This project sets up a Docker container equipped with NVIDIA CUDA, Rclone for cloud storage synchronization, and Tailscale for secure networking. It's designed to create a GPU-accelerated remote workspace with access to files stored on SharePoint via Rclone and secure network access through Tailscale. This setup is mainly intended to facilitate remote work scenarios, such as data science and machine learning, where access to GPU resources and cloud storage is required. The primary focus would be on vast.ai, but later in the development, we can expand to other cloud providers.

• NVIDIA CUDA Base: Utilizes an official NVIDIA CUDA image for GPU-accelerated applications.
• Rclone Integration: Syncs files with cloud storage services, configured for SharePoint in this example.
• Tailscale VPN: Secures network access to the container, enabling remote work scenarios.
• Startup Script: Automates the setup process, including Tailscale authentication and Rclone mounts.

.
├── .env                # Environment variables, including Tailscale auth key
├── .gitignore          # Specifies intentionally untracked files to ignore
├── .vscode             # VS Code settings (excluded from this setup)
│   └── settings.json
├── Dockerfile          # Docker configuration for building the image
├── build_run.sh        # Script to build and run the Docker container
├── rclone.conf         # Rclone configuration (sensitive, not included in the repo)
└── startup.sh          # Startup script executed within the container

• Docker installed on your machine.
• A Tailscale account and an auth key.
• Rclone configured for your cloud storage provider (SharePoint in this example).

1. Clone the repository

   git clone https://github.com/yourusername/remote-workspace.git
   cd remote-workspace
   

2. Rclone Configuration

   Place your rclone.conf file in the root directory. This file contains the configuration for your cloud storage provider, such as SharePoint. It's recommended to exclude this file from the repository by adding it to your .gitignore file.

   If you haven't configured Rclone yet, you can do so by running the following command:

   rclone config
   

   This will guide you through the process of setting up Rclone for your cloud storage provider. Once you've configured Rclone, use the command below to generate a base64-encoded string of your rclone.conf file:

   cat rclone.conf | base64
   

   and then add the output to your .env file as the value for RCLONE_CONFIG_CONTENT.

   We use the base64-encoded string to avoid exposing the rclone.conf file in the repository andd allows us to use it as an environment variable in the Docker container.

3. Environment Variables

   Create a .env file in the root directory with the following content, replacing YOUR_TAILSCALE_AUTHKEY with your actual Tailscale auth key:

   TAILSCALE_AUTHKEY=YOUR_TAILSCALE_AUTHKEY
   RCLONE_CONFIG_CONTENT=<base64-encoded-rclone.conf>
   

4. Build and Run the Docker Container

   Execute the build_run.sh script to build the Docker image and run the container:

   ./build_run.sh
   

   This script will stop and remove any existing container from a previous build, then build a new image and start a container with the necessary environment variables and privileges.

The Dockerfile is configured to use nvidia/cuda:12.3.1-base-ubuntu22.04 as the base image. It's crucial that the CUDA version in this Dockerfile (12.3.1 in this case) matches the CUDA version installed on your host machine to avoid compatibility issues during the container runtime.

Once the container is running, it will execute the startup.sh script, which starts Tailscale and mounts the configured Rclone remotes. You can then access your container over the Tailscale network and interact with your cloud storage through the mounted directories.

• Rclone Configuration: Ensure your rclone.conf file is kept secure and not included in the repository. It's recommended to add it to your .gitignore file.
• Tailscale Auth Key: Treat your Tailscale auth key as sensitive information. Only include it in your .env file, which should also be excluded from the repository.

Contributions are welcome! Please feel free to submit a pull request or open an issue for any improvements or bug fixes.

This project is open-source and available under the MIT License.