Merge pull request #1702 from major/rspeed-3017-fix-response-duration-buckets

RSPEED-3017: use custom buckets for response duration histogram

Author: tisnik

src/metrics/__init__.py

@@ -32,7 +32,10 @@
 # Histogram to measure response durations
 # This will be used to track how long it takes to handle requests
 response_duration_seconds = Histogram(
-    "ls_response_duration_seconds", "Response durations", ["path"]
+    "ls_response_duration_seconds",
+    "Response durations",
+    ["path"],
+    buckets=LLM_INFERENCE_DURATION_BUCKETS,
 )
 
 # Metric that indicates what provider + model customers are using so we can