test_google_llm.py

# Copyright 2026 Google LLC
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import os
import sys
from typing import Optional
from unittest import mock
from unittest.mock import AsyncMock

from google.adk import version as adk_version
from google.adk.agents.context_cache_config import ContextCacheConfig
from google.adk.models.cache_metadata import CacheMetadata
from google.adk.models.gemini_llm_connection import GeminiLlmConnection
from google.adk.models.google_llm import _build_function_declaration_log
from google.adk.models.google_llm import _build_request_log
from google.adk.models.google_llm import _RESOURCE_EXHAUSTED_POSSIBLE_FIX_MESSAGE
from google.adk.models.google_llm import _ResourceExhaustedError
from google.adk.models.google_llm import Gemini
from google.adk.models.llm_request import LlmRequest
from google.adk.models.llm_response import LlmResponse
from google.adk.utils._client_labels_utils import _AGENT_ENGINE_TELEMETRY_ENV_VARIABLE_NAME
from google.adk.utils._client_labels_utils import _AGENT_ENGINE_TELEMETRY_TAG
from google.adk.utils._google_client_headers import get_tracking_headers
from google.adk.utils.variant_utils import GoogleLLMVariant
from google.genai import types
from google.genai.errors import ClientError
from google.genai.types import Content
from google.genai.types import Part
import pytest


class MockAsyncIterator:
  """Mock for async iterator."""

  def __init__(self, seq):
    self.iter = iter(seq)

  def __aiter__(self):
    return self

  async def __anext__(self):
    try:
      return next(self.iter)
    except StopIteration as exc:
      raise StopAsyncIteration from exc

  async def aclose(self):
    pass


@pytest.fixture
def generate_content_response():
  return types.GenerateContentResponse(
      candidates=[
          types.Candidate(
              content=Content(
                  role="model",
                  parts=[Part.from_text(text="Hello, how can I help you?")],
              ),
              finish_reason=types.FinishReason.STOP,
          )
      ]
  )


@pytest.fixture
def gemini_llm():
  return Gemini(model="gemini-1.5-flash")


@pytest.fixture
def llm_request():
  return LlmRequest(
      model="gemini-1.5-flash",
      contents=[Content(role="user", parts=[Part.from_text(text="Hello")])],
      config=types.GenerateContentConfig(
          temperature=0.1,
          response_modalities=[types.Modality.TEXT],
          system_instruction="You are a helpful assistant",
      ),
  )


@pytest.fixture
def cache_metadata():
  import time

  return CacheMetadata(
      cache_name="projects/test/locations/us-central1/cachedContents/test123",
      expire_time=time.time() + 3600,
      fingerprint="test_fingerprint",
      invocations_used=2,
      contents_count=3,
      created_at=time.time() - 600,
  )


@pytest.fixture
def llm_request_with_cache(cache_metadata):
  return LlmRequest(
      model="gemini-1.5-flash",
      contents=[Content(role="user", parts=[Part.from_text(text="Hello")])],
      config=types.GenerateContentConfig(
          temperature=0.1,
          response_modalities=[types.Modality.TEXT],
          system_instruction="You are a helpful assistant",
      ),
      cache_config=ContextCacheConfig(
          cache_intervals=10, ttl_seconds=3600, min_tokens=100
      ),
      cache_metadata=cache_metadata,
  )


@pytest.fixture
def llm_request_with_computer_use():
  return LlmRequest(
      model="gemini-1.5-flash",
      contents=[Content(role="user", parts=[Part.from_text(text="Hello")])],
      config=types.GenerateContentConfig(
          temperature=0.1,
          response_modalities=[types.Modality.TEXT],
          system_instruction="You are a helpful assistant",
          tools=[
              types.Tool(
                  computer_use=types.ComputerUse(
                      environment=types.Environment.ENVIRONMENT_BROWSER
                  )
              )
          ],
      ),
  )


def test_supported_models():
  models = Gemini.supported_models()
  assert len(models) == 5
  assert models[0] == r"gemini-.*"
  assert models[1] == r"gemma-4.*"
  assert models[2] == r"model-optimizer-.*"
  assert models[3] == r"projects\/.+\/locations\/.+\/endpoints\/.+"
  assert (
      models[4]
      == r"projects\/.+\/locations\/.+\/publishers\/google\/models\/gemini.+"
  )


def test_client_version_header():
  model = Gemini(model="gemini-1.5-flash")
  client = model.api_client

  # Check that ADK version and Python version are present in headers
  adk_version_string = f"google-adk/{adk_version.__version__}"
  python_version_string = f"gl-python/{sys.version.split()[0]}"

  x_goog_api_client_header = client._api_client._http_options.headers[
      "x-goog-api-client"
  ]
  user_agent_header = client._api_client._http_options.headers["user-agent"]

  # Verify ADK version is present
  assert adk_version_string in x_goog_api_client_header
  assert adk_version_string in user_agent_header

  # Verify Python version is present
  assert python_version_string in x_goog_api_client_header
  assert python_version_string in user_agent_header

  # Verify some Google SDK version is present (could be genai-sdk or vertex-genai-modules)
  assert any(
      sdk in x_goog_api_client_header
      for sdk in ["google-genai-sdk/", "vertex-genai-modules/"]
  )
  assert any(
      sdk in user_agent_header
      for sdk in ["google-genai-sdk/", "vertex-genai-modules/"]
  )


def test_client_version_header_with_agent_engine(monkeypatch):
  monkeypatch.setenv(
      _AGENT_ENGINE_TELEMETRY_ENV_VARIABLE_NAME, "my_test_project"
  )
  model = Gemini(model="gemini-1.5-flash")
  client = model.api_client

  # Check that ADK version with telemetry tag and Python version are present in
  # headers
  adk_version_with_telemetry = (
      f"google-adk/{adk_version.__version__}+{_AGENT_ENGINE_TELEMETRY_TAG}"
  )
  python_version_string = f"gl-python/{sys.version.split()[0]}"

  x_goog_api_client_header = client._api_client._http_options.headers[
      "x-goog-api-client"
  ]
  user_agent_header = client._api_client._http_options.headers["user-agent"]

  # Verify ADK version with telemetry tag is present
  assert adk_version_with_telemetry in x_goog_api_client_header
  assert adk_version_with_telemetry in user_agent_header

  # Verify Python version is present
  assert python_version_string in x_goog_api_client_header
  assert python_version_string in user_agent_header

  # Verify some Google SDK version is present (could be genai-sdk or vertex-genai-modules)
  assert any(
      sdk in x_goog_api_client_header
      for sdk in ["google-genai-sdk/", "vertex-genai-modules/"]
  )
  assert any(
      sdk in user_agent_header
      for sdk in ["google-genai-sdk/", "vertex-genai-modules/"]
  )


def test_maybe_append_user_content(gemini_llm, llm_request):
  # Test with user content already present
  gemini_llm._maybe_append_user_content(llm_request)
  assert len(llm_request.contents) == 1

  # Test with model content as the last message
  llm_request.contents.append(
      Content(role="model", parts=[Part.from_text(text="Response")])
  )
  gemini_llm._maybe_append_user_content(llm_request)
  assert len(llm_request.contents) == 3
  assert llm_request.contents[-1].role == "user"
  assert "Continue processing" in llm_request.contents[-1].parts[0].text


@pytest.mark.asyncio
async def test_generate_content_async(
    gemini_llm, llm_request, generate_content_response
):
  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    # Create a mock coroutine that returns the generate_content_response
    async def mock_coro():
      return generate_content_response

    # Assign the coroutine to the mocked method
    mock_client.aio.models.generate_content.return_value = mock_coro()

    responses = [
        resp
        async for resp in gemini_llm.generate_content_async(
            llm_request, stream=False
        )
    ]

    assert len(responses) == 1
    assert isinstance(responses[0], LlmResponse)
    assert responses[0].content.parts[0].text == "Hello, how can I help you?"
    mock_client.aio.models.generate_content.assert_called_once()


@pytest.mark.asyncio
async def test_generate_content_async_stream(gemini_llm, llm_request):
  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    mock_responses = [
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model", parts=[Part.from_text(text="Hello")]
                    ),
                    finish_reason=None,
                )
            ]
        ),
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model", parts=[Part.from_text(text=", how")]
                    ),
                    finish_reason=None,
                )
            ]
        ),
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model",
                        parts=[Part.from_text(text=" can I help you?")],
                    ),
                    finish_reason=types.FinishReason.STOP,
                )
            ]
        ),
    ]

    # Create a mock coroutine that returns the MockAsyncIterator
    async def mock_coro():
      return MockAsyncIterator(mock_responses)

    # Set the mock to return the coroutine
    mock_client.aio.models.generate_content_stream.return_value = mock_coro()

    responses = [
        resp
        async for resp in gemini_llm.generate_content_async(
            llm_request, stream=True
        )
    ]

    # Assertions remain the same
    assert len(responses) == 4
    assert responses[0].partial is True
    assert responses[1].partial is True
    assert responses[2].partial is True
    assert responses[3].content.parts[0].text == "Hello, how can I help you?"
    mock_client.aio.models.generate_content_stream.assert_called_once()


@pytest.mark.asyncio
async def test_generate_content_async_stream_preserves_thinking_and_text_parts(
    gemini_llm, llm_request
):
  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    response1 = types.GenerateContentResponse(
        candidates=[
            types.Candidate(
                content=Content(
                    role="model",
                    parts=[Part(text="Think1", thought=True)],
                ),
                finish_reason=None,
            )
        ]
    )
    response2 = types.GenerateContentResponse(
        candidates=[
            types.Candidate(
                content=Content(
                    role="model",
                    parts=[Part(text="Think2", thought=True)],
                ),
                finish_reason=None,
            )
        ]
    )
    response3 = types.GenerateContentResponse(
        candidates=[
            types.Candidate(
                content=Content(
                    role="model",
                    parts=[Part.from_text(text="Answer.")],
                ),
                finish_reason=types.FinishReason.STOP,
            )
        ]
    )

    async def mock_coro():
      return MockAsyncIterator([response1, response2, response3])

    mock_client.aio.models.generate_content_stream.return_value = mock_coro()

    responses = [
        resp
        async for resp in gemini_llm.generate_content_async(
            llm_request, stream=True
        )
    ]

    assert len(responses) == 4
    assert responses[0].partial is True
    assert responses[1].partial is True
    assert responses[2].partial is True
    assert responses[3].content.parts[0].text == "Think1Think2"
    assert responses[3].content.parts[0].thought is True
    assert responses[3].content.parts[1].text == "Answer."
    mock_client.aio.models.generate_content_stream.assert_called_once()


@pytest.mark.parametrize("stream", [True, False])
@pytest.mark.asyncio
async def test_generate_content_async_resource_exhausted_error(
    stream, gemini_llm, llm_request
):
  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    err = ClientError(code=429, response_json={})
    err.code = 429
    if stream:
      mock_client.aio.models.generate_content_stream.side_effect = err
    else:
      mock_client.aio.models.generate_content.side_effect = err

    with pytest.raises(_ResourceExhaustedError) as excinfo:
      responses = []
      async for resp in gemini_llm.generate_content_async(
          llm_request, stream=stream
      ):
        responses.append(resp)
    assert _RESOURCE_EXHAUSTED_POSSIBLE_FIX_MESSAGE in str(excinfo.value)
    assert excinfo.value.code == 429
    if stream:
      mock_client.aio.models.generate_content_stream.assert_called_once()
    else:
      mock_client.aio.models.generate_content.assert_called_once()


@pytest.mark.parametrize("stream", [True, False])
@pytest.mark.asyncio
async def test_generate_content_async_other_client_error(
    stream, gemini_llm, llm_request
):
  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    err = ClientError(code=500, response_json={})
    err.code = 500
    if stream:
      mock_client.aio.models.generate_content_stream.side_effect = err
    else:
      mock_client.aio.models.generate_content.side_effect = err

    with pytest.raises(ClientError) as excinfo:
      responses = []
      async for resp in gemini_llm.generate_content_async(
          llm_request, stream=stream
      ):
        responses.append(resp)
    assert excinfo.value.code == 500
    assert not isinstance(excinfo.value, _ResourceExhaustedError)
    if stream:
      mock_client.aio.models.generate_content_stream.assert_called_once()
    else:
      mock_client.aio.models.generate_content.assert_called_once()


@pytest.mark.asyncio
async def test_connect(gemini_llm, llm_request):
  # Create a mock connection
  mock_connection = mock.MagicMock(spec=GeminiLlmConnection)

  # Create a mock context manager
  class MockContextManager:

    async def __aenter__(self):
      return mock_connection

    async def __aexit__(self, *args):
      pass

  # Mock the connect method at the class level
  with mock.patch(
      "google.adk.models.google_llm.Gemini.connect",
      return_value=MockContextManager(),
  ):
    async with gemini_llm.connect(llm_request) as connection:
      assert connection is mock_connection


@pytest.mark.asyncio
async def test_generate_content_async_with_custom_headers(
    gemini_llm, llm_request, generate_content_response
):
  """Test that tracking headers are updated when custom headers are provided."""
  # Add custom headers to the request config
  custom_headers = {"custom-header": "custom-value"}
  tracking_headers = get_tracking_headers()
  for key in tracking_headers:
    custom_headers[key] = "custom " + tracking_headers[key]
  llm_request.config.http_options = types.HttpOptions(headers=custom_headers)

  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    # Create a mock coroutine that returns the generate_content_response
    async def mock_coro():
      return generate_content_response

    mock_client.aio.models.generate_content.return_value = mock_coro()

    responses = [
        resp
        async for resp in gemini_llm.generate_content_async(
            llm_request, stream=False
        )
    ]

    # Verify that the config passed to generate_content contains merged headers
    mock_client.aio.models.generate_content.assert_called_once()
    call_args = mock_client.aio.models.generate_content.call_args
    config_arg = call_args.kwargs["config"]

    for key, value in config_arg.http_options.headers.items():
      tracking_headers = get_tracking_headers()
      if key in tracking_headers:
        assert value == tracking_headers[key] + " custom"
      else:
        assert value == custom_headers[key]

    assert len(responses) == 1
    assert isinstance(responses[0], LlmResponse)


@pytest.mark.asyncio
async def test_generate_content_async_stream_with_custom_headers(
    gemini_llm, llm_request
):
  """Test that tracking headers are updated when custom headers are provided in streaming mode."""
  # Add custom headers to the request config
  custom_headers = {"custom-header": "custom-value"}
  llm_request.config.http_options = types.HttpOptions(headers=custom_headers)

  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    mock_responses = [
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model", parts=[Part.from_text(text="Hello")]
                    ),
                    finish_reason=types.FinishReason.STOP,
                )
            ]
        )
    ]

    async def mock_coro():
      return MockAsyncIterator(mock_responses)

    mock_client.aio.models.generate_content_stream.return_value = mock_coro()

    responses = [
        resp
        async for resp in gemini_llm.generate_content_async(
            llm_request, stream=True
        )
    ]

    # Verify that the config passed to generate_content_stream contains merged headers
    mock_client.aio.models.generate_content_stream.assert_called_once()
    call_args = mock_client.aio.models.generate_content_stream.call_args
    config_arg = call_args.kwargs["config"]

    expected_headers = custom_headers.copy()
    expected_headers.update(get_tracking_headers())
    assert config_arg.http_options.headers == expected_headers

    assert len(responses) == 2


@pytest.mark.parametrize("stream", [True, False])
@pytest.mark.asyncio
async def test_generate_content_async_patches_tracking_headers(
    stream, gemini_llm, llm_request, generate_content_response
):
  """Tests that tracking headers are added to the request config."""
  # Set the request's config.http_options to None.
  llm_request.config.http_options = None

  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    if stream:
      # Create a mock coroutine that returns the mock_responses.
      async def mock_coro():
        return MockAsyncIterator([generate_content_response])

      # Mock for streaming response.
      mock_client.aio.models.generate_content_stream.return_value = mock_coro()
    else:
      # Create a mock coroutine that returns the generate_content_response.
      async def mock_coro():
        return generate_content_response

      # Mock for non-streaming response.
      mock_client.aio.models.generate_content.return_value = mock_coro()

    # Call the generate_content_async method.
    responses = [
        resp
        async for resp in gemini_llm.generate_content_async(
            llm_request, stream=stream
        )
    ]

    # Assert that the config passed to the generate_content or
    # generate_content_stream method contains the tracking headers.
    if stream:
      mock_client.aio.models.generate_content_stream.assert_called_once()
      call_args = mock_client.aio.models.generate_content_stream.call_args
    else:
      mock_client.aio.models.generate_content.assert_called_once()
      call_args = mock_client.aio.models.generate_content.call_args

    final_config = call_args.kwargs["config"]

    assert final_config is not None
    assert final_config.http_options is not None
    assert (
        final_config.http_options.headers["x-goog-api-client"]
        == get_tracking_headers()["x-goog-api-client"]
    )

    assert len(responses) == 2 if stream else 1


def test_live_api_version_vertex_ai(gemini_llm):
  """Test that _live_api_version returns 'v1beta1' for Vertex AI backend."""
  with mock.patch.object(
      gemini_llm, "_api_backend", GoogleLLMVariant.VERTEX_AI
  ):
    assert gemini_llm._live_api_version == "v1beta1"


def test_live_api_version_gemini_api(gemini_llm):
  """Test that _live_api_version returns 'v1alpha' for Gemini API backend."""
  with mock.patch.object(
      gemini_llm, "_api_backend", GoogleLLMVariant.GEMINI_API
  ):
    assert gemini_llm._live_api_version == "v1alpha"


def test_live_api_client_properties(gemini_llm):
  """Test that _live_api_client is properly configured with tracking headers and API version."""
  with mock.patch.object(
      gemini_llm, "_api_backend", GoogleLLMVariant.VERTEX_AI
  ):
    client = gemini_llm._live_api_client

    # Verify that the client has the correct headers and API version
    http_options = client._api_client._http_options
    assert http_options.api_version == "v1beta1"

    # Check that tracking headers are included
    tracking_headers = get_tracking_headers()
    for key, value in tracking_headers.items():
      assert key in http_options.headers
      assert value in http_options.headers[key]


@pytest.mark.asyncio
async def test_connect_with_custom_headers(gemini_llm, llm_request):
  """Test that connect method updates tracking headers and API version when custom headers are provided."""
  # Setup request with live connect config and custom headers
  custom_headers = {"custom-live-header": "live-value"}
  llm_request.live_connect_config = types.LiveConnectConfig(
      http_options=types.HttpOptions(headers=custom_headers)
  )

  mock_live_session = mock.AsyncMock()

  # Mock the _live_api_client to return a mock client
  with mock.patch.object(gemini_llm, "_live_api_client") as mock_live_client:
    # Create a mock context manager
    class MockLiveConnect:

      async def __aenter__(self):
        return mock_live_session

      async def __aexit__(self, *args):
        pass

    mock_live_client.aio.live.connect.return_value = MockLiveConnect()

    async with gemini_llm.connect(llm_request) as connection:
      # Verify that the connect method was called with the right config
      mock_live_client.aio.live.connect.assert_called_once()
      call_args = mock_live_client.aio.live.connect.call_args
      config_arg = call_args.kwargs["config"]

      # Verify that tracking headers were merged with custom headers
      expected_headers = custom_headers.copy()
      expected_headers.update(get_tracking_headers())
      assert config_arg.http_options.headers == expected_headers

      # Verify that API version was set
      assert config_arg.http_options.api_version == gemini_llm._live_api_version

      # Verify that system instruction and tools were set
      assert config_arg.system_instruction is not None
      assert config_arg.tools == llm_request.config.tools

      # Verify connection is properly wrapped
      assert isinstance(connection, GeminiLlmConnection)


@pytest.mark.asyncio
async def test_connect_without_custom_headers(gemini_llm, llm_request):
  """Test that connect method works properly when no custom headers are provided."""
  # Setup request with live connect config but no custom headers
  llm_request.live_connect_config = types.LiveConnectConfig()

  mock_live_session = mock.AsyncMock()

  with mock.patch.object(gemini_llm, "_live_api_client") as mock_live_client:

    class MockLiveConnect:

      async def __aenter__(self):
        return mock_live_session

      async def __aexit__(self, *args):
        pass

    mock_live_client.aio.live.connect.return_value = MockLiveConnect()

    with mock.patch(
        "google.adk.models.google_llm.GeminiLlmConnection"
    ) as MockGeminiLlmConnection:
      async with gemini_llm.connect(llm_request) as connection:
        # Verify that the connect method was called with the right config
        mock_live_client.aio.live.connect.assert_called_once()
        call_args = mock_live_client.aio.live.connect.call_args
        config_arg = call_args.kwargs["config"]

        # Verify that http_options remains None since no custom headers were provided
        assert config_arg.http_options is None

        # Verify that system instruction and tools were still set
        assert config_arg.system_instruction is not None
        assert config_arg.tools == llm_request.config.tools

        MockGeminiLlmConnection.assert_called_once_with(
            mock_live_session,
            api_backend=gemini_llm._api_backend,
            model_version=llm_request.model,
        )


@pytest.mark.parametrize(
    (
        "api_backend, "
        "expected_file_display_name, "
        "expected_inline_display_name, "
        "expected_labels"
    ),
    [
        (
            GoogleLLMVariant.GEMINI_API,
            None,
            None,
            None,
        ),
        (
            GoogleLLMVariant.VERTEX_AI,
            "My Test PDF",
            "My Test Image",
            {"key": "value"},
        ),
    ],
)
@pytest.mark.asyncio
async def test_preprocess_request_handles_backend_specific_fields(
    gemini_llm: Gemini,
    api_backend: GoogleLLMVariant,
    expected_file_display_name: Optional[str],
    expected_inline_display_name: Optional[str],
    expected_labels: Optional[str],
):
  """Tests that _preprocess_request correctly sanitizes fields based on the API backend.

  - For GEMINI_API, it should remove 'display_name' from file/inline data
    and remove 'labels' from the config.
  - For VERTEX_AI, it should leave these fields untouched.
  """
  # Arrange: Create a request with fields that need to be preprocessed.
  llm_request_with_files = LlmRequest(
      model="gemini-1.5-flash",
      contents=[
          Content(
              role="user",
              parts=[
                  Part(
                      file_data=types.FileData(
                          file_uri="gs://bucket/file.pdf",
                          mime_type="application/pdf",
                          display_name="My Test PDF",
                      )
                  ),
                  Part(
                      inline_data=types.Blob(
                          data=b"some_bytes",
                          mime_type="image/png",
                          display_name="My Test Image",
                      )
                  ),
              ],
          )
      ],
      config=types.GenerateContentConfig(labels={"key": "value"}),
  )

  # Mock the _api_backend property to control the test scenario
  with mock.patch.object(
      Gemini, "_api_backend", new_callable=mock.PropertyMock
  ) as mock_backend:
    mock_backend.return_value = api_backend

    # Act: Run the preprocessing method
    await gemini_llm._preprocess_request(llm_request_with_files)

    # Assert: Check if the fields were correctly processed
    file_part = llm_request_with_files.contents[0].parts[0]
    inline_part = llm_request_with_files.contents[0].parts[1]

    assert file_part.file_data.display_name == expected_file_display_name
    assert inline_part.inline_data.display_name == expected_inline_display_name
    assert llm_request_with_files.config.labels == expected_labels


@pytest.mark.asyncio
async def test_generate_content_async_stream_aggregated_content_regardless_of_finish_reason():
  """Test that aggregated content is generated regardless of finish_reason."""
  gemini_llm = Gemini(model="gemini-1.5-flash")
  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      contents=[Content(role="user", parts=[Part.from_text(text="Hello")])],
      config=types.GenerateContentConfig(
          temperature=0.1,
          response_modalities=[types.Modality.TEXT],
          system_instruction="You are a helpful assistant",
      ),
  )

  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    # Test with different finish reasons
    test_cases = [
        types.FinishReason.MAX_TOKENS,
        types.FinishReason.SAFETY,
        types.FinishReason.RECITATION,
        types.FinishReason.OTHER,
    ]

    for finish_reason in test_cases:
      mock_responses = [
          types.GenerateContentResponse(
              candidates=[
                  types.Candidate(
                      content=Content(
                          role="model", parts=[Part.from_text(text="Hello")]
                      ),
                      finish_reason=None,
                  )
              ]
          ),
          types.GenerateContentResponse(
              candidates=[
                  types.Candidate(
                      content=Content(
                          role="model", parts=[Part.from_text(text=" world")]
                      ),
                      finish_reason=finish_reason,
                      finish_message=f"Finished with {finish_reason}",
                  )
              ]
          ),
      ]

      async def mock_coro():
        return MockAsyncIterator(mock_responses)

      mock_client.aio.models.generate_content_stream.return_value = mock_coro()

      responses = [
          resp
          async for resp in gemini_llm.generate_content_async(
              llm_request, stream=True
          )
      ]

      # Should have 3 responses: 2 partial and 1 final aggregated
      assert len(responses) == 3
      assert responses[0].partial is True
      assert responses[1].partial is True

      # Final response should have aggregated content with error info
      final_response = responses[2]
      assert final_response.content.parts[0].text == "Hello world"
      # After the code changes, error_code and error_message are set for non-STOP finish reasons
      assert final_response.error_code == finish_reason
      assert final_response.error_message == f"Finished with {finish_reason}"


@pytest.mark.asyncio
async def test_generate_content_async_stream_with_thought_and_text_error_handling():
  """Test that aggregated content with thought and text preserves error information."""
  gemini_llm = Gemini(model="gemini-1.5-flash")
  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      contents=[Content(role="user", parts=[Part.from_text(text="Hello")])],
      config=types.GenerateContentConfig(
          temperature=0.1,
          response_modalities=[types.Modality.TEXT],
          system_instruction="You are a helpful assistant",
      ),
  )

  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    mock_responses = [
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model", parts=[Part(text="Think1", thought=True)]
                    ),
                    finish_reason=None,
                )
            ]
        ),
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model", parts=[Part.from_text(text="Answer")]
                    ),
                    finish_reason=types.FinishReason.MAX_TOKENS,
                    finish_message="Maximum tokens reached",
                )
            ]
        ),
    ]

    async def mock_coro():
      return MockAsyncIterator(mock_responses)

    mock_client.aio.models.generate_content_stream.return_value = mock_coro()

    responses = [
        resp
        async for resp in gemini_llm.generate_content_async(
            llm_request, stream=True
        )
    ]

    # Should have 3 responses: 2 partial and 1 final aggregated
    assert len(responses) == 3
    assert responses[0].partial is True
    assert responses[1].partial is True

    # Final response should have aggregated content with both thought and text
    final_response = responses[2]
    assert len(final_response.content.parts) == 2
    assert final_response.content.parts[0].text == "Think1"
    assert final_response.content.parts[0].thought is True
    assert final_response.content.parts[1].text == "Answer"
    # After the code changes, error_code and error_message are set for non-STOP finish reasons
    assert final_response.error_code == types.FinishReason.MAX_TOKENS
    assert final_response.error_message == "Maximum tokens reached"


@pytest.mark.asyncio
async def test_generate_content_async_stream_error_info_none_for_stop_finish_reason():
  """Test that error_code and error_message are None when finish_reason is STOP."""
  gemini_llm = Gemini(model="gemini-1.5-flash")
  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      contents=[Content(role="user", parts=[Part.from_text(text="Hello")])],
      config=types.GenerateContentConfig(
          temperature=0.1,
          response_modalities=[types.Modality.TEXT],
          system_instruction="You are a helpful assistant",
      ),
  )

  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    mock_responses = [
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model", parts=[Part.from_text(text="Hello")]
                    ),
                    finish_reason=None,
                )
            ]
        ),
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model", parts=[Part.from_text(text=" world")]
                    ),
                    finish_reason=types.FinishReason.STOP,
                    finish_message="Successfully completed",
                )
            ]
        ),
    ]

    async def mock_coro():
      return MockAsyncIterator(mock_responses)

    mock_client.aio.models.generate_content_stream.return_value = mock_coro()

    responses = [
        resp
        async for resp in gemini_llm.generate_content_async(
            llm_request, stream=True
        )
    ]

    # Should have 3 responses: 2 partial and 1 final aggregated
    assert len(responses) == 3
    assert responses[0].partial is True
    assert responses[1].partial is True

    # Final response should have aggregated content with error info None for STOP finish reason
    final_response = responses[2]
    assert final_response.content.parts[0].text == "Hello world"
    assert final_response.error_code is None
    assert final_response.error_message is None


@pytest.mark.asyncio
async def test_generate_content_async_stream_error_info_set_for_non_stop_finish_reason():
  """Test that error_code and error_message are set for non-STOP finish reasons."""
  gemini_llm = Gemini(model="gemini-1.5-flash")
  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      contents=[Content(role="user", parts=[Part.from_text(text="Hello")])],
      config=types.GenerateContentConfig(
          temperature=0.1,
          response_modalities=[types.Modality.TEXT],
          system_instruction="You are a helpful assistant",
      ),
  )

  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    mock_responses = [
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model", parts=[Part.from_text(text="Hello")]
                    ),
                    finish_reason=None,
                )
            ]
        ),
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model", parts=[Part.from_text(text=" world")]
                    ),
                    finish_reason=types.FinishReason.MAX_TOKENS,
                    finish_message="Maximum tokens reached",
                )
            ]
        ),
    ]

    async def mock_coro():
      return MockAsyncIterator(mock_responses)

    mock_client.aio.models.generate_content_stream.return_value = mock_coro()

    responses = [
        resp
        async for resp in gemini_llm.generate_content_async(
            llm_request, stream=True
        )
    ]

    # Should have 3 responses: 2 partial and 1 final aggregated
    assert len(responses) == 3
    assert responses[0].partial is True
    assert responses[1].partial is True

    # Final response should have aggregated content with error info set for non-STOP finish reason
    final_response = responses[2]
    assert final_response.content.parts[0].text == "Hello world"
    assert final_response.error_code == types.FinishReason.MAX_TOKENS
    assert final_response.error_message == "Maximum tokens reached"


@pytest.mark.asyncio
async def test_generate_content_async_stream_no_aggregated_content_without_text():
  """Test that no aggregated content is generated when there's no accumulated text."""
  gemini_llm = Gemini(model="gemini-1.5-flash")
  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      contents=[Content(role="user", parts=[Part.from_text(text="Hello")])],
      config=types.GenerateContentConfig(
          temperature=0.1,
          response_modalities=[types.Modality.TEXT],
          system_instruction="You are a helpful assistant",
      ),
  )

  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    # Mock response with no text content
    mock_responses = [
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model",
                        parts=[
                            Part(
                                function_call=types.FunctionCall(
                                    name="test", args={}
                                )
                            )
                        ],
                    ),
                    finish_reason=types.FinishReason.STOP,
                )
            ]
        ),
    ]

    async def mock_coro():
      return MockAsyncIterator(mock_responses)

    mock_client.aio.models.generate_content_stream.return_value = mock_coro()

    responses = [
        resp
        async for resp in gemini_llm.generate_content_async(
            llm_request, stream=True
        )
    ]

    # With progressive SSE streaming enabled by default, we get 2 responses:
    # 1. Partial response with function call
    # 2. Final aggregated response with function call
    assert len(responses) == 2
    # First response is partial
    assert responses[0].partial is True
    assert responses[0].content.parts[0].function_call is not None
    # Second response is the final aggregated response
    assert responses[1].partial is False
    assert responses[1].content.parts[0].function_call is not None


@pytest.mark.asyncio
async def test_generate_content_async_stream_mixed_text_function_call_text():
  """Test streaming with pattern: [text, function_call, text] to verify proper aggregation."""
  gemini_llm = Gemini(model="gemini-1.5-flash")
  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      contents=[Content(role="user", parts=[Part.from_text(text="Hello")])],
      config=types.GenerateContentConfig(
          temperature=0.1,
          response_modalities=[types.Modality.TEXT],
          system_instruction="You are a helpful assistant",
      ),
  )

  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    # Create responses with pattern: text -> function_call -> text
    mock_responses = [
        # First text chunk
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model", parts=[Part.from_text(text="First text")]
                    ),
                    finish_reason=None,
                )
            ]
        ),
        # Function call interrupts the text flow
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model",
                        parts=[
                            Part(
                                function_call=types.FunctionCall(
                                    name="test_func", args={}
                                )
                            )
                        ],
                    ),
                    finish_reason=None,
                )
            ]
        ),
        # More text after function call
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model",
                        parts=[Part.from_text(text=" second text")],
                    ),
                    finish_reason=types.FinishReason.STOP,
                )
            ]
        ),
    ]

    async def mock_coro():
      return MockAsyncIterator(mock_responses)

    mock_client.aio.models.generate_content_stream.return_value = mock_coro()

    responses = [
        resp
        async for resp in gemini_llm.generate_content_async(
            llm_request, stream=True
        )
    ]

    # With progressive SSE streaming enabled, we get 4 responses:
    # 1. Partial text "First text"
    # 2. Partial function call
    # 3. Partial text " second text"
    # 4. Final aggregated response with all parts (text + FC + text)
    assert len(responses) == 4

    # First partial text
    assert responses[0].partial is True
    assert responses[0].content.parts[0].text == "First text"

    # Partial function call
    assert responses[1].partial is True
    assert responses[1].content.parts[0].function_call is not None
    assert responses[1].content.parts[0].function_call.name == "test_func"

    # Partial second text
    assert responses[2].partial is True
    assert responses[2].content.parts[0].text == " second text"

    # Final aggregated response with all parts
    assert responses[3].partial is False
    assert len(responses[3].content.parts) == 3
    assert responses[3].content.parts[0].text == "First text"
    assert responses[3].content.parts[1].function_call.name == "test_func"
    assert responses[3].content.parts[2].text == " second text"
    assert responses[3].error_code is None  # STOP finish reason


@pytest.mark.asyncio
async def test_generate_content_async_stream_multiple_text_parts_in_single_response():
  """Test streaming with multiple text parts in a single response."""
  gemini_llm = Gemini(model="gemini-1.5-flash")
  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      contents=[Content(role="user", parts=[Part.from_text(text="Hello")])],
      config=types.GenerateContentConfig(
          temperature=0.1,
          response_modalities=[types.Modality.TEXT],
          system_instruction="You are a helpful assistant",
      ),
  )

  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    # Create a response with multiple text parts
    mock_responses = [
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model",
                        parts=[
                            Part.from_text(text="First part"),
                            Part.from_text(text=" second part"),
                        ],
                    ),
                    finish_reason=types.FinishReason.STOP,
                )
            ]
        ),
    ]

    async def mock_coro():
      return MockAsyncIterator(mock_responses)

    mock_client.aio.models.generate_content_stream.return_value = mock_coro()

    responses = [
        resp
        async for resp in gemini_llm.generate_content_async(
            llm_request, stream=True
        )
    ]

    # Should handle only the first text part in current implementation
    # Note: This test documents current behavior - the implementation only
    # looks at parts[0].text, so it would only process "First part"
    assert len(responses) >= 1
    assert responses[0].content.parts[0].text == "First part"


@pytest.mark.asyncio
async def test_generate_content_async_stream_complex_mixed_thought_text_function():
  """Test complex streaming with thought, text, and function calls mixed."""
  gemini_llm = Gemini(model="gemini-1.5-flash")
  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      contents=[Content(role="user", parts=[Part.from_text(text="Hello")])],
      config=types.GenerateContentConfig(
          temperature=0.1,
          response_modalities=[types.Modality.TEXT],
          system_instruction="You are a helpful assistant",
      ),
  )

  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    # Complex pattern: thought -> text -> function_call -> thought -> text
    mock_responses = [
        # Thought
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model",
                        parts=[Part(text="Thinking...", thought=True)],
                    ),
                    finish_reason=None,
                )
            ]
        ),
        # Regular text
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model",
                        parts=[Part.from_text(text="Here's my answer")],
                    ),
                    finish_reason=None,
                )
            ]
        ),
        # Function call
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model",
                        parts=[
                            Part(
                                function_call=types.FunctionCall(
                                    name="lookup", args={}
                                )
                            )
                        ],
                    ),
                    finish_reason=None,
                )
            ]
        ),
        # More thought
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model",
                        parts=[Part(text="More thinking...", thought=True)],
                    ),
                    finish_reason=None,
                )
            ]
        ),
        # Final text
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model",
                        parts=[Part.from_text(text=" and conclusion")],
                    ),
                    finish_reason=types.FinishReason.STOP,
                )
            ]
        ),
    ]

    async def mock_coro():
      return MockAsyncIterator(mock_responses)

    mock_client.aio.models.generate_content_stream.return_value = mock_coro()

    responses = [
        resp
        async for resp in gemini_llm.generate_content_async(
            llm_request, stream=True
        )
    ]

    # With progressive SSE streaming, we get 6 responses:
    # 5 partial responses + 1 final aggregated response
    assert len(responses) == 6

    # All but the last should be partial
    for i in range(5):
      assert responses[i].partial is True

    # Final aggregated response should have all parts
    final_response = responses[-1]
    assert final_response.partial is False
    assert final_response.error_code is None  # STOP finish reason
    # Final response aggregates: thought + text + FC + thought + text
    assert len(final_response.content.parts) == 5
    assert final_response.content.parts[0].thought is True
    assert "Thinking..." in final_response.content.parts[0].text
    assert final_response.content.parts[1].text == "Here's my answer"
    assert final_response.content.parts[2].function_call.name == "lookup"
    assert final_response.content.parts[3].thought is True
    assert "More thinking..." in final_response.content.parts[3].text
    assert final_response.content.parts[4].text == " and conclusion"


@pytest.mark.asyncio
async def test_generate_content_async_stream_two_separate_text_aggregations():
  """Test that [text, function_call, text] results in two separate text aggregations."""
  gemini_llm = Gemini(model="gemini-1.5-flash")
  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      contents=[Content(role="user", parts=[Part.from_text(text="Hello")])],
      config=types.GenerateContentConfig(
          temperature=0.1,
          response_modalities=[types.Modality.TEXT],
          system_instruction="You are a helpful assistant",
      ),
  )

  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    # Create responses: multiple text chunks -> function_call -> multiple text chunks
    mock_responses = [
        # First text accumulation (multiple chunks)
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model", parts=[Part.from_text(text="First")]
                    ),
                    finish_reason=None,
                )
            ]
        ),
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model", parts=[Part.from_text(text=" chunk")]
                    ),
                    finish_reason=None,
                )
            ]
        ),
        # Function call interrupts
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model",
                        parts=[
                            Part(
                                function_call=types.FunctionCall(
                                    name="divide", args={}
                                )
                            )
                        ],
                    ),
                    finish_reason=None,
                )
            ]
        ),
        # Second text accumulation (multiple chunks)
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model", parts=[Part.from_text(text="Second")]
                    ),
                    finish_reason=None,
                )
            ]
        ),
        types.GenerateContentResponse(
            candidates=[
                types.Candidate(
                    content=Content(
                        role="model", parts=[Part.from_text(text=" chunk")]
                    ),
                    finish_reason=types.FinishReason.STOP,
                )
            ]
        ),
    ]

    async def mock_coro():
      return MockAsyncIterator(mock_responses)

    mock_client.aio.models.generate_content_stream.return_value = mock_coro()

    responses = [
        resp
        async for resp in gemini_llm.generate_content_async(
            llm_request, stream=True
        )
    ]

    # With progressive SSE streaming, we get 6 responses:
    # 5 partial responses + 1 final aggregated response
    assert len(responses) == 6

    # All but the last should be partial
    for i in range(5):
      assert responses[i].partial is True

    # Final response should be aggregated with all parts
    final_response = responses[-1]
    assert final_response.partial is False
    assert final_response.error_code is None  # STOP finish reason
    # Final response aggregates: text1 + text2 + FC + text3 + text4
    assert len(final_response.content.parts) == 3
    assert final_response.content.parts[0].text == "First chunk"
    assert final_response.content.parts[1].function_call.name == "divide"
    assert final_response.content.parts[2].text == "Second chunk"


@pytest.mark.asyncio
async def test_computer_use_removes_system_instruction():
  """Test that system instruction is set to None when computer use is configured."""
  llm = Gemini()

  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      contents=[
          types.Content(role="user", parts=[types.Part.from_text(text="Hello")])
      ],
      config=types.GenerateContentConfig(
          system_instruction="You are a helpful assistant",
          tools=[
              types.Tool(
                  computer_use=types.ComputerUse(
                      environment=types.Environment.ENVIRONMENT_BROWSER
                  )
              )
          ],
      ),
  )

  await llm._preprocess_request(llm_request)

  # System instruction should be set to None when computer use is configured
  assert llm_request.config.system_instruction is None


@pytest.mark.asyncio
async def test_computer_use_preserves_system_instruction_when_no_computer_use():
  """Test that system instruction is preserved when computer use is not configured."""
  llm = Gemini()

  original_instruction = "You are a helpful assistant"
  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      contents=[
          types.Content(role="user", parts=[types.Part.from_text(text="Hello")])
      ],
      config=types.GenerateContentConfig(
          system_instruction=original_instruction,
          tools=[
              types.Tool(
                  function_declarations=[
                      types.FunctionDeclaration(name="test", description="test")
                  ]
              )
          ],
      ),
  )

  await llm._preprocess_request(llm_request)

  # System instruction should be preserved when no computer use
  assert llm_request.config.system_instruction == original_instruction


@pytest.mark.asyncio
async def test_computer_use_with_no_config():
  """Test that preprocessing works when config is None."""
  llm = Gemini()

  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      contents=[
          types.Content(role="user", parts=[types.Part.from_text(text="Hello")])
      ],
  )

  # Should not raise an exception
  await llm._preprocess_request(llm_request)


@pytest.mark.asyncio
async def test_computer_use_with_no_tools():
  """Test that preprocessing works when config.tools is None."""
  llm = Gemini()

  original_instruction = "You are a helpful assistant"
  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      contents=[
          types.Content(role="user", parts=[types.Part.from_text(text="Hello")])
      ],
      config=types.GenerateContentConfig(
          system_instruction=original_instruction,
          tools=None,
      ),
  )

  await llm._preprocess_request(llm_request)

  # System instruction should be preserved when no tools
  assert llm_request.config.system_instruction == original_instruction


@pytest.mark.asyncio
async def test_adapt_computer_use_tool_wait():
  """Test that _adapt_computer_use_tool correctly adapts wait to wait_5_seconds."""
  from google.adk.tools.computer_use.computer_use_tool import ComputerUseTool

  llm = Gemini()

  # Create a mock wait tool
  mock_wait_func = AsyncMock()
  mock_wait_func.return_value = "mock_result"

  original_wait_tool = ComputerUseTool(
      func=mock_wait_func,
      screen_size=(1920, 1080),
      virtual_screen_size=(1000, 1000),
  )

  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      config=types.GenerateContentConfig(),
  )

  # Add wait to tools_dict
  llm_request.tools_dict["wait"] = original_wait_tool

  # Call the adaptation method (now async)
  await llm._adapt_computer_use_tool(llm_request)

  # Verify wait was removed and wait_5_seconds was added
  assert "wait" not in llm_request.tools_dict
  assert "wait_5_seconds" in llm_request.tools_dict

  # Verify the new tool has correct properties
  wait_5_seconds_tool = llm_request.tools_dict["wait_5_seconds"]
  assert isinstance(wait_5_seconds_tool, ComputerUseTool)
  assert wait_5_seconds_tool._screen_size == (1920, 1080)
  assert wait_5_seconds_tool._coordinate_space == (1000, 1000)

  # Verify calling the new tool calls the original with 5 seconds
  result = await wait_5_seconds_tool.func()
  assert result == "mock_result"
  mock_wait_func.assert_awaited_once_with(5)


@pytest.mark.asyncio
async def test_adapt_computer_use_tool_no_wait():
  """Test that _adapt_computer_use_tool does nothing when wait is not present."""
  llm = Gemini()

  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      config=types.GenerateContentConfig(),
  )

  # Don't add any tools
  original_tools_dict = llm_request.tools_dict.copy()

  # Call the adaptation method (now async)
  await llm._adapt_computer_use_tool(llm_request)

  # Verify tools_dict is unchanged
  assert llm_request.tools_dict == original_tools_dict
  assert "wait_5_seconds" not in llm_request.tools_dict


@pytest.mark.asyncio
async def test_generate_content_async_with_cache_metadata_integration(
    gemini_llm, llm_request_with_cache, cache_metadata
):
  """Test integration between Google LLM and cache manager with proper parameter order.

  This test specifically validates that the cache manager's
  populate_cache_metadata_in_response
  method is called with the correct parameter order: (llm_response,
  cache_metadata).

  This test would have caught the parameter order bug where cache_metadata and
  llm_response
  were passed in the wrong order, causing 'CacheMetadata' object has no
  attribute 'usage_metadata' errors.
  """

  # Create a mock response with usage metadata including cached tokens
  generate_content_response = types.GenerateContentResponse(
      candidates=[
          types.Candidate(
              content=Content(
                  role="model",
                  parts=[Part.from_text(text="Hello, how can I help you?")],
              ),
              finish_reason=types.FinishReason.STOP,
          )
      ],
      usage_metadata=types.GenerateContentResponseUsageMetadata(
          prompt_token_count=1500,
          candidates_token_count=150,
          cached_content_token_count=800,  # This is the key field that was always 0 due to the bug
          total_token_count=1650,
      ),
  )

  with mock.patch.object(gemini_llm, "api_client") as mock_client:
    # Create a mock coroutine that returns the generate_content_response
    async def mock_coro():
      return generate_content_response

    mock_client.aio.models.generate_content.return_value = mock_coro()

    # Mock the cache manager module to verify correct method call
    with mock.patch(
        "google.adk.models.gemini_context_cache_manager.GeminiContextCacheManager"
    ) as MockCacheManagerClass:
      mock_cache_manager = MockCacheManagerClass.return_value
      # Configure cache manager to handle context caching
      mock_cache_manager.handle_context_caching = AsyncMock(
          return_value=cache_metadata
      )

      responses = [
          resp
          async for resp in gemini_llm.generate_content_async(
              llm_request_with_cache, stream=False
          )
      ]

      # Verify the response was processed
      assert len(responses) == 1
      response = responses[0]
      assert isinstance(response, LlmResponse)
      assert response.content.parts[0].text == "Hello, how can I help you?"

      # CRITICAL TEST: Verify populate_cache_metadata_in_response was called with correct parameter order
      mock_cache_manager.populate_cache_metadata_in_response.assert_called_once()
      call_args = (
          mock_cache_manager.populate_cache_metadata_in_response.call_args
      )

      # The first argument should be the LlmResponse (not CacheMetadata)
      first_arg = call_args[0][0]  # First positional argument
      second_arg = call_args[0][1]  # Second positional argument

      # Verify correct parameter order: (llm_response, cache_metadata)
      assert isinstance(first_arg, LlmResponse), (
          f"First parameter should be LlmResponse, got {type(first_arg)}. "
          "This indicates parameters are in wrong order."
      )
      assert isinstance(second_arg, CacheMetadata), (
          f"Second parameter should be CacheMetadata, got {type(second_arg)}. "
          "This indicates parameters are in wrong order."
      )

      # Verify the LlmResponse has the expected usage metadata
      assert first_arg.usage_metadata is not None
      assert first_arg.usage_metadata.cached_content_token_count == 800
      assert first_arg.usage_metadata.prompt_token_count == 1500
      assert first_arg.usage_metadata.candidates_token_count == 150

      # Verify cache metadata is preserved
      assert second_arg.cache_name == cache_metadata.cache_name
      assert second_arg.invocations_used == cache_metadata.invocations_used


def test_build_function_declaration_log():
  """Test that _build_function_declaration_log formats function declarations correctly."""
  # Test case 1: Function with parameters and response
  func_decl1 = types.FunctionDeclaration(
      name="test_func1",
      description="Test function 1",
      parameters=types.Schema(
          type=types.Type.OBJECT,
          properties={
              "param1": types.Schema(
                  type=types.Type.STRING, description="param1 desc"
              )
          },
      ),
      response=types.Schema(type=types.Type.BOOLEAN, description="return bool"),
  )
  log1 = _build_function_declaration_log(func_decl1)
  assert log1 == (
      "test_func1: {'param1': {'description': 'param1 desc', 'type':"
      " <Type.STRING: 'STRING'>}} -> {'description': 'return bool', 'type':"
      " <Type.BOOLEAN: 'BOOLEAN'>}"
  )

  # Test case 2: Function with JSON schema parameters and response
  func_decl2 = types.FunctionDeclaration(
      name="test_func2",
      description="Test function 2",
      parameters_json_schema={
          "type": "object",
          "properties": {"param2": {"type": "integer"}},
      },
      response_json_schema={"type": "string"},
  )
  log2 = _build_function_declaration_log(func_decl2)
  assert log2 == (
      "test_func2: {'type': 'object', 'properties': {'param2': {'type':"
      " 'integer'}}} -> {'type': 'string'}"
  )

  # Test case 3: Function with no parameters and no response
  func_decl3 = types.FunctionDeclaration(
      name="test_func3",
      description="Test function 3",
  )
  log3 = _build_function_declaration_log(func_decl3)
  assert log3 == "test_func3: {} "


def test_build_request_log_with_config_multiple_tool_types():
  """Test that _build_request_log includes config with multiple tool types."""
  func_decl = types.FunctionDeclaration(
      name="test_function",
      description="A test function",
      parameters={"type": "object", "properties": {}},
  )

  tool = types.Tool(
      function_declarations=[func_decl],
      google_search=types.GoogleSearch(),
      code_execution=types.ToolCodeExecution(),
  )

  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      contents=[Content(role="user", parts=[Part.from_text(text="Hello")])],
      config=types.GenerateContentConfig(
          temperature=0.7,
          max_output_tokens=500,
          system_instruction="You are a helpful assistant",
          tools=[tool],
      ),
  )

  log_output = _build_request_log(llm_request)

  # Verify config section exists
  assert "Config:" in log_output

  # Verify config contains expected fields (using Python dict format with single quotes)
  assert "'temperature': 0.7" in log_output
  assert "'max_output_tokens': 500" in log_output

  # Verify config contains other tool types (not function_declarations)
  assert "'google_search'" in log_output
  assert "'code_execution'" in log_output

  # Verify function_declarations is NOT in config section
  # (it should only be in the Functions section)
  config_section = log_output.split("Functions:")[0]
  assert "'function_declarations'" not in config_section

  # Verify function is in Functions section
  assert "Functions:" in log_output
  assert "test_function" in log_output

  # Verify system instruction is NOT in config section
  assert (
      "'system_instruction'"
      not in log_output.split("Contents:")[0].split("Config:")[1]
  )


def test_build_request_log_function_declarations_in_second_tool():
  """Test that function_declarations in non-first tool are handled correctly."""
  func_decl = types.FunctionDeclaration(
      name="my_function",
      description="A test function",
      parameters={"type": "object", "properties": {}},
  )

  # First tool has only google_search
  tool1 = types.Tool(google_search=types.GoogleSearch())

  # Second tool has function_declarations
  tool2 = types.Tool(
      function_declarations=[func_decl],
      code_execution=types.ToolCodeExecution(),
  )

  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      contents=[Content(role="user", parts=[Part.from_text(text="Hello")])],
      config=types.GenerateContentConfig(
          temperature=0.5,
          system_instruction="You are a helpful assistant",
          tools=[tool1, tool2],
      ),
  )

  log_output = _build_request_log(llm_request)

  # Verify function is in Functions section
  assert "Functions:" in log_output
  assert "my_function" in log_output

  # Verify function_declarations is NOT in config section
  config_section = log_output.split("Functions:")[0]
  assert "'function_declarations'" not in config_section

  # Verify both tools are in config but without function_declarations (Python dict format)
  assert "'google_search'" in log_output
  assert "'code_execution'" in log_output

  # Verify config has the expected structure without parsing
  config_section = log_output.split("Config:")[1].split("---")[0]
  # Should have 2 tools (two dict entries in the tools list)
  assert config_section.count("'google_search'") == 1
  assert config_section.count("'code_execution'") == 1
  # Function declarations should NOT be in config section
  assert "'function_declarations'" not in config_section


def test_build_request_log_fallback_to_repr_on_all_failures(monkeypatch):
  """Test that _build_request_log falls back to repr() if model_dump fails."""

  llm_request = LlmRequest(
      model="gemini-1.5-flash",
      contents=[Content(role="user", parts=[Part.from_text(text="Hello")])],
      config=types.GenerateContentConfig(
          temperature=0.7,
          system_instruction="You are a helpful assistant",
      ),
  )

  # Mock model_dump at class level to raise exception
  def mock_model_dump(*args, **kwargs):
    raise Exception("dump failed")

  monkeypatch.setattr(
      types.GenerateContentConfig, "model_dump", mock_model_dump
  )

  log_output = _build_request_log(llm_request)

  # Should still succeed using repr()
  assert "Config:" in log_output
  assert "GenerateContentConfig" in log_output


@pytest.mark.asyncio
async def test_connect_uses_gemini_speech_config_when_request_is_none(
    gemini_llm, llm_request
):
  """Tests that Gemini's speech_config is used when live_connect_config's is None."""
  # Arrange: Set a speech_config on the Gemini instance with the voice "Kore"
  gemini_llm.speech_config = types.SpeechConfig(
      voice_config=types.VoiceConfig(
          prebuilt_voice_config=types.PrebuiltVoiceConfig(
              voice_name="Kore",
          )
      )
  )
  llm_request.live_connect_config = (
      types.LiveConnectConfig()
  )  # speech_config is None

  mock_live_session = mock.AsyncMock()

  with mock.patch.object(gemini_llm, "_live_api_client") as mock_live_client:

    class MockLiveConnect:

      async def __aenter__(self):
        return mock_live_session

      async def __aexit__(self, *args):
        pass

    mock_live_client.aio.live.connect.return_value = MockLiveConnect()

    # Act
    async with gemini_llm.connect(llm_request) as connection:
      # Assert
      mock_live_client.aio.live.connect.assert_called_once()
      call_args = mock_live_client.aio.live.connect.call_args
      config_arg = call_args.kwargs["config"]

      # Verify the speech_config from the Gemini instance was used
      assert config_arg.speech_config is not None
      assert (
          config_arg.speech_config.voice_config.prebuilt_voice_config.voice_name
          == "Kore"
      )
      assert isinstance(connection, GeminiLlmConnection)


@pytest.mark.asyncio
async def test_connect_uses_request_speech_config_when_gemini_is_none(
    gemini_llm, llm_request
):
  """Tests that request's speech_config is used when Gemini's is None."""
  # Arrange: Set a speech_config on the request instance with the voice "Kore"
  gemini_llm.speech_config = None
  request_speech_config = types.SpeechConfig(
      voice_config=types.VoiceConfig(
          prebuilt_voice_config=types.PrebuiltVoiceConfig(
              voice_name="Kore",
          )
      )
  )
  llm_request.live_connect_config = types.LiveConnectConfig(
      speech_config=request_speech_config
  )

  mock_live_session = mock.AsyncMock()

  with mock.patch.object(gemini_llm, "_live_api_client") as mock_live_client:

    class MockLiveConnect:

      async def __aenter__(self):
        return mock_live_session

      async def __aexit__(self, *args):
        pass

    mock_live_client.aio.live.connect.return_value = MockLiveConnect()

    # Act
    async with gemini_llm.connect(llm_request) as connection:
      # Assert
      mock_live_client.aio.live.connect.assert_called_once()
      call_args = mock_live_client.aio.live.connect.call_args
      config_arg = call_args.kwargs["config"]

      # Verify the speech_config from the request instance was used
      assert config_arg.speech_config is not None
      assert (
          config_arg.speech_config.voice_config.prebuilt_voice_config.voice_name
          == "Kore"
      )
      assert isinstance(connection, GeminiLlmConnection)


@pytest.mark.asyncio
async def test_connect_request_gemini_config_overrides_speech_config(
    gemini_llm, llm_request
):
  """Tests that live_connect_config's speech_config is preserved even if Gemini has one."""
  # Arrange: Set different speech_configs on both the Gemini instance ("Puck") and the request ("Zephyr")
  gemini_llm.speech_config = types.SpeechConfig(
      voice_config=types.VoiceConfig(
          prebuilt_voice_config=types.PrebuiltVoiceConfig(
              voice_name="Puck",
          )
      )
  )
  request_speech_config = types.SpeechConfig(
      voice_config=types.VoiceConfig(
          prebuilt_voice_config=types.PrebuiltVoiceConfig(
              voice_name="Zephyr",
          )
      )
  )
  llm_request.live_connect_config = types.LiveConnectConfig(
      speech_config=request_speech_config
  )

  mock_live_session = mock.AsyncMock()

  with mock.patch.object(gemini_llm, "_live_api_client") as mock_live_client:

    class MockLiveConnect:

      async def __aenter__(self):
        return mock_live_session

      async def __aexit__(self, *args):
        pass

    mock_live_client.aio.live.connect.return_value = MockLiveConnect()

    # Act
    async with gemini_llm.connect(llm_request) as connection:
      # Assert
      mock_live_client.aio.live.connect.assert_called_once()
      call_args = mock_live_client.aio.live.connect.call_args
      config_arg = call_args.kwargs["config"]

      # Verify the speech_config from the request ("Zephyr") was overwritten by Gemini's speech_config ("Puck")
      assert config_arg.speech_config is not None
      assert (
          config_arg.speech_config.voice_config.prebuilt_voice_config.voice_name
          == "Puck"
      )
      assert isinstance(connection, GeminiLlmConnection)


@pytest.mark.asyncio
async def test_connect_speech_config_remains_none_when_both_are_none(
    gemini_llm, llm_request
):
  """Tests that speech_config is None when neither Gemini nor the request has it."""
  # Arrange: Ensure both Gemini instance and request have no speech_config
  gemini_llm.speech_config = None
  llm_request.live_connect_config = (
      types.LiveConnectConfig()
  )  # speech_config is None

  mock_live_session = mock.AsyncMock()

  with mock.patch.object(gemini_llm, "_live_api_client") as mock_live_client:

    class MockLiveConnect:

      async def __aenter__(self):
        return mock_live_session

      async def __aexit__(self, *args):
        pass

    mock_live_client.aio.live.connect.return_value = MockLiveConnect()

    # Act
    async with gemini_llm.connect(llm_request) as connection:
      # Assert
      mock_live_client.aio.live.connect.assert_called_once()
      call_args = mock_live_client.aio.live.connect.call_args
      config_arg = call_args.kwargs["config"]

      # Verify the final speech_config is still None
      assert config_arg.speech_config is None
      assert isinstance(connection, GeminiLlmConnection)