From 7a4c6344ea19e0df44d630d65ddf8d9c2f232521 Mon Sep 17 00:00:00 2001
From: Ishaan Jaffer <ishaanjaffer0324@gmail.com>
Date: Fri, 10 Oct 2025 11:22:15 -0700
Subject: [PATCH] fix gemma: stream param

---
 .../vertex_gemma_models/transformation.py     | 25 +++++++++++--------
 .../test_vertex_gemma_transformation.py       |  3 +++
 2 files changed, 18 insertions(+), 10 deletions(-)

diff --git a/litellm/llms/vertex_ai/vertex_gemma_models/transformation.py b/litellm/llms/vertex_ai/vertex_gemma_models/transformation.py
index 779eb69c760..ece1cef37fc 100644
--- a/litellm/llms/vertex_ai/vertex_gemma_models/transformation.py
+++ b/litellm/llms/vertex_ai/vertex_gemma_models/transformation.py
@@ -30,6 +30,18 @@ class VertexGemmaConfig(OpenAIGPTConfig):
     def __init__(self) -> None:
         super().__init__()
 
+    def should_fake_stream(
+        self,
+        model: Optional[str],
+        stream: Optional[bool],
+        custom_llm_provider: Optional[str] = None,
+    ) -> bool:
+        """
+        Vertex AI Gemma models do not support streaming.
+        Return True to enable fake streaming on the client side.
+        """
+        return True
+
     def transform_request(
         self,
         model: str,
@@ -53,8 +65,9 @@ class VertexGemmaConfig(OpenAIGPTConfig):
             headers=headers,
         )
         
-        # Remove 'model' from the request as it's not needed in the instance
+        # Remove params not needed/supported by Vertex Gemma
         openai_request.pop("model", None)
+        openai_request.pop("stream", None)  # Streaming not supported, will be faked client-side
         
         # Wrap in Vertex Gemma format
         return {
@@ -105,16 +118,8 @@ class VertexGemmaConfig(OpenAIGPTConfig):
     ):
         """
         Make completion request to Vertex Gemma endpoint.
-        Supports both sync and async requests.
+        Supports both sync and async requests with fake streaming.
         """
-        # Handle streaming
-        stream = optional_params.get("stream", False)
-        if stream:
-            raise BaseLLMException(
-                status_code=400,
-                message="Streaming is not yet supported for Vertex AI Gemma models",
-            )
-
         if acompletion:
             return self._async_completion(
                 model=model,
diff --git a/tests/test_litellm/llms/vertex_ai/vertex_gemma_models/test_vertex_gemma_transformation.py b/tests/test_litellm/llms/vertex_ai/vertex_gemma_models/test_vertex_gemma_transformation.py
index fb988485791..f40024665de 100644
--- a/tests/test_litellm/llms/vertex_ai/vertex_gemma_models/test_vertex_gemma_transformation.py
+++ b/tests/test_litellm/llms/vertex_ai/vertex_gemma_models/test_vertex_gemma_transformation.py
@@ -155,6 +155,9 @@ class TestVertexGemmaCompletion:
             assert instance["messages"][0]["content"] == "What is machine learning?"
             assert instance["max_tokens"] == 100
             
+            # Verify stream parameter is NOT sent to Vertex (will be faked client-side)
+            assert "stream" not in instance
+            
             # Validate LiteLLM Response (OpenAI format)
             assert response.id == "chatcmpl-aaa4288f-2b8e-4bc0-8b14-4e444decd2c4"
             assert response.object == "chat.completion"