Monitoring Metrics (Prometheus-style)

Day 77 of 100 Days of MLOps. Evidently gives you batch drift reports; production also needs live, continuous metrics scraped every few seconds. Today you expose your model service's metrics in the standard Prometheus format from a /metrics endpoint — a counter for predictions served, a histogram for latency, a gauge for the last predicted price — using prometheus_client. You'll hit the service and read the real exposition-format output, and see how your ML service plugs into the same monitoring stack as the rest of your infrastructure. Runs 100% locally on any OS.