(Fixes) OpenAI Streaming Token Counting + Fixes usage track when litellm.turn_off_message_logging=True (#8156)

* working streaming usage tracking

* fix test_async_chat_openai_stream_options

* fix await asyncio.sleep(1)

* test_async_chat_azure

* fix s3 logging

* fix get_stream_options

* fix get_stream_options

* fix streaming handler

* test_stream_token_counting_with_redaction

* fix codeql concern

litellm/llms/openai/openai.py

@@ -14,6 +14,7 @@ from typing import (
     Union,
     cast,
 )
+from urllib.parse import urlparse
 
 import httpx
 import openai
@@ -833,8 +834,9 @@ class OpenAIChatCompletion(BaseLLM):
         stream_options: Optional[dict] = None,
     ):
         data["stream"] = True
-        if stream_options is not None:
-            data["stream_options"] = stream_options
+        data.update(
+            self.get_stream_options(stream_options=stream_options, api_base=api_base)
+        )
 
         openai_client: OpenAI = self._get_openai_client(  # type: ignore
             is_async=False,
@@ -893,8 +895,9 @@ class OpenAIChatCompletion(BaseLLM):
     ):
         response = None
         data["stream"] = True
-        if stream_options is not None:
-            data["stream_options"] = stream_options
+        data.update(
+            self.get_stream_options(stream_options=stream_options, api_base=api_base)
+        )
         for _ in range(2):
             try:
                 openai_aclient: AsyncOpenAI = self._get_openai_client(  # type: ignore
@@ -977,6 +980,20 @@ class OpenAIChatCompletion(BaseLLM):
                             status_code=500, message=f"{str(e)}", headers=error_headers
                         )
 
+    def get_stream_options(
+        self, stream_options: Optional[dict], api_base: Optional[str]
+    ) -> dict:
+        """
+        Pass `stream_options` to the data dict for OpenAI requests
+        """
+        if stream_options is not None:
+            return {"stream_options": stream_options}
+        else:
+            # by default litellm will include usage for openai endpoints
+            if api_base is None or urlparse(api_base).hostname == "api.openai.com":
+                return {"stream_options": {"include_usage": True}}
+        return {}
+
     # Embedding
     @track_llm_api_timing()
     async def make_openai_embedding_request(