feat: 实现日历提醒完整功能（操作执行、通知服务重构、归档）

- 新增 ReminderActionExecutor 处理取消/稍后提醒操作 - 新增 ReminderOutboxStore 本地存储待处理操作 - 重构 LocalNotificationService 支持聚合提醒和交互操作 - 新增 event_color_resolver 工具类统一颜色解析 - 新增 CalendarService.archiveEvent 归档方法 - 增强 ModelTracking 支持缓存命中、推理token和成本追踪 - 添加 qwen3.5-35b-a3b 模型配置 - 更新 AndroidManifest 全屏intent权限 - 补充相关单元测试和文档
2026-03-18 19:12:47 +08:00
parent 257cb0f5d5
commit 00f37d7e19
35 changed files with 2676 additions and 244 deletions
@@ -0,0 +1,126 @@
+from __future__ import annotations
+
+from typing import Any, cast
+
+from core.agentscope.runtime.model_tracking import TrackingChatModel
+
+
+class _FakeMetadata:
+    def model_dump(self) -> dict[str, object]:
+        return {
+            "prompt_tokens": 120,
+            "completion_tokens": 30,
+            "total_tokens": 150,
+            "prompt_tokens_details": {
+                "cached_tokens": 80,
+            },
+            "prompt_cache_hit_tokens": 80,
+            "prompt_cache_miss_tokens": 40,
+            "completion_tokens_details": {
+                "reasoning_tokens": 5,
+            },
+        }
+
+
+class _FakeUsage:
+    input_tokens = 120
+    output_tokens = 30
+    time = 1.234
+    metadata = _FakeMetadata()
+
+
+class _FakeResponse:
+    usage = _FakeUsage()
+
+
+class _FakeModel:
+    stream = False
+
+    async def __call__(self, *args: object, **kwargs: object) -> _FakeResponse:
+        return _FakeResponse()
+
+
+class _FakeUsageWithProviderCost:
+    input_tokens = 50
+    output_tokens = 10
+    time = 0.5
+    cost = 0.0123
+    metadata = _FakeMetadata()
+
+
+class _FakeResponseWithProviderCost:
+    usage = _FakeUsageWithProviderCost()
+
+
+class _FakeModelWithProviderCost:
+    stream = False
+
+    async def __call__(
+        self, *args: object, **kwargs: object
+    ) -> _FakeResponseWithProviderCost:
+        return _FakeResponseWithProviderCost()
+
+
+class _FakeResponseWithoutUsage:
+    usage = None
+
+
+class _FakeModelWithoutUsage:
+    stream = False
+
+    async def __call__(
+        self, *args: object, **kwargs: object
+    ) -> _FakeResponseWithoutUsage:
+        return _FakeResponseWithoutUsage()
+
+
+async def test_tracking_chat_model_collects_primary_usage_fields() -> None:
+    model = TrackingChatModel(cast(Any, _FakeModel()))
+
+    await model("prompt")
+
+    summary = model.usage_summary()
+    assert summary["input_tokens"] == 120
+    assert summary["output_tokens"] == 30
+    assert summary["total_tokens"] == 150
+    assert summary["latency_ms"] == 1234
+    assert summary["cached_prompt_tokens"] == 80
+    assert summary["prompt_cache_hit_tokens"] == 80
+    assert summary["prompt_cache_miss_tokens"] == 40
+    assert summary["reasoning_tokens"] == 5
+    assert summary["direct_cost"] == 0.0
+    assert summary["direct_cost_observed"] == 0
+    assert summary["direct_cost_complete"] == 0
+    assert summary["model_call_records"] == 1
+    assert summary["usage_records"] == 1
+    assert summary["direct_cost_records"] == 0
+    assert summary["cost_source"] == "catalog_fallback"
+
+
+async def test_tracking_chat_model_prefers_provider_cost_when_available() -> None:
+    model = TrackingChatModel(cast(Any, _FakeModelWithProviderCost()))
+
+    await model("prompt")
+
+    summary = model.usage_summary()
+    assert summary["direct_cost"] == 0.0123
+    assert summary["direct_cost_observed"] == 1
+    assert summary["direct_cost_complete"] == 1
+    assert summary["model_call_records"] == 1
+    assert summary["usage_records"] == 1
+    assert summary["direct_cost_records"] == 1
+    assert summary["cost_source"] == "provider"
+
+
+async def test_tracking_chat_model_marks_direct_cost_incomplete_when_usage_missing() -> (
+    None
+):
+    model = TrackingChatModel(cast(Any, _FakeModelWithoutUsage()))
+
+    await model("prompt")
+
+    summary = model.usage_summary()
+    assert summary["model_call_records"] == 1
+    assert summary["usage_records"] == 0
+    assert summary["direct_cost_records"] == 0
+    assert summary["direct_cost_complete"] == 0
@@ -44,10 +44,75 @@ def test_build_usage_metadata_calculates_cost_from_usage_summary() -> None:
        },
    )

-    assert metadata == {
-        "model": "dashscope/qwen3.5-flash",
-        "inputTokens": 2000,
-        "outputTokens": 100,
-        "cost": pytest.approx(0.00051),
-        "latencyMs": 321,
-    }
+    assert metadata["model"] == "dashscope/qwen3.5-flash"
+    assert metadata["inputTokens"] == 2000
+    assert metadata["outputTokens"] == 100
+    assert metadata["totalTokens"] == 2100
+    assert metadata["cachedPromptTokens"] == 500
+    assert metadata["promptCacheHitTokens"] == 500
+    assert metadata["promptCacheMissTokens"] == 1500
+    assert metadata["reasoningTokens"] == 0
+    assert metadata["cost"] == pytest.approx(0.00051)
+    assert metadata["costSource"] == "catalog_fallback"
+    assert metadata["usageComplete"] is True
+    assert metadata["latencyMs"] == 321
+
+
+def test_build_usage_metadata_prefers_provider_direct_cost() -> None:
+    service = LiteLLMService()
+
+    metadata = service.build_usage_metadata(
+        model="deepseek-chat",
+        usage_summary={
+            "input_tokens": 1000,
+            "output_tokens": 100,
+            "latency_ms": 100,
+            "cached_prompt_tokens": 0,
+            "direct_cost": 0.1234,
+            "direct_cost_observed": 1,
+            "direct_cost_complete": 1,
+        },
+    )
+
+    assert metadata["cost"] == pytest.approx(0.1234)
+    assert metadata["costSource"] == "provider"
+    assert metadata["usageComplete"] is True
+
+
+def test_build_usage_metadata_falls_back_when_provider_cost_incomplete() -> None:
+    service = LiteLLMService()
+
+    metadata = service.build_usage_metadata(
+        model="deepseek-chat",
+        usage_summary={
+            "input_tokens": 1000,
+            "output_tokens": 100,
+            "latency_ms": 100,
+            "cached_prompt_tokens": 0,
+            "direct_cost": 0.1234,
+            "direct_cost_observed": 1,
+            "direct_cost_complete": 0,
+        },
+    )
+
+    assert metadata["cost"] == pytest.approx(0.0023)
+    assert metadata["costSource"] == "catalog_fallback_incomplete_provider_cost"
+
+
+def test_build_usage_metadata_marks_incomplete_usage_fallback() -> None:
+    service = LiteLLMService()
+
+    metadata = service.build_usage_metadata(
+        model="deepseek-chat",
+        usage_summary={
+            "input_tokens": 0,
+            "output_tokens": 0,
+            "latency_ms": 0,
+            "cached_prompt_tokens": 0,
+            "model_call_records": 1,
+            "usage_records": 0,
+        },
+    )
+
+    assert metadata["costSource"] == "incomplete_usage_fallback"
+    assert metadata["usageComplete"] is False
@@ -4,6 +4,7 @@ from uuid import UUID, uuid4

 import pytest
 from fastapi import HTTPException
+from sqlalchemy.exc import SQLAlchemyError

 from core.auth.models import CurrentUser
 from models.schedule_items import (
@@ -13,6 +14,7 @@ from models.schedule_items import (
 )
 from v1.schedule_items.schemas import (
    ScheduleItemCreateRequest,
+    ScheduleItemListRequest,
    ScheduleItemMetadata,
    ScheduleItemUpdateRequest,
 )
@@ -43,6 +45,7 @@ def _create_mock_schedule_item(
 class FakeRepo:
    def __init__(self, item: ScheduleItem | None) -> None:
        self._item = item
+        self.archive_expired_called = 0

    async def get_by_item_id(
        self, item_id: UUID, owner_id: UUID
@@ -89,8 +92,9 @@ class FakeRepo:
        *,
        page: int,
        page_size: int,
+        query: str | None = None,
    ) -> tuple[list[ScheduleItem], int]:
-        del owner_id, page, page_size
+        del owner_id, page, page_size, query
        return ([self._item] if self._item else [], 1 if self._item else 0)

    async def create_subscription(self, data: dict):
@@ -104,7 +108,20 @@ class FakeRepo:
        end_at: datetime,
    ):
        del subscriber_id, start_at, end_at
-        return []
+        if self._item is None:
+            return []
+        subscription = MagicMock()
+        subscription.permission = 1
+        return [(self._item, subscription)]
+
+    async def archive_expired_subscribed_items(
+        self,
+        subscriber_id: UUID,
+        now_at: datetime,
+    ) -> int:
+        del subscriber_id, now_at
+        self.archive_expired_called += 1
+        return 0

    async def get_user_subscriptions(self, subscriber_id: UUID):
        del subscriber_id
@@ -376,3 +393,110 @@ async def test_update_maps_null_metadata_to_extra_metadata_null(
    assert "extra_metadata" in captured
    assert captured["extra_metadata"] is None
    assert "metadata" not in captured
+
+
+@pytest.mark.asyncio
+async def test_list_by_date_range_archives_expired_before_query(
+    mock_session: AsyncMock,
+    mock_inbox_repository: MagicMock,
+) -> None:
+    user_id = UUID("00000000-0000-0000-0000-000000000001")
+    item = _create_mock_schedule_item()
+    repo = FakeRepo(item)
+    service = ScheduleItemService(
+        repository=repo,
+        session=mock_session,
+        current_user=CurrentUser(id=user_id),
+        inbox_repository=mock_inbox_repository,
+    )
+
+    await service.list_by_date_range(
+        request=ScheduleItemListRequest(
+            start_at=datetime(2026, 2, 1, 0, 0, tzinfo=timezone.utc),
+            end_at=datetime(2026, 3, 1, 0, 0, tzinfo=timezone.utc),
+        ),
+    )
+
+    assert repo.archive_expired_called == 1
+
+
+@pytest.mark.asyncio
+async def test_list_by_date_range_commits_when_archived_changed(
+    mock_session: AsyncMock,
+    mock_inbox_repository: MagicMock,
+) -> None:
+    user_id = UUID("00000000-0000-0000-0000-000000000001")
+    item = _create_mock_schedule_item()
+
+    class ArchiveRepo(FakeRepo):
+        async def archive_expired_subscribed_items(
+            self,
+            subscriber_id: UUID,
+            now_at: datetime,
+        ) -> int:
+            del subscriber_id, now_at
+            self.archive_expired_called += 1
+            return 2
+
+    repo = ArchiveRepo(item)
+    service = ScheduleItemService(
+        repository=repo,
+        session=mock_session,
+        current_user=CurrentUser(id=user_id),
+        inbox_repository=mock_inbox_repository,
+    )
+
+    await service.list_by_date_range(
+        request=ScheduleItemListRequest(
+            start_at=datetime(2026, 2, 1, 0, 0, tzinfo=timezone.utc),
+            end_at=datetime(2026, 3, 1, 0, 0, tzinfo=timezone.utc),
+        ),
+    )
+
+    mock_session.commit.assert_awaited_once()
+
+
+@pytest.mark.asyncio
+async def test_list_by_date_range_rolls_back_when_query_fails_after_archive(
+    mock_session: AsyncMock,
+    mock_inbox_repository: MagicMock,
+) -> None:
+    user_id = UUID("00000000-0000-0000-0000-000000000001")
+    item = _create_mock_schedule_item()
+
+    class FailingRepo(FakeRepo):
+        async def archive_expired_subscribed_items(
+            self,
+            subscriber_id: UUID,
+            now_at: datetime,
+        ) -> int:
+            del subscriber_id, now_at
+            return 1
+
+        async def list_subscribed_items_by_date_range(
+            self,
+            subscriber_id: UUID,
+            start_at: datetime,
+            end_at: datetime,
+        ):
+            del subscriber_id, start_at, end_at
+            raise SQLAlchemyError("db unavailable")
+
+    service = ScheduleItemService(
+        repository=FailingRepo(item),
+        session=mock_session,
+        current_user=CurrentUser(id=user_id),
+        inbox_repository=mock_inbox_repository,
+    )
+
+    with pytest.raises(HTTPException) as exc_info:
+        await service.list_by_date_range(
+            request=ScheduleItemListRequest(
+                start_at=datetime(2026, 2, 1, 0, 0, tzinfo=timezone.utc),
+                end_at=datetime(2026, 3, 1, 0, 0, tzinfo=timezone.utc),
+            ),
+        )
+
+    assert exc_info.value.status_code == 503
+    mock_session.rollback.assert_awaited_once()
+    mock_session.commit.assert_not_awaited()