The cluster, right now

Live numbers from the BharatCode serving VM: vLLM throughput, queue depth, model list, and GPU telemetry. Current counters come directly from the serving VM, while the graph uses a short rolling average so shared capacity is easier to read.