​

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.

​

Install

npx research-agent-skills install serving-llms-vllm
npx skills add KalarisLabs/research-agent-skills --skill serving-llms-vllm

​

When to use it

​

Full playbook