How to Install Newelle

Newelle is an open-source AI assistant built for the GNOME desktop. It gives you a chat interface for talking to LLMs (both local models and cloud providers like OpenAI or Anthropic), with extensions for running commands, browsing the web, and reading files, right from a native Linux app. This guide covers installing it on a regular Linux desktop and on a VPS running a desktop environment.

Before You Start

Newelle is a graphical (GTK4/Adwaita) application, so it needs a display to run on. That means:

• A Linux desktop (any distro with Flatpak support), or
• A VPS with a desktop environment and a remote desktop protocol (VNC or RDP) installed, since Newelle isn't a headless/CLI tool.

If you're installing on a fresh VPS, provision it with a lightweight desktop like XFCE first, then connect over VNC before installing Newelle. Ubuntu or Debian images work well for this.

Installing Newelle via Flatpak

Flatpak is the official and recommended distribution method for Newelle, and it works identically across distros.

1. Install Flatpak if it isn't already present:

sudo apt update
sudo apt install flatpak

2. Add the Flathub repository, which hosts Newelle:

flatpak remote-add --if-not-exists flathub https://flathub.org/repo/flathub.flatpakrepo

3. Install Newelle:

flatpak install flathub io.github.qwersyk.Newelle

4. Launch it:

flatpak run io.github.qwersyk.Newelle

On a desktop environment, Newelle will also appear in your applications menu after installation, no need to launch it from the terminal every time.

First-Time Setup

When Newelle opens for the first time, it walks you through choosing a model provider:

1. Pick a provider. You can connect an API key for a cloud model (OpenAI, Anthropic, Google, and others are supported), or point Newelle at a local model server if you're running one (for example, Ollama or a llama.cpp server on the same machine).
2. Enter your credentials. For cloud providers, paste the API key into the settings panel. For local models, enter the endpoint URL.
3. Test the connection. Send a message from the chat window; a response confirms everything is wired up correctly.
4. Enable extensions (optional). Newelle supports extensions for running shell commands, reading local files, and web search. Enable only what you need, since command execution extensions have real access to your system.

Running Newelle on a VPS

Since Newelle needs a graphical session, running it on a headless VPS takes an extra step:

1. Deploy a VPS and install a lightweight desktop environment (XFCE is a good balance of speed and compatibility):

sudo apt install xfce4 xfce4-goodies -y

2. Install and configure a VNC server, such as TightVNC or TigerVNC, so you can connect to the graphical session remotely.
3. Connect over VNC, then follow the Flatpak installation steps above inside that session.

On CloudBlast, a small VPS plan is enough to run Newelle comfortably as long as you're connecting to cloud model providers rather than running large local models on the same box. For local inference, size RAM to the model you plan to load.

Frequently Asked Questions

Does Newelle require an internet connection?

Only if you're using a cloud model provider. If you connect it to a local model server on the same machine or network, Newelle can work fully offline.

Can I run Newelle without a desktop environment?

No. Newelle is a GTK application and needs a display server (X11 or Wayland) to run, either locally or through a remote desktop session like VNC.

Is Newelle free?

Yes, Newelle is open source and free to install. Any cost comes from the model provider you connect it to, cloud API usage is billed by that provider, while local models are free to run beyond your own hardware.

Get Started

If you want a graphical Linux environment to run Newelle on, without touching your own machine, deploy a VPS on CloudBlast with a desktop image, connect over VNC, and install it in a few minutes. See plans and pricing to pick the right size for the models you plan to use.