Additional Fixes (09/17/2024) (#5759)

* fix(auth_checks.py): check if key has all model access via wildcard routing

Fixes issue where key with `openai/*` couldn't call gpt models

* fix(slack_alerting.py): expose flag for disabling failed spend tracking alerts

litellm/integrations/SlackAlerting/slack_alerting.py

@@ -69,6 +69,7 @@ class SlackAlerting(CustomBatchLogger):
             "cooldown_deployment",
             "new_model_added",
             "outage_alerts",
+            "failed_tracking_spend",
         ],
         alert_to_webhook_url: Optional[
             Dict[AlertType, Union[List[str], str]]
@@ -598,6 +599,12 @@ class SlackAlerting(CustomBatchLogger):
 
     async def failed_tracking_alert(self, error_message: str):
         """Raise alert when tracking failed for specific model"""
+        if self.alerting is None or self.alert_types is None:
+            # do nothing if alerting is not switched on
+            return
+        if "failed_tracking_spend" not in self.alert_types:
+            return
+
         _cache: DualCache = self.internal_usage_cache
         message = "Failed Tracking Cost for " + error_message
         _cache_key = "budget_alerts:failed_tracking:{}".format(message)

litellm/proxy/_new_secret_config.yaml

@@ -19,16 +19,4 @@ model_list:
   - model_name: o1-preview
     litellm_params:
       model: o1-preview
-  
+ 
-guardrails:
-  - guardrail_name: "hide-secrets"
-    litellm_params:
-      guardrail: "hide-secrets"  # supported values: "aporia", "lakera"
-      mode: "pre_call"
-      # detect_secrets_config: {
-      #    "plugins_used": [
-      #     {"name": "SoftlayerDetector"},
-      #     {"name": "StripeDetector"},
-      #     {"name": "NpmDetector"}
-      #   ]
-      # }

litellm/proxy/_types.py

@@ -123,6 +123,7 @@ AlertType = Literal[
     "outage_alerts",
     "region_outage_alerts",
     "fallback_reports",
+    "failed_tracking_spend",
 ]
 
 

litellm/proxy/auth/auth_checks.py

@@ -607,11 +607,17 @@ async def can_key_call_model(
 
     filtered_models += models_in_current_access_groups
     verbose_proxy_logger.debug(f"model: {model}; allowed_models: {filtered_models}")
+
+    all_model_access: bool = False
+
     if (
-        model is not None
+        len(filtered_models) == 0
-        and model not in filtered_models
+        or "*" in filtered_models
-        and "*" not in filtered_models
+        or "openai/*" in filtered_models
     ):
+        all_model_access = True
+
+    if model is not None and model not in filtered_models and all_model_access is False:
         raise ValueError(
             f"API Key not allowed to access model. This token can only access models={valid_token.models}. Tried to access {model}"
         )