Sandboxing

Note

This page is minimalistic because we expect that users want to be able to run this extension in a sandboxed environment, but we didn’t initially take this into account.

We plan on expanding options to allow local Neuro API server compatibility with remote sandboxed environments, without requiring it to be exposed to the internet. You can track its progress on issue #32 on the NeuroPilot repository.

NeuroPilot is very powerful, and inherently it is very destructive, so running it in an isolated environment is recommended for safety. This page will go over how to run NeuroPilot in a sandboxed environment.

Overview

The reason why NeuroPilot can be sandboxed easily, compared to most other Neuro integrations, is due to the fact that it extends on VS Code, which provides various versions you can connect to and use as your development environment, in addition to your desktop client.

Sandboxing via SSH or Tunnels on a local VM

We’ll be using Oracle VirtualBox for this example. We’ll assume that you have set up a virtual machine and figured out a way to connect to it; whether that be SSH or Tunnels. You can find guides online if you don’t know how.

In order to not have to expose the Neuro API server to the internet, we’ll use a Host-Only Adapter network, which essentially creates a local system.

1. Open your VM’s settings and navigate to the Network section.

Tip

If you want to use multiple network adapters (e.g. one for connecting to the internet, and one for the Host-Only Adapter), you’ll need to switch to Expert mode.

If you’re comfortable with not having internet access in the VM, you can stay in Basic mode.

2. Change to another adapter (if using Expert mode), and enable the adapter (if it’s not already enabled). Then, click on the dropdown next to Attached to: and select Host-Only Adapter.

3. After selecting that, make sure you note down the name of the adapter.

4. On your host machine, open your console and run ipconfig /all (Windows) or ip a (macOS/Linux). Then, take note of what your assigned IP is for the adapter that has the description containing the name you noted down earlier.

5. Take note of the IP listed there. If you’re on Windows and ran ipconfig /all, note down the IPv4 address that has (Preferred) next to it.

6. Start your Neuro API server on that address. If you’re using one of the testing tools on the official API repo’s README, you should be able to pass a CLI flag to spawn it on a certain host.

7. Now, connect to your virtual machine and install NeuroPilot.

8. Open your settings, and go down to the NeuroPilot > Websocket Url.

9. Add your IP address and host that you spawned your Neuro API server on, in the scheme of ws://(host):(port).

10. Now, attempt to reconnect. This should now connect to your Neuro API server as expected.