bump pyproject version

litellm/tests/test_streaming.py

@@ -214,6 +214,31 @@ def test_completion_cohere_stream():
     except Exception as e:
         pytest.fail(f"Error occurred: {e}")
 
+def test_completion_claude_stream():
+    try:
+        messages = [
+            {"role": "system", "content": "You are a helpful assistant."},
+            {
+                "role": "user",
+                "content": "how does a court case get to the Supreme Court?",
+            },
+        ]
+        response = completion(
+            model="claude-instant-1", messages=messages, stream=True, max_tokens=50
+        )
+        complete_response = ""
+        # Add any assertions here to check the response
+        for idx, chunk in enumerate(response):
+            chunk, finished = streaming_format_tests(idx, chunk)
+            if finished:
+                break
+            complete_response += chunk
+        if complete_response.strip() == "": 
+            raise Exception("Empty response received")
+        print(f"completion_response: {complete_response}")
+    except Exception as e:
+        pytest.fail(f"Error occurred: {e}")
+# test_completion_claude_stream()
 
 def test_completion_bedrock_ai21_stream():
     try:
@@ -327,28 +352,6 @@ def test_together_ai_completion_call_starcoder():
         print(f"error occurred: {traceback.format_exc()}")
         pass
 
-def test_completion_nlp_cloud_streaming():
-    try:
-        messages = [
-            {"role": "system", "content": "You are a helpful assistant."},
-            {
-                "role": "user",
-                "content": "how does a court case get to the Supreme Court?",
-            },
-        ]
-        response = completion(model="dolphin", messages=messages, stream=True, logger_fn=logger_fn)
-        complete_response = "" 
-        # Add any assertions here to check the response
-        for idx, chunk in enumerate(response):
-            chunk, finished = streaming_format_tests(idx, chunk)
-            if finished:
-                break
-            complete_response += chunk
-        if complete_response == "":
-            raise Exception("Empty response received")
-    except Exception as e:
-        pytest.fail(f"Error occurred: {e}")
-
 #### Test Function calling + streaming ####
 
 def test_completion_openai_with_functions():

litellm/utils.py

@@ -2506,14 +2506,13 @@ class CustomStreamWrapper:
                 return chunk_data['outputText']
         return ""
 
+    ## needs to handle the empty string case (even starting chunk can be an empty string)
     def __next__(self):
         model_response = ModelResponse(stream=True, model=self.model)
         try:
+            while True: # loop until a non-empty string is found
                 # return this for all models
                 completion_obj = {"content": ""}
-            if self.sent_first_chunk == False:
-                completion_obj["role"] = "assistant"
-                self.sent_first_chunk = True
                 if self.custom_llm_provider and self.custom_llm_provider == "anthropic":
                     chunk = next(self.completion_stream)
                     completion_obj["content"] = self.handle_anthropic_chunk(chunk)
@@ -2567,6 +2566,9 @@ class CustomStreamWrapper:
                 threading.Thread(target=self.logging_obj.success_handler, args=(completion_obj,)).start()
                 model_response.model = self.model
                 if len(completion_obj["content"]) > 0: # cannot set content of an OpenAI Object to be an empty string
+                    if self.sent_first_chunk == False:
+                        completion_obj["role"] = "assistant"
+                        self.sent_first_chunk = True
                     model_response.choices[0].delta = Delta(**completion_obj)
                     return model_response
         except StopIteration:

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "litellm"
-version = "0.1.685"
+version = "0.1.686"
 description = "Library to easily interface with LLM API providers"
 authors = ["BerriAI"]
 license = "MIT License"