using custom LPU hardware. Get responses in under a second, completely free on the standard tier.

→ Get Groq API Key
⭐ Best context window

Google Gemini API

Massive context window and deep multi-turn reasoning. Ideal for long conversations and complex tasks. Free via AI Studio.

→ Get Gemini API Key
🔒 100% Offline & Private

Local Models (llama.cpp)

Run NOVA completely offline with local AI models. Auto-detects your hardware and recommends the best model. Download models directly from Hugging Face within the app.

→ Installed During Setup

// Gemini is prioritized over Groq when both cloud keys are provided. Local models can be enabled from settings.

03Step 2 — Installation

Up and running
in under 5 minutes.

Choose your platform and follow the steps. A single script handles everything else.

⚠️

Pre-built installer is still being worked on — The .exe installer and .zip portable are being added soon but to maintain stablity it's being tested. For now, the script install below takes less than 5 minutes and works on any machine.

⚠️

Development build warning — NOVA v0.5 is still under active development. Some features may have bugs while the systems are being upgraded.

01

Get the source code

Option A — Clone with Git:

git clone https://github.com/Aetex/nova.git
cd N.O.V.A.-AI-Assistant
📦

Option B: Download ZIP archive — extract the folder, open it.

02

Run the installer

Double-click install_windows.bat and wait for "INSTALLATION COMPLETE!". It sets up your virtual environment, automatically creates the configuration files, and installs all dependencies.

03

Launch NOVA

Double-click launch_nova.bat. Once the holographic HUD boots up, you can paste and configure your Groq or Gemini API keys directly inside the settings panel.

🍎

Untested on macOS — I don't own a Mac, so this install path hasn't been verified on real hardware. If you hit a bug (or want to help, Mac or not), see Contribute below — everyone who helps gets permanent credit.

01

Get the source code

Option A — Clone with Git:

git clone https://github.com/Aetex/nova.git
cd N.O.V.A.-AI-Assistant
📦

Option B: Download ZIP archive — extract and open in terminal.

02

Run the installer

sh install_mac.sh
03

Launch NOVA

Launch the HUD, then paste and configure your API keys directly inside the settings panel:

sh launch_nova.sh
01

Clone the repository

git clone https://github.com/Aetex/nova.git
cd N.O.V.A.-AI-Assistant
02

Run the installer

sh install_linux.sh
03

Launch NOVA

Launch the HUD, then paste and configure your API keys directly inside the settings panel:

sh launch_nova.sh
04Step 3 — How to Use

Talk to it.
It listens.

NOVA is designed to feel natural. No menus, no configuration panels — just your voice.

"NOVA, it's time for a little upgrade." — Tony Stark
// activate

Wake Word

Say "Nova" to get its attention, then speak your command naturally. No push-to-talk button required.

// compact mode

Mini HUD

Hit the [MINI] button to shrink the HUD into a slim overlay. Perfect for multitasking without losing NOVA.

// stop output

Interrupt Anytime

Click the central NOVA core orb at any time to immediately cut off a long response mid-sentence.

// developer mode

Live Debug Logs

Say "Nova, enter debugging mode" to see real-time system logs — useful for tweaking or troubleshooting.

05Known Issues

Something not
working?

Here are the most common issues and how to fix them in under a minute.

"I'm going to need a little time to work on that." — Tony Stark

Microphone not detected

Go to your OS sound settings and set your microphone as the default input device. NOVA uses the system default — no manual device selection yet.

Backend fails to start / API error

Make sure you have configured at least one key (Groq or Google Gemini) inside the settings panel in the HUD, and that there are no invalid characters or trailing spaces.

HUD doesn't open / blank window

Run npm install inside the ui/ folder first. Electron dependencies must be installed separately from the Python backend.

PyAudio install fails on Linux

Re-run sh install_linux.sh. The installer now detects your distro and installs the required PyAudio and PortAudio system packages automatically.

06Contributing

Help build NOVA.
Mac or not.

NOVA was built and tested on Windows and Linux. I don't own a Mac, so the macOS path is untested on real hardware — that's the biggest gap right now. But any contribution, on any platform, gets you the same thing: real credit.

"Sometimes the best solutions are the simplest ones." — Tony Stark
🍎

macOS especially wanted — test the install, file a bug, or fix one, and you'll be credited as macOS maintainer, both in this repo and in NOVA's in-HUD credits.

No Mac? Still welcome — money donation, bug fixes, features, docs, and design contributions on any platform all get you added to the Contributors list below.

🧪

Test macOS

Run the install script on your Mac and report back — even "it worked fine" is useful. Open an issue.

🐛

Fix a Bug

macOS, Windows, or Linux — found and fixed something? Open a Pull Request and it'll get reviewed and merged.

🔊

Mic & Audio

macOS handles audio permissions very differently — this is probably the highest-impact area to dig into.

📦

Packaging

Help get the Electron build signed and notarized for a smoother, warning-free experience on any OS.

Add a Feature

UI polish, wake-word tweaks, a local model you got running well — if it's useful, it's welcome.

📝

Docs & Design

Clearer install steps, typo fixes, HUD themes, accessibility tweaks — small contributions count too.

Contributors

👋 Waiting
Be the first

No contributors yet

  • This spot — and full credit — is waiting for whoever ships the first fix, feature, or fully-tested macOS run, on any platform.
07Version History

What's new
in NOVA?

Latest updates and improvements to the system.

"Sometimes you gotta run before you can walk." — Tony Stark
v0.5 — Latest
2026-06-30

🚀 Major Update: Local Model Support

  • Replaced LM Studio with llama.cpp — More efficient, portable local model support
  • Hardware Detection — Automatically scans system capabilities and recommends optimal models
  • Model Management UI — Download, delete, and switch models directly from settings
  • Hugging Face Integration — Browse and download models with one click
  • Smart Recommendations — Get personalized model suggestions based on your RAM and GPU
  • Portable Installation — Everything stays in the project folder, fully portable
  • Low-End PC Support — Lightweight models available for systems with 8GB RAM
  • Offline Capability — Run NOVA completely offline with local models
v0.4
Previous

🔧 System Improvements

  • Added Google Gemini API support with massive context window
  • Implemented automatic update system
  • Added voice activity detection for wake word
  • Improved error handling and fail-safe mechanisms