vllm-ascend-pd-serving ยท diff
git:20260910.55adacd to git:20260911.9f92063
1 added, 1 removed. Audit A to A.
---
name: "vllm-ascend-pd-serving"
description: "Plan, start, inspect, smoke-test, and stop a vLLM Ascend prefill/decode deployment as one coordinator topology execution. Use for PD disaggregation with NIXL, Mooncake, or another KV connector. Do not use for one colocated service, generic Ray clusters, correctness matrices, performance regression decisions, or distributed root-cause diagnosis."
---
<!-- Generated from .agents/skills/vllm-ascend-pd-serving/SKILL.md. Do not edit. -->
- # vLLM Ascend PD Serving
+ # vllm-ascend-pd-serving
Read `.agents/skills/vllm-ascend-pd-serving/SKILL.md` and only the references needed
for the current task. The canonical skill owns the workflow.