feat(grok): 吸收调度阈值、配额解析、批量用量与 CLI 身份

从 personal-dev 取优并接线到 feat/grok-complete-integration：

- 调度阈值：grok_sched_* 写入、ApplyAccountSchedulingThreshold、
  account_scheduling_thresholds 设置与 Settings UI、网关过滤
- 配额头：x-rate-limit 别名、相对秒 reset、tier/entitlement 扩展
- 批量 usage：GetUsageBatch + POST /usage/batch + AccountsView 合并加载
- CLI 身份：xai.cli_identity 统一 pin/UA；service 层 applyGrokCLIHeaders
  与 transport 最终改写对齐；media eligibility 模块落地
- UsageCell：月度 cents→USD 与 30d 进度条展示

保留既有 failure taxonomy / sticky / free recovery 等 HEAD cools。

frontend/src/api/admin/accounts.ts

@@ -317,6 +317,19 @@ export async function getUsage(id: number, source?: 'passive' | 'active', force?
   return data
 }
 
+export interface BatchAccountUsageResponse {
+  usage: Record<string, AccountUsageInfo>
+  errors: Record<string, string>
+}
+
+export async function getBatchUsage(accountIds: number[], force?: boolean): Promise<BatchAccountUsageResponse> {
+  const { data } = await apiClient.post<BatchAccountUsageResponse>('/admin/accounts/usage/batch', {
+    account_ids: accountIds,
+    force: force === true
+  })
+  return data
+}
+
 /**
  * Clear account rate limit status
  * @param id - Account ID
@@ -986,6 +999,7 @@ export const accountsAPI = {
   getStats,
   clearError,
   getUsage,
+  getBatchUsage,
   getTodayStats,
   getBatchTodayStats,
   clearRateLimit,

frontend/src/api/admin/settings.ts

@@ -32,6 +32,38 @@ export type DefaultPlatformQuotasMap = Partial<Record<PlatformType, PlatformQuot
 
 const PLATFORMS: PlatformType[] = ["anthropic", "openai", "gemini", "antigravity", "grok"]
 
+export type SchedulingThresholdPlatformType =
+  | "openai"
+  | "anthropic"
+  | "grok"
+
+export type AccountSchedulingThresholdsMap = Record<SchedulingThresholdPlatformType, number>
+
+export const SCHEDULING_THRESHOLD_PLATFORMS: SchedulingThresholdPlatformType[] = [
+  "openai",
+  "anthropic",
+  "grok",
+]
+
+export function normalizeAccountSchedulingThresholdsMap(
+  input?: Partial<Record<SchedulingThresholdPlatformType, number>> | null,
+): AccountSchedulingThresholdsMap {
+  const result = {} as AccountSchedulingThresholdsMap
+  for (const platform of SCHEDULING_THRESHOLD_PLATFORMS) {
+    const value = input?.[platform]
+    result[platform] = typeof value === "number" && Number.isFinite(value)
+      ? Math.min(100, Math.max(1, Math.trunc(value)))
+      : 100
+  }
+  return result
+}
+
+export function sanitizeAccountSchedulingThresholdsMap(
+  input?: Partial<Record<SchedulingThresholdPlatformType, number>> | null,
+): AccountSchedulingThresholdsMap {
+  return normalizeAccountSchedulingThresholdsMap(input)
+}
+
 /** 归一化为全 4 平台 × 3 窗口（缺失填 null），供模板非空绑定 */
 export function normalizePlatformQuotasMap(input?: DefaultPlatformQuotasMap | null): DefaultPlatformQuotasMap {
   const result: DefaultPlatformQuotasMap = {}
@@ -559,6 +591,9 @@ export interface SystemSettings {
   grok_default_text_model: string;
   grok_cross_client_model_map_enabled: boolean;
 
+  // Per-platform account auto-pause thresholds (100 = disabled)
+  account_scheduling_thresholds: AccountSchedulingThresholdsMap;
+
   // Identity patch configuration (Claude -> Gemini)
   enable_identity_patch: boolean;
   identity_patch_prompt: string;
@@ -876,6 +911,7 @@ export interface UpdateSettingsRequest {
   fallback_model_antigravity?: string;
   grok_default_text_model?: string;
   grok_cross_client_model_map_enabled?: boolean;
+  account_scheduling_thresholds?: AccountSchedulingThresholdsMap;
   enable_identity_patch?: boolean;
   identity_patch_prompt?: string;
   ops_monitoring_enabled?: boolean;

frontend/src/components/account/AccountUsageCell.vue

@@ -393,6 +393,29 @@
           :show-now-when-idle="true"
           color="indigo"
         />
+        <UsageProgressBar
+          v-if="grokMonthlyBillingBar"
+          label="30d"
+          :utilization="grokMonthlyBillingBar.utilization"
+          :resets-at="grokMonthlyBillingBar.resetsAt"
+          :show-now-when-idle="true"
+          color="indigo"
+        />
+        <div
+          v-if="grokBillingMoneySummary"
+          class="flex flex-wrap items-center gap-1 text-[10px] text-gray-500 dark:text-gray-400"
+        >
+          <span :title="t('admin.accounts.usageWindow.grokMonthlyLimit')">
+            {{ t('admin.accounts.usageWindow.grokUsed') }}
+            {{ grokBillingMoneySummary.used }}/{{ grokBillingMoneySummary.limit }}
+          </span>
+          <span
+            v-if="grokBillingMoneySummary.usedPercent != null"
+            class="rounded bg-gray-100 px-1 py-0.5 dark:bg-gray-800"
+          >
+            {{ grokBillingMoneySummary.usedPercent }}%
+          </span>
+        </div>
         <UsageProgressBar
           v-if="!grokWeeklyBillingBar && !grokIsFree && grokRequestQuotaBar"
           :label="t('admin.accounts.usageWindow.grokRequests')"
@@ -652,16 +675,25 @@ const props = withDefaults(
     todayStats?: WindowStats | null
     todayStatsLoading?: boolean
     manualRefreshToken?: number
+    batchedUsage?: AccountUsageInfo | null
+    batchedUsageError?: string | null
+    batchedUsageLoading?: boolean
+    requestBatchedUsage?: ((account: Account, options?: { force?: boolean }) => void) | null
   }>(),
   {
     todayStats: null,
     todayStatsLoading: false,
-    manualRefreshToken: 0
+    manualRefreshToken: 0,
+    batchedUsage: null,
+    batchedUsageError: null,
+    batchedUsageLoading: false,
+    requestBatchedUsage: null
   }
 )
 
 const emit = defineEmits<{
   'account-updated': [account: Account]
+  'usage-loaded': [usage: AccountUsageInfo]
 }>()
 
 const { t } = useI18n()
@@ -674,6 +706,9 @@ const loading = ref(false)
 const activeQueryLoading = ref(false)
 const error = ref<string | null>(null)
 const usageInfo = ref<AccountUsageInfo | null>(null)
+watch(usageInfo, (usage) => {
+  if (usage) emit('usage-loaded', usage)
+})
 const suppressOpenAIUsageRefreshUntil = ref(0)
 const rootRef = ref<HTMLElement | null>(null)
 const isDesktopViewport = ref(
@@ -713,6 +748,8 @@ const shouldFetchUsage = computed(() => {
   return false
 })
 
+const isBatchManaged = computed(() => typeof props.requestBatchedUsage === 'function')
+
 const showGeminiTodayStats = computed(() => {
   return props.account.platform === 'gemini' && props.account.type === 'service_account'
 })
@@ -1094,6 +1131,58 @@ const grokWeeklyBillingBar = computed((): GrokQuotaBarInfo | null => {
     resetsAt: billing.period_end || null
   }
 })
+// Monthly used/limit % from billing probe (used_percent or derived from cents).
+const grokMonthlyBillingBar = computed((): GrokQuotaBarInfo | null => {
+  const billing = grokBilling.value
+  if (!billing) return null
+  let utilization: number | null = null
+  if (billing.used_percent != null && Number.isFinite(billing.used_percent)) {
+    utilization = billing.used_percent
+  } else if (
+    billing.monthly_limit_cents != null &&
+    billing.monthly_limit_cents > 0 &&
+    billing.used_cents != null
+  ) {
+    utilization = (billing.used_cents / billing.monthly_limit_cents) * 100
+  }
+  if (utilization == null) return null
+  // Avoid duplicating the weekly bar when period_type is weekly-only without monthly.
+  if (billing.period_type?.toLowerCase() === 'weekly' && billing.monthly_limit_cents == null) {
+    return null
+  }
+  return {
+    utilization: Math.min(100, Math.max(0, utilization)),
+    resetsAt: billing.billing_period_end || billing.period_end || null
+  }
+})
+const formatGrokMoneyFromCents = (cents?: number | null) => {
+  if (cents == null || Number.isNaN(cents)) return '0'
+  const dollars = cents / 100
+  if (dollars >= 1000) return formatCompactNumber(dollars)
+  if (dollars >= 100) return dollars.toFixed(0)
+  if (dollars >= 10) return dollars.toFixed(1)
+  return dollars.toFixed(2)
+}
+// Absolute monthly used/limit derived from cents fields already on GrokBillingSummary.
+// (personal-dev has separate prepaid/on_demand fields; HEAD only has cents.)
+const grokBillingMoneySummary = computed(() => {
+  const billing = grokBilling.value
+  if (!billing) return null
+  const limit = billing.monthly_limit_cents
+  const used = billing.used_cents
+  if ((limit == null || limit <= 0) && (used == null || used <= 0)) return null
+  let usedPercent: number | null = null
+  if (billing.used_percent != null && Number.isFinite(billing.used_percent)) {
+    usedPercent = Math.round(Math.min(100, Math.max(0, billing.used_percent)))
+  } else if (limit != null && limit > 0 && used != null) {
+    usedPercent = Math.round(Math.min(100, Math.max(0, (used / limit) * 100)))
+  }
+  return {
+    used: formatGrokMoneyFromCents(used ?? 0),
+    limit: formatGrokMoneyFromCents(limit ?? 0),
+    usedPercent
+  }
+})
 const grokPlanLabelIsFree = (value: string) => value.includes('free') || value.includes('basic')
 const grokPlanLabelIsPaid = (value: string) => {
   return value !== '' && !grokPlanLabelIsFree(value) && !value.includes('unknown')
@@ -1136,7 +1225,16 @@ const grokFreeTokenBar = computed(() => {
 })
 const grokQuotaUnknown = computed(() => {
   if (props.account.platform !== 'grok') return false
-  if (grokBilling.value || grokFreeTokenBar.value || grokRequestQuotaBar.value || grokTokenQuotaBar.value) return false
+  if (
+    grokBilling.value ||
+    grokBillingMoneySummary.value ||
+    grokMonthlyBillingBar.value ||
+    grokFreeTokenBar.value ||
+    grokRequestQuotaBar.value ||
+    grokTokenQuotaBar.value
+  ) {
+    return false
+  }
   return usageInfo.value?.grok_quota_snapshot_state !== 'observed'
 })
 const grokQuotaUnknownLabel = computed(() => {
@@ -1273,8 +1371,24 @@ const isAnthropicOAuthOrSetupToken = computed(() => {
   return props.account.platform === 'anthropic' && (props.account.type === 'oauth' || props.account.type === 'setup-token')
 })
 
+const requestParentBatchUsage = (options?: { force?: boolean }) => {
+  if (!isBatchManaged.value || !shouldFetchUsage.value) return
+  props.requestBatchedUsage?.(props.account, options)
+}
+
+const syncManagedUsageState = () => {
+  if (!isBatchManaged.value) return
+  usageInfo.value = props.batchedUsage ?? null
+  error.value = props.batchedUsageError ?? null
+  loading.value = props.batchedUsageLoading === true
+}
+
 const loadUsage = async (options?: { source?: 'passive' | 'active'; bypassCache?: boolean }) => {
   if (!shouldFetchUsage.value) return
+  if (isBatchManaged.value) {
+    requestParentBatchUsage({ force: options?.bypassCache === true })
+    return
+  }
 
   // Check cache
   if (!options?.bypassCache) {
@@ -1529,11 +1643,49 @@ onMounted(() => {
     }
   }
 
+  if (isBatchManaged.value) {
+    syncManagedUsageState()
+    requestParentBatchUsage()
+    return
+  }
+
   if (!shouldAutoLoadUsageOnMount.value) return
   const source = isAnthropicOAuthOrSetupToken.value ? 'passive' : undefined
   requestAutoLoad(source)
 })
 
+watch(
+  () => [props.batchedUsage, props.batchedUsageError, props.batchedUsageLoading, isBatchManaged.value] as const,
+  () => {
+    syncManagedUsageState()
+  },
+  { immediate: true, deep: true }
+)
+
+watch(isBatchManaged, (managed, wasManaged) => {
+  if (managed && !wasManaged) {
+    syncManagedUsageState()
+    requestParentBatchUsage()
+  }
+})
+
+watch(
+  () => [props.account.id, props.account.platform, props.account.type, isBatchManaged.value] as const,
+  ([accountID, platform, accountType, managed], [previousAccountID, previousPlatform, previousAccountType]) => {
+    if (
+      accountID === previousAccountID &&
+      platform === previousPlatform &&
+      accountType === previousAccountType
+    ) {
+      return
+    }
+    if (!managed || !shouldFetchUsage.value) return
+    syncManagedUsageState()
+    requestParentBatchUsage()
+  },
+  { flush: 'post' }
+)
+
 watch(openAIUsageRefreshKey, (nextKey, prevKey) => {
   if (!prevKey || nextKey === prevKey) return
   if (props.account.platform !== 'openai' || props.account.type !== 'oauth') return
@@ -1542,6 +1694,11 @@ watch(openAIUsageRefreshKey, (nextKey, prevKey) => {
     return
   }
 
+  if (isBatchManaged.value) {
+    requestParentBatchUsage({ force: true })
+    return
+  }
+
   _usageCache.delete(props.account.id)
   requestAutoLoad()
 })
@@ -1552,6 +1709,11 @@ watch(
     if (nextToken === prevToken) return
     if (!shouldFetchUsage.value) return
 
+    if (isBatchManaged.value) {
+      requestParentBatchUsage({ force: true })
+      return
+    }
+
     const source = isAnthropicOAuthOrSetupToken.value ? 'passive' : undefined
     _usageCache.delete(props.account.id)
     loadUsage({ source, bypassCache: true }).catch((e) => {

frontend/src/i18n/locales/en/admin/accounts.ts

@@ -1357,6 +1357,8 @@ export default {
         grokTokens: 'Tok',
         grokFreeQuota24hHint: 'Estimated from local token usage over the rolling 24-hour window ({limit} limit)',
         grokWeeklyUsage: 'Weekly {percent}%',
+        grokUsed: 'Used $',
+        grokMonthlyLimit: 'Monthly used / limit (USD from billing cents)',
         grokUnknown: 'Grok quota is unknown until the first upstream response includes xAI rate-limit headers.',
         grokRetryAfter: 'Retry after {time}',
         grokProbe: 'Probe',

frontend/src/i18n/locales/en/admin/settings.ts

@@ -409,7 +409,12 @@ export default {
         title: 'Gateway Scheduling Settings',
         description: 'Control API Key scheduling behavior',
         allowUngroupedKey: 'Allow Ungrouped Key Scheduling',
-        allowUngroupedKeyHint: 'When disabled, API Keys not assigned to any group cannot make requests (403 Forbidden). Keep disabled to ensure all Keys belong to a specific group.'
+        allowUngroupedKeyHint: 'When disabled, API Keys not assigned to any group cannot make requests (403 Forbidden). Keep disabled to ensure all Keys belong to a specific group.',
+        accountSchedulingThresholdsTitle: 'Platform Account Auto-Pause Thresholds',
+        accountSchedulingThresholdsDescription: 'Set per-platform thresholds that automatically pause scheduling for accounts on that platform when their current native usage window reaches the configured percentage.',
+        accountSchedulingThresholdsGlobalHint: 'This is a system-wide global setting and applies to all accounts on that platform.',
+        accountSchedulingThresholdsDisabledHint: 'A value of 100 disables the auto-pause threshold for that platform.',
+        accountSchedulingThresholdsRangeHint: 'Range 1-100, entered as a percentage.'
       },
       upstreamBillingProbe: {
         title: 'Upstream Rate Auto Detection',

frontend/src/i18n/locales/zh/admin/accounts.ts

@@ -401,6 +401,8 @@ export default {
         grokTokens: 'Token',
         grokFreeQuota24hHint: '按 sub2api 近 24 小时本地 Token 用量估算（上限 {limit}）',
         grokWeeklyUsage: '周额度已用 {percent}%',
+        grokUsed: '已用 $',
+        grokMonthlyLimit: '月度已用/上限（由 billing cents 换算 USD）',
         grokUnknown: 'Grok 配额需等待首次上游响应返回 xAI rate-limit 头后显示。',
         grokRetryAfter: '{time} 后重试',
         grokProbe: '探测',

frontend/src/i18n/locales/zh/admin/settings.ts

@@ -402,7 +402,12 @@ export default {
         title: '网关调度设置',
         description: '控制 API Key 的调度行为',
         allowUngroupedKey: '允许未分组 Key 调度',
-        allowUngroupedKeyHint: '关闭后，未分配到任何分组的 API Key 将无法发起请求（返回 403）。建议保持关闭以确保所有 Key 都归属明确的分组。'
+        allowUngroupedKeyHint: '关闭后，未分配到任何分组的 API Key 将无法发起请求（返回 403）。建议保持关闭以确保所有 Key 都归属明确的分组。',
+        accountSchedulingThresholdsTitle: '平台账号自动停调阈值',
+        accountSchedulingThresholdsDescription: '按平台设置账号自动停调阈值，当账号当前原生使用窗口达到配置百分比时，自动停调该平台账号。',
+        accountSchedulingThresholdsGlobalHint: '这是系统级全局设置，对该平台全部账号生效。',
+        accountSchedulingThresholdsDisabledHint: '100 表示禁用该平台的自动停调阈值。',
+        accountSchedulingThresholdsRangeHint: '范围 1-100，按百分比填写。'
       },
       upstreamBillingProbe: {
         title: '上游倍率自动探测',

frontend/src/views/admin/AccountsView.vue

@@ -318,7 +318,12 @@
               :today-stats="todayStatsByAccountId[String(row.id)] ?? null"
               :today-stats-loading="todayStatsLoading"
               :manual-refresh-token="usageManualRefreshToken"
+              :batched-usage="usageBatchByAccountId[String(row.id)] ?? null"
+              :batched-usage-error="usageBatchErrorByAccountId[String(row.id)] ?? null"
+              :batched-usage-loading="usageBatchLoadingByAccountId[String(row.id)] === true"
+              :request-batched-usage="isDesktopViewport ? queueBatchedUsage : null"
               @account-updated="handleAccountUpdated"
+              @usage-loaded="handleAccountUsageLoaded(row.id, $event)"
             />
           </template>
           <template #cell-proxy="{ row }">
@@ -527,7 +532,7 @@ import { extractApiErrorMessage } from '@/utils/apiError'
 import { sanitizeUrl } from '@/utils/url'
 import { getFloatingPanelPosition } from '@/utils/floatingPanel'
 import { formatMultiplier } from '@/utils/formatters'
-import type { Account, AccountPlatform, AccountSchedulerGroupScore, AccountType, Proxy as AccountProxy, AdminGroup, WindowStats, ClaudeModel, UpstreamBillingProbeSnapshot } from '@/types'
+import type { Account, AccountPlatform, AccountSchedulerGroupScore, AccountType, AccountUsageInfo, Proxy as AccountProxy, AdminGroup, WindowStats, ClaudeModel, UpstreamBillingProbeSnapshot } from '@/types'
 
 const { t } = useI18n()
 const appStore = useAppStore()
@@ -692,6 +697,24 @@ const todayStatsReqSeq = ref(0)
 const pendingTodayStatsRefresh = ref(false)
 const usageManualRefreshToken = ref(0)
 
+const desktopViewportQuery = '(min-width: 768px)'
+const isDesktopViewport = ref(
+  typeof window === 'undefined' ? true : window.matchMedia(desktopViewportQuery).matches
+)
+let desktopViewportMediaQuery: MediaQueryList | null = null
+let desktopViewportListener: ((event: MediaQueryListEvent) => void) | null = null
+
+const usageBatchByAccountId = ref<Record<string, AccountUsageInfo | null>>({})
+const usageBatchErrorByAccountId = ref<Record<string, string | null>>({})
+const usageBatchLoadingByAccountId = ref<Record<string, boolean>>({})
+const usageBatchRequestTokenByAccountId = ref<Record<string, number>>({})
+const usageBatchCache = new Map<number, { data: AccountUsageInfo; ts: number }>()
+const USAGE_BATCH_CACHE_TTL = 5 * 60 * 1000
+const pendingUsageBatchIds = new Set<number>()
+let usageBatchFlushTimer: ReturnType<typeof setTimeout> | null = null
+let queuedUsageBatchForce = false
+let usageBatchRequestToken = 0
+
 const buildDefaultTodayStats = (): WindowStats => ({
   requests: 0,
   tokens: 0,
@@ -700,6 +723,138 @@ const buildDefaultTodayStats = (): WindowStats => ({
   user_cost: 0
 })
 
+const accountSupportsBatchUsage = (account: Account) => {
+  if (account.platform === 'anthropic') {
+    return account.type === 'oauth' || account.type === 'setup-token'
+  }
+  if (account.platform === 'gemini') return true
+  if (account.platform === 'antigravity') return account.type === 'oauth'
+  if (account.platform === 'openai') return account.type === 'oauth'
+  if (account.platform === 'grok') return account.type === 'oauth'
+  return false
+}
+
+const setUsageBatchLoading = (accountID: number, loadingState: boolean) => {
+  usageBatchLoadingByAccountId.value = {
+    ...usageBatchLoadingByAccountId.value,
+    [String(accountID)]: loadingState
+  }
+}
+
+const setUsageBatchState = (accountID: number, usage: AccountUsageInfo | null, error: string | null) => {
+  const key = String(accountID)
+  usageBatchByAccountId.value = {
+    ...usageBatchByAccountId.value,
+    [key]: usage
+  }
+  usageBatchErrorByAccountId.value = {
+    ...usageBatchErrorByAccountId.value,
+    [key]: error
+  }
+}
+
+const handleAccountUsageLoaded = (accountID: number, usage: AccountUsageInfo) => {
+  if (usageBatchByAccountId.value[String(accountID)] === usage) return
+  setUsageBatchState(accountID, usage, null)
+}
+
+const flushQueuedUsageBatch = async () => {
+  usageBatchFlushTimer = null
+  const accountIDs = Array.from(pendingUsageBatchIds)
+  const force = queuedUsageBatchForce
+  pendingUsageBatchIds.clear()
+  queuedUsageBatchForce = false
+
+  if (accountIDs.length === 0) return
+
+  const requestTokensByAccount = accountIDs.reduce<Record<string, number>>((acc, accountID) => {
+    acc[String(accountID)] = usageBatchRequestTokenByAccountId.value[String(accountID)] ?? 0
+    return acc
+  }, {})
+
+  try {
+    const result = await adminAPI.accounts.getBatchUsage(accountIDs, force)
+
+    const usageMap = result.usage ?? {}
+    const errorMap = result.errors ?? {}
+    const now = Date.now()
+    const nextUsage = { ...usageBatchByAccountId.value }
+    const nextErrors = { ...usageBatchErrorByAccountId.value }
+    const nextLoading = { ...usageBatchLoadingByAccountId.value }
+
+    for (const accountID of accountIDs) {
+      const key = String(accountID)
+      if ((usageBatchRequestTokenByAccountId.value[key] ?? 0) !== requestTokensByAccount[key]) {
+        continue
+      }
+      const usage = usageMap[key] ?? null
+      nextUsage[key] = usage
+      nextErrors[key] = errorMap[key] ?? null
+      nextLoading[key] = false
+      if (usage) {
+        usageBatchCache.set(accountID, { data: usage, ts: now })
+      } else {
+        usageBatchCache.delete(accountID)
+      }
+    }
+
+    usageBatchByAccountId.value = nextUsage
+    usageBatchErrorByAccountId.value = nextErrors
+    usageBatchLoadingByAccountId.value = nextLoading
+  } catch (error) {
+    const nextErrors = { ...usageBatchErrorByAccountId.value }
+    const nextLoading = { ...usageBatchLoadingByAccountId.value }
+    for (const accountID of accountIDs) {
+      const key = String(accountID)
+      if ((usageBatchRequestTokenByAccountId.value[key] ?? 0) !== requestTokensByAccount[key]) {
+        continue
+      }
+      nextErrors[key] = 'Failed'
+      nextLoading[key] = false
+    }
+    usageBatchErrorByAccountId.value = nextErrors
+    usageBatchLoadingByAccountId.value = nextLoading
+    console.error('Failed to load account usage batch:', error)
+  }
+}
+
+const queueBatchedUsage = (account: Account, options?: { force?: boolean }) => {
+  if (!isDesktopViewport.value) return
+  if (!accountSupportsBatchUsage(account)) return
+
+  const force = options?.force === true
+  const cacheKey = account.id
+  const key = String(cacheKey)
+
+  if (force) {
+    usageBatchCache.delete(cacheKey)
+  } else {
+    const cached = usageBatchCache.get(cacheKey)
+    if (cached && Date.now() - cached.ts < USAGE_BATCH_CACHE_TTL) {
+      setUsageBatchState(cacheKey, cached.data, null)
+      setUsageBatchLoading(cacheKey, false)
+      return
+    }
+  }
+
+  usageBatchErrorByAccountId.value = {
+    ...usageBatchErrorByAccountId.value,
+    [key]: null
+  }
+  usageBatchRequestTokenByAccountId.value = {
+    ...usageBatchRequestTokenByAccountId.value,
+    [key]: ++usageBatchRequestToken
+  }
+  setUsageBatchLoading(cacheKey, true)
+  pendingUsageBatchIds.add(cacheKey)
+  queuedUsageBatchForce = queuedUsageBatchForce || force
+
+  if (usageBatchFlushTimer !== null) return
+  usageBatchFlushTimer = setTimeout(() => {
+    void flushQueuedUsageBatch()
+  }, 0)
+}
+
 const refreshTodayStatsBatch = async () => {
   // Why this checks both columns:
   // - today_stats column shows dedicated today's metrics.
@@ -1084,6 +1239,22 @@ watch(loading, (isLoading, wasLoading) => {
   }
 })
 
+watch(accounts, (rows) => {
+  const visibleIDs = new Set(rows.map((row) => String(row.id)))
+  usageBatchByAccountId.value = Object.fromEntries(
+    Object.entries(usageBatchByAccountId.value).filter(([key]) => visibleIDs.has(key))
+  )
+  usageBatchErrorByAccountId.value = Object.fromEntries(
+    Object.entries(usageBatchErrorByAccountId.value).filter(([key]) => visibleIDs.has(key))
+  )
+  usageBatchLoadingByAccountId.value = Object.fromEntries(
+    Object.entries(usageBatchLoadingByAccountId.value).filter(([key]) => visibleIDs.has(key))
+  )
+  usageBatchRequestTokenByAccountId.value = Object.fromEntries(
+    Object.entries(usageBatchRequestTokenByAccountId.value).filter(([key]) => visibleIDs.has(key))
+  )
+})
+
 watch(upstreamBillingNow, () => {
   if (sortState.sort_by !== 'upstream_billing_rate' || loading.value) return
   if (typeof document !== 'undefined' && document.hidden) return
@@ -2183,6 +2354,19 @@ const handleClickOutside = (event: MouseEvent) => {
 }
 
 onMounted(async () => {
+  if (typeof window !== 'undefined') {
+    desktopViewportMediaQuery = window.matchMedia(desktopViewportQuery)
+    isDesktopViewport.value = desktopViewportMediaQuery.matches
+    desktopViewportListener = (event: MediaQueryListEvent) => {
+      isDesktopViewport.value = event.matches
+    }
+    if (typeof desktopViewportMediaQuery.addEventListener === 'function') {
+      desktopViewportMediaQuery.addEventListener('change', desktopViewportListener)
+    } else {
+      desktopViewportMediaQuery.addListener(desktopViewportListener)
+    }
+  }
+
   load()
   loadUpstreamBillingProbeGlobalState()
   try {
@@ -2205,9 +2389,23 @@ onMounted(async () => {
 })
 
 onUnmounted(() => {
+  if (usageBatchFlushTimer !== null) {
+    clearTimeout(usageBatchFlushTimer)
+    usageBatchFlushTimer = null
+  }
+  pendingUsageBatchIds.clear()
   window.removeEventListener('scroll', handleScroll, true)
   window.removeEventListener('resize', handleViewportResize)
   document.removeEventListener('click', handleClickOutside)
+  if (desktopViewportMediaQuery && desktopViewportListener) {
+    if (typeof desktopViewportMediaQuery.removeEventListener === 'function') {
+      desktopViewportMediaQuery.removeEventListener('change', desktopViewportListener)
+    } else {
+      desktopViewportMediaQuery.removeListener(desktopViewportListener)
+    }
+  }
+  desktopViewportListener = null
+  desktopViewportMediaQuery = null
 })
 </script>
 

frontend/src/views/admin/SettingsView.vue

@@ -4871,6 +4871,78 @@
                 <Toggle v-model="form.allow_ungrouped_key_scheduling" />
               </div>
 
+              <div class="border-t border-gray-100 pt-4 dark:border-dark-700">
+                <div class="mb-3">
+                  <label class="font-medium text-gray-900 dark:text-white">
+                    {{
+                      t(
+                        "admin.settings.scheduling.accountSchedulingThresholdsTitle",
+                      )
+                    }}
+                  </label>
+                  <p class="mt-1 text-sm text-gray-500 dark:text-gray-400">
+                    {{
+                      t(
+                        "admin.settings.scheduling.accountSchedulingThresholdsDescription",
+                      )
+                    }}
+                  </p>
+                  <p class="mt-0.5 text-xs text-gray-500 dark:text-gray-400">
+                    {{
+                      t(
+                        "admin.settings.scheduling.accountSchedulingThresholdsGlobalHint",
+                      )
+                    }}
+                  </p>
+                  <p class="mt-0.5 text-xs text-amber-600 dark:text-amber-400">
+                    {{
+                      t(
+                        "admin.settings.scheduling.accountSchedulingThresholdsDisabledHint",
+                      )
+                    }}
+                  </p>
+                </div>
+                <div class="grid grid-cols-1 gap-4 sm:grid-cols-2 xl:grid-cols-3">
+                  <div
+                    v-for="platform in schedulingThresholdPlatforms"
+                    :key="platform"
+                    class="rounded-lg border border-gray-200 p-4 dark:border-dark-700"
+                  >
+                    <div class="flex items-start justify-between gap-3">
+                      <div>
+                        <label
+                          class="font-mono text-sm font-medium text-gray-900 dark:text-white"
+                        >
+                          {{ platform }}
+                        </label>
+                        <p class="mt-0.5 text-xs text-gray-500 dark:text-gray-400">
+                          {{
+                            t(
+                              "admin.settings.scheduling.accountSchedulingThresholdsRangeHint",
+                            )
+                          }}
+                        </p>
+                      </div>
+                      <span
+                        class="rounded bg-gray-100 px-2 py-0.5 text-[11px] font-medium text-gray-600 dark:bg-dark-700 dark:text-gray-300"
+                      >
+                        %
+                      </span>
+                    </div>
+                    <input
+                      v-model.number="form.account_scheduling_thresholds[platform]"
+                      type="number"
+                      min="1"
+                      max="100"
+                      step="1"
+                      class="input mt-3"
+                      :data-testid="`account-scheduling-threshold-${platform}`"
+                      placeholder="100"
+                    />
+                  </div>
+                </div>
+              </div>
+
               <div
                 v-if="!form.openai_advanced_scheduler_enabled"
                 class="flex items-center justify-between border-t border-gray-100 pt-5 dark:border-dark-700"
@@ -8491,8 +8563,11 @@ import { adminAPI } from "@/api";
 import {
   appendAuthSourceDefaultsToUpdateRequest,
   buildAuthSourceDefaultsState,
+  normalizeAccountSchedulingThresholdsMap,
   normalizePlatformQuotasMap,
+  sanitizeAccountSchedulingThresholdsMap,
   sanitizePlatformQuotasMap,
+  SCHEDULING_THRESHOLD_PLATFORMS,
   defaultWeChatConnectScopesForMode,
   deriveWeChatConnectStoredMode,
   normalizeDefaultSubscriptionSettings,
@@ -9236,8 +9311,11 @@ type SettingsForm = Omit<
   openai_advanced_scheduler_weight_session_sticky: string;
   // 系统全局平台限额 map；form 内始终归一化为全 4 平台对象（模板非空绑定依赖此不变量）
   default_platform_quotas: DefaultPlatformQuotasMap;
+  account_scheduling_thresholds: ReturnType<typeof normalizeAccountSchedulingThresholdsMap>;
 };
 
+const schedulingThresholdPlatforms = SCHEDULING_THRESHOLD_PLATFORMS;
+
 const form = reactive<SettingsForm>({
   registration_enabled: true,
   email_verify_enabled: false,
@@ -9260,6 +9338,7 @@ const form = reactive<SettingsForm>({
   login_agreement_documents: defaultLoginAgreementDocuments(),
   default_balance: 0,
   default_platform_quotas: normalizePlatformQuotasMap() as DefaultPlatformQuotasMap,
+  account_scheduling_thresholds: normalizeAccountSchedulingThresholdsMap(),
   affiliate_rebate_rate: 20,
   affiliate_rebate_freeze_hours: 0,
   affiliate_rebate_duration_days: 0,
@@ -10512,6 +10591,9 @@ async function loadSettings() {
         : defaultLoginAgreementDocuments();
     Object.assign(authSourceDefaults, buildAuthSourceDefaultsState(settings));
     form.default_platform_quotas = normalizePlatformQuotasMap(settings.default_platform_quotas);
+    form.account_scheduling_thresholds = normalizeAccountSchedulingThresholdsMap(
+      settings.account_scheduling_thresholds,
+    );
     form.backend_mode_enabled = settings.backend_mode_enabled;
     form.default_subscriptions = normalizeDefaultSubscriptionSettings(
       settings.default_subscriptions,
@@ -11186,6 +11268,9 @@ async function saveSettings() {
     }
 
     payload.default_platform_quotas = sanitizePlatformQuotasMap(form.default_platform_quotas);
+    payload.account_scheduling_thresholds = sanitizeAccountSchedulingThresholdsMap(
+      form.account_scheduling_thresholds,
+    );
     appendAuthSourceDefaultsToUpdateRequest(payload, authSourceDefaults);
 
     const updated = await settingsStepUp.run(() =>
@@ -11199,6 +11284,9 @@ async function saveSettings() {
     }
     Object.assign(authSourceDefaults, buildAuthSourceDefaultsState(updated));
     form.default_platform_quotas = normalizePlatformQuotasMap(updated.default_platform_quotas);
+    form.account_scheduling_thresholds = normalizeAccountSchedulingThresholdsMap(
+      updated.account_scheduling_thresholds,
+    );
     registrationEmailSuffixWhitelistTags.value =
       normalizeRegistrationEmailSuffixDomains(
         updated.registration_email_suffix_whitelist,