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

backend/tests/unit/core/agentscope/runtime/test_model_tracking.py

@@ -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