Deploying on Linux

On Linux, AI Studio comes as a Flatpak, which we recommend, or as an AppImage. In both cases, policy files tell AI Studio where its configuration is:

flowchart LR
    MGMT["Your configuration management<br/>e.g. Ansible, Puppet, Salt"] --> POL["Policy files<br/>Flatpak: provisioning extension<br/>AppImage: /etc/xdg/mindwork-ai-studio"]
    POL --> APP["AI Studio<br/>on each device"]
    APP -- "GET {server}/{id}.zip" --> SRV["Configuration server<br/>in your intranet"]

Loading

Read How Central Configuration Works first if you haven't yet.

Flatpak

Installing

Our Flatpak bundles are on the release page, for Intel/AMD (MindWork.AI.Studio_x86_64.flatpak) and ARM (MindWork.AI.Studio_aarch64.flatpak), together with an extension for Pandoc. To install them for all users of a device:

sudo flatpak install --system ./MindWork.AI.Studio_x86_64.flatpak
sudo flatpak install --system ./MindWork.AI.Studio.Plugin.Pandoc_x86_64.flatpak

The app needs the GNOME runtime from Flathub. Set up the Flathub remote on the device, so that Flatpak can install the runtime along with the app.

Updates: AI Studio never updates a Flatpak itself. AI Studio isn't on Flathub yet, so you deliver each new version yourself, either by installing the new bundle with --reinstall or through a Flatpak repository of your own.

Providing the policy files

Inside its sandbox, AI Studio doesn't see the configuration folders of the host. Instead, it reads its policy files from /app/etc/MindWorkAI/ first. That folder belongs to the Flatpak extension org.mindworkai.AIStudio.provisioning. You can fill it without building an extension, by placing the files on the host in the folder of the extension:

For	Folder on the host
All users of the device	/var/lib/flatpak/extension/org.mindworkai.AIStudio.provisioning/<arch>/<branch>/
One user	~/.local/share/flatpak/extension/org.mindworkai.AIStudio.provisioning/<arch>/<branch>/

<arch> is x86_64 or aarch64. <branch> is the branch of the installed app, which is stable for our bundles. flatpak info org.mindworkai.AIStudio shows both.

Flatpak mounts the content of this folder into the sandbox at /app/etc/MindWorkAI/. Place there:

• config_00000.yaml, config_00001.yaml, …: one file per configuration;
• config_encryption_secret.yaml, if you encrypt API keys;
• external_http_custom_root_certificates.yaml and your PEM bundle, if your servers use certificates of your own certificate authority. The sandbox can't read the certificates of the host, so this is usually needed. Point bundle_path to the path inside the sandbox, for example, /app/etc/MindWorkAI/company-root-cas.pem.

The format of these files is described in Policy files and Your own certificate authorities.

After /app/etc/MindWorkAI/, AI Studio also reads the folders from $XDG_CONFIG_DIRS inside the sandbox, as described for the AppImage below.

Environment variables

To set an environment variable for the Flatpak, use an override:

sudo flatpak override --system --env=MINDWORK_AI_STUDIO_ENTERPRISE_CONFIG_ID_00000=9072b77d-ca81-40da-be6a-861da525ef7b org.mindworkai.AIStudio

Policy files are easier to maintain, and AI Studio reads them first.

Good to know

• API keys need the Secret Service. AI Studio keeps API keys, including the encrypted keys of your configuration, in a password manager that offers the FreeDesktop Secret Service, such as GNOME Keyring or KeePassXC. It needs an unlocked default collection.
• Files that your configuration names must be readable inside the sandbox. This concerns, for example, the PowerPoint template of the Slide Assistant, its picture library, or certificate bundles. AI Studio sees the home folder of the user only read-only. The safest place for files of your organization is the provisioning folder, /app/etc/MindWorkAI/.

AppImage

Installing

Our AppImages are on the release page, for Intel/AMD (MindWork.AI.Studio_amd64.AppImage) and ARM (MindWork.AI.Studio_aarch64.AppImage). Place the AppImage in a folder that users can't write to, such as /opt/mindwork-ai-studio/, and make it executable. AI Studio then recognizes the installation as managed by you and never updates it itself.

The marker file managed-installation, which works on Windows, can't be used with an AppImage: AI Studio would look for it inside the image.

Providing the policy files

AI Studio reads its policy files from the subfolder mindwork-ai-studio of every folder in $XDG_CONFIG_DIRS, in the order given there. Only when $XDG_CONFIG_DIRS is empty or not set does it read /etc/xdg/mindwork-ai-studio/.

So for most systems, place the files in:

/etc/xdg/mindwork-ai-studio/

Check $XDG_CONFIG_DIRS in the desktop sessions of your users: when it is set but doesn't contain /etc/xdg, AI Studio doesn't read /etc/xdg/mindwork-ai-studio/. Then place the files below one of the listed folders.

The files and their format are the same as for the Flatpak, see Policy files. After the policy files, AI Studio reads environment variables.

Where AI Studio keeps its data

What	AppImage	Flatpak
Data, plugins, and your configurations	~/.local/share/org.mindworkai.AIStudio/data/	~/.var/app/org.mindworkai.AIStudio/data/org.mindworkai.AIStudio/data/
Settings	~/.config/org.mindworkai.AIStudio/	~/.var/app/org.mindworkai.AIStudio/config/org.mindworkai.AIStudio/
API keys	The Secret Service	The Secret Service

For the AppImage, $XDG_DATA_HOME and $XDG_CONFIG_HOME replace ~/.local/share and ~/.config when they are set. The "Information" page shows the folders as AI Studio resolved them on the device at hand, the path of the log file, and the state of your configurations.

Where to go next

• Configuration Server
• Deploying on Windows
• Deploying on macOS