Skip to content

Manage Terminus

This page introduces Terminus management tasks in Terminus Space, such as monitoring system data, adding worker nodes, and handling cloud services.

View Panel Data

To view system data through Terminus Space:

  1. In your Terminus, navigate to Settings > Integration, bind your Terminus Name with your Terminus Space account. This action will authorize Terminus Space to access your system data.

  2. Log into Terminus Space.

  3. You can find the system panel on the My Terminus page, where you get insight into storage usage and traffic consumption.

    alt text

View Integration for more information.

INFO

If you're using Self-Hosted Terminus, you need to pay attention to the traffic data for intranet penetration and the storage usage by backup services. This is because if you are using these services, you may be charged based on usage.

Add Worker Nodes

If you are using Terminus in the cloud, you can add Worker Nodes for better performance.

  1. Click the More Options (...) button in the upper right corner, and select Add Worker.
  2. In the following guide page, choose the hardware configuration option as needed.
  3. Review fees for storage and traffic.
  4. Confirm the order and submit.

Return Terminus

If you no longer need your Terminus services in the cloud, you can return Terminus. To return your Terminus:

  1. Click the More Options (...) button in the upper right corner, select Destroy Terminus.
  2. Confirm the action and settle your current usage:
    • If a refund is applicable, the corresponding amount will be quickly returned to your account balance.
    • If an additional payment is required, please confirm and settle the payment.

Shared GPU

We currently do not provide instances that include GPUs. If you have such a need, please get in touch with us.

We offer individual users a shared GPU solution based on rCuda. This solution is effective for applications like Stable Diffusion and costs approximately $0.02 per image. However, it still needs further enhancements for Large Language Models (LLMs).