Merge pull request #3112 from BerriAI/litellm_add_alert_types

[Feat] Allow user to select slack alert types to Opt In to

litellm/proxy/_types.py

@@ -697,6 +697,21 @@ class ConfigGeneralSettings(LiteLLMBase):
         None,
         description="List of alerting integrations. Today, just slack - `alerting: ['slack']`",
     )
+    alert_types: Optional[
+        List[
+            Literal[
+                "llm_exceptions",
+                "llm_too_slow",
+                "llm_requests_hanging",
+                "budget_alerts",
+                "db_exceptions",
+            ]
+        ]
+    ] = Field(
+        None,
+        description="List of alerting types. By default it is all alerts",
+    )
+
     alerting_threshold: Optional[int] = Field(
         None,
         description="sends alerts if requests hang for 5min+",

litellm/proxy/proxy_server.py

@@ -2283,6 +2283,7 @@ class ProxyConfig:
             proxy_logging_obj.update_values(
                 alerting=general_settings.get("alerting", None),
                 alerting_threshold=general_settings.get("alerting_threshold", 600),
+                alert_types=general_settings.get("alert_types", None),
                 redis_cache=redis_usage_cache,
             )
             ### CONNECT TO DATABASE ###
@@ -2619,6 +2620,9 @@ class ProxyConfig:
             if "alerting" in _general_settings:
                 general_settings["alerting"] = _general_settings["alerting"]
                 proxy_logging_obj.alerting = general_settings["alerting"]
+            if "alert_types" in _general_settings:
+                general_settings["alert_types"] = _general_settings["alert_types"]
+                proxy_logging_obj.alert_types = general_settings["alert_types"]
 
             # router settings
             _router_settings = config_data.get("router_settings", {})
@@ -8196,10 +8200,12 @@ async def update_config(config_info: ConfigYAML):
             updated_general_settings = config_info.general_settings.dict(
                 exclude_none=True
             )
-            config["general_settings"] = {
+
-                **updated_general_settings,
+            _existing_settings = config["general_settings"]
-                **config["general_settings"],
+            for k, v in updated_general_settings.items():
-            }
+                # overwrite existing settings with updated values
+                _existing_settings[k] = v
+            config["general_settings"] = _existing_settings
 
         if config_info.environment_variables is not None:
             config.setdefault("environment_variables", {})
@@ -8372,7 +8378,15 @@ async def get_config():
                     )
                     _slack_env_vars[_var] = _decrypted_value
 
-            _data_to_return.append({"name": "slack", "variables": _slack_env_vars})
+            _alerting_types = proxy_logging_obj.alert_types
+
+            _data_to_return.append(
+                {
+                    "name": "slack",
+                    "variables": _slack_env_vars,
+                    "alerting_types": _alerting_types,
+                }
+            )
 
         _router_settings = llm_router.get_settings()
         return {

litellm/proxy/utils.py

@@ -64,12 +64,38 @@ class ProxyLogging:
         self.cache_control_check = _PROXY_CacheControlCheck()
         self.alerting: Optional[List] = None
         self.alerting_threshold: float = 300  # default to 5 min. threshold
+        self.alert_types: List[
+            Literal[
+                "llm_exceptions",
+                "llm_too_slow",
+                "llm_requests_hanging",
+                "budget_alerts",
+                "db_exceptions",
+            ]
+        ] = [
+            "llm_exceptions",
+            "llm_too_slow",
+            "llm_requests_hanging",
+            "budget_alerts",
+            "db_exceptions",
+        ]
 
     def update_values(
         self,
         alerting: Optional[List],
         alerting_threshold: Optional[float],
         redis_cache: Optional[RedisCache],
+        alert_types: Optional[
+            List[
+                Literal[
+                    "llm_exceptions",
+                    "llm_too_slow",
+                    "llm_requests_hanging",
+                    "budget_alerts",
+                    "db_exceptions",
+                ]
+            ]
+        ] = None,
     ):
         self.alerting = alerting
         if alerting_threshold is not None:
@@ -78,6 +104,9 @@ class ProxyLogging:
         if redis_cache is not None:
             self.internal_usage_cache.redis_cache = redis_cache
 
+        if alert_types is not None:
+            self.alert_types = alert_types
+
     def _init_litellm_callbacks(self):
         print_verbose(f"INITIALIZING LITELLM CALLBACKS!")
         litellm.callbacks.append(self.max_parallel_request_limiter)
@@ -213,6 +242,8 @@ class ProxyLogging:
     ):
         if self.alerting is None:
             return
+        if "llm_too_slow" not in self.alert_types:
+            return
         time_difference_float, model, api_base, messages = (
             self._response_taking_too_long_callback(
                 kwargs=kwargs,
@@ -259,6 +290,8 @@ class ProxyLogging:
 
         if type == "hanging_request":
             # Simulate a long-running operation that could take more than 5 minutes
+            if "llm_requests_hanging" not in self.alert_types:
+                return
             await asyncio.sleep(
                 self.alerting_threshold
             )  # Set it to 5 minutes - i'd imagine this might be different for streaming, non-streaming, non-completion (embedding + img) requests
@@ -307,6 +340,8 @@ class ProxyLogging:
         if self.alerting is None:
             # do nothing if alerting is not switched on
             return
+        if "budget_alerts" not in self.alert_types:
+            return
         _id: str = "default_id"  # used for caching
         if type == "user_and_proxy_budget":
             user_info = dict(user_info)
@@ -465,6 +500,8 @@ class ProxyLogging:
         Currently only logs exceptions to sentry
         """
         ### ALERTING ###
+        if "db_exceptions" not in self.alert_types:
+            return
         if isinstance(original_exception, HTTPException):
             if isinstance(original_exception.detail, str):
                 error_message = original_exception.detail
@@ -499,6 +536,8 @@ class ProxyLogging:
         """
 
         ### ALERTING ###
+        if "llm_exceptions" not in self.alert_types:
+            return
         asyncio.create_task(
             self.alerting_handler(
                 message=f"LLM API call failed: {str(original_exception)}", level="High"

ui/litellm-dashboard/src/components/settings.tsx

@@ -36,6 +36,7 @@ const Settings: React.FC<SettingsPageProps> = ({
   const [isModalVisible, setIsModalVisible] = useState(false);
   const [form] = Form.useForm();
   const [selectedCallback, setSelectedCallback] = useState<string | null>(null);
+  const [selectedAlertValues, setSelectedAlertValues] = useState([]);
 
   useEffect(() => {
     if (!accessToken || !userRole || !userID) {
@@ -59,6 +60,12 @@ const Settings: React.FC<SettingsPageProps> = ({
     setSelectedCallback(null);
   };
 
+  const handleChange = (values) => {
+    setSelectedAlertValues(values);
+    // Here, you can perform any additional logic with the selected values
+    console.log('Selected values:', values);
+  };
+
   const handleSaveChanges = (callback: any) => {
     if (!accessToken) {
       return;
@@ -68,8 +75,14 @@ const Settings: React.FC<SettingsPageProps> = ({
       Object.entries(callback.variables).map(([key, value]) => [key, (document.querySelector(`input[name="${key}"]`) as HTMLInputElement)?.value || value])
     );
 
+    console.log("updatedVariables", updatedVariables);
+    console.log("updateAlertTypes", selectedAlertValues);
+
     const payload = {
       environment_variables: updatedVariables,
+      general_settings: {
+        alert_types: selectedAlertValues
+      }
     };
 
     try {
@@ -169,6 +182,25 @@ const Settings: React.FC<SettingsPageProps> = ({
   </li>
 ))}
         </ul>
+        {callback.alerting_types && (
+          <div>
+            <Text className="mt-2">Alerting Types</Text>
+            <Select
+              mode="multiple"
+              style={{ width: '100%' }}
+              placeholder="Select Alerting Types"
+              optionLabelProp="label"
+              onChange={handleChange}
+              defaultValue={callback.alerting_types}
+            >
+              {callback.alerting_types.map((type: string) => (
+                <Select.Option key={type} value={type} label={type}>
+                  {type}
+                </Select.Option>
+              ))}
+            </Select>
+          </div>
+        )}
         <Button className="mt-2" onClick={() => handleSaveChanges(callback)}>
           Save Changes
         </Button>