serving-llms-vllm
Serves LLMs with high throughput using vLLM's PagedAttention and continuous batching. Use when deploying production LLM APIs, optimizing inference latency/throughput, or serving models with limited GPU memory. Supports OpenAI-compatible endpoints, quantization (GPTQ/AWQ/FP8), and tensor parallelism.
- Score
- Unranked
- Evidence
- No signals yet
- Last commit
- Not known
- Listed
Install
armory install serving-llms-vllm --cli claudewrites the skill into.claude/skills/serving-llms-vllm/SKILL.mdListed as compatible
# fetches the source and writes it to:
.claude/skills/serving-llms-vllm/SKILL.mdNeeds the armory CLI · not on npm yet, build it from cli/ in the repository
What it is
Serves LLMs with high throughput using vLLM's PagedAttention and continuous batching. Use when deploying production LLM APIs, optimizing inference latency/throughput, or serving models with limited GPU memory. Supports OpenAI-compatible endpoints, quantization (GPTQ/AWQ/FP8), and tensor parallelism.
When to use it
Serves LLMs with high throughput using vLLM's PagedAttention and continuous batching. Use when deploying production LLM APIs, optimizing inference latency/throughput, or serving models with limited GPU memory. Supports OpenAI-compatible endpoints, quantization (GPTQ/AWQ/FP8), and tensor parallelism.
How to install / invoke
# Copy the skill into your .claude/skills/ directory
curl -sL https://raw.githubusercontent.com/davila7/claude-code-templates/main/cli-tool/components/skills/ai-research/inference-serving-vllm/SKILL.md -o .claude/skills/inference-serving-vllm/SKILL.md
Notes
Extracted from davila7/claude-code-templates, inference-serving-vllm category.