omh-inference-serving

[omh] OMH Inference Serving workflow: choose the serving engine and quantization from decision tables, prepare deployment as an idempotent runbook with observed-only verification, and measure the endpoint with the standard TTFT/TPOT/goodput protocol. Use when the user says: inference-serving, inference serving, serve this model, serve the model, model serving, serving endpoint, vllm, llama.cpp.