Agent Skills
› NeverSight/learn-skills.dev
› distributed-inference-orchestrator
distributed-inference-orchestrator
GitHub管理AI推理任务在多个全球GPU提供商间的分发，通过实时监控价格与延迟实现智能路由、故障转移及负载均衡，优化大规模并发请求处理及冷启动时间。
Trigger Scenarios
需要跨云供应商分配AI计算任务
监控或优化GPU推理延迟与成本
处理高并发AI请求的负载均衡
Install
npx skills add NeverSight/learn-skills.dev --skill distributed-inference-orchestrator -g -y
SKILL.md
Frontmatter
{
"name": "distributed-inference-orchestrator",
"description": "Manage AI processing across multiple global GPU providers."
}
Distributed Inference Orchestrator
This skill manages the 'Brain Latency' of the system by routing AI queries to the closest or cheapest GPU provider (RunPod, Lambda, AWS) in real-time.
Instructions
- Monitor live prices and queue times for 5+ GPU cloud providers.
- Implement 'Failover' logic: if Provider A is slow, route to Provider B.
- Use 'Load Balancing' to handle bursts of 10,000+ simultaneous AI requests.
- Optimize cold-start times for custom model weights.
Examples
- "Route all non-urgent 8K video renders to the cheapest available GPU node globally."
Version History
- e0220ca Current 2026-07-05 23:32
