ai-workspace-services/litellm
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix(prometheus.py): fix spend metrics

Browse Source
This commit is contained in:
Krrish Dholakia 2025-09-18 19:12:07 -07:00
parent aa7839e4cb
commit 4d87199266
2 changed files with 24 additions and 25 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

41
enterprise/litellm_enterprise/integrations/prometheus.py
View File

@ -102,7 +102,9 @@ class PrometheusLogger(CustomLogger):
# "team",
# "team_alias",
# ],
labelnames=self.get_labels_for_metric("litellm_llm_api_time_to_first_token_metric"),
labelnames=self.get_labels_for_metric(
"litellm_llm_api_time_to_first_token_metric"
),
buckets=LATENCY_BUCKETS,
)
@ -240,14 +242,14 @@ class PrometheusLogger(CustomLogger):
self.litellm_deployment_state = self._gauge_factory(
"litellm_deployment_state",
"LLM Deployment Analytics - The state of the deployment: 0 = healthy, 1 = partial outage, 2 = complete outage",
labelnames=self.get_labels_for_metric("litellm_deployment_state")
labelnames=self.get_labels_for_metric("litellm_deployment_state"),
)
self.litellm_deployment_cooled_down = self._counter_factory(
"litellm_deployment_cooled_down",
"LLM Deployment Analytics - Number of times a deployment has been cooled down by LiteLLM load balancing logic. exception_status is the status of the exception that caused the deployment to be cooled down",
# labelnames=_logged_llm_labels + [EXCEPTION_STATUS],
labelnames=self.get_labels_for_metric("litellm_deployment_cooled_down")
labelnames=self.get_labels_for_metric("litellm_deployment_cooled_down"),
)
self.litellm_deployment_success_responses = self._counter_factory(
@ -1039,20 +1041,12 @@ class PrometheusLogger(CustomLogger):
_labels = prometheus_label_factory(
supported_enum_labels=self.get_labels_for_metric(
metric_name="litellm_proxy_total_requests_metric"
metric_name="litellm_spend_metric"
),
enum_values=enum_values,
)
self.litellm_spend_metric.labels(
end_user_id,
user_api_key,
user_api_key_alias,
model,
user_api_team,
user_api_team_alias,
user_id,
).inc(response_cost)
self.litellm_spend_metric.labels(**_labels).inc(response_cost)
def _set_virtual_key_rate_limit_metrics(
self,
@ -2280,7 +2274,9 @@ def get_custom_labels_from_metadata(metadata: dict) -> Dict[str, str]:
return result
def _tag_matches_wildcard_configured_pattern(tags: List[str], configured_tag: str) -> bool:
def _tag_matches_wildcard_configured_pattern(
tags: List[str], configured_tag: str
) -> bool:
"""
Check if any of the request tags matches a wildcard configured pattern
@ -2305,6 +2301,7 @@ def _tag_matches_wildcard_configured_pattern(tags: List[str], configured_tag: st
import re
from litellm.router_utils.pattern_match_deployments import PatternMatchRouter
pattern_router = PatternMatchRouter()
regex_pattern = pattern_router._pattern_to_regex(configured_tag)
return any(re.match(pattern=regex_pattern, string=tag) for tag in tags)
@ -2313,11 +2310,11 @@ def _tag_matches_wildcard_configured_pattern(tags: List[str], configured_tag: st
def get_custom_labels_from_tags(tags: List[str]) -> Dict[str, str]:
"""
Get custom labels from tags based on admin configuration.
Supports both exact matches and wildcard patterns:
- Exact match: "prod" matches "prod" exactly
- Wildcard pattern: "User-Agent: curl/*" matches "User-Agent: curl/7.68.0"
- Wildcard pattern: "User-Agent: curl/*" matches "User-Agent: curl/7.68.0"
Reuses PatternMatchRouter for wildcard pattern matching.
Returns dict of label_name: "true" if the tag matches the configured tag, "false" otherwise
@ -2345,17 +2342,19 @@ def get_custom_labels_from_tags(tags: List[str]) -> Dict[str, str]:
for configured_tag in configured_tags:
label_name = _sanitize_prometheus_label_name(f"tag_{configured_tag}")
# Check for exact match first (backwards compatibility)
if configured_tag in tags:
result[label_name] = "true"
continue
# Use PatternMatchRouter for wildcard pattern matching
if "*" in configured_tag and _tag_matches_wildcard_configured_pattern(tags=tags, configured_tag=configured_tag):
if "*" in configured_tag and _tag_matches_wildcard_configured_pattern(
tags=tags, configured_tag=configured_tag
):
result[label_name] = "true"
continue
# No match found
result[label_name] = "false"

8
tests/otel_tests/test_prometheus.py
View File

@ -616,10 +616,10 @@ async def test_user_email_in_all_required_metrics():
# Check that user_email appears in all the required metrics
required_metrics_with_user_email = [
"litellm_proxy_total_requests_metric_total",
"litellm_input_tokens_metric_total",
"litellm_output_tokens_metric_total",
"litellm_requests_metric_total",
# "litellm_proxy_total_requests_metric_total",
# "litellm_input_tokens_metric_total",
# "litellm_output_tokens_metric_total",
# "litellm_requests_metric_total",
"litellm_spend_metric_total",
]