Start with one DGX Spark
Team AI Box is an open-source team workspace from Weerocks LLC, purpose-built for NVIDIA DGX Spark. It brings AI tools, team files, messaging, and virtual devices into one browser-based workspace. Rather than configuring a separate AI environment for every colleague, a team can access a shared installation without deploying an additional server.
The product is built around one-box deployment and local processing, not around sending every task to an external AI API. Teams can choose whether to connect external model services. That choice makes it important to distinguish work processed locally from services that you deliberately enable outside the box.
Bring everyday work together
A useful workflow rarely ends with a single question. A meeting produces a recording, a proposal includes images and documents, and the results need to reach colleagues. Team AI Box brings related capabilities together:
- Think and create: discuss ideas and organize content with the AI assistant, or explore visual directions with image generation.
- Work with information: recognize text in images, transcribe speech, and translate text or files.
- Collaborate: use messaging for conversations and the team drive to store and share files.
- Present and use devices: cast from Apple devices for colleagues to watch in their browsers, and access virtual Android phones and Windows or Linux computers.
These tools support practical workflows; they do not promise universal compatibility with every application or file. Casting excludes DRM-protected content, and third-party applications and services retain their own requirements. Evaluate important business applications against the actual software and documents your team uses.
Keep virtual devices under your control
Virtual phones provide independent spaces for applications and data, accessed through a browser. Virtual computers provide a virtual display that can show the boot process and operating-system installation screens.
Windows or Linux does not have to be embedded in the product image. Upload an ARM64 ISO you are entitled to use, or select one from the team drive, mount it in the virtual optical drive, and install through the virtual display. Users provide the necessary operating-system and application licenses. Team AI Box does not include a Windows license.
Prepare for a fresh installation
Use a fresh DGX Spark with Ubuntu 24.04 ARM64, working Docker/Compose, the NVIDIA runtime and KVM. Install Git to clone the source. Reserve several hundred GB of temporary disk space and an internet connection for model and build downloads. Run the following commands on the Spark itself, not on your laptop.
First, distinguish a clean installation target from a machine carrying existing business data. A fresh-install script is not automatically an upgrade tool. Read the included instructions before changing existing services or data.
Clone, then run one installation command
git clone https://github.com/weerocksllc/teamaibox.git
cd teamaibox
sudo bash install/from-source.sh
The script downloads the required public inputs, builds the application containers, prepares models and installation dependencies, assembles a complete package, and installs it. This is an online source build followed by offline installation. Large downloads, compilation and model preparation take time; cloning the repository alone does not produce a ready system.
If interrupted, keep the build workspace and rerun the same command with the same source version. Verified downloads and completed stages can be reused. Do not delete the cache to retry, disable integrity checks, or copy another appliance's account databases or virtual-machine disks. See the public-source installation guide for progress files, troubleshooting, download-source options and changing source versions safely. Third-party mirrors are optional and their availability is not guaranteed.
First login and disconnected installations
When installation completes, open the HTTPS address reported by the installer on your local network. The factory administrator is admin with password admin; change this password before sharing access and create individual team accounts. Use the Spark's current LAN address, not its Docker or VPN address. Review any locally generated certificate warning before trusting the appliance.
For a disconnected target, run sudo bash install/from-source.sh --build-only on a compatible connected Spark. Copy the resulting complete package and checksum to the target, verify and extract the package, read its instructions, then run sudo ./install.sh inside it. A model download cache alone is not a complete offline installer.
Offline installation does not mean every later activity is offline. Local features do not require an external model API, but third-party services, online content, and some applications may need an internet connection when you choose to use them.
Understand what is open source
Original Weerocks LLC code and documentation are MIT-licensed. You may use, modify, and distribute those materials while preserving the required license and copyright notices.
Third-party code, model weights, firmware, operating systems, and components from providers such as Google and NVIDIA keep their own terms. Distribution of the complete offline package requires separate consideration; the project's MIT license does not automatically permit redistribution of everything in that package.
The source is public on GitHub: weerocksllc/teamaibox, with original code and documentation under MIT. The source repository does not include container images, model weights or offline installation packages. Contributions, translations and reproducible bug reports are welcome; read the contribution guide before opening a pull request. Contact Weerocks LLC at teamaibox@weerocks.llc for deployment questions; third-party components and models retain their own distribution terms. Visit the Team AI Box website for the product overview and deployment entry point. DGX Spark is an NVIDIA product; Team AI Box is not an official NVIDIA project.