vllm:kv_cache_usage_perc
Gauge for KV-cache usage where 1 means 100 percent usage.
MetricPrometheusvLLM v0.21.0
Semantic coverage
gap · Availability & namegap · Structural identitygap · Measurement semanticsgap · Causal/runtime semanticsgap · Interpretation & compositiongap · Guarded guidance