[Metrics] Add group-aware KV cache capacity to vllm:cache_config_info (#42206)

The startup log already reports the correct group-aware KV cache capacity for
hybrid models, but Prometheus did not expose matching info in 'vllm:cache_config_info`.

This PR adds kv_cache_size_tokens and kv_cache_max_concurrency.

Signed-off-by: Ethan Feng <ethan.fengch@gmail.com>
This commit is contained in:
Ethan Feng
2026-06-12 11:49:44 +00:00
committed by GitHub
parent 8af550b399
commit b7f9b6ab27
7 changed files with 75 additions and 26 deletions
+3
View File
@@ -78,6 +78,9 @@ class EngineCoreReadyResponse:
dp_stats_address: str | None
dtype: str
vllm_version: str
# KV cache capacity (None for encoder-only/attention-free models).
kv_cache_size_tokens: int | None = None
kv_cache_max_concurrency: float | None = None
class EngineCoreRequest(