fix(litellm_logging.py): fix calling success callback w/ stream_options true

Fixes https://github.com/BerriAI/litellm/issues/5118

litellm/integrations/openmeter.py

@@ -1,12 +1,18 @@
 # What is this?
 ## On Success events log cost to OpenMeter - https://github.com/BerriAI/litellm/issues/1268
 
-import dotenv, os, json
-import litellm
+import json
+import os
 import traceback
+import uuid
+
+import dotenv
+import httpx
+
+import litellm
+from litellm import verbose_logger
 from litellm.integrations.custom_logger import CustomLogger
 from litellm.llms.custom_httpx.http_handler import AsyncHTTPHandler, HTTPHandler
-import uuid
 
 
 def get_utc_datetime():
@@ -122,7 +128,11 @@ class OpenMeterLogger(CustomLogger):
             )
 
             response.raise_for_status()
-        except Exception as e:
-            if hasattr(response, "text"):
-                litellm.print_verbose(f"\nError Message: {response.text}")
+        except httpx.HTTPStatusError as e:
+            verbose_logger.error(
+                "Failed OpenMeter logging - {}".format(e.response.text)
+            )
+            raise e
+        except Exception as e:
+            verbose_logger.error("Failed OpenMeter logging - {}".format(str(e)))
             raise e

litellm/litellm_core_utils/litellm_logging.py

@@ -638,6 +638,8 @@ class Logging:
             verbose_logger.debug(f"success callbacks: {litellm.success_callback}")
             ## BUILD COMPLETE STREAMED RESPONSE
             complete_streaming_response = None
+            if "complete_streaming_response" in self.model_call_details:
+                return  # break out of this.
             if self.stream and isinstance(result, ModelResponse):
                 if (
                     result.choices[0].finish_reason is not None
@@ -1279,6 +1281,8 @@ class Logging:
         )
         ## BUILD COMPLETE STREAMED RESPONSE
         complete_streaming_response = None
+        if "async_complete_streaming_response" in self.model_call_details:
+            return  # break out of this.
         if self.stream:
             if result.choices[0].finish_reason is not None:  # if it's the last chunk
                 self.streaming_chunks.append(result)
@@ -1302,6 +1306,7 @@ class Logging:
                 self.streaming_chunks.append(result)
         if complete_streaming_response is not None:
             print_verbose("Async success callbacks: Got a complete streaming response")
+
             self.model_call_details["async_complete_streaming_response"] = (
                 complete_streaming_response
             )
@@ -1431,7 +1436,7 @@ class Logging:
                             end_time=end_time,
                         )
                 if isinstance(callback, CustomLogger):  # custom logger class
-                    if self.stream == True:
+                    if self.stream is True:
                         if (
                             "async_complete_streaming_response"
                             in self.model_call_details

litellm/proxy/_new_secret_config.yaml

@@ -1,24 +1,7 @@
-general_settings:
-  store_model_in_db: true
-  database_connection_pool_limit: 20
-
 model_list:
-  - model_name: fake-openai-endpoint
+  - model_name: "*"
     litellm_params:
-      model: openai/my-fake-model
-      api_key: my-fake-key
-      api_base: https://exampleopenaiendpoint-production.up.railway.app/
-litellm_settings:
-  drop_params: True
-  success_callback: ["prometheus"]
-  failure_callback: ["prometheus"]
-  service_callback: ["prometheus_system"]
-  _langfuse_default_tags: ["user_api_key_alias", "user_api_key_user_id", "user_api_key_user_email", "user_api_key_team_alias", "semantic-similarity", "proxy_base_url"]
+      model: "*"
 
-router_settings:
-  routing_strategy: "latency-based-routing"
-  routing_strategy_args: {"ttl": 86400} # Average the last 10 calls to compute avg latency per model
-  allowed_fails: 1
-  num_retries: 3
-  retry_after: 5 # seconds to wait before retrying a failed request
-  cooldown_time: 30 # seconds to cooldown a deployment after failure
+litellm_settings:
+  callbacks: ["openmeter"]

litellm/tests/test_custom_callback_input.py

@@ -14,6 +14,7 @@ from pydantic import BaseModel
 
 sys.path.insert(0, os.path.abspath("../.."))
 from typing import List, Literal, Optional, Union
+from unittest.mock import AsyncMock, MagicMock, patch
 
 import litellm
 from litellm import Cache, completion, embedding
@@ -518,6 +519,29 @@ async def test_async_chat_azure_stream():
 # asyncio.run(test_async_chat_azure_stream())
 
 
+@pytest.mark.asyncio
+async def test_async_chat_openai_stream_options():
+    try:
+        customHandler = CompletionCustomHandler()
+        litellm.callbacks = [customHandler]
+        with patch.object(
+            customHandler, "async_log_success_event", new=AsyncMock()
+        ) as mock_client:
+            response = await litellm.acompletion(
+                model="gpt-3.5-turbo",
+                messages=[{"role": "user", "content": "Hi 👋 - i'm async openai"}],
+                stream=True,
+                stream_options={"include_usage": True},
+            )
+
+            async for chunk in response:
+                continue
+
+            mock_client.assert_awaited_once()
+    except Exception as e:
+        pytest.fail(f"An exception occurred: {str(e)}")
+
+
 ## Test Bedrock + sync
 def test_chat_bedrock_stream():
     try: