Use NVIDIA Sync with AI Workbench#
Overview#
- NVIDIA Sync is an application for managing SSH connections to remote Linux systems.**
NVIDIA Sync runs on Windows, macOS, and Ubuntu to streamline SSH connections, launch IDEs and applications, and configure Tailscale access. It also has an AI Workbench integration that handles installing Workbench on a remote and adding it to the Desktop App as a remote location.
NVIDIA Sync creates and manages a single private/public key pair to provide access to all of the Sync-managed devices. For full NVIDIA Sync documentation, see here.
- Use NVIDIA Sync to create a Workbench location without manually installing and adding.
Once NVIDIA Sync is connected to an Ubuntu remote, you can launch AI Workbench on the remote through the Sync UI.
This automatically checks for AI Workbench on the remote, and if not yet installed, NVIDIA Sync will install it, add it as a remote location, and then launch that location.
This is similar to how the NVIDIA Brev integration works to bootstrap the AI Workbench install.
- Use NVIDIA Sync when you have password-based access to a remote device.
AI Workbench doesn’t support password-based SSH access to a remote machine. It requires a configured private/public key pair.
However, you can add a device to NVIDIA Sync with password-based SSH access, and Sync will configure the managed private/public key pair for the remote. This allows AI Workbench to use that private/public key pair.
- The AI Workbench integration with NVIDIA Sync is still a work in progress.
The integration requires coordinating and managing a variety of SSH configuration files and Workbench location configurations. Overall, the integration works well, but there are some idiosyncrasies to be aware of.
For example, adding a remote location to Workbench via Sync will name the location sync-<ip> address instead of the device name in Sync.
Key Concepts#
- NVIDIA Sync:
An application for Windows, macOS, and Ubuntu that manages SSH connections and launches applications and containers on remote Linux systems.
- Sync Managed Device
A device that you have added to NVIDIA Sync on your laptop.
- Sync SSH Configuration Type
Designation in the Workbench contexts.json configuration file that indicates the SSH config is handled by Sync.
Prerequisites#
- Verify the following before using NVIDIA Sync with AI Workbench.
NVIDIA Sync installed on your local machine (see the installation instructions)
AI Workbench Desktop App installed on your local machine
A remote Ubuntu system accessible over the network
SSH access to the remote system as a non-root user with sudo privileges
Container Runtime#
- NVIDIA Sync defaults to Docker as the container runtime on the remote system.
If you want to use Podman instead, you need to switch it manually after the Sync-managed install completes. SSH into the remote system and reconfigure the container runtime.
This is a limitation of the Sync integration — the manual remote install lets you choose the runtime during installation.
Using Tailscale with AI Workbench#
- NVIDIA Sync supports Tailscale for secure remote access from outside the local network.
Tailscale creates a peer-to-peer mesh connection so you can access a remote system without being on the same network. AI Workbench dynamically detects Tailscale enrollment when the location is created by Sync.
- See Use Tailscale with AI Workbench for the current AI Workbench guidance on using Tailscale with Sync-managed locations.
The page explains the user-facing workflow and when manual recovery steps are appropriate for older locations.
SSH Configuration on Windows#
- NVIDIA Sync, AI Workbench, and Brev each use different SSH configuration files on Windows.
This can cause connection issues when these tools interact because they may not see each other’s SSH entries.
NVIDIA Sync operates on the Windows side and uses the main Windows SSH config file (
C:\Users\<username>\.ssh\config).AI Workbench can use SSH configs on either the Windows side or the WSL side, depending on how the location was added.
Brev uses its own SSH config in
~/.brevon the WSL side, which is also imported to the Windows side.
- If a remote location was added through Sync, AI Workbench may not find the SSH configuration.
This happens when Sync writes to the Windows SSH config but AI Workbench looks for the entry on the WSL side.
Be aware of which SSH config file each tool is using.
When troubleshooting connection issues, check both the Windows SSH config (C:\Users\<username>\.ssh\config) and the WSL SSH config (~/.ssh/config inside WSL).
If you are also using Brev, check ~/.brev on the WSL side as well.