fix(azure.py): raise streaming exceptions

litellm/llms/azure.py

@@ -398,6 +398,7 @@ class AzureChatCompletion(BaseLLM):
         azure_ad_token: Optional[str] = None,
         client=None,
     ):
+        try:
             # init AzureOpenAI Client
             azure_client_params = {
                 "api_version": api_version,
@@ -435,6 +436,11 @@ class AzureChatCompletion(BaseLLM):
                 logging_obj=logging_obj,
             )
             return streamwrapper  ## DO NOT make this into an async for ... loop, it will yield an async generator, which won't raise errors if the response fails
+        except Exception as e:
+            if hasattr(e, "status_code"):
+                raise AzureOpenAIError(status_code=e.status_code, message=str(e))
+            else:
+                raise AzureOpenAIError(status_code=500, message=str(e))
 
     async def aembedding(
         self,

litellm/tests/test_custom_callback_input.py

@@ -290,6 +290,7 @@ class CompletionCustomHandler(
                     kwargs["original_response"], (str, litellm.CustomStreamWrapper)
                 )
                 or inspect.isasyncgen(kwargs["original_response"])
+                or inspect.iscoroutine(kwargs["original_response"])
                 or kwargs["original_response"] == None
             )
             assert isinstance(kwargs["additional_args"], (dict, type(None)))
@@ -439,7 +440,7 @@ async def test_async_chat_azure_stream():
         )
         async for chunk in response:
             continue
-        ## test failure callback
+        # test failure callback
         try:
             response = await litellm.acompletion(
                 model="azure/chatgpt-v-2",
@@ -459,7 +460,7 @@ async def test_async_chat_azure_stream():
         pytest.fail(f"An exception occurred: {str(e)}")
 
 
-# asyncio.run(test_async_chat_azure_stream())
+asyncio.run(test_async_chat_azure_stream())
 
 
 ## Test Bedrock + sync