https://github.com/jc1122/hetzner-delegation

Agent-facing skill for delegating workloads through the ray-hetzner Ray cluster on Hetzner Cloud.
https://github.com/jc1122/hetzner-delegation

ai-agents claude-code hetzner leaf-skill metaopt ray ray-hetzner remote-execution skill

Last synced: 3 days ago
JSON representation

Agent-facing skill for delegating workloads through the ray-hetzner Ray cluster on Hetzner Cloud.

• Host: GitHub
• URL: https://github.com/jc1122/hetzner-delegation
• Owner: jc1122
• Created: 2026-04-03T15:07:28.000Z (2 months ago)
• Default Branch: master
• Last Pushed: 2026-04-10T21:15:37.000Z (about 2 months ago)
• Last Synced: 2026-04-10T22:06:50.253Z (about 2 months ago)
• Topics: ai-agents, claude-code, hetzner, leaf-skill, metaopt, ray, ray-hetzner, remote-execution, skill
• Size: 37.1 KB
• Stars: 1
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# hetzner-delegation

Mirror repository for the agent-facing `hetzner-delegation` skill.

This repo keeps the skill contract in version control. The actual execution backend lives in [`ray-hetzner`](https://github.com/jc1122/ray-hetzner).

Current model:

- Aorus is the permanent Ray head and queue control plane

- Hetzner nodes are autoscaled disposable worker capacity

- queue-backed `metaopt` execution is the primary path

- direct one-off Ray jobs are the secondary path

The old manual head and worker lifecycle has been removed from the backend. The supported surface is the Aorus bootstrap path, queue thin-client path, direct Ray Jobs helper, autoscaler configuration, and supporting queue/runtime code.

The skill definition is in [SKILL.md](SKILL.md). For the actual runtime behavior, start from the [`ray-hetzner` README](https://github.com/jc1122/ray-hetzner/blob/master/README.md) and `docs/OPERATIONS.md` in that repository.

Design specs live under `docs/superpowers/specs/`.