test(google_genai): cover the limiter's Google config precedence

Covers an empty config next to a filled generationConfig and both config spellings in one request, for the cap check and the injected cap, and types the new test params as Mapping.

tests/unit/google_genai/test_google_genai_adapter.py

@@ -5,6 +5,7 @@ Test to verify the Google GenAI generate_content adapter functionality
 
 import json
 import unittest
+from collections.abc import Mapping
 from typing import Final
 
 import pytest
@@ -273,7 +274,7 @@ def test_list_contents_transformation():
     ],
     ids=["camelCase", "snake_case"],
 )
-def test_config_parameter_mapping(config: dict[str, object]):
+def test_config_parameter_mapping(config: Mapping[str, object]):
     """Test that config parameters are correctly mapped"""
     from litellm.google_genai.adapters.transformation import GoogleGenAIAdapter
 

tests/unit/google_genai/test_google_genai_main.py

@@ -4,7 +4,7 @@ Test to verify the Google GenAI generate_content adapter functionality
 """
 
 import json
-from collections.abc import Iterator
+from collections.abc import Iterator, Mapping
 from typing import Final
 
 import httpx
@@ -418,7 +418,7 @@ def _httpx_transport(monkeypatch: pytest.MonkeyPatch) -> None:
     monkeypatch.setattr(litellm, "disable_aiohttp_transport", True)
 
 
-async def _call_entrypoint(entrypoint: str, output_cap: dict[str, dict[str, int]]) -> None:
+async def _call_entrypoint(entrypoint: str, output_cap: Mapping[str, Mapping[str, int]]) -> None:
     from litellm.google_genai.main import (
         agenerate_content,
         agenerate_content_stream,
@@ -451,7 +451,7 @@ async def _call_entrypoint(entrypoint: str, output_cap: dict[str, dict[str, int]
     "entrypoint", ["generate_content", "agenerate_content", "generate_content_stream", "agenerate_content_stream"]
 )
 async def test_adapter_path_sends_the_output_cap_in_every_spelling(
-    _httpx_transport: None, entrypoint: str, output_cap: dict[str, dict[str, int]]
+    _httpx_transport: None, entrypoint: str, output_cap: Mapping[str, Mapping[str, int]]
 ) -> None:
     with respx.mock() as upstream:
         route: Final = upstream.post(f"{_CHAT_UPSTREAM}/chat/completions").mock(side_effect=_chat_reply)
@@ -463,7 +463,7 @@ async def test_adapter_path_sends_the_output_cap_in_every_spelling(
 @pytest.mark.asyncio
 @pytest.mark.parametrize("output_cap", _OUTPUT_CAP_SPELLINGS)
 async def test_native_gemini_request_carries_the_output_cap_in_every_spelling(
-    _httpx_transport: None, output_cap: dict[str, dict[str, int]]
+    _httpx_transport: None, output_cap: Mapping[str, Mapping[str, int]]
 ) -> None:
     from litellm.google_genai.main import agenerate_content
 

tests/unit/proxy/hooks/test_tpm_concurrent.py

@@ -1354,6 +1354,63 @@ async def test_project_otpm_injects_google_genai_native_output_cap(
     assert "max_tokens" not in data
 
 
+@pytest.mark.asyncio
+@pytest.mark.parametrize(
+    "configs",
+    [
+        {"config": {}, "generationConfig": {"maxOutputTokens": 100}},
+        {"generationConfig": {"maxOutputTokens": 100}, "generation_config": {"max_output_tokens": 10}},
+        {"config": {"maxOutputTokens": 100}, "generationConfig": {"maxOutputTokens": 10}},
+    ],
+)
+async def test_project_otpm_checks_the_first_non_empty_google_genai_config(rate_limiter, configs):
+    handler, cache = rate_limiter
+    model = "gemini/gemini-3-flash-preview"
+    user_api_key_dict = UserAPIKeyAuth(
+        api_key=hash_token("sk-google-genai-config-precedence"),
+        project_id="project-google-genai-config-precedence",
+        project_metadata={"model_otpm_limit": {model: 50}},
+    )
+
+    with pytest.raises(Exception, match="Rate limit exceeded for model_per_project_otpm") as exc_info:
+        await handler.async_pre_call_hook(
+            user_api_key_dict=user_api_key_dict,
+            cache=cache,
+            data={"model": model, "contents": [{"role": "user", "parts": [{"text": "Hello"}]}], **configs},
+            call_type="agenerate_content",
+        )
+
+    assert getattr(exc_info.value, "status_code", None) == 429
+
+
+@pytest.mark.asyncio
+async def test_project_otpm_injects_cap_into_the_first_non_empty_google_genai_config(rate_limiter):
+    handler, cache = rate_limiter
+    model = "gemini/gemini-3-flash-preview"
+    user_api_key_dict = UserAPIKeyAuth(
+        api_key=hash_token("sk-google-genai-config-inject-precedence"),
+        project_id="project-google-genai-config-inject-precedence",
+        project_metadata={"model_otpm_limit": {model: 40}},
+    )
+    data: Final = {
+        "model": model,
+        "contents": [{"role": "user", "parts": [{"text": "Hello"}]}],
+        "config": {},
+        "generationConfig": {"temperature": 0},
+        "generation_config": {"max_output_tokens": 5000},
+    }
+
+    await handler.async_pre_call_hook(
+        user_api_key_dict=user_api_key_dict,
+        cache=cache,
+        data=data,
+        call_type="agenerate_content",
+    )
+
+    assert data["generationConfig"] == {"temperature": 0, "maxOutputTokens": 10}
+    assert data["config"] == {}
+
+
 @pytest.mark.asyncio
 async def test_project_otpm_over_limit_rolls_back_itpm_reservation(rate_limiter):
     """