feat(factor): tears报告方案A原生重构+「加入对比」做实(用户08-29拍板) [vps]

①后端tears序列化:sanguo_factor/tears_data.py新模块,alphalens已算分层序列(日度IC/月度聚合/十分组累计净值/多空Q10−Q1净值+最大回撤/去重叠年化/因子秩自相关)在分析时序列化为{factor}_tears.json,与tearsheet同源;FactorReport加tears_paths,GET /task/{id}/tears/{factor}端点(token header),_persist_factor同步落DB
②前端tears页:TearsPanel.vue按设计稿tab①——指标条7格+月度IC柱(红正绿负)/累计IC线双轴+月度IC热力图(年×月,CSS格)+分组累计净值Q1/Q5/Q10+多空净值(琥珀+面积+○最大回撤标注)+十分组年化(±5%虚线)+IC衰减(1/5/10D),1/5/10D全页联动;Result.vue的iframe→原生渲染;旧任务404自动回退iframe旧alphalens报告
③加入对比(设计稿tab②纯前端):factorCompare store(localStorage持久化,2~6个)+排行榜行内「＋对比/✓已选」列+详情页死按钮做实(选中青色态)+全局底部托盘CompareTray(chips可删/清空/对比N因子→)+对比页Compare.vue(指标并排·行最优青色高亮/累计IC叠加多线/月度IC序列Pearson相关性矩阵前端算/十分组小倍数SVG)

测试:tears_data纯函数5+全链真实alphalens6(合成因子IC>0/分层单调/JSON可序列化)+analyzer写盘/容错2+端点401/404/200共3;factor+api+orchestrator 317全绿;npm run build绿

frontend/src/api/factor.ts

@@ -34,6 +34,40 @@ export function reportUrl(taskId: string, factor: string): string {
   return `/api/v1/task/${taskId}/report/${factor}?token=${encodeURIComponent(auth.token ?? '')}`
 }
 
+// —— Tears 分层序列(方案A:分析时序列化,前端 ECharts 暗色渲染) ——
+
+export interface TearsPeriodData {
+  count: number
+  ic_mean: number | null
+  ic_std: number | null
+  icir: number | null
+  t_stat: number | null
+  win_rate: number | null
+  ic_dates: string[]
+  ic_values: number[]
+  monthly_ic: MonthlyIcPoint[]
+  quantile_keys: string[]
+  nav_dates: string[]
+  quantile_nav: Record<string, number[]>
+  quantile_annual: Record<string, number>
+  ls_nav: number[]
+  ls_annual: number
+  ls_max_dd: number
+}
+
+export interface TearsData {
+  factor: string
+  generated_at?: string
+  factor_autocorr: number | null
+  periods: Record<string, TearsPeriodData>
+}
+
+export async function getTearsData(taskId: string, factor: string): Promise<TearsData> {
+  const { data } = await apiClient.get<TearsData>(`/task/${taskId}/tears/${factor}`)
+  return data
+}
+
+
 // —— Factor Evaluation API ——
 
 export interface EvalRun {