fix streaming error

litellm/tests/test_streaming.py

@@ -121,13 +121,14 @@ def test_openai_chat_completion_call():
             print(chunk)
             if "content" in chunk["choices"][0]["delta"]:
                 complete_response += chunk["choices"][0]["delta"]["content"]
-        if complete_response == "": 
+            print(f'complete_chunk: {complete_response}')
+        if complete_response.strip() == "": 
             raise Exception("Empty response received")
     except:
         print(f"error occurred: {traceback.format_exc()}")
         pass
 
-# test_openai_chat_completion_call()
+test_openai_chat_completion_call()
 async def completion_call():
     try:
         response = completion(

litellm/utils.py

@@ -2451,7 +2451,11 @@ class CustomStreamWrapper:
 
     def __next__(self):
         try:
+<<<<<<< HEAD
             completion_obj = {"content": "", "role": ""} # default to role being assistant
+=======
+            completion_obj = {"content": ""} # default to role being assistant
+>>>>>>> 31d771b (fix streaming error)
             if self.model in litellm.anthropic_models:
                 chunk = next(self.completion_stream)
                 completion_obj["content"] = self.handle_anthropic_chunk(chunk)
@@ -2496,15 +2500,12 @@ class CustomStreamWrapper:
             else: # openai chat/azure models
                 chunk = next(self.completion_stream)
                 completion_obj["content"] = chunk["choices"][0]["delta"]["content"]
-                completion_obj["role"] = chunk["choices"][0]["delta"]["role"]
-                # return chunk # open ai returns finish_reason, we should just return the openai chunk
 
-                #completion_obj["content"] = self.handle_openai_chat_completion_chunk(chunk)
             # LOGGING
             threading.Thread(target=self.logging_obj.success_handler, args=(completion_obj,)).start()
             # return this for all models
             model_response = ModelResponse(stream=True)
-            model_response.choices[0].delta = completion_obj
+            model_response.choices[0].delta.content = completion_obj["content"]
             return model_response
         except StopIteration:
             raise StopIteration

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "litellm"
-version = "0.1.644"
+version = "0.1.645"
 description = "Library to easily interface with LLM API providers"
 authors = ["BerriAI"]
 license = "MIT License"