Merge pull request #3829 from BerriAI/litellm_send_alerts_making_new_key

[Feat] - send Email alerts when making new key

litellm/integrations/slack_alerting.py

@@ -14,6 +14,7 @@ from pydantic import BaseModel
 from enum import Enum
 from datetime import datetime as dt, timedelta, timezone
 from litellm.integrations.custom_logger import CustomLogger
+from litellm.proxy._types import WebhookEvent
 import random
 from typing import TypedDict
 from openai import APIError
@@ -30,6 +31,13 @@ class OutageModel(TypedDict):
     major_alert_sent: bool
     last_updated_at: float
 
+# we use this for the email header, please send a test email if you change this. verify it looks good on email
+LITELLM_LOGO_URL = "https://litellm-listing.s3.amazonaws.com/litellm_logo.png"
+EMAIL_LOGO_URL = os.getenv(
+    "SMTP_SENDER_LOGO", "https://litellm-listing.s3.amazonaws.com/litellm_logo.png"
+)
+EMAIL_SUPPORT_CONTACT = os.getenv("EMAIL_SUPPORT_CONTACT", "support@berri.ai")
+
 
 class LiteLLMBase(BaseModel):
     """
@@ -57,12 +65,6 @@ class SlackAlertingArgs(LiteLLMBase):
     max_outage_alert_list_size: int = 10  # prevent memory leak
 
 
-class WebhookEvent(CallInfo):
-    event: Literal["budget_crossed", "threshold_crossed", "projected_limit_exceeded"]
-    event_group: Literal["user", "key", "team", "proxy"]
-    event_message: str  # human-readable description of event
-
-
 class DeploymentMetrics(LiteLLMBase):
     """
     Metrics per deployment, stored in cache
@@ -664,6 +666,9 @@ class SlackAlerting(CustomLogger):
             _id = user_info.token
 
         # percent of max_budget left to spend
+        if user_info.max_budget is None:
+            return
+
         if user_info.max_budget > 0:
             percent_left = (
                 user_info.max_budget - user_info.spend
@@ -946,6 +951,103 @@ Model Info:
 
         return False
 
+    async def send_key_created_email(self, webhook_event: WebhookEvent) -> bool:
+        from litellm.proxy.utils import send_email
+
+        if self.alerting is None or "email" not in self.alerting:
+            # do nothing if user does not want email alerts
+            return False
+
+        # make sure this is a premium user
+        from litellm.proxy.proxy_server import premium_user
+        from litellm.proxy.proxy_server import CommonProxyErrors, prisma_client
+
+        if premium_user != True:
+            raise Exception(
+                f"Trying to use Email Alerting on key creation\n {CommonProxyErrors.not_premium_user.value}"
+            )
+
+        event_name = webhook_event.event_message
+        recipient_email = webhook_event.user_email
+        recipient_user_id = webhook_event.user_id
+        if (
+            recipient_email is None
+            and recipient_user_id is not None
+            and prisma_client is not None
+        ):
+            user_row = await prisma_client.db.litellm_usertable.find_unique(
+                where={"user_id": recipient_user_id}
+            )
+
+            if user_row is not None:
+                recipient_email = user_row.user_email
+
+        key_name = webhook_event.key_alias
+        key_token = webhook_event.token
+        key_budget = webhook_event.max_budget
+
+        email_html_content = "Alert from LiteLLM Server"
+        if recipient_email is None:
+            verbose_proxy_logger.error(
+                "Trying to send email alert to no recipient", extra=webhook_event.dict()
+            )
+        email_html_content = f"""
+            <img src="{EMAIL_LOGO_URL}" alt="LiteLLM Logo" width="150" height="50" />
+
+            <p> Hi {recipient_email}, <br/>
+ 
+            I'm happy to provide you with an OpenAI Proxy API Key, loaded with ${key_budget} per month. <br /> <br />
+
+            <b>
+            Key: <pre>{key_token}</pre> <br>
+            </b>
+
+            <h2>Usage Example</h2>
+
+            Detailed Documentation on <a href="https://docs.litellm.ai/docs/proxy/user_keys">Usage with OpenAI Python SDK, Langchain, LlamaIndex, Curl</a>
+
+            <pre>
+
+            import openai
+            client = openai.OpenAI(
+                api_key="{key_token}",
+                base_url={os.getenv("PROXY_BASE_URL", "http://0.0.0.0:4000")}
+            )
+
+            response = client.chat.completions.create(
+                model="gpt-3.5-turbo", # model to send to the proxy
+                messages = [
+                    {{
+                        "role": "user",
+                        "content": "this is a test request, write a short poem"
+                    }}
+                ]
+            )
+
+            </pre>
+
+
+            If you have any questions, please send an email to {EMAIL_SUPPORT_CONTACT} <br /> <br />
+
+            Best, <br />
+            The LiteLLM team <br />
+            """
+
+        payload = webhook_event.model_dump_json()
+        email_event = {
+            "to": recipient_email,
+            "subject": f"LiteLLM: {event_name}",
+            "html": email_html_content,
+        }
+
+        response = await send_email(
+            receiver_email=email_event["to"],
+            subject=email_event["subject"],
+            html=email_event["html"],
+        )
+
+        return False
+
     async def send_email_alert_using_smtp(self, webhook_event: WebhookEvent) -> bool:
         """
         Sends structured Email alert to an SMTP server
@@ -968,15 +1070,15 @@ Model Info:
 
         if webhook_event.event == "budget_crossed":
             email_html_content = f"""
-            <h1>LiteLLM</h1>
+            <img src="{EMAIL_LOGO_URL}" alt="LiteLLM Logo" width="150" height="50" />
 
             <p> Hi {user_name}, <br/>
 
-            Your LLM API usage this month has reached your account's monthly budget of ${max_budget} <br /> <br />
+            Your LLM API usage this month has reached your account's <b> monthly budget of ${max_budget} </b> <br /> <br />
 
             API requests will be rejected until either (a) you increase your monthly budget or (b) your monthly usage resets at the beginning of the next calendar month. <br /> <br />
 
-            If you have any questions, please send an email to support@berri.ai <br /> <br />
+            If you have any questions, please send an email to {EMAIL_SUPPORT_CONTACT} <br /> <br />
 
             Best, <br />
             The LiteLLM team <br />

litellm/proxy/_types.py

@@ -1096,7 +1096,7 @@ class CallInfo(LiteLLMBase):
     """Used for slack budget alerting"""
 
     spend: float
-    max_budget: float
+    max_budget: Optional[float] = None
     token: str = Field(description="Hashed value of that key")
     user_id: Optional[str] = None
     team_id: Optional[str] = None
@@ -1104,3 +1104,11 @@ class CallInfo(LiteLLMBase):
     key_alias: Optional[str] = None
     projected_exceeded_date: Optional[str] = None
     projected_spend: Optional[float] = None
+
+
+class WebhookEvent(CallInfo):
+    event: Literal[
+        "budget_crossed", "threshold_crossed", "projected_limit_exceeded", "key_created"
+    ]
+    event_group: Literal["user", "key", "team", "proxy"]
+    event_message: str  # human-readable description of event

litellm/proxy/proxy_server.py

@@ -5402,6 +5402,25 @@ async def generate_key_fn(
         response["soft_budget"] = (
             data.soft_budget
         )  # include the user-input soft budget in the response
+        event = WebhookEvent(
+            event="key_created",
+            event_group="key",
+            event_message=f"API Key Created",
+            token=response.get("token", ""),
+            spend=response.get("spend", 0.0),
+            max_budget=response.get("max_budget", 0.0),
+            user_id=response.get("user_id", None),
+            team_id=response.get("team_id", "Default Team"),
+            key_alias=response.get("key_alias", None),
+        )
+
+        # If user configured email alerting - send an Email letting their end-user know the key was created
+        asyncio.create_task(
+            proxy_logging_obj.slack_alerting_instance.send_key_created_email(
+                webhook_event=event,
+            )
+        )
+
         return GenerateKeyResponse(**response)
     except Exception as e:
         traceback.print_exc()

litellm/proxy/utils.py

@@ -12,6 +12,7 @@ from litellm.proxy._types import (
     LiteLLM_TeamTable,
     Member,
     CallInfo,
+    WebhookEvent,
     AlertType,
 )
 from litellm.caching import DualCache, RedisCache