fix re-add virtual key auth checks on vertex ai pass thru endpoints (#5827)

docs/my-website/docs/proxy/configs.md

@@ -792,7 +792,8 @@ general_settings:
     "alerting": [
       "string"
     ],
-    "alerting_threshold": 0
+    "alerting_threshold": 0,
+    "use_client_credentials_pass_through_routes" : "boolean", # use client credentials for all pass through routes like "/vertex-ai", /bedrock/. When this is True Virtual Key auth will not be applied on these endpoints" https://docs.litellm.ai/docs/pass_through/vertex_ai
   }
 }
 ```

litellm/proxy/auth/auth_utils.py

@@ -354,9 +354,26 @@ def is_pass_through_provider_route(route: str) -> bool:
 def should_run_auth_on_pass_through_provider_route(route: str) -> bool:
     """
     Use this to decide if the rest of the LiteLLM Virtual Key auth checks should run on /vertex-ai/{endpoint} routes
+    Use this to decide if the rest of the LiteLLM Virtual Key auth checks should run on provider pass through routes
+    ex /vertex-ai/{endpoint} routes
+    Run virtual key auth if the following is try:
+    - User is premium_user
+    - User has enabled litellm_setting.use_client_credentials_pass_through_routes
     """
-    # by default we do not run virtual key auth checks on /vertex-ai/{endpoint} routes
+    from litellm.proxy.proxy_server import general_settings, premium_user
-    return False
+
+    if premium_user is not True:
+        return False
+
+    # premium use has opted into using client credentials
+    if (
+        general_settings.get("use_client_credentials_pass_through_routes", False)
+        is True
+    ):
+        return False
+
+    # only enabled for LiteLLM Enterprise
+    return True
 
 
 def _has_user_setup_sso():

litellm/tests/test_key_generate_prisma.py

@@ -3221,3 +3221,31 @@ async def test_key_list_unsupported_params(prisma_client):
         error_str = str(e.message)
         assert "Unsupported parameter" in error_str
         pass
+
+
+@pytest.mark.asyncio
+async def test_auth_vertex_ai_route(prisma_client):
+    """
+    If user is premium user and vertex-ai route is used. Assert Virtual Key checks are run
+    """
+    litellm.set_verbose = True
+    setattr(litellm.proxy.proxy_server, "prisma_client", prisma_client)
+    setattr(litellm.proxy.proxy_server, "premium_user", True)
+    setattr(litellm.proxy.proxy_server, "master_key", "sk-1234")
+    await litellm.proxy.proxy_server.prisma_client.connect()
+
+    route = "/vertex-ai/publishers/google/models/gemini-1.5-flash-001:generateContent"
+    request = Request(scope={"type": "http"})
+    request._url = URL(url=route)
+    request._headers = {"Authorization": "Bearer sk-12345"}
+    try:
+        await user_api_key_auth(request=request, api_key="Bearer " + "sk-12345")
+        pytest.fail("Expected this call to fail. User is over limit.")
+    except Exception as e:
+        print(vars(e))
+        print("error str=", str(e.message))
+        error_str = str(e.message)
+        assert e.code == "401"
+        assert "Invalid proxy server token passed" in error_str
+
+        pass