From f35f3a8a33a067f08d13206c43f2ed3e5ee72261 Mon Sep 17 00:00:00 2001
From: medisean <6093631+medisean@users.noreply.github.com>
Date: Fri, 14 Aug 2026 12:44:47 +0800
Subject: [PATCH] fix: normalize Prometheus latency datetimes

---
 litellm/integrations/prometheus.py            |  8 ++++++--
 ...est_prometheus_overhead_with_guardrails.py | 20 +++++++++++++++++++
 2 files changed, 26 insertions(+), 2 deletions(-)

diff --git a/litellm/integrations/prometheus.py b/litellm/integrations/prometheus.py
index a9056aaf4e1..e433c24be3c 100644
--- a/litellm/integrations/prometheus.py
+++ b/litellm/integrations/prometheus.py
@@ -8,7 +8,7 @@ import math
 import os
 import sys
 from collections.abc import Awaitable, Callable, Mapping, Sequence
-from datetime import datetime, timedelta
+from datetime import datetime, timedelta, timezone
 from typing import TYPE_CHECKING, Any, Final, Literal, cast
 
 import litellm
@@ -3990,7 +3990,11 @@ class PrometheusLogger(CustomLogger):
         otherwise returns None.
         """
         if isinstance(start_time, datetime) and isinstance(end_time, datetime):
-            return (end_time - start_time).total_seconds()
+            normalized_start: Final = (
+                start_time if start_time.tzinfo is not None else start_time.replace(tzinfo=timezone.utc)
+            )
+            normalized_end: Final = end_time if end_time.tzinfo is not None else end_time.replace(tzinfo=timezone.utc)
+            return (normalized_end - normalized_start).total_seconds()
         return None
 
     @staticmethod
diff --git a/tests/test_litellm/integrations/test_prometheus_overhead_with_guardrails.py b/tests/test_litellm/integrations/test_prometheus_overhead_with_guardrails.py
index 56c9702189e..e7af8f7b517 100644
--- a/tests/test_litellm/integrations/test_prometheus_overhead_with_guardrails.py
+++ b/tests/test_litellm/integrations/test_prometheus_overhead_with_guardrails.py
@@ -7,6 +7,7 @@ durations. During-call (moderation) guardrails run concurrently with the LLM
 call and are excluded so they don't inflate the overhead.
 """
 
+from datetime import datetime, timedelta, timezone
 from unittest.mock import MagicMock
 
 import pytest
@@ -160,6 +161,25 @@ def test_get_guardrail_overhead_seconds_counts_single_pre_call_dict():
     assert abs(PrometheusLogger._get_guardrail_overhead_seconds(payload) - 0.3) < 1e-6
 
 
+@pytest.mark.parametrize(
+    ("start_time", "end_time"),
+    (
+        (
+            datetime(2026, 8, 14, 12, 0, 0, tzinfo=timezone.utc).replace(tzinfo=None),
+            datetime(2026, 8, 14, 12, 0, 1, tzinfo=timezone.utc),
+        ),
+        (
+            datetime(2026, 8, 14, 14, 0, 0, tzinfo=timezone(timedelta(hours=2))),
+            datetime(2026, 8, 14, 12, 0, 1, tzinfo=timezone.utc).replace(tzinfo=None),
+        ),
+    ),
+)
+def test_safe_duration_seconds_normalizes_mixed_timezone_awareness(start_time, end_time):
+    logger = PrometheusLogger()
+
+    assert logger._safe_duration_seconds(start_time, end_time) == 1.0
+
+
 def _patch_label_factory(monkeypatch):
     monkeypatch.setattr(
         "litellm.integrations.prometheus.prometheus_label_factory",