Inference serving

vLLM Inference Serving Pack

vLLM capability pack for high-throughput LLM inference serving, API deployment checks, memory budgets, and rollback boundaries.