docker

Jules Endpoint Agent: Docker Installation

This directory contains the necessary files to run the Jules Endpoint Agent in a Docker container.

This method is highly recommended for users who want a standardized, isolated, and high-performance Linux environment for the agent, regardless of their host operating system (Windows, macOS, or Linux).

File Descriptions

• Dockerfile: This file defines the Docker image. It builds a standard Ubuntu environment, installs all necessary dependencies (git, cloudflared, shell2http), and sets up the agent's runner script and entrypoint.
• docker-compose.yml: This is the easiest way to run the agent. It defines the service, manages environment variables for credentials, and includes a pre-configured section for enabling GPU acceleration.
• runner.sh: A copy of the Linux runner script, included here to be copied into the Docker image during the build process.

Design Choices & Technical Details

Why Docker?

• Consistency: The agent runs in the exact same Ubuntu-based environment every time, no matter what the host OS is. This eliminates "it works on my machine" problems.
• Isolation & Security: The agent and any commands it runs are sandboxed inside the container. This provides a strong layer of security, preventing accidental or malicious changes to the host system.
• Performance & Hardware Acceleration: Unlike traditional VMs, Docker can provide near-native performance. Crucially, it allows for direct access to the host's GPU (on Windows and Linux), which is essential for machine learning and other hardware-accelerated tasks.

Acknowledging Limitations

• Native Testing: This method runs a Linux container. It cannot be used to test projects that require a native Windows or macOS environment. For that, please use the native installers.
• GPU on macOS: GPU acceleration for Docker containers on macOS is not well-supported by Docker Desktop at this time. The GPU instructions below apply primarily to Windows (with WSL2) and Linux hosts.

Installation Instructions

Prerequisites

• Docker Desktop (for Windows/macOS) or Docker Engine (for Linux) must be installed.
• (Optional, for GPU) For NVIDIA GPUs, you must have the latest drivers installed. On Linux, you must also install the NVIDIA Container Toolkit. On Windows, GPU support must be enabled in Docker Desktop's settings.

Running the Agent

1. Clone the Repository: You will need these files locally. Clone the repository and navigate into this docker directory.
2. Configure Credentials: Open the docker-compose.yml file in a text editor. You must fill in the following environment variables:
   • CLOUDFLARE_TOKEN: Get this from the Cloudflare Zero Trust dashboard by creating a new tunnel and copying the token string.
   • JULES_USERNAME: The username for the agent.
   • JULES_PASSWORD: The password for the agent.
3. Build and Run: From your terminal in this docker directory, run the following command:

   docker-compose up --build -d

   This will build the Docker image, create a container, and run it in the background.
4. Get Your URL: The public URL will be visible in the Cloudflare Zero Trust dashboard for the tunnel you created.

Enabling GPU Access (Optional)

If you need the agent to have access to your NVIDIA GPU:

1. Ensure you have met the prerequisites mentioned above.
2. Open the docker-compose.yml file.
3. Uncomment the deploy section at the bottom of the file.
4. Restart the container: docker-compose up --build -d.

The agent running inside the container will now have access to the GPU.