106 lines
3.8 KiB
Python
106 lines
3.8 KiB
Python
"""Tests for CommitDaemon and MultiModelLlamaClient queue integration.
|
|
|
|
Verifies that auto-commit-service sends batch priority, stay_warm=0,
|
|
and cooldown=60 to the model-boss coordinator via InferenceClient.
|
|
"""
|
|
|
|
from unittest.mock import AsyncMock, MagicMock, patch
|
|
|
|
import pytest
|
|
|
|
from auto_commit_service.config import AutoCommitSettings
|
|
|
|
|
|
@pytest.fixture
|
|
def settings():
|
|
"""Minimal settings for testing."""
|
|
return AutoCommitSettings(
|
|
service_name="test-daemon",
|
|
reasoning_model_id="ministral-14b-reasoning",
|
|
instruct_model_id="ministral-3b-instruct",
|
|
llm_timeout=30.0,
|
|
cycle_interval_seconds=1,
|
|
enabled=False,
|
|
)
|
|
|
|
|
|
class TestMultiModelClientQueueParams:
|
|
"""MultiModelLlamaClient passes queue params to InferenceClient."""
|
|
|
|
def test_default_client_id(self) -> None:
|
|
from auto_commit_service.llm.multi_model_client import MultiModelLlamaClient
|
|
|
|
client = MultiModelLlamaClient()
|
|
assert client._client_id == "auto-commit-service"
|
|
assert client._client._client_id == "auto-commit-service"
|
|
|
|
def test_custom_client_id(self) -> None:
|
|
from auto_commit_service.llm.multi_model_client import MultiModelLlamaClient
|
|
|
|
client = MultiModelLlamaClient(client_id="test-service")
|
|
assert client._client_id == "test-service"
|
|
assert client._client._client_id == "test-service"
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_chat_passes_queue_params(self) -> None:
|
|
"""_chat() sends keep_alive and context through to InferenceClient."""
|
|
from auto_commit_service.llm.multi_model_client import MultiModelLlamaClient
|
|
|
|
client = MultiModelLlamaClient(client_id="test-svc")
|
|
client._client = MagicMock()
|
|
client._client.chat = AsyncMock(return_value="response text")
|
|
|
|
result = await client._chat(
|
|
"ministral-3b-instruct",
|
|
[{"role": "user", "content": "test"}],
|
|
max_tokens=100,
|
|
)
|
|
|
|
assert result == "response text"
|
|
client._client.chat.assert_called_once_with(
|
|
model="ministral-3b-instruct",
|
|
messages=[{"role": "user", "content": "test"}],
|
|
max_tokens=100,
|
|
temperature=client._temperature,
|
|
keep_alive=300,
|
|
context=None,
|
|
)
|
|
|
|
|
|
class TestDaemonSimplifiedLoop:
|
|
"""Daemon loop no longer manages model lifecycle directly."""
|
|
|
|
def _make_daemon(self, settings):
|
|
mock_client = MagicMock()
|
|
mock_client.is_available = AsyncMock(return_value=True)
|
|
mock_client.ensure_services = AsyncMock()
|
|
mock_client.reasoning_model_id = settings.reasoning_model_id
|
|
mock_client.instruct_model_id = settings.instruct_model_id
|
|
|
|
from auto_commit_service.scheduler.daemon import CommitDaemon
|
|
|
|
return CommitDaemon(settings=settings, llm_client=mock_client)
|
|
|
|
def test_no_lifecycle_attrs(self, settings) -> None:
|
|
"""Daemon no longer has _models_loaded_since or _models_released_at."""
|
|
daemon = self._make_daemon(settings)
|
|
assert not hasattr(daemon, "_models_loaded_since")
|
|
assert not hasattr(daemon, "_models_released_at")
|
|
|
|
def test_no_lifecycle_methods(self, settings) -> None:
|
|
"""Daemon no longer has _should_force_release or _is_in_cooldown."""
|
|
daemon = self._make_daemon(settings)
|
|
assert not hasattr(daemon, "_should_force_release")
|
|
assert not hasattr(daemon, "_is_in_cooldown")
|
|
|
|
|
|
class TestConfigRemovedFields:
|
|
"""model_max_alive_seconds and model_cooldown_seconds are removed."""
|
|
|
|
def test_no_model_max_alive_seconds(self) -> None:
|
|
settings = AutoCommitSettings(service_name="test")
|
|
assert not hasattr(settings, "model_max_alive_seconds")
|
|
|
|
def test_no_model_cooldown_seconds(self) -> None:
|
|
settings = AutoCommitSettings(service_name="test")
|
|
assert not hasattr(settings, "model_cooldown_seconds")
|