From d8418fff08f1a3f59dbe78318ac9d214fe59a8a5 Mon Sep 17 00:00:00 2001
From: StatPan <statpan@naver.com>
Date: Fri, 1 May 2026 21:11:57 +0900
Subject: [PATCH] fix(gemini): avoid duplicate model route for full api_base

---
 litellm/llms/vertex_ai/vertex_llm_base.py     |  9 ++++-
 .../llms/vertex_ai/test_vertex_llm_base.py    | 39 +++++++++++++++++++
 2 files changed, 47 insertions(+), 1 deletion(-)

diff --git a/litellm/llms/vertex_ai/vertex_llm_base.py b/litellm/llms/vertex_ai/vertex_llm_base.py
index 6f687dae7e8..80f1a96f854 100644
--- a/litellm/llms/vertex_ai/vertex_llm_base.py
+++ b/litellm/llms/vertex_ai/vertex_llm_base.py
@@ -412,7 +412,14 @@ class VertexBase:
                     raise ValueError(
                         "Model parameter is required for Gemini custom API base URLs"
                     )
-                url = "{}/models/{}:{}".format(api_base, model, endpoint)
+                if api_base.endswith(f"/models/{model}:{endpoint}"):
+                    url = api_base
+                elif api_base.endswith(f"/models/{model}"):
+                    url = f"{api_base}:{endpoint}"
+                else:
+                    url = "{}/models/{}:{}".format(
+                        api_base.rstrip("/"), model, endpoint
+                    )
                 if gemini_api_key is None:
                     raise ValueError(
                         "Missing Gemini API key. Set the GEMINI_API_KEY or GOOGLE_API_KEY environment variable."
diff --git a/tests/test_litellm/llms/vertex_ai/test_vertex_llm_base.py b/tests/test_litellm/llms/vertex_ai/test_vertex_llm_base.py
index 88aac07a0c9..88212ded130 100644
--- a/tests/test_litellm/llms/vertex_ai/test_vertex_llm_base.py
+++ b/tests/test_litellm/llms/vertex_ai/test_vertex_llm_base.py
@@ -891,6 +891,45 @@ class TestVertexBase:
                 result_url == expected_url
             ), f"Expected {expected_url}, got {result_url} for model {model}"
 
+    @pytest.mark.parametrize(
+        "api_base, endpoint, expected_url",
+        [
+            (
+                "https://proxy.example.com/generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-lite:generateContent",
+                "generateContent",
+                "https://proxy.example.com/generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-lite:generateContent",
+            ),
+            (
+                "https://proxy.example.com/generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-lite",
+                "generateContent",
+                "https://proxy.example.com/generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-lite:generateContent",
+            ),
+            (
+                "https://proxy.example.com/generativelanguage.googleapis.com/v1beta/",
+                "generateContent",
+                "https://proxy.example.com/generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-lite:generateContent",
+            ),
+        ],
+    )
+    def test_check_custom_proxy_gemini_prebuilt_route_handling(
+        self, api_base, endpoint, expected_url
+    ):
+        """Test that Gemini custom api_base values are not double-appended when they already include model routing."""
+        vertex_base = VertexBase()
+
+        _, result_url = vertex_base._check_custom_proxy(
+            api_base=api_base,
+            custom_llm_provider="gemini",
+            gemini_api_key="test-api-key",
+            endpoint=endpoint,
+            stream=False,
+            auth_header=None,
+            url="https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-lite:generateContent",
+            model="gemini-2.5-flash-lite",
+        )
+
+        assert result_url == expected_url
+
     def test_check_custom_proxy_streaming_parameter(self):
         """Test that streaming parameter correctly adds ?alt=sse to URLs"""
         vertex_base = VertexBase()