diff --git a/ROADMAP.md b/ROADMAP.md
index 2db92b4..419f4a0 100644
--- a/ROADMAP.md
+++ b/ROADMAP.md
@@ -507,6 +507,7 @@ services/api.ts (1324L)
 | Q33 | 前端巨型组件拆分（DashboardRecent 518L→25L (-95%)，提取 dashboard-recent/ 7 个文件 (types 48L + shared-components(StatusDot/SeverityBadge) 35L + AlertMessagesCard 72L + CertificateExpiryCard 99L + ReleaseHistoryCard 99L + OpsOverviewCard 77L + KubernetesClusterCard 97L)，主文件改为 barrel re-export 保持向后兼容，tsc -b=0，npm build✓，ESLint=0/0/0，go build=0，17/17 测试通过） | P0 | **done** |
 | Q34 | 前端巨型组件拆分（AIOpsAssistant 505L→288L (-43%)，提取 aiops-assistant/ 5 个文件 (types 39L Message/QuickAction/defaultConfig/fmtTime + context-loaders.ts 69L loadAlerts/loadMetrics/loadReport 3 个平台数据加载纯函数 + quick-actions.tsx 17L + ChatMessage 47L 消息气泡(含 streaming/thinking/error 状态) + AssistantSidebar 76L 快捷操作+能力卡片)，tsc -b=0，npm build✓，ESLint=0/0/0，go build=0，17/17 测试通过） | P0 | **done** |
 | Q35 | 前端巨型组件拆分（AgentCLI 521L→399L (-23%)，提取 agent-cli/ 2 个文件 (event-handler.ts 149L createWsEventHandler 工厂函数封装 8 种 WS 事件分发+高危命令确认 + ChatInputArea.tsx 44L 受控输入区组件)，导出 ConfirmOptions 接口供跨模块引用，tsc -b=0，npm build✓，ESLint=0/66/0，go build=0，17/17 测试通过。最后一个 ≥500L 业务组件拆分完成，剩余最大业务组件 UnifiedQueryPanel 529L(已从 682L 拆分)） | P0 | **done** |
+| Q36 | 前端巨型工具文件拆分（predictUtils.ts 716L→49L barrel re-export (-93%)，提取 predict-types.ts 140L 类型+常量 + capacity-predict.ts 194L 容量预测函数(linearRegression+computePredictions+buildCapacityInsightPrompt) + cost-predict.ts 434L 云成本预测函数(computeCostPredictions+buildCostInsightPrompt)，3 个消费者 import 路径无需修改，tsc -b=0，npm build✓ (17.26s)，ESLint=0/66/0，go build=0，17/17 测试通过） | P0 | **done** |
 
 ---
 
diff --git a/frontend/src/components/aiops/capacity-predict.ts b/frontend/src/components/aiops/capacity-predict.ts
new file mode 100644
index 0000000..3169bf8
--- /dev/null
+++ b/frontend/src/components/aiops/capacity-predict.ts
@@ -0,0 +1,194 @@
+import type { JsonData } from '@/types/common';
+import i18n from '@/i18n';
+import {
+  METRIC_DEFS,
+  TIME_RANGE_LABELS,
+  TIME_RANGE_HOURS,
+  type PredictionMetric,
+  type CapacityForecast,
+  type ComputePredictionsInput,
+  type ComputedPredictions,
+} from './predict-types';
+
+/**
+ * 容量预测 — 线性回归 + 指标预测 + 容量规划 + AI 解读 prompt
+ *
+ * 从 predictUtils.ts (716L) 提取。
+ * 所有函数均为纯函数（输入 → 输出，无副作用），便于单元测试和复用。
+ * // Logan.Li
+ */
+
+// ============ 纯函数工具 ============
+
+/** 简单最小二乘线性回归，返回斜率和 R² */
+export function linearRegression(values: number[]): { slope: number; r2: number } {
+  const n = values.length;
+  if (n < 2) return { slope: 0, r2: 0 };
+  const xs = values.map((_, i) => i);
+  const meanX = xs.reduce((a, b) => a + b, 0) / n;
+  const meanY = values.reduce((a, b) => a + b, 0) / n;
+  let num = 0,
+    den = 0,
+    ssTot = 0,
+    ssRes = 0;
+  for (let i = 0; i < n; i++) {
+    num += (xs[i] - meanX) * (values[i] - meanY);
+    den += (xs[i] - meanX) ** 2;
+    ssTot += (values[i] - meanY) ** 2;
+  }
+  const slope = den === 0 ? 0 : num / den;
+  const intercept = meanY - slope * meanX;
+  for (let i = 0; i < n; i++) {
+    const pred = intercept + slope * xs[i];
+    ssRes += (values[i] - pred) ** 2;
+  }
+  const r2 = ssTot === 0 ? 1 : Math.max(0, 1 - ssRes / ssTot);
+  return { slope, r2 };
+}
+
+/** 从时序指标中提取按时间排序的数值序列 */
+export function extractSeries(metrics: JsonData[], name: string): number[] {
+  return metrics
+    .filter((m) => m.name === name)
+    .sort((a, b) => new Date(a.timestamp).getTime() - new Date(b.timestamp).getTime())
+    .map((m) => Number(m.value) || 0);
+}
+
+/** 容量规划建议 */
+export function capacityRecommendation(metricKey: string, days: number): string {
+  const level = days <= 7 ? 'urgent' : days <= 14 ? 'warning' : 'normal';
+  const recs: Record<string, Record<string, string>> = {
+    urgent: {
+      cpu: i18n.t('aiops.predictutilsts.s4d0bd643'),
+      memory: i18n.t('aiops.predictutilsts.sb4593f3a'),
+      disk: i18n.t('aiops.predictutilsts.s6c25383f'),
+    },
+    warning: {
+      cpu: i18n.t('aiops.predictutilsts.s8ea826f4'),
+      memory: i18n.t('aiops.predictutilsts.sb08bd5fc'),
+      disk: i18n.t('aiops.predictutilsts.s4cdda5dc'),
+    },
+    normal: {
+      cpu: i18n.t('aiops.predictutilsts.sfa295c44'),
+      memory: i18n.t('aiops.predictutilsts.s7156a4c0'),
+      disk: i18n.t('aiops.predictutilsts.s0073f3a3'),
+    },
+  };
+  return recs[level]?.[metricKey] || i18n.t('aiops.predictutilsts.sfeb368e6');
+}
+
+// ============ 综合预测推算 ============
+
+/**
+ * 综合预测推算：基于时序指标计算指标预测与容量规划。
+ * 主组件 AIPredictiveAnalytics.tsx 直接调用此函数获取展示数据。
+ * // Logan.Li
+ */
+export function computePredictions(input: ComputePredictionsInput): ComputedPredictions {
+  const { rawMetrics, serverStats, selectedTimeRange } = input;
+  const horizonHours = TIME_RANGE_HOURS[selectedTimeRange] || 24;
+  const timeRangeLabel = TIME_RANGE_LABELS[selectedTimeRange] || i18n.t('aiops.predictutilsts.sfc78646e');
+
+  // ---- 指标预测 ----
+  const predictions: PredictionMetric[] = METRIC_DEFS.map((def) => {
+    const series = extractSeries(rawMetrics, def.key);
+    const { slope, r2 } = linearRegression(series);
+    const avgFallback =
+      def.key === 'cpu' ? serverStats.avg_cpu : def.key === 'memory' ? serverStats.avg_memory : serverStats.avg_disk;
+    const current = series.length > 0 ? series[series.length - 1] : avgFallback || 0;
+    let predicted = current + slope * horizonHours;
+    predicted = Math.max(0, Math.min(100, predicted));
+    const trend: 'up' | 'down' | 'stable' = slope > 0.05 ? 'up' : slope < -0.05 ? 'down' : 'stable';
+    const confidence = Math.round(60 + r2 * 35);
+    return {
+      name: def.label,
+      current: Math.round(current),
+      predicted: Math.round(predicted),
+      unit: '%',
+      trend,
+      confidence,
+      timeRange: timeRangeLabel,
+    };
+  });
+
+  // ---- 容量规划 ----
+  const capacityForecasts: CapacityForecast[] = METRIC_DEFS.map((def) => {
+    const series = extractSeries(rawMetrics, def.key);
+    const { slope } = linearRegression(series);
+    const avgFallback =
+      def.key === 'cpu' ? serverStats.avg_cpu : def.key === 'memory' ? serverStats.avg_memory : serverStats.avg_disk;
+    const currentUsage = Math.round(series.length > 0 ? series[series.length - 1] : avgFallback || 0);
+    const slopePerDay = slope * 24;
+    const predictedUsage = Math.round(Math.min(100, currentUsage + slopePerDay * 7));
+    const threshold = def.threshold;
+    let daysUntilThreshold = 999;
+    if (slopePerDay > 0.01) {
+      daysUntilThreshold = Math.round((threshold - currentUsage) / slopePerDay);
+      daysUntilThreshold = daysUntilThreshold < 1 ? 1 : daysUntilThreshold;
+    }
+    const resourceNames: Record<string, string> = {
+      cpu: '服务器CPU算力',
+      memory: '服务器内存',
+      disk: '存储磁盘空间',
+    };
+    return {
+      resource: resourceNames[def.key] || def.label,
+      metricKey: def.key,
+      currentUsage,
+      predictedUsage,
+      threshold,
+      daysUntilThreshold,
+      recommendation: capacityRecommendation(def.key, daysUntilThreshold),
+    };
+  }).filter((f) => f.currentUsage > 0);
+
+  // ---- 概览统计 ----
+  const avgR2 = METRIC_DEFS.length > 0
+    ? METRIC_DEFS.reduce((sum, def) => sum + linearRegression(extractSeries(rawMetrics, def.key)).r2, 0) /
+      METRIC_DEFS.length
+    : 0;
+  const accuracy = Math.round(78 + avgR2 * 17);
+  const capacityWarningCount = capacityForecasts.filter((f) => f.daysUntilThreshold <= 14).length;
+
+  return {
+    predictions,
+    capacityForecasts,
+    accuracy,
+    capacityWarningCount,
+  };
+}
+
+/** 趋势图标颜色工具，供主组件 render 用 */
+export function trendColorClass(trend: string): string {
+  switch (trend) {
+    case 'up':
+      return 'dg-text-red';
+    case 'down':
+      return 'dg-text-green';
+    default:
+      return 'dg-text-secondary';
+  }
+}
+
+/**
+ * 构建容量预测「AI 深度解读」的 prompt —— 把硬数据序列化为结构化上下文喂给 LLM。 // Logan.Li
+ * 纯函数，供 PredictiveCapacityTab 调 aiApi.chat 使用。无数据返回空串（调用方据此禁用按钮）。
+ */
+export function buildCapacityInsightPrompt(forecasts: CapacityForecast[]): string {
+  if (forecasts.length === 0) return '';
+  const lines = forecasts.map((f) => {
+    const risk = f.daysUntilThreshold >= 999 ? i18n.t('aiops.predictutilsts.s179fd5a9') : i18n.t('aiops.predictutilsts.daysUntilThreshold', { days: f.daysUntilThreshold });
+    return `- ${f.resource}：当前 ${f.currentUsage}% → 7天后预测 ${f.predictedUsage}%（阈值 ${f.threshold}%），${risk}`;
+  });
+  return [
+    i18n.t('aiops.predictutilsts.s3b46e5a0'),
+    '',
+    i18n.t('aiops.predictutilsts.s586bac5d'),
+    ...lines,
+    '',
+    i18n.t('aiops.predictutilsts.sdf6ebee3'),
+    i18n.t('aiops.predictutilsts.s534b6893'),
+    i18n.t('aiops.predictutilsts.saf173e7a'),
+    i18n.t('aiops.predictutilsts.s4ed1d035'),
+  ].join('\n');
+}
diff --git a/frontend/src/components/aiops/cost-predict.ts b/frontend/src/components/aiops/cost-predict.ts
new file mode 100644
index 0000000..23c9a2f
--- /dev/null
+++ b/frontend/src/components/aiops/cost-predict.ts
@@ -0,0 +1,434 @@
+import type { JsonData } from '@/types/common';
+import i18n from '@/i18n';
+import { linearRegression } from './capacity-predict';
+import type {
+  DailyCostPoint,
+  CostDimensionForecast,
+  BudgetRiskForecast,
+  CostPredictionSummary,
+  ComputedCostPredictions,
+  ComputeCostPredictionsInput,
+} from './predict-types';
+
+/**
+ * 云成本预测 — 月末费用预测 + 预算触线风险 + AI 解读 prompt
+ *
+ * 从 predictUtils.ts (716L) 提取。
+ * 所有函数均为纯函数（输入 → 输出，无副作用），便于单元测试和复用。
+ * // Logan.Li
+ */
+
+// ============ 私有常量 ============
+
+const PROVIDER_LABELS: Record<string, string> = {
+  aliyun: i18n.t('aiops.predictutilsts.providerAliyun'),
+  tencent: i18n.t('aiops.predictutilsts.providerTencent'),
+  aws: 'AWS',
+  huawei: i18n.t('aiops.predictutilsts.providerHuawei'),
+  azure: 'Azure',
+};
+
+// ============ 日期工具 ============
+
+/** 本月第一天 YYYY-MM-DD */
+export function monthStartDate(ref = new Date()): string {
+  return `${ref.getFullYear()}-${String(ref.getMonth() + 1).padStart(2, '0')}-01`;
+}
+
+/** 本月剩余天数（不含今天） */
+export function daysRemainingInMonth(ref = new Date()): number {
+  const lastDay = new Date(ref.getFullYear(), ref.getMonth() + 1, 0).getDate();
+  return Math.max(0, lastDay - ref.getDate());
+}
+
+// ============ 账单聚合 ============
+
+/** 账单明细按日聚合 */
+export function aggregateDailyCosts(
+  records: JsonData[],
+  filter?: { startDate?: string; provider?: string },
+): DailyCostPoint[] {
+  const map = new Map<string, number>();
+  for (const r of records) {
+    const date = String(r.billing_date || '');
+    if (!date) continue;
+    if (filter?.startDate && date < filter.startDate) continue;
+    if (filter?.provider && r.provider !== filter.provider) continue;
+    map.set(date, (map.get(date) || 0) + (Number(r.cost) || 0));
+  }
+  return Array.from(map.entries())
+    .sort((a, b) => a[0].localeCompare(b[0]))
+    .map(([date, cost]) => ({ date, cost }));
+}
+
+/** 本月累计费用 */
+export function sumMonthCost(records: JsonData[], provider?: string): number {
+  const start = monthStartDate();
+  return records.reduce((sum, r) => {
+    const date = String(r.billing_date || '');
+    if (date < start) return sum;
+    if (provider && r.provider !== provider) return sum;
+    return sum + (Number(r.cost) || 0);
+  }, 0);
+}
+
+// ============ 私有工具 ============
+
+function costTrendFromSlope(slope: number): 'up' | 'down' | 'stable' {
+  if (slope > 1) return 'up';
+  if (slope < -1) return 'down';
+  return 'stable';
+}
+
+/** 成本优化建议 */
+export function costRecommendation(trend: 'up' | 'down' | 'stable', ratio: number): string {
+  if (trend === 'up' && ratio >= 0.85) {
+    return i18n.t('aiops.predictutilsts.sa7d4e3c1');
+  }
+  if (trend === 'up') {
+    return i18n.t('aiops.predictutilsts.sb3f5a2d8');
+  }
+  if (trend === 'down') {
+    return i18n.t('aiops.predictutilsts.sc1a9b4e7');
+  }
+  return i18n.t('aiops.predictutilsts.sd8c6f2a3');
+}
+
+/** 预算风险建议 */
+function budgetRiskAction(metric: 'daily' | 'monthly', provider: string): string {
+  const scope = provider && provider !== i18n.t('aiops.predictutilsts.sg5a3d8e1') ? `${provider} ` : '';
+  if (metric === 'daily') {
+    return i18n.t('aiops.predictutilsts.se4b7a5d2', { scope });
+  }
+  return i18n.t('aiops.predictutilsts.sf2d8c6b4', { scope });
+}
+
+interface MonthEndForecast {
+  currentMonth: number;
+  predictedMonthEnd: number;
+  dailyAvg: number;
+  recentDailyAvg: number;
+  billingPattern: 'steady' | 'front_loaded' | 'sparse';
+  trend: 'up' | 'down' | 'stable';
+  confidence: number;
+  series: DailyCostPoint[];
+}
+
+/** 识别账单形态（预付费常表现为月初大额、后续近零） // Logan.Li */
+function detectBillingPattern(series: DailyCostPoint[]): 'steady' | 'front_loaded' | 'sparse' {
+  const active = series.filter((d) => d.cost > 0);
+  if (active.length < 2) return 'sparse';
+  const total = series.reduce((s, d) => s + d.cost, 0);
+  const recent = series.slice(-Math.min(7, series.length));
+  const recentAvg = recent.reduce((s, d) => s + d.cost, 0) / recent.length;
+  const overallAvg = total / series.length;
+  if (overallAvg > 0 && recentAvg < overallAvg * 0.25) return 'front_loaded';
+  return 'steady';
+}
+
+function recentRunRate(series: DailyCostPoint[], days = 7): number {
+  const recent = series.slice(-Math.min(days, series.length));
+  if (recent.length === 0) return 0;
+  return recent.reduce((s, d) => s + d.cost, 0) / recent.length;
+}
+
+/** 本月 Top 服务消费 */
+function topServicesByCost(records: JsonData[], limit = 5): { service: string; cost: number }[] {
+  const start = monthStartDate();
+  const map = new Map<string, number>();
+  records.forEach((r) => {
+    const date = String(r.billing_date || '');
+    if (date < start) return;
+    const svc = String(r.service || i18n.t('aiops.predictutilsts.sh7c2b9d6'));
+    map.set(svc, (map.get(svc) || 0) + (Number(r.cost) || 0));
+  });
+  return [...map.entries()]
+    .sort((a, b) => b[1] - a[1])
+    .slice(0, limit)
+    .map(([service, cost]) => ({ service, cost: Math.round(cost * 100) / 100 }));
+}
+
+/** 基于日费用序列预测月末总费用 */
+function forecastMonthEnd(
+  records: JsonData[],
+  statsMonthCost: number,
+  fallbackTrend: DailyCostPoint[],
+  provider?: string,
+): MonthEndForecast {
+  const start = monthStartDate();
+  const monthSeries = aggregateDailyCosts(records, { startDate: start, provider });
+  const series = monthSeries.length >= 2 ? monthSeries : fallbackTrend;
+  const costs = series.map((d) => d.cost);
+  const { slope, r2 } = linearRegression(costs);
+  const currentMonth =
+    monthSeries.length > 0 ? monthSeries.reduce((s, d) => s + d.cost, 0) : statsMonthCost;
+  const daysLeft = daysRemainingInMonth();
+  const dailyAvg = costs.length > 0 ? costs.reduce((a, b) => a + b, 0) / costs.length : 0;
+  const recentDailyAvg = recentRunRate(series);
+  const billingPattern = detectBillingPattern(series);
+  // 预付费/前高后平：用近 7 日运行费率外推，避免线性回归高估或矛盾 // Logan.Li
+  const extrapolation =
+    billingPattern === 'front_loaded' ? recentDailyAvg * daysLeft : slope * daysLeft;
+  const predictedMonthEnd = Math.max(0, Math.round((currentMonth + extrapolation) * 100) / 100);
+  let confidence = Math.round(55 + r2 * 40);
+  if (billingPattern === 'front_loaded') confidence = Math.max(45, confidence - 15);
+  if (billingPattern === 'sparse') confidence = Math.min(confidence, 50);
+  return {
+    currentMonth,
+    predictedMonthEnd,
+    dailyAvg: Math.round(dailyAvg * 100) / 100,
+    recentDailyAvg: Math.round(recentDailyAvg * 100) / 100,
+    billingPattern,
+    trend: costTrendFromSlope(billingPattern === 'front_loaded' ? recentDailyAvg - dailyAvg : slope),
+    confidence,
+    series: monthSeries.length > 0 ? monthSeries : fallbackTrend,
+  };
+}
+
+// ============ 综合成本预测 ============
+
+/**
+ * 云成本预测：基于账单日粒度时序 + 预算预警规则，推算月末费用与触线风险。
+ * // Logan.Li
+ */
+export function computeCostPredictions(input: ComputeCostPredictionsInput): ComputedCostPredictions {
+  const { billingStats, billingRecords, billingAlertRules } = input;
+  const empty: ComputedCostPredictions = {
+    summary: {
+      monthCost: 0,
+      predictedMonthEnd: 0,
+      todayCost: 0,
+      yesterdayCost: 0,
+      dailyAvg: 0,
+      daysRemaining: daysRemainingInMonth(),
+      trend: 'stable',
+      confidence: 0,
+      monthTrend: [],
+      providerDist: [],
+    },
+    dimensionForecasts: [],
+    budgetRisks: [],
+    budgetRiskCount: 0,
+    recentDailyAvg: 0,
+    billingPattern: 'sparse',
+    topServices: [],
+  };
+
+  if (!billingStats) return empty;
+
+  const monthTrend: DailyCostPoint[] = Array.isArray(billingStats.month_trend)
+    ? billingStats.month_trend.map((d: JsonData) => ({
+        date: String(d.date || ''),
+        cost: Number(d.cost) || 0,
+      }))
+    : [];
+  const providerDist: { provider: string; cost: number; percent: number }[] = Array.isArray(
+    billingStats.provider_dist,
+  )
+    ? billingStats.provider_dist.map((p: JsonData) => ({
+        provider: String(p.provider || ''),
+        cost: Number(p.cost) || 0,
+        percent: Number(p.percent) || 0,
+      }))
+    : [];
+
+  const overall = forecastMonthEnd(
+    billingRecords,
+    Number(billingStats.month_cost) || 0,
+    monthTrend,
+  );
+
+  const summary: CostPredictionSummary = {
+    monthCost: Number(billingStats.month_cost) || overall.currentMonth,
+    predictedMonthEnd: overall.predictedMonthEnd,
+    todayCost: Number(billingStats.today_cost) || 0,
+    yesterdayCost: Number(billingStats.yesterday_cost) || 0,
+    dailyAvg: overall.dailyAvg,
+    daysRemaining: daysRemainingInMonth(),
+    trend: overall.trend,
+    confidence: overall.confidence,
+    monthTrend: overall.series.length > 0 ? overall.series : monthTrend,
+    providerDist,
+  };
+
+  // 全部 + 各云厂商维度预测
+  const dimensionForecasts: CostDimensionForecast[] = [];
+  const allRatio =
+    summary.predictedMonthEnd > 0 ? summary.monthCost / summary.predictedMonthEnd : 0;
+  dimensionForecasts.push({
+    dimension: i18n.t('aiops.predictutilsts.sa1b2c3d4'),
+    currentMonth: summary.monthCost,
+    predictedMonthEnd: summary.predictedMonthEnd,
+    dailyAvg: summary.dailyAvg,
+    trend: summary.trend,
+    confidence: summary.confidence,
+    recommendation: costRecommendation(summary.trend, allRatio),
+  });
+
+  const providers = new Set<string>();
+  providerDist.forEach((p) => providers.add(p.provider));
+  billingRecords.forEach((r) => {
+    if (r.provider) providers.add(String(r.provider));
+  });
+
+  for (const provider of providers) {
+    if (!provider) continue;
+    const provForecast = forecastMonthEnd(billingRecords, 0, monthTrend, provider);
+    const provMonth =
+      provForecast.currentMonth > 0 ? provForecast.currentMonth : sumMonthCost(billingRecords, provider);
+    const ratio = provForecast.predictedMonthEnd > 0 ? provMonth / provForecast.predictedMonthEnd : 0;
+    dimensionForecasts.push({
+      dimension: PROVIDER_LABELS[provider] || provider,
+      provider,
+      currentMonth: provMonth,
+      predictedMonthEnd: provForecast.predictedMonthEnd,
+      dailyAvg: provForecast.dailyAvg,
+      trend: provForecast.trend,
+      confidence: provForecast.confidence,
+      recommendation: costRecommendation(provForecast.trend, ratio),
+    });
+  }
+
+  dimensionForecasts.sort((a, b) => b.predictedMonthEnd - a.predictedMonthEnd);
+
+  // 预算触线风险
+  const budgetRisks: BudgetRiskForecast[] = [];
+  const activeAlerts: JsonData[] = Array.isArray(billingStats.alerts) ? billingStats.alerts : [];
+
+  for (const rule of billingAlertRules) {
+    if (!rule.enabled) continue;
+    const threshold = Number(rule.threshold) || 0;
+    if (threshold <= 0) continue;
+
+    const metric = (rule.metric === 'daily' ? 'daily' : 'monthly') as 'daily' | 'monthly';
+    const ruleProvider = String(rule.provider || '');
+    const providerLabel = ruleProvider ? PROVIDER_LABELS[ruleProvider] || ruleProvider : i18n.t('aiops.predictutilsts.sb2c3d4e5');
+
+    let current = 0;
+    let predicted = 0;
+    let slope = 0;
+
+    if (metric === 'daily') {
+      current = Number(billingStats.today_cost) || 0;
+      if (ruleProvider) {
+        const today = new Date().toISOString().slice(0, 10);
+        current = billingRecords
+          .filter((r) => r.billing_date === today && r.provider === ruleProvider)
+          .reduce((s, r) => s + (Number(r.cost) || 0), 0);
+      }
+      const recent = aggregateDailyCosts(billingRecords, ruleProvider ? { provider: ruleProvider } : undefined);
+      const costs = recent.slice(-7).map((d) => d.cost);
+      slope = linearRegression(costs).slope;
+      predicted = Math.max(0, current + slope);
+    } else {
+      const provForecast = ruleProvider
+        ? forecastMonthEnd(billingRecords, 0, monthTrend, ruleProvider)
+        : overall;
+      current = ruleProvider ? sumMonthCost(billingRecords, ruleProvider) : summary.monthCost;
+      predicted = provForecast.predictedMonthEnd;
+      const costs = provForecast.series.map((d) => d.cost);
+      slope = linearRegression(costs).slope;
+    }
+
+    const active = activeAlerts.find((a) => a.rule_id === rule.id || a.rule_name === rule.name);
+    if (active) current = Number(active.current) || current;
+
+    const ratio = threshold > 0 ? predicted / threshold : 0;
+    let probability = Math.round(ratio * 50);
+    if (slope > 0.5) probability += 15;
+    if (current >= threshold) probability += 25;
+    probability = Math.max(5, Math.min(92, probability));
+
+    let timeframe = metric === 'daily' ? i18n.t('aiops.predictutilsts.sc3d4e5f6') : i18n.t('aiops.predictutilsts.sd4e5f6g7');
+    if (slope > 0.01 && threshold > current) {
+      const days = (threshold - current) / slope;
+      if (days < 1) timeframe = metric === 'daily' ? i18n.t('aiops.predictutilsts.se5f6g7h8') : i18n.t('aiops.predictutilsts.sf6g7h8i9');
+      else if (days <= 7) timeframe = i18n.t('aiops.predictutilsts.sg7h8i9j0', { days: Math.ceil(days) });
+      else timeframe = i18n.t('aiops.predictutilsts.sh8i9j0k1', { days: Math.ceil(days) });
+    }
+
+    if (probability >= 30 || current >= threshold * 0.7) {
+      budgetRisks.push({
+        ruleName: String(rule.name || i18n.t('aiops.predictutilsts.si9j0k1l2')),
+        provider: providerLabel,
+        metric,
+        threshold,
+        current: Math.round(current * 100) / 100,
+        predicted: Math.round(predicted * 100) / 100,
+        probability,
+        timeframe,
+        suggestedAction: budgetRiskAction(metric, providerLabel),
+      });
+    }
+  }
+
+  budgetRisks.sort((a, b) => b.probability - a.probability);
+  const budgetRiskCount = budgetRisks.filter((r) => r.probability >= 50).length;
+
+  return {
+    summary,
+    dimensionForecasts,
+    budgetRisks,
+    budgetRiskCount,
+    recentDailyAvg: overall.recentDailyAvg,
+    billingPattern: overall.billingPattern,
+    topServices: topServicesByCost(billingRecords),
+  };
+}
+
+/**
+ * 构建成本预测「AI 深度解读」prompt。
+ * // Logan.Li
+ */
+export function buildCostInsightPrompt(data: ComputedCostPredictions): string {
+  const { summary, dimensionForecasts, budgetRisks, recentDailyAvg, billingPattern, topServices } = data;
+  if (summary.monthCost <= 0 && dimensionForecasts.length === 0) return '';
+
+  const patternHint =
+    billingPattern === 'front_loaded'
+      ? i18n.t('aiops.predictutilsts.sj0k1l2m3')
+      : billingPattern === 'sparse'
+        ? i18n.t('aiops.predictutilsts.sk1l2m3n4')
+        : i18n.t('aiops.predictutilsts.sl2m3n4o5');
+
+  const dimLines = dimensionForecasts.slice(0, 6).map(
+    (d) =>
+      `- ${d.dimension}：本月 ${d.currentMonth.toFixed(2)} 元 → 预测月末 ${d.predictedMonthEnd.toFixed(2)} 元（趋势 ${d.trend}）`,
+  );
+  const riskLines =
+    budgetRisks.length > 0
+      ? budgetRisks.slice(0, 5).map(
+          (r) =>
+            `- ${r.ruleName}（${r.provider}）：当前 ${r.current} / 阈值 ${r.threshold}，预测 ${r.predicted}，触线概率 ${r.probability}%`,
+        )
+      : [`- ${i18n.t('aiops.predictutilsts.sm3n4o5p6')}`];
+  const serviceLines =
+    topServices.length > 0
+      ? topServices.map((s) => `- ${s.service}：${s.cost.toFixed(2)} 元`)
+      : [`- ${i18n.t('aiops.predictutilsts.sn4o5p6q7')}`];
+
+  return [
+    i18n.t('aiops.predictutilsts.so5p6q7r8'),
+    '',
+    i18n.t('aiops.predictutilsts.sp6q7r8s9'),
+    `- 本月已花：${summary.monthCost.toFixed(2)} 元`,
+    `- 预测月末：${summary.predictedMonthEnd.toFixed(2)} 元（置信 ${summary.confidence}%）`,
+    `- 全月日均：${summary.dailyAvg.toFixed(2)} 元；近7日日均：${recentDailyAvg.toFixed(2)} 元；剩余 ${summary.daysRemaining} 天`,
+    `- 账单形态：${patternHint}`,
+    '',
+    i18n.t('aiops.predictutilsts.sq7r8s9t0'),
+    ...serviceLines,
+    '',
+    i18n.t('aiops.predictutilsts.sr8s9t0u1'),
+    ...dimLines,
+    '',
+    i18n.t('aiops.predictutilsts.ss9t0u1v2'),
+    ...riskLines,
+    '',
+    i18n.t('aiops.predictutilsts.st0u1v2w3'),
+    i18n.t('aiops.predictutilsts.su1v2w3x4'),
+    i18n.t('aiops.predictutilsts.sv2w3x4y5'),
+    i18n.t('aiops.predictutilsts.sw3x4y5z6'),
+    i18n.t('aiops.predictutilsts.sx4y5z6a7'),
+  ].join('\n');
+}
diff --git a/frontend/src/components/aiops/predict-types.ts b/frontend/src/components/aiops/predict-types.ts
new file mode 100644
index 0000000..96cf34f
--- /dev/null
+++ b/frontend/src/components/aiops/predict-types.ts
@@ -0,0 +1,140 @@
+import type { JsonData } from '@/types/common';
+import i18n from '@/i18n';
+
+/**
+ * AI 预测分析模块 — 共享类型定义 + 常量
+ *
+ * 从 predictUtils.ts (716L) 提取。
+ * // Logan.Li
+ */
+
+// ============ 容量预测类型 ============
+
+export interface PredictionMetric {
+  name: string;
+  current: number;
+  predicted: number;
+  unit: string;
+  trend: 'up' | 'down' | 'stable';
+  confidence: number;
+  timeRange: string;
+}
+
+export interface CapacityForecast {
+  resource: string;
+  metricKey: string;
+  currentUsage: number;
+  predictedUsage: number;
+  threshold: number;
+  daysUntilThreshold: number;
+  recommendation: string;
+}
+
+// ============ 云成本预测类型 ============
+
+/** 日费用数据点 */
+export interface DailyCostPoint {
+  date: string;
+  cost: number;
+}
+
+/** 单维度（全部 / 云厂商）成本预测 */
+export interface CostDimensionForecast {
+  dimension: string;
+  provider?: string;
+  currentMonth: number;
+  predictedMonthEnd: number;
+  dailyAvg: number;
+  trend: 'up' | 'down' | 'stable';
+  confidence: number;
+  recommendation: string;
+}
+
+/** 预算触线风险（对接 BillingAlertRule） */
+export interface BudgetRiskForecast {
+  ruleName: string;
+  provider: string;
+  metric: 'daily' | 'monthly';
+  threshold: number;
+  current: number;
+  predicted: number;
+  probability: number;
+  timeframe: string;
+  suggestedAction: string;
+}
+
+/** 成本预测汇总 */
+export interface CostPredictionSummary {
+  monthCost: number;
+  predictedMonthEnd: number;
+  todayCost: number;
+  yesterdayCost: number;
+  dailyAvg: number;
+  daysRemaining: number;
+  trend: 'up' | 'down' | 'stable';
+  confidence: number;
+  monthTrend: DailyCostPoint[];
+  providerDist: { provider: string; cost: number; percent: number }[];
+}
+
+export interface ComputedCostPredictions {
+  summary: CostPredictionSummary;
+  dimensionForecasts: CostDimensionForecast[];
+  budgetRisks: BudgetRiskForecast[];
+  budgetRiskCount: number;
+  /** 近 7 日日均（用于识别预付费后按量趋零） // Logan.Li */
+  recentDailyAvg: number;
+  /** 账单形态：steady 平稳 / front_loaded 月初集中 / sparse 样本不足 */
+  billingPattern: 'steady' | 'front_loaded' | 'sparse';
+  /** 本月 Top 服务消费 */
+  topServices: { service: string; cost: number }[];
+}
+
+// ============ 容量预测入参/出参 ============
+
+/** computePredictions 的输入参数 */
+export interface ComputePredictionsInput {
+  rawMetrics: JsonData[];
+  serverStats?: JsonData;
+  selectedTimeRange: string; // '1h' | '24h' | '7d' | '30d'
+}
+
+/** computePredictions 的聚合输出 */
+export interface ComputedPredictions {
+  predictions: PredictionMetric[];
+  capacityForecasts: CapacityForecast[];
+  accuracy: number;
+  capacityWarningCount: number;
+}
+
+// ============ 成本预测入参/出参 ============
+
+/** computeCostPredictions 输入 */
+export interface ComputeCostPredictionsInput {
+  billingStats: JsonData | null;
+  billingRecords: JsonData[];
+  billingAlertRules: JsonData[];
+}
+
+// ============ 共享常量 ============
+
+/** 指标定义：后端 metric name → 展示标签 / 告警表达式前缀 / 阈值 */
+export const METRIC_DEFS = [
+  { key: 'cpu', label: i18n.t('aiops.predictutilsts.s7054bc34'), exprKey: 'cpu_usage', threshold: 85 },
+  { key: 'memory', label: i18n.t('aiops.predictutilsts.s82c0d1b0'), exprKey: 'memory_usage', threshold: 85 },
+  { key: 'disk', label: i18n.t('aiops.predictutilsts.s4e2d8f91'), exprKey: 'disk_usage', threshold: 90 },
+] as const;
+
+export const TIME_RANGE_LABELS: Record<string, string> = {
+  '1h': i18n.t('aiops.predictutilsts.s75719bb8'),
+  '24h': i18n.t('aiops.predictutilsts.sfc78646e'),
+  '7d': i18n.t('aiops.predictutilsts.s5b5a820c'),
+  '30d': i18n.t('aiops.predictutilsts.s933ab0a6'),
+};
+
+export const TIME_RANGE_HOURS: Record<string, number> = {
+  '1h': 1,
+  '24h': 24,
+  '7d': 168,
+  '30d': 720,
+};
diff --git a/frontend/src/components/aiops/predictUtils.ts b/frontend/src/components/aiops/predictUtils.ts
index fa15f6b..5d5bc45 100644
--- a/frontend/src/components/aiops/predictUtils.ts
+++ b/frontend/src/components/aiops/predictUtils.ts
@@ -1,716 +1,49 @@
-import type { JsonData } from "@/types/common";
-import i18n from '@/i18n';
 /**
- * AI 预测分析模块的纯计算工具 — 类型定义、常量、线性回归、预测推算
+ * AI 预测分析模块 — Barrel re-export（向后兼容）
  *
- * 从 AIPredictiveAnalytics.tsx(670行) 提取，避免主组件超 700 行硬上限。
- * 所有函数均为纯函数（输入 → 输出，无副作用），便于单元测试和复用。
+ * 原 716L 文件已按域拆分为 3 个文件：
+ * - predict-types.ts   — 类型定义 + 共享常量 (METRIC_DEFS / TIME_RANGE_*)
+ * - capacity-predict.ts — 容量预测（线性回归 + computePredictions + 容量规划 + AI prompt）
+ * - cost-predict.ts     — 云成本预测（月末预测 + 预算触线风险 + AI prompt）
+ *
+ * 现有 3 个消费者的 import 路径 (`from './predictUtils'`) 无需修改。
  * // Logan.Li
  */
 
-// ============ 类型定义 ============
-
-export interface PredictionMetric {
-  name: string;
-  current: number;
-  predicted: number;
-  unit: string;
-  trend: 'up' | 'down' | 'stable';
-  confidence: number;
-  timeRange: string;
-}
-
-export interface CapacityForecast {
-  resource: string;
-  metricKey: string;
-  currentUsage: number;
-  predictedUsage: number;
-  threshold: number;
-  daysUntilThreshold: number;
-  recommendation: string;
-}
-
-/** 日费用数据点 */
-export interface DailyCostPoint {
-  date: string;
-  cost: number;
-}
-
-/** 单维度（全部 / 云厂商）成本预测 */
-export interface CostDimensionForecast {
-  dimension: string;
-  provider?: string;
-  currentMonth: number;
-  predictedMonthEnd: number;
-  dailyAvg: number;
-  trend: 'up' | 'down' | 'stable';
-  confidence: number;
-  recommendation: string;
-}
-
-/** 预算触线风险（对接 BillingAlertRule） */
-export interface BudgetRiskForecast {
-  ruleName: string;
-  provider: string;
-  metric: 'daily' | 'monthly';
-  threshold: number;
-  current: number;
-  predicted: number;
-  probability: number;
-  timeframe: string;
-  suggestedAction: string;
-}
-
-/** 成本预测汇总 */
-export interface CostPredictionSummary {
-  monthCost: number;
-  predictedMonthEnd: number;
-  todayCost: number;
-  yesterdayCost: number;
-  dailyAvg: number;
-  daysRemaining: number;
-  trend: 'up' | 'down' | 'stable';
-  confidence: number;
-  monthTrend: DailyCostPoint[];
-  providerDist: { provider: string; cost: number; percent: number }[];
-}
-
-export interface ComputedCostPredictions {
-  summary: CostPredictionSummary;
-  dimensionForecasts: CostDimensionForecast[];
-  budgetRisks: BudgetRiskForecast[];
-  budgetRiskCount: number;
-  /** 近 7 日日均（用于识别预付费后按量趋零） // Logan.Li */
-  recentDailyAvg: number;
-  /** 账单形态：steady 平稳 / front_loaded 月初集中 / sparse 样本不足 */
-  billingPattern: 'steady' | 'front_loaded' | 'sparse';
-  /** 本月 Top 服务消费 */
-  topServices: { service: string; cost: number }[];
-}
+// ============ 类型 ============
+export type {
+  PredictionMetric,
+  CapacityForecast,
+  DailyCostPoint,
+  CostDimensionForecast,
+  BudgetRiskForecast,
+  CostPredictionSummary,
+  ComputedCostPredictions,
+  ComputePredictionsInput,
+  ComputedPredictions,
+  ComputeCostPredictionsInput,
+} from './predict-types';
 
 // ============ 常量 ============
+export { METRIC_DEFS, TIME_RANGE_LABELS, TIME_RANGE_HOURS } from './predict-types';
 
-/** 指标定义：后端 metric name → 展示标签 / 告警表达式前缀 / 阈值 */
-export const METRIC_DEFS = [
-  { key: 'cpu', label: i18n.t('aiops.predictutilsts.s7054bc34'), exprKey: 'cpu_usage', threshold: 85 },
-  { key: 'memory', label: i18n.t('aiops.predictutilsts.s82c0d1b0'), exprKey: 'memory_usage', threshold: 85 },
-  { key: 'disk', label: i18n.t('aiops.predictutilsts.s4e2d8f91'), exprKey: 'disk_usage', threshold: 90 },
-] as const;
+// ============ 容量预测函数 ============
+export {
+  linearRegression,
+  extractSeries,
+  capacityRecommendation,
+  computePredictions,
+  trendColorClass,
+  buildCapacityInsightPrompt,
+} from './capacity-predict';
 
-export const TIME_RANGE_LABELS: Record<string, string> = {
-  '1h': i18n.t('aiops.predictutilsts.s75719bb8'),
-  '24h': i18n.t('aiops.predictutilsts.sfc78646e'),
-  '7d': i18n.t('aiops.predictutilsts.s5b5a820c'),
-  '30d': i18n.t('aiops.predictutilsts.s933ab0a6'),
-};
-
-export const TIME_RANGE_HOURS: Record<string, number> = {
-  '1h': 1,
-  '24h': 24,
-  '7d': 168,
-  '30d': 720,
-};
-
-// ============ 纯函数工具 ============
-
-/** 简单最小二乘线性回归，返回斜率和 R² */
-export function linearRegression(values: number[]): { slope: number; r2: number } {
-  const n = values.length;
-  if (n < 2) return { slope: 0, r2: 0 };
-  const xs = values.map((_, i) => i);
-  const meanX = xs.reduce((a, b) => a + b, 0) / n;
-  const meanY = values.reduce((a, b) => a + b, 0) / n;
-  let num = 0,
-    den = 0,
-    ssTot = 0,
-    ssRes = 0;
-  for (let i = 0; i < n; i++) {
-    num += (xs[i] - meanX) * (values[i] - meanY);
-    den += (xs[i] - meanX) ** 2;
-    ssTot += (values[i] - meanY) ** 2;
-  }
-  const slope = den === 0 ? 0 : num / den;
-  const intercept = meanY - slope * meanX;
-  for (let i = 0; i < n; i++) {
-    const pred = intercept + slope * xs[i];
-    ssRes += (values[i] - pred) ** 2;
-  }
-  const r2 = ssTot === 0 ? 1 : Math.max(0, 1 - ssRes / ssTot);
-  return { slope, r2 };
-}
-
-/** 从时序指标中提取按时间排序的数值序列 */
-export function extractSeries( metrics: JsonData[], name: string): number[] {
-  return metrics
-    .filter((m) => m.name === name)
-    .sort((a, b) => new Date(a.timestamp).getTime() - new Date(b.timestamp).getTime())
-    .map((m) => Number(m.value) || 0);
-}
-
-/** 容量规划建议 */
-export function capacityRecommendation(metricKey: string, days: number): string {
-  const level = days <= 7 ? 'urgent' : days <= 14 ? 'warning' : 'normal';
-  const recs: Record<string, Record<string, string>> = {
-    urgent: {
-      cpu: i18n.t('aiops.predictutilsts.s4d0bd643'),
-      memory: i18n.t('aiops.predictutilsts.sb4593f3a'),
-      disk: i18n.t('aiops.predictutilsts.s6c25383f'),
-    },
-    warning: {
-      cpu: i18n.t('aiops.predictutilsts.s8ea826f4'),
-      memory: i18n.t('aiops.predictutilsts.sb08bd5fc'),
-      disk: i18n.t('aiops.predictutilsts.s4cdda5dc'),
-    },
-    normal: {
-      cpu: i18n.t('aiops.predictutilsts.sfa295c44'),
-      memory: i18n.t('aiops.predictutilsts.s7156a4c0'),
-      disk: i18n.t('aiops.predictutilsts.s0073f3a3'),
-    },
-  };
-  return recs[level]?.[metricKey] || i18n.t('aiops.predictutilsts.sfeb368e6');
-}
-
-// ============ 综合预测推算 ============
-
-/** computePredictions 的输入参数 */
-export interface ComputePredictionsInput {
-  rawMetrics: JsonData[];
-  serverStats?: JsonData;
-  selectedTimeRange: string; // '1h' | '24h' | '7d' | '30d'
-}
-
-/** computePredictions 的聚合输出 */
-export interface ComputedPredictions {
-  predictions: PredictionMetric[];
-  capacityForecasts: CapacityForecast[];
-  accuracy: number;
-  capacityWarningCount: number;
-}
-
-/**
- * 综合预测推算：基于时序指标计算指标预测与容量规划。
- * 主组件 AIPredictiveAnalytics.tsx 直接调用此函数获取展示数据。
- * // Logan.Li
- */
-export function computePredictions(input: ComputePredictionsInput): ComputedPredictions {
-  const { rawMetrics, serverStats, selectedTimeRange } = input;
-  const horizonHours = TIME_RANGE_HOURS[selectedTimeRange] || 24;
-  const timeRangeLabel = TIME_RANGE_LABELS[selectedTimeRange] || i18n.t('aiops.predictutilsts.sfc78646e');
-
-  // ---- 指标预测 ----
-  const predictions: PredictionMetric[] = METRIC_DEFS.map((def) => {
-    const series = extractSeries(rawMetrics, def.key);
-    const { slope, r2 } = linearRegression(series);
-    const avgFallback =
-      def.key === 'cpu' ? serverStats.avg_cpu : def.key === 'memory' ? serverStats.avg_memory : serverStats.avg_disk;
-    const current = series.length > 0 ? series[series.length - 1] : avgFallback || 0;
-    let predicted = current + slope * horizonHours;
-    predicted = Math.max(0, Math.min(100, predicted));
-    const trend: 'up' | 'down' | 'stable' = slope > 0.05 ? 'up' : slope < -0.05 ? 'down' : 'stable';
-    const confidence = Math.round(60 + r2 * 35);
-    return {
-      name: def.label,
-      current: Math.round(current),
-      predicted: Math.round(predicted),
-      unit: '%',
-      trend,
-      confidence,
-      timeRange: timeRangeLabel,
-    };
-  });
-
-  // ---- 容量规划 ----
-  const capacityForecasts: CapacityForecast[] = METRIC_DEFS.map((def) => {
-    const series = extractSeries(rawMetrics, def.key);
-    const { slope } = linearRegression(series);
-    const avgFallback =
-      def.key === 'cpu' ? serverStats.avg_cpu : def.key === 'memory' ? serverStats.avg_memory : serverStats.avg_disk;
-    const currentUsage = Math.round(series.length > 0 ? series[series.length - 1] : avgFallback || 0);
-    const slopePerDay = slope * 24;
-    const predictedUsage = Math.round(Math.min(100, currentUsage + slopePerDay * 7));
-    const threshold = def.threshold;
-    let daysUntilThreshold = 999;
-    if (slopePerDay > 0.01) {
-      daysUntilThreshold = Math.round((threshold - currentUsage) / slopePerDay);
-      daysUntilThreshold = daysUntilThreshold < 1 ? 1 : daysUntilThreshold;
-    }
-    const resourceNames: Record<string, string> = {
-      cpu: '服务器CPU算力',
-      memory: '服务器内存',
-      disk: '存储磁盘空间',
-    };
-    return {
-      resource: resourceNames[def.key] || def.label,
-      metricKey: def.key,
-      currentUsage,
-      predictedUsage,
-      threshold,
-      daysUntilThreshold,
-      recommendation: capacityRecommendation(def.key, daysUntilThreshold),
-    };
-  }).filter((f) => f.currentUsage > 0);
-
-  // ---- 概览统计 ----
-  const avgR2 = METRIC_DEFS.length > 0
-    ? METRIC_DEFS.reduce((sum, def) => sum + linearRegression(extractSeries(rawMetrics, def.key)).r2, 0) /
-      METRIC_DEFS.length
-    : 0;
-  const accuracy = Math.round(78 + avgR2 * 17);
-  const capacityWarningCount = capacityForecasts.filter((f) => f.daysUntilThreshold <= 14).length;
-
-  return {
-    predictions,
-    capacityForecasts,
-    accuracy,
-    capacityWarningCount,
-  };
-}
-
-/** 趋势图标颜色工具，供主组件 render 用 */
-export function trendColorClass(trend: string): string {
-  switch (trend) {
-    case 'up':
-      return 'dg-text-red';
-    case 'down':
-      return 'dg-text-green';
-    default:
-      return 'dg-text-secondary';
-  }
-}
-
-/**
- * 构建容量预测「AI 深度解读」的 prompt —— 把硬数据序列化为结构化上下文喂给 LLM。 // Logan.Li
- * 纯函数，供 PredictiveCapacityTab 调 aiApi.chat 使用。无数据返回空串（调用方据此禁用按钮）。
- */
-export function buildCapacityInsightPrompt(forecasts: CapacityForecast[]): string {
-  if (forecasts.length === 0) return '';
-  const lines = forecasts.map((f) => {
-    const risk = f.daysUntilThreshold >= 999 ? i18n.t('aiops.predictutilsts.s179fd5a9') : i18n.t('aiops.predictutilsts.daysUntilThreshold', { days: f.daysUntilThreshold });
-    return `- ${f.resource}：当前 ${f.currentUsage}% → 7天后预测 ${f.predictedUsage}%（阈值 ${f.threshold}%），${risk}`;
-  });
-  return [
-    i18n.t('aiops.predictutilsts.s3b46e5a0'),
-    '',
-    i18n.t('aiops.predictutilsts.s586bac5d'),
-    ...lines,
-    '',
-    i18n.t('aiops.predictutilsts.sdf6ebee3'),
-    i18n.t('aiops.predictutilsts.s534b6893'),
-    i18n.t('aiops.predictutilsts.saf173e7a'),
-    i18n.t('aiops.predictutilsts.s4ed1d035'),
-  ].join('\n');
-}
-
-// ============ 云成本预测 ============
-
-const PROVIDER_LABELS: Record<string, string> = {
-  aliyun: i18n.t('aiops.predictutilsts.providerAliyun'),
-  tencent: i18n.t('aiops.predictutilsts.providerTencent'),
-  aws: 'AWS',
-  huawei: i18n.t('aiops.predictutilsts.providerHuawei'),
-  azure: 'Azure',
-};
-
-/** 本月第一天 YYYY-MM-DD */
-export function monthStartDate(ref = new Date()): string {
-  return `${ref.getFullYear()}-${String(ref.getMonth() + 1).padStart(2, '0')}-01`;
-}
-
-/** 本月剩余天数（不含今天） */
-export function daysRemainingInMonth(ref = new Date()): number {
-  const lastDay = new Date(ref.getFullYear(), ref.getMonth() + 1, 0).getDate();
-  return Math.max(0, lastDay - ref.getDate());
-}
-
-/** 账单明细按日聚合 */
-export function aggregateDailyCosts(
-  records: JsonData[],
-  filter?: { startDate?: string; provider?: string },
-): DailyCostPoint[] {
-  const map = new Map<string, number>();
-  for (const r of records) {
-    const date = String(r.billing_date || '');
-    if (!date) continue;
-    if (filter?.startDate && date < filter.startDate) continue;
-    if (filter?.provider && r.provider !== filter.provider) continue;
-    map.set(date, (map.get(date) || 0) + (Number(r.cost) || 0));
-  }
-  return Array.from(map.entries())
-    .sort((a, b) => a[0].localeCompare(b[0]))
-    .map(([date, cost]) => ({ date, cost }));
-}
-
-/** 本月累计费用 */
-export function sumMonthCost(records: JsonData[], provider?: string): number {
-  const start = monthStartDate();
-  return records.reduce((sum, r) => {
-    const date = String(r.billing_date || '');
-    if (date < start) return sum;
-    if (provider && r.provider !== provider) return sum;
-    return sum + (Number(r.cost) || 0);
-  }, 0);
-}
-
-function costTrendFromSlope(slope: number): 'up' | 'down' | 'stable' {
-  if (slope > 1) return 'up';
-  if (slope < -1) return 'down';
-  return 'stable';
-}
-
-/** 成本优化建议 */
-export function costRecommendation(trend: 'up' | 'down' | 'stable', ratio: number): string {
-  if (trend === 'up' && ratio >= 0.85) {
-    return i18n.t('aiops.predictutilsts.sa7d4e3c1');
-  }
-  if (trend === 'up') {
-    return i18n.t('aiops.predictutilsts.sb3f5a2d8');
-  }
-  if (trend === 'down') {
-    return i18n.t('aiops.predictutilsts.sc1a9b4e7');
-  }
-  return i18n.t('aiops.predictutilsts.sd8c6f2a3');
-}
-
-/** 预算风险建议 */
-function budgetRiskAction(metric: 'daily' | 'monthly', provider: string): string {
-  const scope = provider && provider !== i18n.t('aiops.predictutilsts.sg5a3d8e1') ? `${provider} ` : '';
-  if (metric === 'daily') {
-    return i18n.t('aiops.predictutilsts.se4b7a5d2', { scope });
-  }
-  return i18n.t('aiops.predictutilsts.sf2d8c6b4', { scope });
-}
-
-interface MonthEndForecast {
-  currentMonth: number;
-  predictedMonthEnd: number;
-  dailyAvg: number;
-  recentDailyAvg: number;
-  billingPattern: 'steady' | 'front_loaded' | 'sparse';
-  trend: 'up' | 'down' | 'stable';
-  confidence: number;
-  series: DailyCostPoint[];
-}
-
-/** 识别账单形态（预付费常表现为月初大额、后续近零） // Logan.Li */
-function detectBillingPattern(series: DailyCostPoint[]): 'steady' | 'front_loaded' | 'sparse' {
-  const active = series.filter((d) => d.cost > 0);
-  if (active.length < 2) return 'sparse';
-  const total = series.reduce((s, d) => s + d.cost, 0);
-  const recent = series.slice(-Math.min(7, series.length));
-  const recentAvg = recent.reduce((s, d) => s + d.cost, 0) / recent.length;
-  const overallAvg = total / series.length;
-  if (overallAvg > 0 && recentAvg < overallAvg * 0.25) return 'front_loaded';
-  return 'steady';
-}
-
-function recentRunRate(series: DailyCostPoint[], days = 7): number {
-  const recent = series.slice(-Math.min(days, series.length));
-  if (recent.length === 0) return 0;
-  return recent.reduce((s, d) => s + d.cost, 0) / recent.length;
-}
-
-/** 本月 Top 服务消费 */
-function topServicesByCost(records: JsonData[], limit = 5): { service: string; cost: number }[] {
-  const start = monthStartDate();
-  const map = new Map<string, number>();
-  records.forEach((r) => {
-    const date = String(r.billing_date || '');
-    if (date < start) return;
-    const svc = String(r.service || i18n.t('aiops.predictutilsts.sh7c2b9d6'));
-    map.set(svc, (map.get(svc) || 0) + (Number(r.cost) || 0));
-  });
-  return [...map.entries()]
-    .sort((a, b) => b[1] - a[1])
-    .slice(0, limit)
-    .map(([service, cost]) => ({ service, cost: Math.round(cost * 100) / 100 }));
-}
-
-/** 基于日费用序列预测月末总费用 */
-function forecastMonthEnd(
-  records: JsonData[],
-  statsMonthCost: number,
-  fallbackTrend: DailyCostPoint[],
-  provider?: string,
-): MonthEndForecast {
-  const start = monthStartDate();
-  const monthSeries = aggregateDailyCosts(records, { startDate: start, provider });
-  const series = monthSeries.length >= 2 ? monthSeries : fallbackTrend;
-  const costs = series.map((d) => d.cost);
-  const { slope, r2 } = linearRegression(costs);
-  const currentMonth =
-    monthSeries.length > 0 ? monthSeries.reduce((s, d) => s + d.cost, 0) : statsMonthCost;
-  const daysLeft = daysRemainingInMonth();
-  const dailyAvg = costs.length > 0 ? costs.reduce((a, b) => a + b, 0) / costs.length : 0;
-  const recentDailyAvg = recentRunRate(series);
-  const billingPattern = detectBillingPattern(series);
-  // 预付费/前高后平：用近 7 日运行费率外推，避免线性回归高估或矛盾 // Logan.Li
-  const extrapolation =
-    billingPattern === 'front_loaded' ? recentDailyAvg * daysLeft : slope * daysLeft;
-  const predictedMonthEnd = Math.max(0, Math.round((currentMonth + extrapolation) * 100) / 100);
-  let confidence = Math.round(55 + r2 * 40);
-  if (billingPattern === 'front_loaded') confidence = Math.max(45, confidence - 15);
-  if (billingPattern === 'sparse') confidence = Math.min(confidence, 50);
-  return {
-    currentMonth,
-    predictedMonthEnd,
-    dailyAvg: Math.round(dailyAvg * 100) / 100,
-    recentDailyAvg: Math.round(recentDailyAvg * 100) / 100,
-    billingPattern,
-    trend: costTrendFromSlope(billingPattern === 'front_loaded' ? recentDailyAvg - dailyAvg : slope),
-    confidence,
-    series: monthSeries.length > 0 ? monthSeries : fallbackTrend,
-  };
-}
-
-/** computeCostPredictions 输入 */
-export interface ComputeCostPredictionsInput {
-  billingStats: JsonData | null;
-  billingRecords: JsonData[];
-  billingAlertRules: JsonData[];
-}
-
-/**
- * 云成本预测：基于账单日粒度时序 + 预算预警规则，推算月末费用与触线风险。
- * // Logan.Li
- */
-export function computeCostPredictions(input: ComputeCostPredictionsInput): ComputedCostPredictions {
-  const { billingStats, billingRecords, billingAlertRules } = input;
-  const empty: ComputedCostPredictions = {
-    summary: {
-      monthCost: 0,
-      predictedMonthEnd: 0,
-      todayCost: 0,
-      yesterdayCost: 0,
-      dailyAvg: 0,
-      daysRemaining: daysRemainingInMonth(),
-      trend: 'stable',
-      confidence: 0,
-      monthTrend: [],
-      providerDist: [],
-    },
-    dimensionForecasts: [],
-    budgetRisks: [],
-    budgetRiskCount: 0,
-    recentDailyAvg: 0,
-    billingPattern: 'sparse',
-    topServices: [],
-  };
-
-  if (!billingStats) return empty;
-
-  const monthTrend: DailyCostPoint[] = Array.isArray(billingStats.month_trend)
-    ? billingStats.month_trend.map((d: JsonData) => ({
-        date: String(d.date || ''),
-        cost: Number(d.cost) || 0,
-      }))
-    : [];
-  const providerDist: { provider: string; cost: number; percent: number }[] = Array.isArray(
-    billingStats.provider_dist,
-  )
-    ? billingStats.provider_dist.map((p: JsonData) => ({
-        provider: String(p.provider || ''),
-        cost: Number(p.cost) || 0,
-        percent: Number(p.percent) || 0,
-      }))
-    : [];
-
-  const overall = forecastMonthEnd(
-    billingRecords,
-    Number(billingStats.month_cost) || 0,
-    monthTrend,
-  );
-
-  const summary: CostPredictionSummary = {
-    monthCost: Number(billingStats.month_cost) || overall.currentMonth,
-    predictedMonthEnd: overall.predictedMonthEnd,
-    todayCost: Number(billingStats.today_cost) || 0,
-    yesterdayCost: Number(billingStats.yesterday_cost) || 0,
-    dailyAvg: overall.dailyAvg,
-    daysRemaining: daysRemainingInMonth(),
-    trend: overall.trend,
-    confidence: overall.confidence,
-    monthTrend: overall.series.length > 0 ? overall.series : monthTrend,
-    providerDist,
-  };
-
-  // 全部 + 各云厂商维度预测
-  const dimensionForecasts: CostDimensionForecast[] = [];
-  const allRatio =
-    summary.predictedMonthEnd > 0 ? summary.monthCost / summary.predictedMonthEnd : 0;
-  dimensionForecasts.push({
-    dimension: i18n.t('aiops.predictutilsts.sa1b2c3d4'),
-    currentMonth: summary.monthCost,
-    predictedMonthEnd: summary.predictedMonthEnd,
-    dailyAvg: summary.dailyAvg,
-    trend: summary.trend,
-    confidence: summary.confidence,
-    recommendation: costRecommendation(summary.trend, allRatio),
-  });
-
-  const providers = new Set<string>();
-  providerDist.forEach((p) => providers.add(p.provider));
-  billingRecords.forEach((r) => {
-    if (r.provider) providers.add(String(r.provider));
-  });
-
-  for (const provider of providers) {
-    if (!provider) continue;
-    const provForecast = forecastMonthEnd(billingRecords, 0, monthTrend, provider);
-    const provMonth =
-      provForecast.currentMonth > 0 ? provForecast.currentMonth : sumMonthCost(billingRecords, provider);
-    const ratio = provForecast.predictedMonthEnd > 0 ? provMonth / provForecast.predictedMonthEnd : 0;
-    dimensionForecasts.push({
-      dimension: PROVIDER_LABELS[provider] || provider,
-      provider,
-      currentMonth: provMonth,
-      predictedMonthEnd: provForecast.predictedMonthEnd,
-      dailyAvg: provForecast.dailyAvg,
-      trend: provForecast.trend,
-      confidence: provForecast.confidence,
-      recommendation: costRecommendation(provForecast.trend, ratio),
-    });
-  }
-
-  dimensionForecasts.sort((a, b) => b.predictedMonthEnd - a.predictedMonthEnd);
-
-  // 预算触线风险
-  const budgetRisks: BudgetRiskForecast[] = [];
-  const activeAlerts: JsonData[] = Array.isArray(billingStats.alerts) ? billingStats.alerts : [];
-
-  for (const rule of billingAlertRules) {
-    if (!rule.enabled) continue;
-    const threshold = Number(rule.threshold) || 0;
-    if (threshold <= 0) continue;
-
-    const metric = (rule.metric === 'daily' ? 'daily' : 'monthly') as 'daily' | 'monthly';
-    const ruleProvider = String(rule.provider || '');
-    const providerLabel = ruleProvider ? PROVIDER_LABELS[ruleProvider] || ruleProvider : i18n.t('aiops.predictutilsts.sb2c3d4e5');
-
-    let current = 0;
-    let predicted = 0;
-    let slope = 0;
-
-    if (metric === 'daily') {
-      current = Number(billingStats.today_cost) || 0;
-      if (ruleProvider) {
-        const today = new Date().toISOString().slice(0, 10);
-        current = billingRecords
-          .filter((r) => r.billing_date === today && r.provider === ruleProvider)
-          .reduce((s, r) => s + (Number(r.cost) || 0), 0);
-      }
-      const recent = aggregateDailyCosts(billingRecords, ruleProvider ? { provider: ruleProvider } : undefined);
-      const costs = recent.slice(-7).map((d) => d.cost);
-      slope = linearRegression(costs).slope;
-      predicted = Math.max(0, current + slope);
-    } else {
-      const provForecast = ruleProvider
-        ? forecastMonthEnd(billingRecords, 0, monthTrend, ruleProvider)
-        : overall;
-      current = ruleProvider ? sumMonthCost(billingRecords, ruleProvider) : summary.monthCost;
-      predicted = provForecast.predictedMonthEnd;
-      const costs = provForecast.series.map((d) => d.cost);
-      slope = linearRegression(costs).slope;
-    }
-
-    const active = activeAlerts.find((a) => a.rule_id === rule.id || a.rule_name === rule.name);
-    if (active) current = Number(active.current) || current;
-
-    const ratio = threshold > 0 ? predicted / threshold : 0;
-    let probability = Math.round(ratio * 50);
-    if (slope > 0.5) probability += 15;
-    if (current >= threshold) probability += 25;
-    probability = Math.max(5, Math.min(92, probability));
-
-    let timeframe = metric === 'daily' ? i18n.t('aiops.predictutilsts.sc3d4e5f6') : i18n.t('aiops.predictutilsts.sd4e5f6g7');
-    if (slope > 0.01 && threshold > current) {
-      const days = (threshold - current) / slope;
-      if (days < 1) timeframe = metric === 'daily' ? i18n.t('aiops.predictutilsts.se5f6g7h8') : i18n.t('aiops.predictutilsts.sf6g7h8i9');
-      else if (days <= 7) timeframe = i18n.t('aiops.predictutilsts.sg7h8i9j0', { days: Math.ceil(days) });
-      else timeframe = i18n.t('aiops.predictutilsts.sh8i9j0k1', { days: Math.ceil(days) });
-    }
-
-    if (probability >= 30 || current >= threshold * 0.7) {
-      budgetRisks.push({
-        ruleName: String(rule.name || i18n.t('aiops.predictutilsts.si9j0k1l2')),
-        provider: providerLabel,
-        metric,
-        threshold,
-        current: Math.round(current * 100) / 100,
-        predicted: Math.round(predicted * 100) / 100,
-        probability,
-        timeframe,
-        suggestedAction: budgetRiskAction(metric, providerLabel),
-      });
-    }
-  }
-
-  budgetRisks.sort((a, b) => b.probability - a.probability);
-  const budgetRiskCount = budgetRisks.filter((r) => r.probability >= 50).length;
-
-  return {
-    summary,
-    dimensionForecasts,
-    budgetRisks,
-    budgetRiskCount,
-    recentDailyAvg: overall.recentDailyAvg,
-    billingPattern: overall.billingPattern,
-    topServices: topServicesByCost(billingRecords),
-  };
-}
-
-/**
- * 构建成本预测「AI 深度解读」prompt。
- * // Logan.Li
- */
-export function buildCostInsightPrompt(data: ComputedCostPredictions): string {
-  const { summary, dimensionForecasts, budgetRisks, recentDailyAvg, billingPattern, topServices } = data;
-  if (summary.monthCost <= 0 && dimensionForecasts.length === 0) return '';
-
-  const patternHint =
-    billingPattern === 'front_loaded'
-      ? i18n.t('aiops.predictutilsts.sj0k1l2m3')
-      : billingPattern === 'sparse'
-        ? i18n.t('aiops.predictutilsts.sk1l2m3n4')
-        : i18n.t('aiops.predictutilsts.sl2m3n4o5');
-
-  const dimLines = dimensionForecasts.slice(0, 6).map(
-    (d) =>
-      `- ${d.dimension}：本月 ${d.currentMonth.toFixed(2)} 元 → 预测月末 ${d.predictedMonthEnd.toFixed(2)} 元（趋势 ${d.trend}）`,
-  );
-  const riskLines =
-    budgetRisks.length > 0
-      ? budgetRisks.slice(0, 5).map(
-          (r) =>
-            `- ${r.ruleName}（${r.provider}）：当前 ${r.current} / 阈值 ${r.threshold}，预测 ${r.predicted}，触线概率 ${r.probability}%`,
-        )
-      : [`- ${i18n.t('aiops.predictutilsts.sm3n4o5p6')}`];
-  const serviceLines =
-    topServices.length > 0
-      ? topServices.map((s) => `- ${s.service}：${s.cost.toFixed(2)} 元`)
-      : [`- ${i18n.t('aiops.predictutilsts.sn4o5p6q7')}`];
-
-  return [
-    i18n.t('aiops.predictutilsts.so5p6q7r8'),
-    '',
-    i18n.t('aiops.predictutilsts.sp6q7r8s9'),
-    `- 本月已花：${summary.monthCost.toFixed(2)} 元`,
-    `- 预测月末：${summary.predictedMonthEnd.toFixed(2)} 元（置信 ${summary.confidence}%）`,
-    `- 全月日均：${summary.dailyAvg.toFixed(2)} 元；近7日日均：${recentDailyAvg.toFixed(2)} 元；剩余 ${summary.daysRemaining} 天`,
-    `- 账单形态：${patternHint}`,
-    '',
-    i18n.t('aiops.predictutilsts.sq7r8s9t0'),
-    ...serviceLines,
-    '',
-    i18n.t('aiops.predictutilsts.sr8s9t0u1'),
-    ...dimLines,
-    '',
-    i18n.t('aiops.predictutilsts.ss9t0u1v2'),
-    ...riskLines,
-    '',
-    i18n.t('aiops.predictutilsts.st0u1v2w3'),
-    i18n.t('aiops.predictutilsts.su1v2w3x4'),
-    i18n.t('aiops.predictutilsts.sv2w3x4y5'),
-    i18n.t('aiops.predictutilsts.sw3x4y5z6'),
-    i18n.t('aiops.predictutilsts.sx4y5z6a7'),
-  ].join('\n');
-}
+// ============ 云成本预测函数 ============
+export {
+  monthStartDate,
+  daysRemainingInMonth,
+  aggregateDailyCosts,
+  sumMonthCost,
+  costRecommendation,
+  computeCostPredictions,
+  buildCostInsightPrompt,
+} from './cost-predict';