test(vertex): cover model garden api base handling

tests/test_litellm/llms/vertex_ai/test_vertex_model_garden_openapi.py

@@ -1,8 +1,12 @@
 """Vertex Model Garden: OpenAPI base URL for publisher/model ids vs per-endpoint path."""
 
+import sys
+from unittest.mock import MagicMock, patch
+
 import pytest
 
 from litellm.llms.vertex_ai.vertex_model_garden.main import (
+    VertexAIModelGardenModels,
     _vertex_model_garden_model_id_in_json_body,
     create_vertex_url,
 )
@@ -37,5 +41,114 @@ def test_create_vertex_url_openapi_vs_deployed_endpoint(
 
 
 def test_model_id_in_json_body_heuristic() -> None:
-    assert _vertex_model_garden_model_id_in_json_body("xai/grok-4.1-fast-reasoning") is True
+    assert (
+        _vertex_model_garden_model_id_in_json_body("xai/grok-4.1-fast-reasoning")
+        is True
+    )
     assert _vertex_model_garden_model_id_in_json_body("5464397967697903616") is False
+
+
+def _mock_vertexai_module() -> MagicMock:
+    mock_vertexai = MagicMock()
+    mock_vertexai.preview = MagicMock()
+    mock_vertexai.preview.language_models = MagicMock()
+    return mock_vertexai
+
+
+def test_completion_preserves_custom_model_garden_api_base() -> None:
+    custom_api_base = (
+        "https://mg-endpoint-123.us-west1-456.prediction.vertexai.goog/v1/"
+        "projects/123/locations/us-west1/endpoints/mg-endpoint-123"
+    )
+    mock_vertexai = _mock_vertexai_module()
+    mock_openai_like_handler = MagicMock()
+    mock_openai_like_handler.completion.return_value = "mock-response"
+
+    with (
+        patch.dict(
+            sys.modules,
+            {"vertexai": mock_vertexai, "vertexai.preview": mock_vertexai.preview},
+        ),
+        patch(
+            "litellm.llms.vertex_ai.gemini.vertex_and_google_ai_studio_gemini."
+            "VertexLLM._ensure_access_token",
+            return_value=("fake-token", "resolved-project"),
+        ),
+        patch(
+            "litellm.llms.openai_like.chat.handler.OpenAILikeChatHandler",
+            return_value=mock_openai_like_handler,
+        ),
+    ):
+        response = VertexAIModelGardenModels().completion(
+            model="gemma4",
+            messages=[{"role": "user", "content": "hi"}],
+            model_response=MagicMock(),
+            print_verbose=MagicMock(),
+            encoding=None,
+            logging_obj=MagicMock(),
+            api_base=custom_api_base,
+            optional_params={},
+            custom_prompt_dict={},
+            headers=None,
+            timeout=30,
+            litellm_params={},
+            vertex_project="configured-project",
+            vertex_location="us-west1",
+        )
+
+    assert response == "mock-response"
+    mock_openai_like_handler.completion.assert_called_once()
+    call_kwargs = mock_openai_like_handler.completion.call_args.kwargs
+    assert call_kwargs["api_base"] == custom_api_base
+    assert call_kwargs["model"] == ""
+    assert call_kwargs["optional_params"]["stream"] is False
+
+
+def test_completion_builds_openapi_base_for_publisher_model_without_custom_api_base() -> (
+    None
+):
+    mock_vertexai = _mock_vertexai_module()
+    mock_openai_like_handler = MagicMock()
+    mock_openai_like_handler.completion.return_value = "mock-response"
+
+    with (
+        patch.dict(
+            sys.modules,
+            {"vertexai": mock_vertexai, "vertexai.preview": mock_vertexai.preview},
+        ),
+        patch(
+            "litellm.llms.vertex_ai.gemini.vertex_and_google_ai_studio_gemini."
+            "VertexLLM._ensure_access_token",
+            return_value=("fake-token", "resolved-project"),
+        ),
+        patch(
+            "litellm.llms.openai_like.chat.handler.OpenAILikeChatHandler",
+            return_value=mock_openai_like_handler,
+        ),
+    ):
+        response = VertexAIModelGardenModels().completion(
+            model="xai/grok-4.1-fast-reasoning",
+            messages=[{"role": "user", "content": "hi"}],
+            model_response=MagicMock(),
+            print_verbose=MagicMock(),
+            encoding=None,
+            logging_obj=MagicMock(),
+            api_base=None,
+            optional_params={},
+            custom_prompt_dict={},
+            headers=None,
+            timeout=30,
+            litellm_params={},
+            vertex_project="configured-project",
+            vertex_location="us-west1",
+        )
+
+    assert response == "mock-response"
+    mock_openai_like_handler.completion.assert_called_once()
+    call_kwargs = mock_openai_like_handler.completion.call_args.kwargs
+    assert call_kwargs["api_base"] == (
+        "https://us-west1-aiplatform.googleapis.com/v1/projects/configured-project/"
+        "locations/us-west1/endpoints/openapi"
+    )
+    assert call_kwargs["model"] == "xai/grok-4.1-fast-reasoning"
+    assert call_kwargs["optional_params"]["stream"] is False