handle vertex pass through separately

litellm/proxy/pass_through_endpoints/streaming_handler.py

@@ -40,8 +40,12 @@ async def chunk_processor(
     - Yields chunks from the response
     - Collect non-empty chunks for post-processing (logging)
     """
-    collected_chunks: List[str] = []  # List to store all chunks
     try:
+        if endpoint_type == EndpointType.VERTEX_AI:
+            async for chunk in response.aiter_bytes():
+                yield chunk
+        else:
+            collected_chunks: List[str] = []  # List to store all chunks
             async for chunk in response.aiter_lines():
                 verbose_proxy_logger.debug(f"Processing chunk: {chunk}")
                 if not chunk: