xiangongyun-api

What This Skill Does

The xiangongyun-api skill serves as a comprehensive interface for the 仙宫云 (Xiangongyun) GPU cloud service platform. It enables users to perform full-lifecycle management of high-performance computing instances directly through OpenClaw. Key capabilities include deploying new GPU instances, monitoring instance status, managing private machine images, and overseeing account finances such as balance checks and recharge operations.

Installation

You can install this skill by running the following command in your terminal or via the OpenClaw management console: clawhub install openclaw/skills/skills/chsengni/xiangongyun-api Ensure that you have the requests library (version 2.28.0 or higher) installed in your Python environment. No additional complex configuration is required beyond the standard OpenClaw skill authorization process.

Use Cases

• Deep Learning Training: Quickly deploy multi-GPU environments with pre-configured images like PyTorch or TensorFlow for model training.
• Cost Optimization: Efficiently manage GPU resources by using the 'shutdown-release-gpu' action to minimize costs during idle periods while preserving your system disk and data.
• Environment Portability: Capture custom configurations or trained models by saving active instances as private images for future rapid deployment.
• Budget Management: Monitor account credits and handle payments programmatically to ensure uninterrupted access to cloud compute resources.

Example Prompts

1. "Check my current Xiangongyun account balance and tell me if I have enough credits to run a 2-GPU instance for 5 hours."
2. "List all my existing GPU instances and show me the status of the one named 'ai-training-node'."
3. "Save my instance 'abc123' as a new private image named 'experiment-v1' and then shut it down to save costs."

Tips & Limitations

• Async Awareness: Most operations (like deployment or destruction) are asynchronous. The API will acknowledge the request immediately, but the actual state change may take time. Always verify current status via list_instances or get_instance.
• Cost Control: Note that shutting down an instance while releasing the GPU still incurs storage fees for the system disk. For maximum savings, destroy instances you no longer need.
• Documentation: For advanced configurations regarding data centers or SSH keys, consult the references/api_reference.md file located in the skill directory.