diff --git a/memgpt/llm_api/openai.py b/memgpt/llm_api/openai.py
index e7001d63e4..653d3f14a2 100644
--- a/memgpt/llm_api/openai.py
+++ b/memgpt/llm_api/openai.py
@@ -252,7 +252,7 @@ def _sse_post(url: str, data: dict, headers: dict) -> Generator[ChatCompletionCh
             # Inspect for errors before iterating (see https://github.com/florimondmanca/httpx-sse/pull/12)
             if not event_source.response.is_success:
                 # handle errors
-                from utils import printd
+                from memgpt.utils import printd
 
                 printd("Caught error before iterating SSE request:", vars(event_source.response))
                 printd(event_source.response.read())
diff --git a/memgpt/server/rest_api/agents/message.py b/memgpt/server/rest_api/agents/message.py
index 45db2defe5..76c651c6d3 100644
--- a/memgpt/server/rest_api/agents/message.py
+++ b/memgpt/server/rest_api/agents/message.py
@@ -95,6 +95,7 @@ async def send_message_to_agent(
 ) -> Union[StreamingResponse, UserMessageResponse]:
     """Split off into a separate function so that it can be imported in the /chat/completion proxy."""
 
+    include_final_message = True
     # handle the legacy mode streaming
     if stream_legacy:
         # NOTE: override