feat(rate-limit): 实现分层 RPM 限速，支持系统默认/用户/独立Key三级配置

- 新增用户级 rate_limit 字段，支持系统默认/用户自定义/不限制三种模式 - 独立 Key 的 rate_limit 语义调整：null=跟随系统默认，0=不限制，>0=自定义 - 实现 UserRpmLimiter 基于 Redis sliding window 的 RPM 限速引擎 - Pipeline 请求流程集成用户级 RPM 检查 - 管理后台和用户面板新增 RPM 限速配置与实时状态查看 - 系统设置新增全局默认 RPM 配置项 - 迁移脚本回填现有 API Key 的 rate_limit 默认值 - 新增用户/Key RPM 状态监控 API 和前端展示 Closes #231 Co-authored-by: LewisPen <LewisPen@nyadoo.com>
2026-09-08 20:20:19 +08:00 · 2026-03-15 14:22:59 +08:00
parent 920a383136
commit f92b0943b5
35 changed files with 2051 additions and 238 deletions
@@ -0,0 +1,54 @@
+"""add user rate_limit and backfill normal api key limits
+
+Revision ID: b7e8f9a0c1d2
+Revises: b7c8d9e0f1a2
+Create Date: 2026-03-13 12:00:00.000000+00:00
+"""
+
+from __future__ import annotations
+
+from collections.abc import Sequence
+
+import sqlalchemy as sa
+from sqlalchemy import inspect
+
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = "b7e8f9a0c1d2"
+down_revision: str | None = "b7c8d9e0f1a2"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+
+def column_exists(table_name: str, column_name: str) -> bool:
+    bind = op.get_bind()
+    inspector = inspect(bind)
+    columns = [c["name"] for c in inspector.get_columns(table_name)]
+    return column_name in columns
+
+
+def upgrade() -> None:
+    if not column_exists("users", "rate_limit"):
+        op.add_column("users", sa.Column("rate_limit", sa.Integer(), nullable=True))
+
+    # 普通 Key 新语义不再允许 NULL；存量 NULL 统一回填为 0（不限制）。
+    op.execute(sa.text("""
+            UPDATE api_keys
+            SET rate_limit = 0
+            WHERE is_standalone = FALSE
+              AND rate_limit IS NULL
+            """))
+
+
+def downgrade() -> None:
+    # 恢复普通 Key 的 rate_limit 为 NULL（与 upgrade 中回填 0 对应）
+    op.execute(sa.text("""
+            UPDATE api_keys
+            SET rate_limit = NULL
+            WHERE is_standalone = FALSE
+              AND rate_limit = 0
+            """))
+
+    if column_exists("users", "rate_limit"):
+        op.drop_column("users", "rate_limit")
@@ -86,6 +86,7 @@ export interface UserExport {
  allowed_providers?: string[] | null
  allowed_api_formats?: string[] | null
  allowed_models?: string[] | null
+  rate_limit?: number | null  // null = 跟随系统默认，0 = 不限制
  model_capability_settings?: Record<string, Record<string, boolean>>
  unlimited?: boolean
  wallet?: BillingSummary | null
@@ -102,7 +103,7 @@ export interface UserApiKeyExport {
  allowed_providers?: string[] | null
  allowed_api_formats?: string[] | null
  allowed_models?: string[] | null
-  rate_limit?: number | null  // null = 无限制
+  rate_limit?: number | null  // legacy/null 兼容；1.3+ standalone null = 跟随系统默认
  concurrent_limit?: number | null
  force_capabilities?: Record<string, boolean>
  is_active: boolean
@@ -349,7 +350,7 @@ export interface AdminApiKey {
  total_requests?: number
  total_tokens?: number
  total_cost_usd?: number
-  rate_limit?: number | null  // null = 无限制
+  rate_limit?: number | null  // null = 跟随系统默认，0 = 不限制
  allowed_providers?: string[] | null  // 允许的提供商列表
  allowed_api_formats?: string[] | null  // 允许的 API 格式列表
  allowed_models?: string[] | null  // 允许的模型列表
@@ -365,7 +366,7 @@ export interface CreateStandaloneApiKeyRequest {
  allowed_providers?: string[] | null
  allowed_api_formats?: string[] | null
  allowed_models?: string[] | null
-  rate_limit?: number | null  // null = 无限制
+  rate_limit?: number | null  // null = 跟随系统默认，0 = 不限制
  expires_at?: string | null  // ISO 日期字符串，如 "2025-12-31"，null = 永不过期
  initial_balance_usd: number | null  // 初始余额，null = 无限制
  unlimited_balance?: boolean | null  // 编辑时仅切换额度模式，不调整余额数值
@@ -142,6 +142,7 @@ export interface ApiKey {
  created_at: string
  total_requests?: number
  total_cost_usd?: number
+  rate_limit?: number | null
  allowed_providers?: ProviderConfig[]
  force_capabilities?: Record<string, boolean> | null  // 强制能力配置
 }
@@ -181,8 +182,8 @@ export const meApi = {
    return response.data
  },

-  async createApiKey(name: string): Promise<ApiKey> {
-    const response = await apiClient.post<ApiKey>('/api/users/me/api-keys', { name })
+  async createApiKey(data: { name: string; rate_limit?: number }): Promise<ApiKey> {
+    const response = await apiClient.post<ApiKey>('/api/users/me/api-keys', data)
    return response.data
  },

@@ -212,6 +213,17 @@ export const meApi = {
    return response.data
  },

+  async updateApiKey(
+    keyId: string,
+    data: { name?: string; rate_limit?: number | null }
+  ): Promise<ApiKey & { message: string }> {
+    const response = await apiClient.put<ApiKey & { message: string }>(
+      `/api/users/me/api-keys/${keyId}`,
+      data
+    )
+    return response.data
+  },
+
  // 使用统计
  async getUsage(params?: {
    start_date?: string
@@ -10,6 +10,7 @@ export interface User {
  allowed_providers: string[] | null  // 允许使用的提供商 ID 列表
  allowed_api_formats: string[] | null  // 允许使用的 API 格式列表
  allowed_models: string[] | null  // 允许使用的模型名称列表
+  rate_limit?: number | null  // null = 跟随系统默认，0 = 不限制
  created_at: string
  updated_at?: string
  last_login_at?: string | null
@@ -25,6 +26,7 @@ export interface CreateUserRequest {
  allowed_providers?: string[] | null
  allowed_api_formats?: string[] | null
  allowed_models?: string[] | null
+  rate_limit?: number | null
 }

 export interface UpdateUserRequest {
@@ -36,6 +38,7 @@ export interface UpdateUserRequest {
  allowed_providers?: string[] | null
  allowed_api_formats?: string[] | null
  allowed_models?: string[] | null
+  rate_limit?: number | null
 }

 export interface ApiKey {
@@ -49,11 +52,16 @@ export interface ApiKey {
  is_active: boolean
  is_locked: boolean  // 管理员锁定标志
  is_standalone: boolean  // 是否为独立余额Key
-  rate_limit?: number  // 速率限制（请求/分钟）
+  rate_limit?: number | null  // 普通Key: 0 = 不限制，历史 null 视为跟随系统默认
  total_requests?: number  // 总请求数
  total_cost_usd?: number  // 总费用
 }

+export interface UpsertUserApiKeyRequest {
+  name?: string
+  rate_limit?: number | null
+}
+
 export const usersApi = {
  async getAllUsers(): Promise<User[]> {
    const response = await apiClient.get<User[]>('/api/admin/users')
@@ -84,8 +92,26 @@ export const usersApi = {
    return response.data.api_keys
  },

-  async createApiKey(userId: string, name?: string): Promise<ApiKey & { key: string }> {
-    const response = await apiClient.post<ApiKey & { key: string }>(`/api/admin/users/${userId}/api-keys`, { name })
+  async createApiKey(
+    userId: string,
+    data: UpsertUserApiKeyRequest
+  ): Promise<ApiKey & { key: string }> {
+    const response = await apiClient.post<ApiKey & { key: string }>(
+      `/api/admin/users/${userId}/api-keys`,
+      data
+    )
+    return response.data
+  },
+
+  async updateApiKey(
+    userId: string,
+    keyId: string,
+    data: UpsertUserApiKeyRequest
+  ): Promise<ApiKey & { message: string }> {
+    const response = await apiClient.put<ApiKey & { message: string }>(
+      `/api/admin/users/${userId}/api-keys/${keyId}`,
+      data
+    )
    return response.data
  },

@@ -96,23 +96,6 @@
              {{ form.expires_at ? '到期后' + (form.auto_delete_on_expiry ? '自动删除' : '仅禁用') + '（当天 23:59 失效）' : '留空表示永不过期' }}
            </p>
          </div>
-
-          <div class="space-y-2">
-            <Label
-              for="form-rate-limit"
-              class="text-sm font-medium"
-            >速率限制 (请求/分钟)</Label>
-            <Input
-              id="form-rate-limit"
-              :model-value="form.rate_limit ?? ''"
-              type="number"
-              min="1"
-              max="10000"
-              placeholder="留空不限制"
-              class="h-10"
-              @update:model-value="(v) => form.rate_limit = parseNumberInput(v, { min: 1, max: 10000 })"
-            />
-          </div>
        </div>

        <!-- 右侧：访问限制 -->
@@ -190,6 +173,36 @@
            </div>
          </div>

+          <div class="space-y-2">
+            <Label
+              for="form-rate-limit"
+              class="text-sm font-medium"
+            >速率限制 (请求/分钟)</Label>
+            <div class="flex items-center gap-3">
+              <div class="flex-1 min-w-0">
+                <Input
+                  v-if="!form.rate_limit_inherited"
+                  id="form-rate-limit"
+                  :model-value="form.rate_limit ?? ''"
+                  type="number"
+                  min="0"
+                  max="10000"
+                  placeholder="0 = 不限速"
+                  class="h-10"
+                  @update:model-value="(v) => form.rate_limit = parseNumberInput(v, { min: 0, max: 10000 })"
+                />
+                <span
+                  v-else
+                  class="flex h-10 w-full items-center rounded-lg border bg-background px-3 text-sm text-muted-foreground opacity-60"
+                >跟随系统</span>
+              </div>
+              <Switch
+                v-model="form.rate_limit_inherited"
+                class="shrink-0"
+              />
+            </div>
+          </div>
+
          <!-- 额度 -->
          <div class="space-y-2">
            <Label class="text-sm font-medium">额度</Label>
@@ -267,7 +280,7 @@ export interface StandaloneKeyFormData {
  initial_balance_usd?: number
  unlimited_balance?: boolean
  expires_at?: string  // ISO 日期字符串，如 "2025-12-31"，undefined = 永不过期
-  rate_limit?: number
+  rate_limit?: number | null
  auto_delete_on_expiry: boolean
  allowed_providers?: string[] | null
  allowed_api_formats?: string[] | null
@@ -280,6 +293,7 @@ interface StandaloneKeyFormState {
  initial_balance_usd?: number
  unlimited_balance?: boolean
  expires_at?: string
+  rate_limit_inherited: boolean
  rate_limit?: number
  auto_delete_on_expiry: boolean
  provider_unrestricted: boolean
@@ -333,6 +347,7 @@ const form = ref<StandaloneKeyFormState>({
  initial_balance_usd: 10,
  unlimited_balance: false,
  expires_at: undefined,
+  rate_limit_inherited: true,
  rate_limit: undefined,
  auto_delete_on_expiry: false,
  provider_unrestricted: true,
@@ -356,6 +371,7 @@ function resetForm() {
    initial_balance_usd: 10,
    unlimited_balance: false,
    expires_at: undefined,
+    rate_limit_inherited: true,
    rate_limit: undefined,
    auto_delete_on_expiry: false,
    provider_unrestricted: true,
@@ -375,7 +391,8 @@ function loadKeyData() {
    initial_balance_usd: props.apiKey.initial_balance_usd,
    unlimited_balance: props.apiKey.initial_balance_usd == null,
    expires_at: props.apiKey.expires_at,
-    rate_limit: props.apiKey.rate_limit,
+    rate_limit_inherited: props.apiKey.rate_limit == null,
+    rate_limit: props.apiKey.rate_limit ?? undefined,
    auto_delete_on_expiry: props.apiKey.auto_delete_on_expiry,
    provider_unrestricted: props.apiKey.allowed_providers == null,
    api_format_unrestricted: props.apiKey.allowed_api_formats == null,
@@ -425,7 +442,7 @@ function handleSubmit() {
    initial_balance_usd: form.value.initial_balance_usd,
    unlimited_balance: form.value.unlimited_balance,
    expires_at: form.value.expires_at,
-    rate_limit: form.value.rate_limit,
+    rate_limit: form.value.rate_limit_inherited ? null : (form.value.rate_limit ?? 0),
    auto_delete_on_expiry: form.value.auto_delete_on_expiry,
    allowed_providers: form.value.provider_unrestricted ? null : [...form.value.allowed_providers],
    allowed_api_formats: form.value.api_format_unrestricted ? null : [...form.value.allowed_api_formats],
@@ -256,6 +256,36 @@
            </div>
          </div>

+          <div class="space-y-2">
+            <Label
+              for="form-rate-limit"
+              class="text-sm font-medium"
+            >速率限制 (请求/分钟)</Label>
+            <div class="flex items-center gap-3">
+              <div class="flex-1 min-w-0">
+                <Input
+                  v-if="!form.rate_limit_inherited"
+                  id="form-rate-limit"
+                  :model-value="form.rate_limit ?? ''"
+                  type="number"
+                  min="0"
+                  max="10000"
+                  placeholder="0 = 不限速"
+                  class="h-10"
+                  @update:model-value="(v) => form.rate_limit = parseNumberInput(v, { min: 0, max: 10000 })"
+                />
+                <span
+                  v-else
+                  class="flex h-10 w-full items-center rounded-lg border bg-background px-3 text-sm text-muted-foreground opacity-60"
+                >跟随系统默认</span>
+              </div>
+              <Switch
+                v-model="form.rate_limit_inherited"
+                class="shrink-0"
+              />
+            </div>
+          </div>
+
          <!-- 额度 -->
          <div class="space-y-2">
            <Label class="text-sm font-medium">额度</Label>
@@ -352,6 +382,7 @@ export interface UserFormData {
  allowed_providers?: string[] | null
  allowed_api_formats?: string[] | null
  allowed_models?: string[] | null
+  rate_limit?: number | null
 }

 const props = defineProps<{
@@ -406,9 +437,11 @@ const form = ref({
  provider_unrestricted: true,
  api_format_unrestricted: true,
  model_unrestricted: true,
+  rate_limit_inherited: true,
  allowed_providers: [] as string[],
  allowed_api_formats: [] as string[],
  allowed_models: [] as string[],
+  rate_limit: undefined as number | undefined,
 })

 function createFieldNonce(): string {
@@ -429,9 +462,11 @@ function resetForm() {
    provider_unrestricted: true,
    api_format_unrestricted: true,
    model_unrestricted: true,
+    rate_limit_inherited: true,
    allowed_providers: [],
    allowed_api_formats: [],
    allowed_models: [],
+    rate_limit: undefined,
  }
 }

@@ -451,9 +486,11 @@ function loadUserData() {
    provider_unrestricted: props.user.allowed_providers == null,
    api_format_unrestricted: props.user.allowed_api_formats == null,
    model_unrestricted: props.user.allowed_models == null,
+    rate_limit_inherited: props.user.rate_limit == null,
    allowed_providers: props.user.allowed_providers ? [...props.user.allowed_providers] : [],
    allowed_api_formats: props.user.allowed_api_formats ? [...props.user.allowed_api_formats] : [],
    allowed_models: props.user.allowed_models ? [...props.user.allowed_models] : [],
+    rate_limit: props.user.rate_limit ?? undefined,
  }
 }

@@ -543,6 +580,7 @@ async function handleSubmit() {
      allowed_models: form.value.model_unrestricted
        ? null
        : [...form.value.allowed_models],
+      rate_limit: form.value.rate_limit_inherited ? null : (form.value.rate_limit ?? 0),
    }

    if (isEditMode.value && props.user?.id) {
@@ -1,6 +1,13 @@
 import { defineStore } from 'pinia'
 import { ref } from 'vue'
-import { usersApi, type User, type CreateUserRequest, type UpdateUserRequest, type ApiKey } from '@/api/users'
+import {
+  usersApi,
+  type User,
+  type CreateUserRequest,
+  type UpdateUserRequest,
+  type ApiKey,
+  type UpsertUserApiKeyRequest
+} from '@/api/users'
 import { parseApiError } from '@/utils/errorParser'

 export const useUsersStore = defineStore('users', () => {
@@ -84,15 +91,28 @@ export const useUsersStore = defineStore('users', () => {
    }
  }

-  async function createApiKey(userId: string, name?: string): Promise<ApiKey> {
+  async function createApiKey(userId: string, data: UpsertUserApiKeyRequest): Promise<ApiKey> {
    try {
-      return await usersApi.createApiKey(userId, name)
+      return await usersApi.createApiKey(userId, data)
    } catch (err: unknown) {
      error.value = parseApiError(err, '创建 API Key 失败')
      throw err
    }
  }

+  async function updateApiKey(
+    userId: string,
+    keyId: string,
+    data: UpsertUserApiKeyRequest
+  ): Promise<ApiKey> {
+    try {
+      return await usersApi.updateApiKey(userId, keyId, data)
+    } catch (err: unknown) {
+      error.value = parseApiError(err, '更新 API Key 失败')
+      throw err
+    }
+  }
+
  async function deleteApiKey(userId: string, keyId: string) {
    try {
      await usersApi.deleteApiKey(userId, keyId)
@@ -121,6 +141,7 @@ export const useUsersStore = defineStore('users', () => {
    deleteUser,
    getUserApiKeys,
    createApiKey,
+    updateApiKey,
    deleteApiKey,
    getFullApiKey
  }
@@ -159,3 +159,25 @@ export function formatHitRate(rate: number | undefined): string {
  if (typeof rate !== 'number' || Number.isNaN(rate)) return '-'
  return `${rate.toFixed(2)}%`
 }
+
+// Rate limit formatting (supports "inherit" semantics: null = inherit system default)
+export function formatRateLimitInheritable(rateLimit?: number | null): string {
+  if (rateLimit == null) return '跟随系统'
+  if (rateLimit === 0) return '不限速'
+  return `${rateLimit}/min`
+}
+
+// Rate limit formatting (simple: null/0 both mean unlimited)
+export function formatRateLimitSimple(rateLimit?: number | null): string {
+  if (rateLimit == null || rateLimit === 0) return '不限速'
+  return `${rateLimit}/min`
+}
+
+// Rate limit state helpers
+export function isRateLimitInherited(rateLimit?: number | null): boolean {
+  return rateLimit == null
+}
+
+export function isRateLimitUnlimited(rateLimit?: number | null): boolean {
+  return rateLimit === 0
+}
@@ -105,8 +105,8 @@
                <TableHead class="w-[240px] h-12 font-semibold">
                  钱包
                </TableHead>
-                <TableHead class="w-[130px] h-12 font-semibold">
-                  使用统计
+                <TableHead class="w-[190px] h-12 font-semibold">
+                  统计/限速
                </TableHead>
                <TableHead class="w-[110px] h-12 font-semibold">
                  有效期
@@ -221,7 +221,23 @@
                      请求: <span class="font-medium text-foreground">{{ (apiKey.total_requests || 0).toLocaleString() }}</span>
                    </div>
                    <div class="text-muted-foreground">
-                      速率: <span class="font-medium text-foreground">{{ apiKey.rate_limit ? `${apiKey.rate_limit}/min` : '未设置' }}</span>
+                      Tokens: <span class="font-medium text-foreground">{{ formatTokens(apiKey.total_tokens || 0) }}</span>
+                    </div>
+                    <div class="flex items-center gap-1 text-muted-foreground">
+                      <span>限速:</span>
+                      <Badge
+                        v-if="isRateLimitInherited(apiKey.rate_limit) || isRateLimitUnlimited(apiKey.rate_limit)"
+                        variant="secondary"
+                        class="h-5 px-1.5 py-0 text-[10px] font-medium"
+                      >
+                        {{ formatRateLimitInheritable(apiKey.rate_limit) }}
+                      </Badge>
+                      <span
+                        v-else
+                        class="font-medium text-foreground"
+                      >
+                        {{ formatRateLimitInheritable(apiKey.rate_limit) }}
+                      </span>
                    </div>
                  </div>
                </TableCell>
@@ -387,6 +403,12 @@
                  >
                    {{ walletStatusLabel(getApiKeyWalletStatus(apiKey.id)) }}
                  </Badge>
+                  <Badge
+                    variant="secondary"
+                    class="h-5 px-1.5 py-0 text-[10px] font-medium"
+                  >
+                    {{ formatRateLimitInheritable(apiKey.rate_limit) }}
+                  </Badge>
                  <Badge
                    v-if="apiKey.auto_delete_on_expiry"
                    variant="secondary"
@@ -429,14 +451,6 @@
                </div>

                <div class="grid grid-cols-2 gap-2.5 text-xs">
-                  <div class="rounded-lg border border-border/50 bg-background/70 p-2.5">
-                    <div class="mb-1 text-muted-foreground">
-                      速率限制
-                    </div>
-                    <div class="font-semibold text-foreground">
-                      {{ apiKey.rate_limit ? `${apiKey.rate_limit}/min` : '未设置' }}
-                    </div>
-                  </div>
                  <div class="rounded-lg border border-border/50 bg-background/70 p-2.5">
                    <div class="mb-1 text-muted-foreground">
                      请求次数
@@ -445,6 +459,14 @@
                      {{ (apiKey.total_requests || 0).toLocaleString() }}
                    </div>
                  </div>
+                  <div class="rounded-lg border border-border/50 bg-background/70 p-2.5">
+                    <div class="mb-1 text-muted-foreground">
+                      Tokens
+                    </div>
+                    <div class="font-semibold text-foreground">
+                      {{ formatTokens(apiKey.total_tokens || 0) }}
+                    </div>
+                  </div>
                  <div class="col-span-2 rounded-lg border border-border/50 bg-background/70 p-2.5">
                    <div class="mb-1 text-muted-foreground">
                      有效期
@@ -663,6 +685,7 @@ import {

 import { StandaloneKeyFormDialog, type StandaloneKeyFormData } from '@/features/api-keys'
 import { parseApiError } from '@/utils/errorParser'
+import { formatTokens, formatRateLimitInheritable, isRateLimitInherited, isRateLimitUnlimited } from '@/utils/format'
 import { log } from '@/utils/logger'

 const { success, error } = useToast()
@@ -1028,7 +1051,7 @@ async function handleKeyFormSubmit(data: StandaloneKeyFormData) {
      const updateData: Partial<CreateStandaloneApiKeyRequest> = {
        name: data.name || undefined,
        unlimited_balance: Boolean(data.unlimited_balance),
-        rate_limit: data.rate_limit ?? null,  // undefined = 无限制，显式传 null
+        rate_limit: data.rate_limit ?? null,  // undefined = 跟随系统默认，显式传 null
        expires_at: data.expires_at || null,  // undefined/空 = 永不过期
        auto_delete_on_expiry: data.auto_delete_on_expiry,
        // 空数组表示清除限制（允许全部），后端会将空数组存为 NULL
@@ -1057,7 +1080,7 @@ async function handleKeyFormSubmit(data: StandaloneKeyFormData) {
      const createData: CreateStandaloneApiKeyRequest = {
        name: data.name || undefined,
        initial_balance_usd: isUnlimited ? null : (data.initial_balance_usd as number),
-        rate_limit: data.rate_limit ?? null,  // undefined = 无限制，显式传 null
+        rate_limit: data.rate_limit ?? null,  // undefined = 跟随系统默认，显式传 null
        expires_at: data.expires_at || null,  // undefined/空 = 永不过期
        auto_delete_on_expiry: data.auto_delete_on_expiry,
        // 空数组表示不设置限制（允许全部），后端会将空数组存为 NULL
@@ -180,7 +180,7 @@
                钱包
              </TableHead>
              <TableHead class="w-[170px] h-12 font-semibold">
-                使用统计
+                统计/限速
              </TableHead>
              <TableHead class="w-[110px] h-12 font-semibold">
                创建时间
@@ -258,26 +258,42 @@
                </div>
              </TableCell>
              <TableCell class="py-4">
-                <div
-                  v-if="userStats[user.id]"
-                  class="space-y-1 text-xs"
-                >
-                  <div class="flex items-center text-muted-foreground">
-                    <span class="w-14">请求:</span>
-                    <span class="font-medium text-foreground">{{ formatNumber(userStats[user.id]?.request_count) }}</span>
+                <div class="space-y-1 text-xs">
+                  <template v-if="userStats[user.id]">
+                    <div class="flex items-center text-muted-foreground">
+                      <span class="w-14">请求:</span>
+                      <span class="font-medium text-foreground">{{ formatNumber(userStats[user.id]?.request_count) }}</span>
+                    </div>
+                    <div class="flex items-center text-muted-foreground">
+                      <span class="w-14">Tokens:</span>
+                      <span class="font-medium text-foreground">{{ formatTokens(userStats[user.id]?.total_tokens ?? 0) }}</span>
+                    </div>
+                  </template>
+                  <div
+                    v-else
+                    class="flex items-center text-muted-foreground"
+                  >
+                    <span class="w-14">统计:</span>
+                    <span v-if="loadingStats">加载中...</span>
+                    <span v-else>无数据</span>
                  </div>
                  <div class="flex items-center text-muted-foreground">
-                    <span class="w-14">Tokens:</span>
-                    <span class="font-medium text-foreground">{{ formatTokens(userStats[user.id]?.total_tokens ?? 0) }}</span>
+                    <span class="w-14">限速:</span>
+                    <Badge
+                      v-if="isRateLimitInherited(user.rate_limit) || isRateLimitUnlimited(user.rate_limit)"
+                      variant="secondary"
+                      class="h-5 px-1.5 py-0 text-[10px] font-medium"
+                    >
+                      {{ formatRateLimitInheritable(user.rate_limit) }}
+                    </Badge>
+                    <span
+                      v-else
+                      class="font-medium text-foreground"
+                    >
+                      {{ formatRateLimitInheritable(user.rate_limit) }}
+                    </span>
                  </div>
                </div>
-                <div
-                  v-else
-                  class="text-xs text-muted-foreground"
-                >
-                  <span v-if="loadingStats">加载中...</span>
-                  <span v-else>无数据</span>
-                </div>
              </TableCell>
              <TableCell class="py-4 text-xs text-muted-foreground">
                {{ formatDate(user.created_at) }}
@@ -436,6 +452,12 @@
                >
                  {{ walletStatusLabel(getUserWalletStatus(user.id)) }}
                </Badge>
+                <Badge
+                  variant="secondary"
+                  class="h-5 px-1.5 py-0 text-[10px] font-medium"
+                >
+                  {{ formatRateLimitInheritable(user.rate_limit) }}
+                </Badge>
              </div>

              <div class="rounded-xl border border-border/60 bg-muted/40 p-3.5">
@@ -633,6 +655,12 @@
                    >
                      独立余额
                    </Badge>
+                    <Badge
+                      variant="secondary"
+                      class="text-xs"
+                    >
+                      {{ formatRateLimitSimple(apiKey.rate_limit) }}
+                    </Badge>
                  </div>
                  <div class="flex items-center gap-1 mt-0.5">
                    <code class="text-xs font-mono text-muted-foreground">
@@ -658,6 +686,15 @@
                    ${{ (apiKey.total_cost_usd || 0).toFixed(4) }}
                  </div>
                </div>
+                <Button
+                  variant="ghost"
+                  size="icon"
+                  class="h-8 w-8"
+                  title="编辑"
+                  @click="openEditUserApiKeyDialog(apiKey)"
+                >
+                  <SquarePen class="h-4 w-4" />
+                </Button>
                <Button
                  variant="ghost"
                  size="icon"
@@ -718,13 +755,87 @@
        <Button
          class="h-10 px-5"
          :disabled="creatingApiKey"
-          @click="createApiKey"
+          @click="openCreateUserApiKeyDialog"
        >
          {{ creatingApiKey ? '创建中...' : '创建' }}
        </Button>
      </template>
    </Dialog>

+    <Dialog
+      v-model="showUserApiKeyFormDialog"
+      size="lg"
+    >
+      <template #header>
+        <div class="border-b border-border px-6 py-4">
+          <div class="flex items-center gap-3">
+            <div class="flex h-9 w-9 items-center justify-center rounded-lg bg-kraft/10 flex-shrink-0">
+              <Key class="h-5 w-5 text-kraft" />
+            </div>
+            <div class="flex-1 min-w-0">
+              <h3 class="text-lg font-semibold text-foreground leading-tight">
+                {{ editingUserApiKey ? '编辑 API Key' : '创建 API Key' }}
+              </h3>
+              <p class="text-xs text-muted-foreground">
+                {{ editingUserApiKey ? '更新用户 API Key 的名称和速率限制' : '为用户创建新的 API Key' }}
+              </p>
+            </div>
+          </div>
+        </div>
+      </template>
+
+      <div class="space-y-4">
+        <div class="space-y-2">
+          <Label
+            for="admin-user-key-name"
+            class="text-sm font-medium"
+          >密钥名称</Label>
+          <Input
+            id="admin-user-key-name"
+            v-model="userApiKeyForm.name"
+            class="h-10"
+            placeholder="例如：生产环境 Key"
+          />
+        </div>
+        <div class="space-y-2">
+          <Label
+            for="admin-user-key-rate-limit"
+            class="text-sm font-medium"
+          >速率限制 (请求/分钟)</Label>
+          <Input
+            id="admin-user-key-rate-limit"
+            :model-value="userApiKeyForm.rate_limit ?? ''"
+            type="number"
+            min="0"
+            max="10000"
+            class="h-10"
+            placeholder="留空不限"
+            @update:model-value="(v) => userApiKeyForm.rate_limit = parseNumberInput(v, { min: 0, max: 10000 })"
+          />
+          <p class="text-xs text-muted-foreground">
+            留空表示不限制
+          </p>
+        </div>
+      </div>
+
+      <template #footer>
+        <Button
+          variant="outline"
+          class="h-10 px-5"
+          @click="closeUserApiKeyFormDialog"
+        >
+          取消
+        </Button>
+        <Button
+          class="h-10 px-5"
+          :disabled="creatingApiKey"
+          @click="submitUserApiKeyForm"
+        >
+          {{ creatingApiKey ? (editingUserApiKey ? '保存中...' : '创建中...') : (editingUserApiKey ? '保存' : '创建') }}
+        </Button>
+      </template>
+    </Dialog>
+
    <WalletOpsDrawer
      :open="showWalletActionDialogState"
      :wallet="walletActionTarget?.wallet || null"
@@ -849,6 +960,8 @@ import {
 import UserFormDialog, { type UserFormData } from '@/features/users/components/UserFormDialog.vue'
 import WalletOpsDrawer from '@/features/wallet/components/WalletOpsDrawer.vue'
 import { parseApiError } from '@/utils/errorParser'
+import { formatTokens, formatRateLimitInheritable, formatRateLimitSimple, isRateLimitInherited, isRateLimitUnlimited } from '@/utils/format'
+import { parseNumberInput } from '@/utils/form'
 import { log } from '@/utils/logger'

 const { success, error } = useToast()
@@ -864,11 +977,17 @@ const userFormDialogRef = ref<InstanceType<typeof UserFormDialog>>()
 // API Keys 对话框状态
 const showApiKeysDialog = ref(false)
 const showNewApiKeyDialog = ref(false)
+const showUserApiKeyFormDialog = ref(false)
 const selectedUser = ref<User | null>(null)
 const userApiKeys = ref<ApiKey[]>([])
 const newApiKey = ref('')
 const creatingApiKey = ref(false)
 const apiKeyInput = ref<HTMLInputElement>()
+const editingUserApiKey = ref<ApiKey | null>(null)
+const userApiKeyForm = ref({
+  name: '',
+  rate_limit: undefined as number | undefined,
+})

 // 用户统计
 const userStats = ref<Record<string, UsageByUser>>({})
@@ -979,15 +1098,6 @@ async function loadUserWallets() {
  }
 }

-function formatTokens(tokens: number): string {
-  if (tokens >= 1000000) {
-    return `${(tokens / 1000000).toFixed(1)}M`
-  } else if (tokens >= 1000) {
-    return `${(tokens / 1000).toFixed(1)}K`
-  }
-  return tokens.toString()
-}
-
 function formatNumber(value?: number | null): string {
  const numericValue = typeof value === 'number' && Number.isFinite(value) ? value : 0
  return numericValue.toLocaleString()
@@ -1071,7 +1181,8 @@ function editUser(user: User) {
    is_active: user.is_active,
    allowed_providers: user.allowed_providers == null ? null : [...user.allowed_providers],
    allowed_api_formats: user.allowed_api_formats == null ? null : [...user.allowed_api_formats],
-    allowed_models: user.allowed_models == null ? null : [...user.allowed_models]
+    allowed_models: user.allowed_models == null ? null : [...user.allowed_models],
+    rate_limit: user.rate_limit ?? null
  }
  showUserFormDialog.value = true
 }
@@ -1093,7 +1204,8 @@ async function handleUserFormSubmit(data: UserFormData & { password?: string; un
        role: data.role,
        allowed_providers: data.allowed_providers,
        allowed_api_formats: data.allowed_api_formats,
-        allowed_models: data.allowed_models
+        allowed_models: data.allowed_models,
+        rate_limit: data.rate_limit ?? null
      }
      if (data.password) {
        updateData.password = data.password
@@ -1112,7 +1224,8 @@ async function handleUserFormSubmit(data: UserFormData & { password?: string; un
        role: data.role,
        allowed_providers: data.allowed_providers,
        allowed_api_formats: data.allowed_api_formats,
-        allowed_models: data.allowed_models
+        allowed_models: data.allowed_models,
+        rate_limit: data.rate_limit ?? null
      })
      // 如果创建时指定为禁用，则更新状态
      if (data.is_active === false && newUser) {
@@ -1145,20 +1258,61 @@ async function loadUserApiKeys(userId: string) {
  }
 }

-async function createApiKey() {
+function openCreateUserApiKeyDialog() {
+  userApiKeyForm.value = {
+    name: `Key-${new Date().toISOString().split('T')[0]}`,
+    rate_limit: undefined,
+  }
+  editingUserApiKey.value = null
+  showUserApiKeyFormDialog.value = true
+}
+
+function openEditUserApiKeyDialog(apiKey: ApiKey) {
+  editingUserApiKey.value = apiKey
+  userApiKeyForm.value = {
+    name: apiKey.name || '',
+    rate_limit: apiKey.rate_limit ?? undefined,
+  }
+  showUserApiKeyFormDialog.value = true
+}
+
+function closeUserApiKeyFormDialog() {
+  showUserApiKeyFormDialog.value = false
+  editingUserApiKey.value = null
+  userApiKeyForm.value = {
+    name: '',
+    rate_limit: undefined,
+  }
+}
+
+async function submitUserApiKeyForm() {
  if (!selectedUser.value) return
+  if (!userApiKeyForm.value.name.trim()) {
+    error('请输入密钥名称', editingUserApiKey.value ? '更新 API Key 失败' : '创建 API Key 失败')
+    return
+  }

  creatingApiKey.value = true
  try {
-    const response = await usersStore.createApiKey(
-      selectedUser.value.id,
-      `Key-${new Date().toISOString().split('T')[0]}`
-    )
-    newApiKey.value = response.key || ''
-    showNewApiKeyDialog.value = true
+    if (editingUserApiKey.value) {
+      await usersStore.updateApiKey(selectedUser.value.id, editingUserApiKey.value.id, {
+        name: userApiKeyForm.value.name,
+        rate_limit: userApiKeyForm.value.rate_limit ?? 0,
+      })
+      success('API Key已更新')
+    } else {
+      const response = await usersStore.createApiKey(selectedUser.value.id, {
+        name: userApiKeyForm.value.name,
+        rate_limit: userApiKeyForm.value.rate_limit ?? 0,
+      })
+      newApiKey.value = response.key || ''
+      showNewApiKeyDialog.value = true
+      success('API Key创建成功')
+    }
    await loadUserApiKeys(selectedUser.value.id)
+    closeUserApiKeyFormDialog()
  } catch (err: unknown) {
-    error(parseApiError(err, '未知错误'), '创建 API Key 失败')
+    error(parseApiError(err, '未知错误'), editingUserApiKey.value ? '更新 API Key 失败' : '创建 API Key 失败')
  } finally {
    creatingApiKey.value = false
  }
@@ -39,7 +39,7 @@
          for="rate-limit"
          class="block text-sm font-medium"
        >
-          每分钟请求限制
+          默认速率限制 (请求/分钟)
        </Label>
        <Input
          id="rate-limit"
@@ -50,7 +50,7 @@
          @update:model-value="$emit('update:rateLimitPerMinute', Number($event))"
        />
        <p class="mt-1 text-xs text-muted-foreground">
-          0 表示不限制
+          0 表示默认不限制；未单独配置的用户和独立 Key 会跟随这里
        </p>
      </div>

@@ -20,7 +20,7 @@
              size="icon"
              class="h-8 w-8"
              title="创建新 API Key"
-              @click="showCreateDialog = true"
+              @click="openCreateApiKeyDialog"
            >
              <Plus class="w-3.5 h-3.5" />
            </Button>
@@ -56,7 +56,7 @@
            <Button
              size="lg"
              class="shadow-lg shadow-primary/20"
-              @click="showCreateDialog = true"
+              @click="openCreateApiKeyDialog"
            >
              <Plus class="mr-2 h-4 w-4" />
              创建新 API Key
@@ -157,17 +157,23 @@
                <div class="flex flex-col items-center gap-1">
                  <Badge
                    :variant="apiKey.is_active ? 'success' : 'secondary'"
-                    class="font-medium px-3 py-1"
+                    class="h-5 px-2 py-0 text-[10px] font-medium"
                  >
                    {{ apiKey.is_active ? '活跃' : '禁用' }}
                  </Badge>
                  <Badge
                    v-if="apiKey.is_locked"
                    variant="warning"
-                    class="font-medium text-[10px]"
+                    class="h-5 px-2 py-0 text-[10px] font-medium"
                  >
                    已锁定
                  </Badge>
+                  <Badge
+                    variant="secondary"
+                    class="h-5 px-2 py-0 text-[10px] font-medium"
+                  >
+                    {{ formatRateLimitSimple(apiKey.rate_limit) }}
+                  </Badge>
                </div>
              </TableCell>

@@ -179,6 +185,16 @@
              <!-- 操作按钮 -->
              <TableCell class="py-4">
                <div class="flex justify-center gap-1">
+                  <Button
+                    variant="ghost"
+                    size="icon"
+                    class="h-8 w-8"
+                    :title="apiKey.is_locked ? '已锁定' : '编辑'"
+                    :disabled="apiKey.is_locked"
+                    @click="openEditApiKeyDialog(apiKey)"
+                  >
+                    <SquarePen class="h-4 w-4" />
+                  </Button>
                  <Button
                    variant="ghost"
                    size="icon"
@@ -237,8 +253,24 @@
                >
                  已锁定
                </Badge>
+                <Badge
+                  variant="secondary"
+                  class="text-[10px] px-1.5 py-0"
+                >
+                  {{ formatRateLimitSimple(apiKey.rate_limit) }}
+                </Badge>
              </div>
              <div class="flex items-center gap-0.5 flex-shrink-0">
+                <Button
+                  variant="ghost"
+                  size="icon"
+                  class="h-7 w-7"
+                  :title="apiKey.is_locked ? '已锁定' : '编辑'"
+                  :disabled="apiKey.is_locked"
+                  @click="openEditApiKeyDialog(apiKey)"
+                >
+                  <SquarePen class="h-3.5 w-3.5" />
+                </Button>
                <Button
                  variant="ghost"
                  size="icon"
@@ -288,6 +320,10 @@
                <span class="text-foreground font-medium">
                  {{ formatNumber(apiKey.total_requests || 0) }} 次
                </span>
+                <span class="text-muted-foreground">•</span>
+                <span class="text-muted-foreground">
+                  {{ formatRateLimitSimple(apiKey.rate_limit) }}
+                </span>
              </div>
            </div>
          </div>
@@ -316,10 +352,10 @@
            </div>
            <div class="flex-1 min-w-0">
              <h3 class="text-lg font-semibold text-foreground leading-tight">
-                创建 API 密钥
+                {{ editingApiKey ? '编辑 API 密钥' : '创建 API 密钥' }}
              </h3>
              <p class="text-xs text-muted-foreground">
-                创建一个新的密钥用于访问 API 服务
+                {{ editingApiKey ? '更新密钥名称和速率限制' : '创建一个新的密钥用于访问 API 服务' }}
              </p>
            </div>
          </div>
@@ -344,26 +380,46 @@
            给密钥起一个有意义的名称方便识别
          </p>
        </div>
+
+        <div class="space-y-2">
+          <Label
+            for="key-rate-limit"
+            class="text-sm font-semibold"
+          >速率限制 (请求/分钟)</Label>
+          <Input
+            id="key-rate-limit"
+            :model-value="newKeyRateLimit ?? ''"
+            type="number"
+            min="0"
+            max="10000"
+            placeholder="留空不限"
+            class="h-11 border-border/60"
+            @update:model-value="(v) => newKeyRateLimit = parseNumberInput(v, { min: 0, max: 10000 })"
+          />
+          <p class="text-xs text-muted-foreground">
+            留空不限
+          </p>
+        </div>
      </div>

      <template #footer>
        <Button
          variant="outline"
          class="h-11 px-6"
-          @click="showCreateDialog = false"
+          @click="closeApiKeyDialog"
        >
          取消
        </Button>
        <Button
          class="h-11 px-6 shadow-lg shadow-primary/20"
          :disabled="creating"
-          @click="createApiKey"
+          @click="saveApiKey"
        >
          <Loader2
            v-if="creating"
            class="animate-spin h-4 w-4 mr-2"
          />
-          {{ creating ? '创建中...' : '创建' }}
+          {{ creating ? (editingApiKey ? '保存中...' : '创建中...') : (editingApiKey ? '保存' : '创建') }}
        </Button>
      </template>
    </Dialog>
@@ -455,10 +511,12 @@ import {
  TableRow
 } from '@/components/ui'
 import RefreshButton from '@/components/ui/refresh-button.vue'
-import { Plus, Key, Copy, Trash2, Loader2, Activity, CheckCircle, Power } from 'lucide-vue-next'
+import { Plus, Key, Copy, Trash2, Loader2, Activity, CheckCircle, Power, SquarePen } from 'lucide-vue-next'
 import { useToast } from '@/composables/useToast'
 import { log } from '@/utils/logger'
 import { parseApiError } from '@/utils/errorParser'
+import { formatRateLimitSimple } from '@/utils/format'
+import { parseNumberInput } from '@/utils/form'
 import { getErrorStatus } from '@/types/api-error'
 import { computed } from 'vue'

@@ -483,8 +541,10 @@ const showKeyDialog = ref(false)
 const showDeleteDialog = ref(false)

 const newKeyName = ref('')
+const newKeyRateLimit = ref<number | undefined>(undefined)
 const newKeyValue = ref('')
 const keyToDelete = ref<ApiKey | null>(null)
+const editingApiKey = ref<ApiKey | null>(null)

 onMounted(() => {
  loadApiKeys()
@@ -509,7 +569,28 @@ async function loadApiKeys() {
  }
 }

-async function createApiKey() {
+function openEditApiKeyDialog(apiKey: ApiKey) {
+  editingApiKey.value = apiKey
+  newKeyName.value = apiKey.name || ''
+  newKeyRateLimit.value = apiKey.rate_limit ?? undefined
+  showCreateDialog.value = true
+}
+
+function openCreateApiKeyDialog() {
+  editingApiKey.value = null
+  newKeyName.value = ''
+  newKeyRateLimit.value = undefined
+  showCreateDialog.value = true
+}
+
+function closeApiKeyDialog() {
+  showCreateDialog.value = false
+  editingApiKey.value = null
+  newKeyName.value = ''
+  newKeyRateLimit.value = undefined
+}
+
+async function saveApiKey() {
  if (!newKeyName.value.trim()) {
    showError('请输入密钥名称')
    return
@@ -517,16 +598,26 @@ async function createApiKey() {

  creating.value = true
  try {
-    const newKey = await meApi.createApiKey(newKeyName.value)
-    newKeyValue.value = newKey.key || ''
-    showCreateDialog.value = false
-    showKeyDialog.value = true
-    newKeyName.value = ''
+    if (editingApiKey.value) {
+      await meApi.updateApiKey(editingApiKey.value.id, {
+        name: newKeyName.value,
+        rate_limit: newKeyRateLimit.value ?? 0,
+      })
+      success('API 密钥更新成功')
+    } else {
+      const newKey = await meApi.createApiKey({
+        name: newKeyName.value,
+        rate_limit: newKeyRateLimit.value ?? 0,
+      })
+      newKeyValue.value = newKey.key || ''
+      showKeyDialog.value = true
+      success('API 密钥创建成功')
+    }
+    closeApiKeyDialog()
    await loadApiKeys()
-    success('API 密钥创建成功')
  } catch (error) {
-    log.error('创建 API 密钥失败:', error)
-    showError('创建 API 密钥失败')
+    log.error(editingApiKey.value ? '更新 API 密钥失败:' : '创建 API 密钥失败:', error)
+    showError(editingApiKey.value ? '更新 API 密钥失败' : '创建 API 密钥失败')
  } finally {
    creating.value = false
  }
@@ -22,7 +22,7 @@ from src.core.exceptions import InvalidRequestException, NotFoundException
 from src.core.logger import logger
 from src.database import get_db, get_db_context
 from src.models.api import CreateApiKeyRequest
-from src.models.database import ApiKey, Wallet
+from src.models.database import ApiKey, Usage, Wallet
 from src.services.user.apikey import ApiKeyService
 from src.services.user.bulk_cleanup import pre_clean_api_key
 from src.services.wallet import WalletService
@@ -70,7 +70,9 @@ router = APIRouter(prefix="/api/admin/api-keys", tags=["Admin - API Keys (Standa
 pipeline = get_pipeline()


-def _serialize_standalone_key_item(api_key: ApiKey) -> dict[str, Any]:
+def _serialize_standalone_key_item(
+    api_key: ApiKey, *, total_tokens: int | None = None
+) -> dict[str, Any]:
    return {
        "id": api_key.id,
        "user_id": api_key.user_id,
@@ -79,6 +81,7 @@ def _serialize_standalone_key_item(api_key: ApiKey) -> dict[str, Any]:
        "is_active": api_key.is_active,
        "is_standalone": api_key.is_standalone,
        "total_requests": api_key.total_requests,
+        "total_tokens": int(total_tokens or 0),
        "total_cost_usd": float(api_key.total_cost_usd or 0),
        "rate_limit": api_key.rate_limit,
        "allowed_providers": api_key.allowed_providers,
@@ -116,8 +119,27 @@ def _list_standalone_api_keys_sync(
            for api_key in api_keys:
                db.refresh(api_key)

+        token_map: dict[str, int] = {}
+        if api_keys:
+            stats_rows = (
+                db.query(
+                    Usage.api_key_id,
+                    func.sum(Usage.total_tokens).label("total_tokens"),
+                )
+                .filter(Usage.api_key_id.in_([api_key.id for api_key in api_keys]))
+                .group_by(Usage.api_key_id)
+                .all()
+            )
+            token_map = {row.api_key_id: int(row.total_tokens or 0) for row in stats_rows}
+
        return {
-            "api_keys": [_serialize_standalone_key_item(api_key) for api_key in api_keys],
+            "api_keys": [
+                _serialize_standalone_key_item(
+                    api_key,
+                    total_tokens=token_map.get(api_key.id, 0),
+                )
+                for api_key in api_keys
+            ],
            "total": total,
            "limit": limit,
            "skip": skip,
@@ -385,7 +407,7 @@ async def create_standalone_api_key(
    - `allowed_providers`: 可选，允许使用的提供商列表
    - `allowed_api_formats`: 可选，允许使用的 API 格式列表
    - `allowed_models`: 可选，允许使用的模型列表
-    - `rate_limit`: 可选，速率限制配置（请求数/秒）
+    - `rate_limit`: 可选，每分钟请求限制（null 表示跟随系统默认，0 表示不限制）
    - `expire_days`: 可选，过期天数（与 expires_at 二选一）
    - `expires_at`: 可选，过期时间（ISO 格式或 YYYY-MM-DD 格式，优先级高于 expire_days）
    - `auto_delete_on_expiry`: 可选，过期后是否自动删除
@@ -421,7 +443,7 @@ async def update_api_key(
    **请求体字段**:
    - `name`: 可选，API Key 的名称
    - `unlimited_balance`: 可选，是否无限余额（true=无限，false=有限，不修改余额数值）
-    - `rate_limit`: 可选，速率限制配置（null 表示无限制）
+    - `rate_limit`: 可选，每分钟请求限制（null 表示跟随系统默认，0 表示不限制）
    - `allowed_providers`: 可选，允许使用的提供商列表
    - `allowed_api_formats`: 可选，允许使用的 API 格式列表
    - `allowed_models`: 可选，允许使用的模型列表
@@ -683,6 +705,14 @@ class AdminGetKeyDetailAdapter(AdminApiAdapter):
            "is_active": api_key.is_active,
            "is_standalone": api_key.is_standalone,
            "total_requests": api_key.total_requests,
+            "total_tokens": int(
+                (
+                    db.query(func.sum(Usage.total_tokens))
+                    .filter(Usage.api_key_id == api_key.id)
+                    .scalar()
+                )
+                or 0
+            ),
            "total_cost_usd": float(api_key.total_cost_usd or 0),
            "rate_limit": api_key.rate_limit,
            "allowed_providers": api_key.allowed_providers,
@@ -2252,6 +2252,7 @@ class AdminExportUsersAdapter(AdminApiAdapter):
                    "allowed_providers": user.allowed_providers,
                    "allowed_api_formats": user.allowed_api_formats,
                    "allowed_models": user.allowed_models,
+                    "rate_limit": user.rate_limit,
                    "model_capability_settings": user.model_capability_settings,
                    "unlimited": wallet_service.is_unlimited_wallet(wallet),
                    "wallet": (wallet_service.serialize_wallet_summary(wallet) if wallet else None),
@@ -2265,7 +2266,7 @@ class AdminExportUsersAdapter(AdminApiAdapter):
        standalone_keys_data = [self._serialize_api_key(key, db=db) for key in standalone_keys]

        return {
-            "version": "1.2",
+            "version": "1.3",
            "exported_at": datetime.now(timezone.utc).isoformat(),
            "users": users_data,
            "standalone_keys": standalone_keys_data,
@@ -2273,6 +2274,17 @@ class AdminExportUsersAdapter(AdminApiAdapter):


 class AdminImportUsersAdapter(AdminApiAdapter):
+    @staticmethod
+    def _is_legacy_users_export(version: object) -> bool:
+        if version is None:
+            return True
+        normalized = str(version).strip()
+        try:
+            parts = normalized.split(".")
+            return (int(parts[0]), int(parts[1])) < (1, 3)
+        except Exception:
+            return True
+
    @staticmethod
    def _resolve_api_key_material(key_data: dict[str, Any]) -> tuple[str | None, str | None]:
        """解析用户 API Key 导入材料，优先使用明文 key。"""
@@ -2289,6 +2301,30 @@ class AdminImportUsersAdapter(AdminApiAdapter):
        key_encrypted = key_data.get("key_encrypted")
        return key_hash, key_encrypted

+    @staticmethod
+    def _normalize_imported_user_rate_limit(user_data: dict[str, Any]) -> int | None:
+        if "rate_limit" not in user_data:
+            return None
+        value = user_data.get("rate_limit")
+        return int(value) if value is not None else None
+
+    @staticmethod
+    def _normalize_imported_api_key_rate_limit(
+        key_data: dict[str, Any],
+        *,
+        is_standalone: bool,
+        legacy_export: bool,
+    ) -> int | None:
+        if "rate_limit" not in key_data:
+            return None if is_standalone and not legacy_export else 0
+
+        value = key_data.get("rate_limit")
+        if value is None:
+            if is_standalone and not legacy_export:
+                return None
+            return 0
+        return int(value)
+
    async def handle(self, context: ApiRequestContext) -> Any:  # type: ignore[override]
        """导入用户数据"""
        import uuid
@@ -2306,6 +2342,7 @@ class AdminImportUsersAdapter(AdminApiAdapter):

        # 获取导入选项
        merge_mode = payload.get("merge_mode", "skip")  # skip, overwrite, error
+        legacy_export = self._is_legacy_users_export(payload.get("version"))
        users_data = payload.get("users", [])
        standalone_keys_data = payload.get("standalone_keys", [])

@@ -2358,7 +2395,11 @@ class AdminImportUsersAdapter(AdminApiAdapter):
                    allowed_providers=key_data.get("allowed_providers"),
                    allowed_api_formats=key_data.get("allowed_api_formats"),
                    allowed_models=key_data.get("allowed_models"),
-                    rate_limit=key_data.get("rate_limit"),
+                    rate_limit=self._normalize_imported_api_key_rate_limit(
+                        key_data,
+                        is_standalone=is_standalone or key_data.get("is_standalone", False),
+                        legacy_export=legacy_export,
+                    ),
                    concurrent_limit=key_data.get("concurrent_limit", 5),
                    force_capabilities=key_data.get("force_capabilities"),
                    is_active=key_data.get("is_active", True),
@@ -2396,6 +2437,7 @@ class AdminImportUsersAdapter(AdminApiAdapter):
                    and wallet_payload.get("limit_mode") in {"finite", "unlimited"}
                    else ("unlimited" if user_data.get("unlimited") else "finite")
                )
+                imported_user_rate_limit = self._normalize_imported_user_rate_limit(user_data)

                if existing_user:
                    user_id = existing_user.id
@@ -2413,6 +2455,7 @@ class AdminImportUsersAdapter(AdminApiAdapter):
                        existing_user.allowed_providers = user_data.get("allowed_providers")
                        existing_user.allowed_api_formats = user_data.get("allowed_api_formats")
                        existing_user.allowed_models = user_data.get("allowed_models")
+                        existing_user.rate_limit = imported_user_rate_limit
                        existing_user.model_capability_settings = user_data.get(
                            "model_capability_settings"
                        )
@@ -2449,6 +2492,7 @@ class AdminImportUsersAdapter(AdminApiAdapter):
                        allowed_providers=user_data.get("allowed_providers"),
                        allowed_api_formats=user_data.get("allowed_api_formats"),
                        allowed_models=user_data.get("allowed_models"),
+                        rate_limit=imported_user_rate_limit,
                        model_capability_settings=user_data.get("model_capability_settings"),
                        is_active=user_data.get("is_active", True),
                    )
@@ -18,7 +18,7 @@ from src.core.exceptions import InvalidRequestException, NotFoundException, tran
 from src.core.logger import logger
 from src.database import get_db, get_db_context
 from src.models.admin_requests import UpdateUserRequest
-from src.models.api import CreateApiKeyRequest, CreateUserRequest
+from src.models.api import CreateApiKeyRequest, CreateUserRequest, UpdateMyApiKeyRequest
 from src.models.database import ApiKey, User, UserRole, Wallet
 from src.services.cache.user_cache import UserCacheService
 from src.services.system.config import SystemConfigService
@@ -57,6 +57,7 @@ def _serialize_user(
        "allowed_providers": user.allowed_providers,
        "allowed_api_formats": user.allowed_api_formats,
        "allowed_models": user.allowed_models,
+        "rate_limit": user.rate_limit,
        "unlimited": WalletService.is_unlimited_wallet(resolved_wallet),
        "is_active": user.is_active,
        "created_at": user.created_at.isoformat(),
@@ -89,6 +90,7 @@ def _create_user_sync(
            allowed_providers=request.allowed_providers,
            allowed_api_formats=request.allowed_api_formats,
            allowed_models=request.allowed_models,
+            rate_limit=request.rate_limit,
        )
        return _serialize_user(db, user), {
            "action": "create_user",
@@ -255,6 +257,61 @@ def _delete_user_key_sync(user_id: str, key_id: str) -> tuple[dict[str, Any], di
        }


+def _update_user_key_sync(
+    user_id: str,
+    key_id: str,
+    request: UpdateMyApiKeyRequest,
+) -> tuple[dict[str, Any], dict[str, Any]]:
+    with get_db_context() as db:
+        api_key = (
+            db.query(ApiKey)
+            .filter(
+                ApiKey.id == key_id,
+                ApiKey.user_id == user_id,
+                ApiKey.is_standalone == False,
+            )
+            .first()
+        )
+
+        if not api_key:
+            raise NotFoundException("API Key不存在或不属于该用户", "api_key")
+
+        update_data = request.model_dump(exclude_unset=True)
+        if "rate_limit" in update_data and update_data["rate_limit"] is None:
+            update_data["rate_limit"] = 0
+
+        updated_key = ApiKeyService.update_api_key(db, key_id, **update_data)
+        if not updated_key:
+            raise NotFoundException("API Key不存在或不属于该用户", "api_key")
+
+        return (
+            {
+                "id": updated_key.id,
+                "name": updated_key.name,
+                "key_display": updated_key.get_display_key(),
+                "is_active": updated_key.is_active,
+                "is_locked": updated_key.is_locked,
+                "total_requests": updated_key.total_requests,
+                "total_cost_usd": float(updated_key.total_cost_usd or 0),
+                "rate_limit": updated_key.rate_limit,
+                "expires_at": (
+                    updated_key.expires_at.isoformat() if updated_key.expires_at else None
+                ),
+                "last_used_at": (
+                    updated_key.last_used_at.isoformat() if updated_key.last_used_at else None
+                ),
+                "created_at": updated_key.created_at.isoformat(),
+                "message": "API Key更新成功",
+            },
+            {
+                "action": "update_user_api_key",
+                "target_user_id": user_id,
+                "key_id": key_id,
+                "updated_fields": list(update_data.keys()),
+            },
+        )
+
+
 def _toggle_user_key_lock_sync(
    user_id: str,
    key_id: str,
@@ -452,6 +509,26 @@ async def delete_user_api_key(
    return await pipeline.run(adapter=adapter, http_request=request, db=db, mode=adapter.mode)


+@router.put("/{user_id}/api-keys/{key_id}")
+async def update_user_api_key(
+    user_id: str,
+    key_id: str,
+    request: Request,
+    db: Session = Depends(get_db),
+) -> Any:
+    """
+    更新用户的 API 密钥
+
+    更新指定用户的普通 API 密钥基础配置。
+
+    **路径参数**:
+    - `user_id`: 用户 ID (UUID)
+    - `key_id`: 密钥 ID
+    """
+    adapter = AdminUpdateUserKeyAdapter(user_id=user_id, key_id=key_id)
+    return await pipeline.run(adapter=adapter, http_request=request, db=db, mode=adapter.mode)
+
+
@router.patch("/{user_id}/api-keys/{key_id}/lock")
 async def toggle_user_api_key_lock(
    user_id: str,
@@ -700,6 +777,33 @@ class AdminDeleteUserKeyAdapter(AdminApiAdapter):
        return response


+class AdminUpdateUserKeyAdapter(AdminApiAdapter):
+    """更新用户的普通 API Key"""
+
+    def __init__(self, user_id: str, key_id: str):
+        self.user_id = user_id
+        self.key_id = key_id
+
+    async def handle(self, context: ApiRequestContext) -> Any:  # type: ignore[override]
+        payload = context.ensure_json_body()
+        try:
+            request = UpdateMyApiKeyRequest.model_validate(payload)
+        except ValidationError as e:
+            errors = e.errors()
+            if errors:
+                raise InvalidRequestException(translate_pydantic_error(errors[0]))
+            raise InvalidRequestException("请求数据验证失败")
+
+        response, audit_meta = await run_in_threadpool(
+            _update_user_key_sync,
+            self.user_id,
+            self.key_id,
+            request,
+        )
+        context.add_audit_metadata(**audit_meta)
+        return response
+
+
 class AdminToggleUserKeyLockAdapter(AdminApiAdapter):
    """切换用户普通 API Key 的锁定状态"""

@@ -19,7 +19,9 @@ from src.core.logger import logger
 from src.database.database import create_session
 from src.models.database import ApiKey, AuditEventType, User
 from src.services.auth.service import AuthService
+from src.services.rate_limit.user_rpm_limiter import SYSTEM_RPM_CONFIG_KEY, get_user_rpm_limiter
 from src.services.system.audit import AuditService
+from src.services.system.config import SystemConfigService
 from src.services.usage.service import UsageService
 from src.services.wallet import WalletService
 from src.utils.perf import PerfRecorder
@@ -123,6 +125,9 @@ class ApiRequestPipeline:
            auth_duration = PerfRecorder.stop(auth_start, "pipeline_auth", labels=perf_labels)
            _record_perf_metric("auth_ms", auth_duration)

+        if mode in {ApiMode.STANDARD, ApiMode.PROXY} and api_key and user:
+            await self._check_user_rate_limit(http_request, db, user, api_key)
+
        raw_body = None
        should_eager_read_body = http_request.method in {"POST", "PUT", "PATCH"} and getattr(
            adapter, "eager_request_body", True
@@ -267,6 +272,55 @@ class ApiRequestPipeline:
    # Internal helpers
    # --------------------------------------------------------------------- #

+    async def _check_user_rate_limit(
+        self,
+        request: Request,
+        db: Session,
+        user: User,
+        api_key: ApiKey,
+    ) -> None:
+        limiter = await get_user_rpm_limiter()
+        system_default_raw = SystemConfigService.get_config(db, SYSTEM_RPM_CONFIG_KEY, default=0)
+        system_default = max(int(system_default_raw or 0), 0)
+
+        if api_key.is_standalone:
+            effective_user_limit = (
+                max(int(api_key.rate_limit or 0), 0)
+                if api_key.rate_limit is not None
+                else system_default
+            )
+            user_rpm_key = limiter.get_standalone_rpm_key(api_key.id)
+            key_rpm_limit = 0
+        else:
+            effective_user_limit = (
+                max(int(user.rate_limit or 0), 0) if user.rate_limit is not None else system_default
+            )
+            user_rpm_key = limiter.get_user_rpm_key(user.id)
+            key_rpm_limit = max(int(api_key.rate_limit or 0), 0)
+
+        result = await limiter.check_and_consume(
+            user_rpm_key=user_rpm_key,
+            user_rpm_limit=effective_user_limit,
+            key_rpm_key=limiter.get_key_rpm_key(api_key.id),
+            key_rpm_limit=key_rpm_limit,
+        )
+
+        if result.allowed:
+            return
+
+        scope = result.scope or "user"
+        limit = result.limit or (effective_user_limit if scope == "user" else key_rpm_limit)
+        retry_after = result.retry_after or limiter.get_retry_after()
+
+        headers = {
+            "Retry-After": str(retry_after),
+            "X-RateLimit-Limit": str(limit),
+            "X-RateLimit-Remaining": "0",
+            "X-RateLimit-Scope": scope,
+        }
+        request.state.rate_limit_scope = scope
+        raise HTTPException(status_code=429, detail="请求过于频繁，请稍后重试", headers=headers)
+
    async def _authenticate_client(
        self, request: Request, db: Session, adapter: ApiAdapter, **_kw: object
    ) -> tuple[User, ApiKey]:
@@ -16,7 +16,8 @@ from src.api.base.pipeline import get_pipeline
 from src.core.logger import logger
 from src.database import get_db
 from src.models.database import ApiKey, AuditLog
-from src.plugins.manager import get_plugin_manager
+from src.services.rate_limit.user_rpm_limiter import SYSTEM_RPM_CONFIG_KEY, get_user_rpm_limiter
+from src.services.system.config import SystemConfigService

 router = APIRouter(prefix="/api/monitoring", tags=["Monitoring"])
 pipeline = get_pipeline()
@@ -146,10 +147,6 @@ class UserRateLimitStatusAdapter(AuthenticatedApiAdapter):
        if not user:
            raise HTTPException(status_code=401, detail="未登录")

-        rate_limiter = _get_rate_limit_plugin()
-        if not rate_limiter or not hasattr(rate_limiter, "get_rate_limit_headers"):
-            raise HTTPException(status_code=503, detail="速率限制插件未启用或不支持状态查询")
-
        api_keys = (
            db.query(ApiKey)
            .filter(ApiKey.user_id == user.id, ApiKey.is_active.is_(True))
@@ -157,31 +154,84 @@ class UserRateLimitStatusAdapter(AuthenticatedApiAdapter):
            .all()
        )

+        try:
+            limiter = await get_user_rpm_limiter()
+            system_default_raw = SystemConfigService.get_config(
+                db, SYSTEM_RPM_CONFIG_KEY, default=0
+            )
+            system_default = max(int(system_default_raw or 0), 0)
+            reset_at = limiter.get_reset_at()
+            window = f"{limiter.bucket_seconds}s"
+        except Exception as exc:
+            logger.warning("读取新 RPM 限流状态失败，回退插件状态接口: {}", exc)
+            limiter = None
+            system_default = 0
+            reset_at = None
+            window = None
+
        rate_limit_info = []
        for key in api_keys:
-            try:
-                headers = rate_limiter.get_rate_limit_headers(key)
-            except Exception as exc:
-                logger.warning(f"无法获取Key {key.id} 的限流信息: {exc}")
-                headers = {}
+            if limiter is not None:
+                if key.is_standalone:
+                    user_limit = key.rate_limit if key.rate_limit is not None else system_default
+                    user_scope_key = limiter.get_standalone_rpm_key(key.id)
+                    key_limit = 0
+                else:
+                    user_limit = user.rate_limit if user.rate_limit is not None else system_default
+                    user_scope_key = limiter.get_user_rpm_key(user.id)
+                    key_limit = max(int(key.rate_limit or 0), 0)
+
+                user_count = (
+                    await limiter.get_scope_count(user_scope_key)
+                    if user_limit and user_limit > 0
+                    else 0
+                )
+                key_count = (
+                    await limiter.get_scope_count(limiter.get_key_rpm_key(key.id))
+                    if key_limit > 0
+                    else 0
+                )
+                user_remaining = max(user_limit - user_count, 0) if user_limit > 0 else None
+                key_remaining = max(key_limit - key_count, 0) if key_limit > 0 else None
+
+                scoped_statuses: list[tuple[str, int, int]] = []
+                if user_limit > 0 and user_remaining is not None:
+                    scoped_statuses.append(("user", user_limit, user_remaining))
+                if key_limit > 0 and key_remaining is not None:
+                    scoped_statuses.append(("key", key_limit, key_remaining))
+
+                primary_scope = (
+                    min(scoped_statuses, key=lambda item: item[2]) if scoped_statuses else None
+                )
+                rate_limit_info.append(
+                    {
+                        "api_key_name": key.name or f"Key-{key.id}",
+                        "limit": primary_scope[1] if primary_scope else None,
+                        "remaining": primary_scope[2] if primary_scope else None,
+                        "scope": primary_scope[0] if primary_scope else None,
+                        "reset_time": reset_at.isoformat() if reset_at else None,
+                        "window": window,
+                        "user_limit": user_limit,
+                        "user_remaining": user_remaining,
+                        "key_limit": key_limit if key_limit > 0 else None,
+                        "key_remaining": key_remaining,
+                    }
+                )
+                continue

            rate_limit_info.append(
                {
                    "api_key_name": key.name or f"Key-{key.id}",
-                    "limit": headers.get("X-RateLimit-Limit"),
-                    "remaining": headers.get("X-RateLimit-Remaining"),
-                    "reset_time": headers.get("X-RateLimit-Reset"),
-                    "window": headers.get("X-RateLimit-Window"),
+                    "limit": None,
+                    "remaining": None,
+                    "scope": None,
+                    "reset_time": None,
+                    "window": None,
+                    "user_limit": None,
+                    "user_remaining": None,
+                    "key_limit": None,
+                    "key_remaining": None,
                }
            )

        return {"user_id": user.id, "api_keys": rate_limit_info}
-
-
-def _get_rate_limit_plugin() -> Any:
-    try:
-        plugin_manager = get_plugin_manager()
-        return plugin_manager.get_plugin("rate_limit")
-    except Exception as exc:
-        logger.warning(f"获取速率限制插件失败: {exc}")
-        return None
@@ -33,6 +33,7 @@ from src.models.api import (
    PublicGlobalModelListResponse,
    PublicGlobalModelResponse,
    UpdateApiKeyProvidersRequest,
+    UpdateMyApiKeyRequest,
    UpdatePreferencesRequest,
    UpdateProfileRequest,
 )
@@ -146,6 +147,7 @@ def _create_my_api_key_sync(user_id: str, request: CreateMyApiKeyRequest) -> dic
                db=db,
                user_id=user_id,
                name=request.name,
+                rate_limit=request.rate_limit,
            )
        except ValueError as exc:
            raise InvalidRequestException(str(exc)) from exc
@@ -154,6 +156,7 @@ def _create_my_api_key_sync(user_id: str, request: CreateMyApiKeyRequest) -> dic
            "name": api_key.name,
            "key": plain_key,
            "key_display": api_key.get_display_key(),
+            "rate_limit": api_key.rate_limit,
            "message": "API密钥创建成功",
        }

@@ -189,6 +192,50 @@ def _toggle_my_api_key_sync(user_id: str, key_id: str) -> dict[str, Any]:
        }


+def _update_my_api_key_sync(
+    user_id: str,
+    key_id: str,
+    request: UpdateMyApiKeyRequest,
+) -> dict[str, Any]:
+    with get_db_context() as db:
+        api_key = (
+            db.query(ApiKey)
+            .filter(
+                ApiKey.id == key_id,
+                ApiKey.user_id == user_id,
+                ApiKey.is_standalone == False,
+            )
+            .first()
+        )
+        if not api_key:
+            raise NotFoundException("API密钥不存在", "api_key")
+        if api_key.is_locked:
+            raise ForbiddenException("该密钥已被管理员锁定，无法修改")
+
+        update_data = request.model_dump(exclude_unset=True)
+        if "rate_limit" in update_data and update_data["rate_limit"] is None:
+            update_data["rate_limit"] = 0
+
+        updated = ApiKeyService.update_api_key(db, key_id, **update_data)
+        if not updated:
+            raise NotFoundException("API密钥不存在", "api_key")
+
+        return {
+            "id": updated.id,
+            "name": updated.name,
+            "key_display": updated.get_display_key(),
+            "is_active": updated.is_active,
+            "is_locked": updated.is_locked,
+            "allowed_providers": updated.allowed_providers,
+            "force_capabilities": updated.force_capabilities,
+            "rate_limit": updated.rate_limit,
+            "last_used_at": updated.last_used_at.isoformat() if updated.last_used_at else None,
+            "expires_at": updated.expires_at.isoformat() if updated.expires_at else None,
+            "created_at": updated.created_at.isoformat(),
+            "message": "API密钥已更新",
+        }
+
+
 def _update_api_key_providers_sync(
    user_id: str,
    api_key_id: str,
@@ -484,6 +531,20 @@ async def delete_my_api_key(key_id: str, request: Request, db: Session = Depends
    return await pipeline.run(adapter=adapter, http_request=request, db=db, mode=adapter.mode)


+@router.put("/api-keys/{key_id}")
+async def update_my_api_key(key_id: str, request: Request, db: Session = Depends(get_db)) -> Any:
+    """
+    更新 API 密钥
+
+    更新指定 API 密钥的基础配置。
+
+    **路径参数**:
+    - `key_id`: 密钥 ID
+    """
+    adapter = UpdateMyApiKeyAdapter(key_id=key_id)
+    return await pipeline.run(adapter=adapter, http_request=request, db=db, mode=adapter.mode)
+
+
@router.patch("/api-keys/{key_id}")
 async def toggle_my_api_key(key_id: str, request: Request, db: Session = Depends(get_db)) -> Any:
    """
@@ -877,6 +938,7 @@ class ListMyApiKeysAdapter(AuthenticatedApiAdapter):
                    "created_at": key.created_at.isoformat(),
                    "total_requests": real_stats["total_requests"],
                    "total_cost_usd": real_stats["total_cost_usd"],
+                    "rate_limit": key.rate_limit,
                    "allowed_providers": key.allowed_providers,
                    "force_capabilities": key.force_capabilities,
                }
@@ -965,6 +1027,27 @@ class GetMyApiKeyDetailAdapter(AuthenticatedApiAdapter):
        }


+@dataclass
+class UpdateMyApiKeyAdapter(AuthenticatedApiAdapter):
+    """更新 API 密钥基础配置的适配器"""
+
+    key_id: str
+
+    async def handle(self, context: ApiRequestContext) -> Any:  # type: ignore[override]
+        payload = context.ensure_json_body()
+        try:
+            request = UpdateMyApiKeyRequest.model_validate(payload)
+        except ValidationError as e:
+            errors = e.errors()
+            if errors:
+                raise InvalidRequestException(translate_pydantic_error(errors[0]))
+            raise InvalidRequestException("请求数据验证失败")
+
+        return await run_in_threadpool(
+            _update_my_api_key_sync, context.user.id, self.key_id, request
+        )
+
+
@dataclass
 class DeleteMyApiKeyAdapter(AuthenticatedApiAdapter):
    """删除 API 密钥的适配器"""
@@ -109,8 +109,6 @@ class Config:
        # 支付回调安全配置（公开回调入口必须携带该共享密钥）
        self.payment_callback_secret = os.getenv("PAYMENT_CALLBACK_SECRET", "").strip()

-        # LLM API 速率限制配置（每分钟请求数）
-        self.llm_api_rate_limit = int(os.getenv("LLM_API_RATE_LIMIT", "100"))
        self.public_api_rate_limit = int(os.getenv("PUBLIC_API_RATE_LIMIT", "60"))

        # 异常处理配置
@@ -45,6 +45,7 @@ if TYPE_CHECKING:
    from src.services.model.fetch_scheduler import ModelFetchScheduler
    from src.services.provider_keys.pool_quota_probe_scheduler import PoolQuotaProbeScheduler
    from src.services.rate_limit.concurrency_manager import ConcurrencyManager
+    from src.services.rate_limit.user_rpm_limiter import UserRpmLimiter
    from src.services.system.maintenance_scheduler import MaintenanceScheduler
    from src.services.system.scheduler import TaskScheduler
    from src.services.task.polling.task_poller import TaskPollerService
@@ -99,6 +100,7 @@ class LifecycleState:

    redis_client: Redis | None = None
    concurrency_manager: ConcurrencyManager | None = None
+    user_rpm_limiter: UserRpmLimiter | None = None
    plugin_manager: PluginManager | None = None
    available_modules: list[ModuleDefinition] = field(default_factory=list)
    task_coordinator: StartupTaskCoordinator | None = None
@@ -181,6 +183,11 @@ async def _initialize_core_infrastructure(state: LifecycleState) -> None:

    state.concurrency_manager = await get_concurrency_manager()

+    logger.info("初始化用户/API Key RPM 限流器...")
+    from src.services.rate_limit.user_rpm_limiter import get_user_rpm_limiter
+
+    state.user_rpm_limiter = await get_user_rpm_limiter()
+
    # 初始化批量提交器（提升数据库并发能力）
    logger.info("初始化批量提交器...")
    from src.core.batch_committer import init_batch_committer
@@ -541,6 +548,10 @@ async def _run_shutdown(state: LifecycleState) -> None:
    if state.concurrency_manager:
        await state.concurrency_manager.close()

+    logger.info("关闭用户/API Key RPM 限流器...")
+    if state.user_rpm_limiter:
+        await state.user_rpm_limiter.close()
+
    # 关闭全局Redis客户端
    logger.info("关闭全局Redis客户端...")
    from src.clients.redis_client import close_redis_client
@@ -8,7 +8,6 @@

 from __future__ import annotations

-import hashlib
 import time
 from typing import TYPE_CHECKING

@@ -50,7 +49,6 @@ class PluginMiddleware:
        self._notification_cache_expires: float = 0.0

        # 从配置读取速率限制值
-        self.llm_api_rate_limit = config.llm_api_rate_limit
        self.public_api_rate_limit = config.public_api_rate_limit

        # 完全跳过限流的路径（静态资源、文档等）
@@ -69,14 +67,6 @@ class PluginMiddleware:
            "/api/monitoring/",  # 监控端点
        ]

-        # LLM API 端点（需要特殊的速率限制策略）
-        self.llm_api_paths = [
-            "/v1/messages",
-            "/v1/chat/completions",
-            "/v1/responses",
-            "/v1/completions",
-        ]
-
    async def __call__(self, scope: Scope, receive: Receive, send: Send) -> None:
        """ASGI 入口点"""
        if scope["type"] != "http":
@@ -291,13 +281,6 @@ class PluginMiddleware:

        return "unknown"

-    def _is_llm_api_path(self, path: str) -> bool:
-        """检查是否为 LLM API 端点"""
-        for llm_path in self.llm_api_paths:
-            if path.startswith(llm_path):
-                return True
-        return False
-
    async def _get_rate_limit_key_and_config(
        self, request: Request
    ) -> tuple[str | None, int | None]:
@@ -305,7 +288,6 @@ class PluginMiddleware:
        获取速率限制的key和配置

        策略说明:
-        - /v1/messages, /v1/chat/completions 等 LLM API: 按 API Key 限流
        - /api/public/* 端点: 使用服务器级别 IP 限制
        - /api/admin/* 端点: 跳过（在 skip_rate_limit_paths 中跳过）
        - /api/auth/* 端点: 跳过（由路由层的 IPRateLimiter 处理）
@@ -315,30 +297,6 @@ class PluginMiddleware:
        """
        path = request.url.path

-        # LLM API 端点: 按 API Key 或 IP 限流
-        if self._is_llm_api_path(path):
-            # 尝试从请求头获取 API Key
-            auth_header = request.headers.get("authorization", "")
-            api_key = request.headers.get("x-api-key", "")
-
-            if auth_header.lower().startswith("bearer "):
-                api_key = auth_header[7:]
-
-            if api_key:
-                # 使用 API Key 的哈希作为限制 key（避免日志泄露完整 key）
-                key_hash = hashlib.sha256(api_key.encode()).hexdigest()[:16]
-                key = f"llm_api_key:{key_hash}"
-                request.state.rate_limit_key_type = "api_key"
-            else:
-                # 无 API Key 时使用 IP 限制（更严格）
-                client_ip = self._get_client_ip(request)
-                key = f"llm_ip:{client_ip}"
-                request.state.rate_limit_key_type = "ip"
-
-            rate_limit = self.llm_api_rate_limit
-            request.state.rate_limit_value = rate_limit
-            return key, rate_limit
-
        # /api/public/* 端点: 使用服务器级别 IP 地址作为限制 key
        if path.startswith("/api/public/"):
            client_ip = self._get_client_ip(request)
@@ -692,6 +692,11 @@ class UpdateUserRequest(BaseModel):
    allowed_providers: list[str] | None = Field(None, description="允许使用的提供商 ID 列表")
    allowed_api_formats: list[str] | None = Field(None, description="允许使用的 API 格式列表")
    allowed_models: list[str] | None = Field(None, description="允许使用的模型名称列表")
+    rate_limit: int | None = Field(
+        None,
+        ge=0,
+        description="每分钟请求限制；null 表示继承系统默认，0 表示不限制",
+    )

    @field_validator("username")
    @classmethod
@@ -252,6 +252,11 @@ class CreateUserRequest(BaseModel):
    allowed_models: list[str] | None = Field(
        default=None, description="允许使用的模型名称列表，null表示无限制"
    )
+    rate_limit: int | None = Field(
+        default=None,
+        ge=0,
+        description="每分钟请求限制；null 表示继承系统默认，0 表示不限制",
+    )

    @field_validator("initial_gift_usd", mode="before")
    @classmethod
@@ -335,6 +340,11 @@ class UpdateUserRequest(BaseModel):
    allowed_providers: list[str] | None = None  # 允许使用的提供商 ID 列表
    allowed_api_formats: list[str] | None = None  # 允许使用的 API 格式列表
    allowed_models: list[str] | None = None  # 允许使用的模型名称列表
+    rate_limit: int | None = Field(
+        default=None,
+        ge=0,
+        description="每分钟请求限制；null 表示继承系统默认，0 表示不限制",
+    )
    is_active: bool | None = None

    @field_validator("allowed_api_formats")
@@ -351,7 +361,11 @@ class CreateApiKeyRequest(BaseModel):
    allowed_providers: list[str] | None = None  # 允许使用的提供商 ID 列表
    allowed_api_formats: list[str] | None = None  # 允许使用的 API 格式列表
    allowed_models: list[str] | None = None  # 允许使用的模型名称列表
-    rate_limit: int | None = None  # None = 无限制
+    rate_limit: int | None = Field(
+        None,
+        ge=0,
+        description="每分钟请求限制；独立Key: null=继承系统默认，0=不限制；普通Key: 0=不限制",
+    )
    expire_days: int | None = None  # None = 永不过期，数字 = 多少天后过期
    expires_at: str | None = None  # ISO 日期字符串，如 "2025-12-31"，优先于 expire_days
    initial_balance_usd: float | None = Field(
@@ -382,6 +396,7 @@ class UserResponse(BaseModel):
    allowed_providers: list[str] | None = None  # 允许使用的提供商 ID 列表
    allowed_api_formats: list[str] | None = None  # 允许使用的 API 格式列表
    allowed_models: list[str] | None = None  # 允许使用的模型名称列表
+    rate_limit: int | None = None
    unlimited: bool = False
    is_active: bool
    created_at: datetime
@@ -402,7 +417,7 @@ class ApiKeyResponse(BaseModel):
    total_cost_usd: float
    allowed_providers: list[str] | None
    allowed_models: list[str] | None
-    rate_limit: int
+    rate_limit: int | None
    is_active: bool
    expires_at: datetime | None = None
    is_standalone: bool = False
@@ -772,6 +787,18 @@ class CreateMyApiKeyRequest(BaseModel):
    """创建我的API密钥请求"""

    name: str
+    rate_limit: int = Field(0, ge=0, description="该 Key 的每分钟请求限制，0 表示不限制")
+
+
+class UpdateMyApiKeyRequest(BaseModel):
+    """更新我的 API 密钥请求"""
+
+    name: str | None = None
+    rate_limit: int | None = Field(
+        None,
+        ge=0,
+        description="该 Key 的每分钟请求限制；0 表示不限制，null 表示不修改",
+    )


 class ProviderConfig(BaseModel):
@@ -110,6 +110,9 @@ class User(Base):
    allowed_providers = Column(JSON, nullable=True)  # 允许使用的提供商 ID 列表
    allowed_api_formats = Column(JSON, nullable=True)  # 允许使用的 API 格式列表
    allowed_models = Column(JSON, nullable=True)  # 允许使用的模型名称列表
+    rate_limit = Column(
+        Integer, nullable=True, default=None
+    )  # 每分钟请求限制，NULL=继承系统默认，0=不限制，N=N RPM

    # Key 能力配置
    model_capability_settings = Column(JSON, nullable=True)  # 用户针对特定模型的能力配置
@@ -209,7 +212,9 @@ class ApiKey(Base):
    allowed_providers = Column(JSON, nullable=True)  # 允许使用的提供商 ID 列表
    allowed_api_formats = Column(JSON, nullable=True)  # 允许使用的 API 格式列表
    allowed_models = Column(JSON, nullable=True)  # 允许使用的模型名称列表
-    rate_limit = Column(Integer, default=None, nullable=True)  # 每分钟请求限制，None = 无限制
+    rate_limit = Column(
+        Integer, default=None, nullable=True
+    )  # 每分钟请求限制；独立Key: NULL=继承系统默认，普通Key: 0=不限制
    concurrent_limit = Column(Integer, default=5, nullable=True)  # 并发请求限制

    # Key 能力配置
@@ -12,6 +12,7 @@ from src.services.rate_limit.adaptive_rpm import (
 from src.services.rate_limit.concurrency_manager import ConcurrencyManager
 from src.services.rate_limit.detector import RateLimitDetector
 from src.services.rate_limit.ip_limiter import IPRateLimiter
+from src.services.rate_limit.user_rpm_limiter import UserRpmLimiter, get_user_rpm_limiter

 __all__ = [
    "AdaptiveConcurrencyManager",  # 向后兼容
@@ -19,5 +20,7 @@ __all__ = [
    "ConcurrencyManager",
    "IPRateLimiter",
    "RateLimitDetector",
+    "UserRpmLimiter",
    "get_adaptive_rpm_manager",
+    "get_user_rpm_limiter",
 ]
@@ -0,0 +1,354 @@
+"""
+用户/API Key RPM 限制器
+
+支持两层叠加限流：
+1. 用户级（或独立 Key 级）总 RPM
+2. 普通 Key 子限制 RPM
+
+实现策略：
+- Redis 可用时使用分钟桶 + Lua 脚本原子检查/消费
+- Redis 不可用时降级为内存计数（仅适用于单实例）
+"""
+
+from __future__ import annotations
+
+import asyncio
+import time
+from dataclasses import dataclass
+from datetime import datetime, timezone
+
+import redis.asyncio as aioredis
+
+from src.config.settings import config
+from src.core.logger import logger
+
+SYSTEM_RPM_CONFIG_KEY = "rate_limit_per_minute"
+
+
+@dataclass(slots=True)
+class RpmCheckResult:
+    """RPM 检查结果。"""
+
+    allowed: bool
+    scope: str | None = None
+    limit: int | None = None
+    remaining: int | None = None
+    retry_after: int | None = None
+
+
+class UserRpmLimiter:
+    """用户/API Key 双层 RPM 限制器。
+
+    通过模块级 ``get_user_rpm_limiter()`` 工厂函数获取唯一实例，
+    不要直接调用构造函数。
+    """
+
+    _CHECK_AND_CONSUME_SCRIPT = """
+    local user_key = KEYS[1]
+    local key_key = KEYS[2]
+    local user_limit = tonumber(ARGV[1])
+    local key_limit = tonumber(ARGV[2])
+    local ttl = tonumber(ARGV[3])
+    local retry_after = tonumber(ARGV[4])
+
+    local user_count = 0
+    if user_limit > 0 then
+        user_count = tonumber(redis.call('GET', user_key) or '0')
+        if user_count >= user_limit then
+            return {0, 1, user_limit, 0, retry_after}
+        end
+    end
+
+    local key_count = 0
+    if key_limit > 0 then
+        key_count = tonumber(redis.call('GET', key_key) or '0')
+        if key_count >= key_limit then
+            return {0, 2, key_limit, 0, retry_after}
+        end
+    end
+
+    local remaining = -1
+    if user_limit > 0 then
+        user_count = redis.call('INCR', user_key)
+        redis.call('EXPIRE', user_key, ttl)
+        remaining = user_limit - user_count
+    end
+
+    if key_limit > 0 then
+        key_count = redis.call('INCR', key_key)
+        redis.call('EXPIRE', key_key, ttl)
+        local key_remaining = key_limit - key_count
+        if remaining == -1 or key_remaining < remaining then
+            remaining = key_remaining
+        end
+    end
+
+    return {1, 0, 0, remaining, 0}
+    """
+
+    def __init__(self) -> None:
+        self._redis: aioredis.Redis | None = None
+        self._bucket_seconds = int(config.rpm_bucket_seconds)
+        self._key_ttl_seconds = int(config.rpm_key_ttl_seconds)
+        self._cleanup_interval_seconds = int(config.rpm_cleanup_interval_seconds)
+        self._memory_lock: asyncio.Lock = asyncio.Lock()
+        self._memory_counts: dict[str, tuple[int, int]] = {}
+        self._cleanup_task: asyncio.Task | None = None
+
+    async def initialize(self) -> None:
+        if self._redis is not None:
+            return
+
+        try:
+            from src.clients.redis_client import get_redis_client
+
+            self._redis = await get_redis_client(require_redis=False)
+            if self._redis:
+                logger.info("[OK] UserRpmLimiter 已复用全局 Redis 客户端")
+                return
+        except Exception as exc:
+            logger.warning("初始化 UserRpmLimiter Redis 客户端失败，降级为内存模式: {}", exc)
+
+        self._redis = None
+        self._start_background_cleanup()
+
+    async def close(self) -> None:
+        if self._cleanup_task is not None:
+            self._cleanup_task.cancel()
+            try:
+                await self._cleanup_task
+            except asyncio.CancelledError:
+                pass
+            self._cleanup_task = None
+
+    @property
+    def bucket_seconds(self) -> int:
+        return self._bucket_seconds
+
+    def get_user_rpm_key(self, user_id: str, bucket: int | None = None) -> str:
+        b = bucket if bucket is not None else self._get_rpm_bucket()
+        return f"rpm:user:{user_id}:{b}"
+
+    def get_standalone_rpm_key(self, api_key_id: str, bucket: int | None = None) -> str:
+        b = bucket if bucket is not None else self._get_rpm_bucket()
+        return f"rpm:ukey:{api_key_id}:{b}"
+
+    def get_key_rpm_key(self, api_key_id: str, bucket: int | None = None) -> str:
+        b = bucket if bucket is not None else self._get_rpm_bucket()
+        return f"rpm:key:{api_key_id}:{b}"
+
+    def get_retry_after(self, now_ts: float | None = None) -> int:
+        ts = now_ts if now_ts is not None else time.time()
+        elapsed = int(ts % self._bucket_seconds)
+        return max(1, self._bucket_seconds - elapsed)
+
+    def get_reset_at(self, now_ts: float | None = None) -> datetime:
+        ts = now_ts if now_ts is not None else time.time()
+        bucket = self._get_rpm_bucket(ts)
+        reset_ts = (bucket + 1) * self._bucket_seconds
+        return datetime.fromtimestamp(reset_ts, tz=timezone.utc)
+
+    async def get_scope_count(self, scope_key: str) -> int:
+        await self.initialize()
+
+        if self._redis is None:
+            async with self._memory_lock:
+                bucket = self._get_rpm_bucket()
+                self._cleanup_expired_memory_counts(bucket)
+                return self._get_memory_count(scope_key)
+
+        try:
+            result = await self._redis.get(scope_key)
+            return int(result) if result else 0
+        except Exception as exc:
+            logger.warning("读取 RPM 计数失败，回退内存模式: {}", exc)
+            if config.rate_limit_fail_open:
+                return 0
+            async with self._memory_lock:
+                bucket = self._get_rpm_bucket()
+                self._cleanup_expired_memory_counts(bucket)
+                return self._get_memory_count(scope_key)
+
+    async def check_and_consume(
+        self,
+        *,
+        user_rpm_key: str,
+        user_rpm_limit: int,
+        key_rpm_key: str,
+        key_rpm_limit: int,
+    ) -> RpmCheckResult:
+        """原子检查并消费两层 RPM 配额。"""
+
+        await self.initialize()
+
+        normalized_user_limit = max(int(user_rpm_limit or 0), 0)
+        normalized_key_limit = max(int(key_rpm_limit or 0), 0)
+
+        if normalized_user_limit <= 0 and normalized_key_limit <= 0:
+            return RpmCheckResult(allowed=True)
+
+        if self._redis is None:
+            return await self._check_and_consume_memory(
+                user_rpm_key=user_rpm_key,
+                user_rpm_limit=normalized_user_limit,
+                key_rpm_key=key_rpm_key,
+                key_rpm_limit=normalized_key_limit,
+            )
+
+        retry_after = self.get_retry_after()
+
+        try:
+            raw_result = await self._redis.eval(
+                self._CHECK_AND_CONSUME_SCRIPT,
+                2,
+                user_rpm_key,
+                key_rpm_key,
+                normalized_user_limit,
+                normalized_key_limit,
+                self._key_ttl_seconds,
+                retry_after,
+            )
+            return self._parse_redis_result(raw_result)
+        except Exception as exc:
+            logger.warning("Redis RPM 检查失败: {}", exc)
+            if config.rate_limit_fail_open:
+                return RpmCheckResult(allowed=True)
+            return await self._check_and_consume_memory(
+                user_rpm_key=user_rpm_key,
+                user_rpm_limit=normalized_user_limit,
+                key_rpm_key=key_rpm_key,
+                key_rpm_limit=normalized_key_limit,
+            )
+
+    def _parse_redis_result(self, raw_result: object) -> RpmCheckResult:
+        values = list(raw_result) if isinstance(raw_result, (list, tuple)) else [raw_result]
+        allowed = int(values[0]) == 1
+        scope_code = int(values[1]) if len(values) > 1 else 0
+        limit = int(values[2]) if len(values) > 2 and values[2] is not None else None
+        remaining = int(values[3]) if len(values) > 3 and values[3] is not None else None
+        retry_after = int(values[4]) if len(values) > 4 and values[4] is not None else None
+        scope = {1: "user", 2: "key"}.get(scope_code)
+        return RpmCheckResult(
+            allowed=allowed,
+            scope=scope,
+            limit=limit,
+            remaining=remaining,
+            retry_after=retry_after,
+        )
+
+    async def _check_and_consume_memory(
+        self,
+        *,
+        user_rpm_key: str,
+        user_rpm_limit: int,
+        key_rpm_key: str,
+        key_rpm_limit: int,
+    ) -> RpmCheckResult:
+        async with self._memory_lock:
+            bucket = self._get_rpm_bucket()
+            self._cleanup_expired_memory_counts(bucket)
+
+            user_count = self._get_memory_count(user_rpm_key)
+            if user_rpm_limit > 0 and user_count >= user_rpm_limit:
+                return RpmCheckResult(
+                    allowed=False,
+                    scope="user",
+                    limit=user_rpm_limit,
+                    remaining=0,
+                    retry_after=self.get_retry_after(),
+                )
+
+            key_count = self._get_memory_count(key_rpm_key)
+            if key_rpm_limit > 0 and key_count >= key_rpm_limit:
+                return RpmCheckResult(
+                    allowed=False,
+                    scope="key",
+                    limit=key_rpm_limit,
+                    remaining=0,
+                    retry_after=self.get_retry_after(),
+                )
+
+            remaining_candidates: list[int] = []
+
+            if user_rpm_limit > 0:
+                user_count += 1
+                self._set_memory_count(user_rpm_key, user_count)
+                remaining_candidates.append(user_rpm_limit - user_count)
+
+            if key_rpm_limit > 0:
+                key_count += 1
+                self._set_memory_count(key_rpm_key, key_count)
+                remaining_candidates.append(key_rpm_limit - key_count)
+
+            remaining = min(remaining_candidates) if remaining_candidates else None
+            return RpmCheckResult(allowed=True, remaining=remaining)
+
+    def _start_background_cleanup(self) -> None:
+        if self._cleanup_task is not None:
+            return
+
+        async def cleanup_loop() -> None:
+            while True:
+                try:
+                    await asyncio.sleep(self._bucket_seconds)
+                    async with self._memory_lock:
+                        self._cleanup_expired_memory_counts(self._get_rpm_bucket())
+                except asyncio.CancelledError:
+                    break
+                except Exception as exc:
+                    logger.debug("UserRpmLimiter 后台清理异常: {}", exc)
+
+        try:
+            self._cleanup_task = asyncio.create_task(cleanup_loop())
+        except RuntimeError:
+            self._cleanup_task = None
+
+    def _get_rpm_bucket(self, now_ts: float | None = None) -> int:
+        ts = now_ts if now_ts is not None else time.time()
+        return int(ts // self._bucket_seconds)
+
+    def _split_scope_key(self, scope_key: str) -> tuple[str, int]:
+        base_key, bucket_str = scope_key.rsplit(":", 1)
+        return base_key, int(bucket_str)
+
+    def _get_memory_count(self, scope_key: str) -> int:
+        base_key, bucket = self._split_scope_key(scope_key)
+        stored = self._memory_counts.get(base_key)
+        if not stored:
+            return 0
+        stored_bucket, count = stored
+        if stored_bucket != bucket:
+            self._memory_counts.pop(base_key, None)
+            return 0
+        return count
+
+    def _set_memory_count(self, scope_key: str, count: int) -> None:
+        base_key, bucket = self._split_scope_key(scope_key)
+        self._memory_counts[base_key] = (bucket, count)
+
+    def _cleanup_expired_memory_counts(self, current_bucket: int) -> None:
+        expired_keys = [
+            base_key
+            for base_key, (bucket, _count) in self._memory_counts.items()
+            if bucket < current_bucket
+        ]
+        for base_key in expired_keys:
+            self._memory_counts.pop(base_key, None)
+
+        if expired_keys:
+            logger.debug(
+                "[CLEANUP] 清理了 {} 个过期的用户/API Key RPM 计数（interval={}s）",
+                len(expired_keys),
+                self._cleanup_interval_seconds,
+            )
+
+
+_user_rpm_limiter: UserRpmLimiter | None = None
+
+
+async def get_user_rpm_limiter() -> UserRpmLimiter:
+    global _user_rpm_limiter
+    if _user_rpm_limiter is None:
+        _user_rpm_limiter = UserRpmLimiter()
+        await _user_rpm_limiter.initialize()
+    return _user_rpm_limiter
@@ -61,6 +61,10 @@ class ApiKeyService:
        if final_expires_at is None and expire_days:
            final_expires_at = datetime.now(timezone.utc) + timedelta(days=expire_days)

+        normalized_rate_limit = rate_limit
+        if not is_standalone and normalized_rate_limit is None:
+            normalized_rate_limit = 0
+
        api_key = ApiKey(
            user_id=user_id,
            key_hash=key_hash,
@@ -69,7 +73,7 @@ class ApiKeyService:
            allowed_providers=allowed_providers,
            allowed_api_formats=allowed_api_formats,
            allowed_models=allowed_models,
-            rate_limit=rate_limit,
+            rate_limit=normalized_rate_limit,
            concurrent_limit=concurrent_limit,
            expires_at=final_expires_at,
            is_standalone=is_standalone,
@@ -144,7 +148,8 @@ class ApiKeyService:
        # 允许显式设置为空数组/None 的字段（NULL=不限制，[]=全部禁用）
        nullable_list_fields = {"allowed_providers", "allowed_api_formats", "allowed_models"}

-        # 允许显式设置为 None 的字段（如 expires_at=None 表示永不过期，rate_limit=None 表示无限制）
+        # 允许显式设置为 None 的字段（如 expires_at=None 表示永不过期；
+        # standalone rate_limit=None 表示继承系统默认）
        nullable_fields = {"expires_at", "rate_limit"}

        for field, value in kwargs.items():
@@ -154,7 +159,9 @@ class ApiKeyService:
            if field in nullable_list_fields:
                setattr(api_key, field, value)
            elif field in nullable_fields:
-                # 这些字段允许显式设置为 None
+                if field == "rate_limit" and not api_key.is_standalone and value is None:
+                    setattr(api_key, field, 0)
+                    continue
                setattr(api_key, field, value)
            elif value is not None:
                setattr(api_key, field, value)
@@ -180,39 +187,6 @@ class ApiKeyService:
        logger.info(f"删除API密钥: ID {key_id}")
        return True

-    @staticmethod
-    def check_rate_limit(db: Session, api_key: ApiKey, window_minutes: int = 1) -> tuple[bool, int]:
-        """检查速率限制
-
-        Returns:
-            (is_allowed, remaining): 是否允许请求，剩余可用次数
-            当 rate_limit 为 None 时表示不限制，返回 (True, -1)
-        """
-        # 如果 rate_limit 为 None，表示不限制
-        if api_key.rate_limit is None:
-            return True, -1  # -1 表示无限制
-
-        # 计算时间窗口
-        window_start = datetime.now(timezone.utc) - timedelta(minutes=window_minutes)
-
-        # 统计窗口内的请求数
-        request_count = (
-            db.query(func.count(Usage.id))
-            .filter(Usage.api_key_id == api_key.id, Usage.created_at >= window_start)
-            .scalar()
-            or 0
-        )
-
-        # 检查是否超限
-        is_allowed = request_count < api_key.rate_limit
-
-        if not is_allowed:
-            logger.warning(
-                f"API密钥速率限制: Key ID {api_key.id}, 请求数 {request_count}/{api_key.rate_limit}"
-            )
-
-        return is_allowed, api_key.rate_limit - request_count
-
    @staticmethod
    def cleanup_expired_keys(db: Session, auto_delete: bool = False) -> int:
        """清理过期的API密钥
@@ -38,6 +38,7 @@ class UserService:
        allowed_providers: list[str] | None = None,
        allowed_api_formats: list[str] | None = None,
        allowed_models: list[str] | None = None,
+        rate_limit: int | None = None,
    ) -> User:
        """创建新用户。"""

@@ -74,6 +75,7 @@ class UserService:
            allowed_providers=allowed_providers,
            allowed_api_formats=allowed_api_formats,
            allowed_models=allowed_models,
+            rate_limit=rate_limit,
        )
        user.set_password(password)

@@ -218,6 +220,7 @@ class UserService:
            "allowed_providers",
            "allowed_api_formats",
            "allowed_models",
+            "rate_limit",
        ]

        # 允许设置为 None 的字段（表示无限制）
@@ -225,6 +228,7 @@ class UserService:
            "allowed_providers",
            "allowed_api_formats",
            "allowed_models",
+            "rate_limit",
        ]

        for field, value in kwargs.items():
@@ -1,7 +1,9 @@
 from __future__ import annotations

+from contextlib import contextmanager
 from datetime import datetime, timezone
 from types import SimpleNamespace
+from typing import Generator
 from unittest.mock import MagicMock

 import pytest
@@ -18,6 +20,7 @@ from src.api.admin.api_keys.routes import router as admin_api_keys_router
 from src.api.admin.users.routes import (
    AdminGetUserKeyFullKeyAdapter,
    AdminToggleUserKeyLockAdapter,
+    AdminUpdateUserKeyAdapter,
 )
 from src.api.admin.users.routes import router as admin_users_router
 from src.core.exceptions import InvalidRequestException, NotFoundException
@@ -25,6 +28,15 @@ from src.database import get_db
 from src.models.api import CreateApiKeyRequest


+def _patch_get_db_context(monkeypatch: pytest.MonkeyPatch, db: MagicMock) -> None:
+    @contextmanager
+    def _fake_ctx() -> Generator[MagicMock, None, None]:
+        yield db
+
+    monkeypatch.setattr("src.api.admin.users.routes.get_db_context", _fake_ctx)
+    monkeypatch.setattr("src.api.admin.api_keys.routes.get_db_context", _fake_ctx)
+
+
 def _build_context(db: MagicMock) -> SimpleNamespace:
    return SimpleNamespace(
        db=db,
@@ -46,11 +58,15 @@ def _build_admin_users_app(db: MagicMock, monkeypatch: pytest.MonkeyPatch) -> Te
        *, adapter: object, http_request: object, db: MagicMock, mode: object
    ) -> object:
        _ = http_request, mode
+        try:
+            payload = await http_request.json()
+        except Exception:
+            payload = {}
        context = SimpleNamespace(
            db=db,
            request=SimpleNamespace(state=SimpleNamespace()),
            user=SimpleNamespace(id="admin-1"),
-            ensure_json_body=lambda: {},
+            ensure_json_body=lambda: payload,
            add_audit_metadata=lambda **_: None,
        )
        return await adapter.handle(context)
@@ -82,10 +98,11 @@ def _build_admin_api_keys_app(db: MagicMock, monkeypatch: pytest.MonkeyPatch) ->


@pytest.mark.asyncio
-async def test_toggle_user_key_lock_adapter_success() -> None:
+async def test_toggle_user_key_lock_adapter_success(monkeypatch: pytest.MonkeyPatch) -> None:
    db = MagicMock()
    api_key = SimpleNamespace(id="key-1", user_id="user-1", is_standalone=False, is_locked=False)
    _mock_query_first(db, api_key)
+    _patch_get_db_context(monkeypatch, db)

    adapter = AdminToggleUserKeyLockAdapter(user_id="user-1", key_id="key-1")
    result = await adapter.handle(_build_context(db))
@@ -98,9 +115,12 @@ async def test_toggle_user_key_lock_adapter_success() -> None:


@pytest.mark.asyncio
-async def test_toggle_user_key_lock_adapter_not_found_for_standalone_or_wrong_owner() -> None:
+async def test_toggle_user_key_lock_adapter_not_found_for_standalone_or_wrong_owner(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
    db = MagicMock()
    _mock_query_first(db, None)
+    _patch_get_db_context(monkeypatch, db)

    adapter = AdminToggleUserKeyLockAdapter(user_id="user-1", key_id="key-standalone")
    with pytest.raises(NotFoundException):
@@ -170,7 +190,9 @@ async def test_get_user_key_full_key_adapter_returns_500_on_decrypt_error(


@pytest.mark.asyncio
-async def test_standalone_toggle_adapters_reject_normal_user_key() -> None:
+async def test_standalone_toggle_adapters_reject_normal_user_key(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
    db = MagicMock()
    normal_key = SimpleNamespace(
        id="key-user",
@@ -182,6 +204,7 @@ async def test_standalone_toggle_adapters_reject_normal_user_key() -> None:
        updated_at=datetime.now(timezone.utc),
    )
    _mock_query_first(db, normal_key)
+    _patch_get_db_context(monkeypatch, db)
    context = _build_context(db)

    with pytest.raises(InvalidRequestException):
@@ -195,6 +218,7 @@ def test_user_key_lock_route_path_smoke(monkeypatch: pytest.MonkeyPatch) -> None
    db = MagicMock()
    api_key = SimpleNamespace(id="key-5", user_id="user-2", is_standalone=False, is_locked=False)
    _mock_query_first(db, api_key)
+    _patch_get_db_context(monkeypatch, db)
    client = _build_admin_users_app(db, monkeypatch)

    response = client.patch("/api/admin/users/user-2/api-keys/key-5/lock")
@@ -220,6 +244,72 @@ def test_user_key_full_key_route_path_smoke(monkeypatch: pytest.MonkeyPatch) ->
    assert response.json() == {"key": "sk-user-route-key"}


+@pytest.mark.asyncio
+async def test_update_user_key_adapter_passes_rate_limit_and_name(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    captured: dict[str, object] = {}
+
+    def _update_user_key_sync(
+        user_id: str, key_id: str, request: object
+    ) -> tuple[dict[str, object], dict[str, object]]:
+        captured["user_id"] = user_id
+        captured["key_id"] = key_id
+        captured["name"] = getattr(request, "name", None)
+        captured["rate_limit"] = getattr(request, "rate_limit", None)
+        return {"id": key_id, "name": captured["name"], "rate_limit": captured["rate_limit"]}, {}
+
+    monkeypatch.setattr("src.api.admin.users.routes._update_user_key_sync", _update_user_key_sync)
+
+    adapter = AdminUpdateUserKeyAdapter(user_id="user-1", key_id="key-7")
+    context = SimpleNamespace(
+        db=MagicMock(),
+        request=SimpleNamespace(state=SimpleNamespace()),
+        ensure_json_body=lambda: {"name": "Renamed Key", "rate_limit": 12},
+        add_audit_metadata=lambda **_: None,
+    )
+
+    result = await adapter.handle(context)
+
+    assert result["id"] == "key-7"
+    assert captured == {
+        "user_id": "user-1",
+        "key_id": "key-7",
+        "name": "Renamed Key",
+        "rate_limit": 12,
+    }
+
+
+def test_update_user_key_route_path_smoke(monkeypatch: pytest.MonkeyPatch) -> None:
+    captured: dict[str, object] = {}
+
+    def _update_user_key_sync(
+        user_id: str, key_id: str, request: object
+    ) -> tuple[dict[str, object], dict[str, object]]:
+        captured["user_id"] = user_id
+        captured["key_id"] = key_id
+        captured["name"] = getattr(request, "name", None)
+        captured["rate_limit"] = getattr(request, "rate_limit", None)
+        return {"id": key_id, "name": captured["name"], "rate_limit": captured["rate_limit"]}, {}
+
+    monkeypatch.setattr("src.api.admin.users.routes._update_user_key_sync", _update_user_key_sync)
+    client = _build_admin_users_app(MagicMock(), monkeypatch)
+
+    response = client.put(
+        "/api/admin/users/user-2/api-keys/key-8",
+        json={"name": "Updated", "rate_limit": 9},
+    )
+
+    assert response.status_code == 200
+    assert response.json()["rate_limit"] == 9
+    assert captured == {
+        "user_id": "user-2",
+        "key_id": "key-8",
+        "name": "Updated",
+        "rate_limit": 9,
+    }
+
+
 def test_standalone_lock_route_removed(monkeypatch: pytest.MonkeyPatch) -> None:
    client = _build_admin_api_keys_app(MagicMock(), monkeypatch)
    response = client.patch("/api/admin/api-keys/key-1/lock")
@@ -228,6 +318,7 @@ def test_standalone_lock_route_removed(monkeypatch: pytest.MonkeyPatch) -> None:

 def test_standalone_list_route_does_not_expose_is_locked(monkeypatch: pytest.MonkeyPatch) -> None:
    db = MagicMock()
+    _patch_get_db_context(monkeypatch, db)
    api_key = SimpleNamespace(
        id="sa-key-1",
        user_id="admin-1",
@@ -306,6 +397,7 @@ async def test_create_standalone_key_adapter_preserves_empty_restriction_lists(
    monkeypatch: pytest.MonkeyPatch,
 ) -> None:
    db = MagicMock()
+    _patch_get_db_context(monkeypatch, db)
    captured: dict[str, object] = {}
    created_key = SimpleNamespace(
        id="sa-key-3",
@@ -365,6 +457,7 @@ async def test_update_standalone_key_adapter_preserves_empty_restriction_lists(
    monkeypatch: pytest.MonkeyPatch,
 ) -> None:
    db = MagicMock()
+    _patch_get_db_context(monkeypatch, db)
    existing_key = SimpleNamespace(id="sa-key-4", is_standalone=True)
    _mock_query_first(db, existing_key)

@@ -49,6 +49,7 @@ def test_list_users_uses_wallet_batch_lookup(monkeypatch: pytest.MonkeyPatch) ->
            allowed_providers=None,
            allowed_api_formats=None,
            allowed_models=None,
+            rate_limit=None,
            is_active=True,
            created_at=now,
            updated_at=now,
@@ -62,6 +63,7 @@ def test_list_users_uses_wallet_batch_lookup(monkeypatch: pytest.MonkeyPatch) ->
            allowed_providers=None,
            allowed_api_formats=None,
            allowed_models=None,
+            rate_limit=None,
            is_active=True,
            created_at=now,
            updated_at=None,
@@ -101,24 +103,14 @@ async def test_create_user_adapter_preserves_empty_restriction_lists(
    db = MagicMock()
    captured: dict[str, Any] = {}

-    def _create_user(**kwargs: Any) -> SimpleNamespace:
-        captured.update(kwargs)
-        return SimpleNamespace(
-            id="user-3",
-            email="u3@example.com",
-            username="user3",
-            role=SimpleNamespace(value="user"),
-            is_active=True,
-            allowed_providers=[],
-            allowed_api_formats=[],
-            allowed_models=[],
-        )
+    def _create_user_sync(request: Any, role: Any) -> tuple[dict[str, Any], dict[str, Any]]:
+        captured["allowed_providers"] = request.allowed_providers
+        captured["allowed_api_formats"] = request.allowed_api_formats
+        captured["allowed_models"] = request.allowed_models
+        captured["role"] = role
+        return {"id": "user-3"}, {}

-    monkeypatch.setattr("src.api.admin.users.routes.UserService.create_user", _create_user)
-    monkeypatch.setattr(
-        "src.api.admin.users.routes._serialize_user",
-        lambda _db, user: {"id": user.id},
-    )
+    monkeypatch.setattr("src.api.admin.users.routes._create_user_sync", _create_user_sync)

    context = SimpleNamespace(
        db=db,
@@ -0,0 +1,120 @@
+from __future__ import annotations
+
+from datetime import datetime, timezone
+from types import SimpleNamespace
+from unittest.mock import AsyncMock, MagicMock
+
+import pytest
+
+from src.api.monitoring.user import UserRateLimitStatusAdapter
+
+
+def _build_query_with_keys(keys: list[object]) -> MagicMock:
+    query = MagicMock()
+    query.filter.return_value.order_by.return_value.all.return_value = keys
+    return query
+
+
+@pytest.mark.asyncio
+async def test_rate_limit_status_adapter_reports_user_and_key_layers(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    now = datetime(2026, 3, 13, 12, 0, tzinfo=timezone.utc)
+    db = MagicMock()
+    user = SimpleNamespace(id="user-1", rate_limit=None)
+    key = SimpleNamespace(id="key-1", name="Primary", is_standalone=False, rate_limit=10)
+    standalone = SimpleNamespace(
+        id="skey-1", name="Standalone", is_standalone=True, rate_limit=None
+    )
+
+    db.query.return_value = _build_query_with_keys([key, standalone])
+
+    limiter = MagicMock()
+    limiter.bucket_seconds = 60
+    limiter.get_reset_at.return_value = now
+    limiter.get_user_rpm_key.return_value = "rpm:user:user-1:bucket"
+    limiter.get_standalone_rpm_key.return_value = "rpm:ukey:skey-1:bucket"
+    limiter.get_key_rpm_key.side_effect = lambda key_id: f"rpm:key:{key_id}:bucket"
+
+    async def _get_scope_count(scope_key: str) -> int:
+        counts = {
+            "rpm:user:user-1:bucket": 55,
+            "rpm:key:key-1:bucket": 7,
+            "rpm:ukey:skey-1:bucket": 12,
+        }
+        return counts[scope_key]
+
+    limiter.get_scope_count = AsyncMock(side_effect=_get_scope_count)
+
+    monkeypatch.setattr(
+        "src.api.monitoring.user.get_user_rpm_limiter",
+        AsyncMock(return_value=limiter),
+    )
+    monkeypatch.setattr(
+        "src.api.monitoring.user.SystemConfigService.get_config",
+        lambda *_a, **_k: 60,
+    )
+
+    context = SimpleNamespace(db=db, user=user)
+    result = await UserRateLimitStatusAdapter().handle(context)
+
+    assert result["user_id"] == "user-1"
+    assert result["api_keys"][0] == {
+        "api_key_name": "Primary",
+        "limit": 10,
+        "remaining": 3,
+        "scope": "key",
+        "reset_time": now.isoformat(),
+        "window": "60s",
+        "user_limit": 60,
+        "user_remaining": 5,
+        "key_limit": 10,
+        "key_remaining": 3,
+    }
+    assert result["api_keys"][1] == {
+        "api_key_name": "Standalone",
+        "limit": 60,
+        "remaining": 48,
+        "scope": "user",
+        "reset_time": now.isoformat(),
+        "window": "60s",
+        "user_limit": 60,
+        "user_remaining": 48,
+        "key_limit": None,
+        "key_remaining": None,
+    }
+
+
+@pytest.mark.asyncio
+async def test_rate_limit_status_adapter_reports_unlimited_key_without_counts(
+    monkeypatch: pytest.MonkeyPatch,
+) -> None:
+    db = MagicMock()
+    user = SimpleNamespace(id="user-1", rate_limit=0)
+    key = SimpleNamespace(id="key-1", name="Unlimited", is_standalone=False, rate_limit=0)
+
+    db.query.return_value = _build_query_with_keys([key])
+
+    limiter = MagicMock()
+    limiter.bucket_seconds = 60
+    limiter.get_reset_at.return_value = datetime.now(timezone.utc)
+    limiter.get_user_rpm_key.return_value = "rpm:user:user-1:bucket"
+    limiter.get_key_rpm_key.return_value = "rpm:key:key-1:bucket"
+    limiter.get_scope_count = AsyncMock()
+
+    monkeypatch.setattr(
+        "src.api.monitoring.user.get_user_rpm_limiter",
+        AsyncMock(return_value=limiter),
+    )
+    monkeypatch.setattr(
+        "src.api.monitoring.user.SystemConfigService.get_config",
+        lambda *_a, **_k: 60,
+    )
+
+    context = SimpleNamespace(db=db, user=user)
+    result = await UserRateLimitStatusAdapter().handle(context)
+
+    assert result["api_keys"][0]["limit"] is None
+    assert result["api_keys"][0]["remaining"] is None
+    assert result["api_keys"][0]["scope"] is None
+    limiter.get_scope_count.assert_not_awaited()
@@ -18,6 +18,7 @@ from src.api.base.adapter import ApiMode
 from src.api.base.pipeline import ApiRequestPipeline
 from src.core.enums import UserRole
 from src.core.modules.hooks import AUTH_TOKEN_PREFIX_AUTHENTICATORS
+from src.services.rate_limit.user_rpm_limiter import RpmCheckResult


 class TestPipelineBalanceCalculation:
@@ -499,6 +500,166 @@ class TestPipelineAuthentication:
        assert "锁定" in str(exc_info.value.detail)


+class TestPipelineUserRateLimit:
+    @pytest.fixture
+    def pipeline(self) -> ApiRequestPipeline:
+        return ApiRequestPipeline()
+
+    @pytest.mark.asyncio
+    async def test_check_user_rate_limit_uses_system_default_for_user_scope(
+        self, pipeline: ApiRequestPipeline, monkeypatch: pytest.MonkeyPatch
+    ) -> None:
+        request = MagicMock()
+        request.state = MagicMock()
+        db = MagicMock()
+        user = MagicMock(id="user-1", rate_limit=None)
+        api_key = MagicMock(id="key-1", is_standalone=False, rate_limit=0)
+
+        limiter = MagicMock()
+        limiter.get_user_rpm_key.return_value = "rpm:user:user-1:1"
+        limiter.get_key_rpm_key.return_value = "rpm:key:key-1:1"
+        limiter.check_and_consume = AsyncMock(return_value=RpmCheckResult(allowed=True))
+
+        monkeypatch.setattr(
+            "src.api.base.pipeline.get_user_rpm_limiter",
+            AsyncMock(return_value=limiter),
+        )
+        monkeypatch.setattr(
+            "src.api.base.pipeline.SystemConfigService.get_config",
+            lambda *_a, **_k: 60,
+        )
+
+        await pipeline._check_user_rate_limit(request, db, user, api_key)
+
+        limiter.check_and_consume.assert_awaited_once_with(
+            user_rpm_key="rpm:user:user-1:1",
+            user_rpm_limit=60,
+            key_rpm_key="rpm:key:key-1:1",
+            key_rpm_limit=0,
+        )
+
+    @pytest.mark.asyncio
+    async def test_check_user_rate_limit_returns_429_with_scope_header(
+        self, pipeline: ApiRequestPipeline, monkeypatch: pytest.MonkeyPatch
+    ) -> None:
+        request = MagicMock()
+        request.state = MagicMock()
+        db = MagicMock()
+        user = MagicMock(id="user-1", rate_limit=100)
+        api_key = MagicMock(id="key-1", is_standalone=False, rate_limit=10)
+
+        limiter = MagicMock()
+        limiter.get_user_rpm_key.return_value = "rpm:user:user-1:1"
+        limiter.get_key_rpm_key.return_value = "rpm:key:key-1:1"
+        limiter.get_retry_after.return_value = 17
+        limiter.check_and_consume = AsyncMock(
+            return_value=RpmCheckResult(
+                allowed=False,
+                scope="key",
+                limit=10,
+                remaining=0,
+                retry_after=17,
+            )
+        )
+
+        monkeypatch.setattr(
+            "src.api.base.pipeline.get_user_rpm_limiter",
+            AsyncMock(return_value=limiter),
+        )
+        monkeypatch.setattr(
+            "src.api.base.pipeline.SystemConfigService.get_config",
+            lambda *_a, **_k: 60,
+        )
+
+        with pytest.raises(HTTPException) as exc_info:
+            await pipeline._check_user_rate_limit(request, db, user, api_key)
+
+        assert exc_info.value.status_code == 429
+        assert exc_info.value.headers == {
+            "Retry-After": "17",
+            "X-RateLimit-Limit": "10",
+            "X-RateLimit-Remaining": "0",
+            "X-RateLimit-Scope": "key",
+        }
+
+    @pytest.mark.asyncio
+    async def test_check_user_rate_limit_uses_system_default_for_standalone_key(
+        self, pipeline: ApiRequestPipeline, monkeypatch: pytest.MonkeyPatch
+    ) -> None:
+        request = MagicMock()
+        request.state = MagicMock()
+        db = MagicMock()
+        user = MagicMock(id="user-1", rate_limit=999)
+        api_key = MagicMock(id="standalone-1", is_standalone=True, rate_limit=None)
+
+        limiter = MagicMock()
+        limiter.get_standalone_rpm_key.return_value = "rpm:ukey:standalone-1:1"
+        limiter.get_key_rpm_key.return_value = "rpm:key:standalone-1:1"
+        limiter.check_and_consume = AsyncMock(return_value=RpmCheckResult(allowed=True))
+
+        monkeypatch.setattr(
+            "src.api.base.pipeline.get_user_rpm_limiter",
+            AsyncMock(return_value=limiter),
+        )
+        monkeypatch.setattr(
+            "src.api.base.pipeline.SystemConfigService.get_config",
+            lambda *_a, **_k: 60,
+        )
+
+        await pipeline._check_user_rate_limit(request, db, user, api_key)
+
+        limiter.check_and_consume.assert_awaited_once_with(
+            user_rpm_key="rpm:ukey:standalone-1:1",
+            user_rpm_limit=60,
+            key_rpm_key="rpm:key:standalone-1:1",
+            key_rpm_limit=0,
+        )
+
+    @pytest.mark.asyncio
+    async def test_check_user_rate_limit_returns_429_with_user_scope_header(
+        self, pipeline: ApiRequestPipeline, monkeypatch: pytest.MonkeyPatch
+    ) -> None:
+        request = MagicMock()
+        request.state = MagicMock()
+        db = MagicMock()
+        user = MagicMock(id="user-1", rate_limit=3)
+        api_key = MagicMock(id="key-1", is_standalone=False, rate_limit=10)
+
+        limiter = MagicMock()
+        limiter.get_user_rpm_key.return_value = "rpm:user:user-1:1"
+        limiter.get_key_rpm_key.return_value = "rpm:key:key-1:1"
+        limiter.get_retry_after.return_value = 23
+        limiter.check_and_consume = AsyncMock(
+            return_value=RpmCheckResult(
+                allowed=False,
+                scope="user",
+                limit=3,
+                remaining=0,
+                retry_after=23,
+            )
+        )
+
+        monkeypatch.setattr(
+            "src.api.base.pipeline.get_user_rpm_limiter",
+            AsyncMock(return_value=limiter),
+        )
+        monkeypatch.setattr(
+            "src.api.base.pipeline.SystemConfigService.get_config",
+            lambda *_a, **_k: 60,
+        )
+
+        with pytest.raises(HTTPException) as exc_info:
+            await pipeline._check_user_rate_limit(request, db, user, api_key)
+
+        assert exc_info.value.status_code == 429
+        assert exc_info.value.headers == {
+            "Retry-After": "23",
+            "X-RateLimit-Limit": "3",
+            "X-RateLimit-Remaining": "0",
+            "X-RateLimit-Scope": "user",
+        }
+
+
 class TestPipelineTokenPrefixAuth:
    """Tests token-prefix auth isolation."""

@@ -0,0 +1,110 @@
+from __future__ import annotations
+
+from types import SimpleNamespace
+from typing import Any
+from unittest.mock import MagicMock
+
+import pytest
+from fastapi import FastAPI
+from fastapi.testclient import TestClient
+
+from src.api.user_me.routes import UpdateMyApiKeyAdapter
+from src.api.user_me.routes import router as me_router
+from src.database import get_db
+
+
+def _build_me_app(db: MagicMock, monkeypatch: Any) -> TestClient:
+    app = FastAPI()
+    app.include_router(me_router)
+    app.dependency_overrides[get_db] = lambda: db
+
+    async def _fake_pipeline_run(
+        *, adapter: object, http_request: object, db: MagicMock, mode: object
+    ) -> object:
+        _ = http_request, mode
+        try:
+            payload = await http_request.json()
+        except Exception:
+            payload = {}
+        context = SimpleNamespace(
+            db=db,
+            user=SimpleNamespace(id="user-1", email="u@example.com"),
+            request=SimpleNamespace(state=SimpleNamespace()),
+            ensure_json_body=lambda: payload,
+            add_audit_metadata=lambda **_: None,
+        )
+        return await adapter.handle(context)
+
+    monkeypatch.setattr("src.api.user_me.routes.pipeline.run", _fake_pipeline_run)
+    return TestClient(app)
+
+
+async def _fake_update_my_api_key_sync(
+    user_id: str,
+    key_id: str,
+    request: object,
+    captured: dict[str, object],
+) -> dict[str, object]:
+    captured["user_id"] = user_id
+    captured["key_id"] = key_id
+    captured["name"] = getattr(request, "name", None)
+    captured["rate_limit"] = getattr(request, "rate_limit", None)
+    return {"id": key_id, "name": captured["name"], "rate_limit": captured["rate_limit"]}
+
+
+def test_update_my_api_key_route_path_smoke(monkeypatch: Any) -> None:
+    captured: dict[str, object] = {}
+
+    def _sync(user_id: str, key_id: str, request: object) -> dict[str, object]:
+        captured["user_id"] = user_id
+        captured["key_id"] = key_id
+        captured["name"] = getattr(request, "name", None)
+        captured["rate_limit"] = getattr(request, "rate_limit", None)
+        return {"id": key_id, "name": captured["name"], "rate_limit": captured["rate_limit"]}
+
+    monkeypatch.setattr("src.api.user_me.routes._update_my_api_key_sync", _sync)
+    client = _build_me_app(MagicMock(), monkeypatch)
+
+    response = client.put("/api/users/me/api-keys/key-1", json={"name": "Edited", "rate_limit": 6})
+
+    assert response.status_code == 200
+    assert response.json()["rate_limit"] == 6
+    assert captured == {
+        "user_id": "user-1",
+        "key_id": "key-1",
+        "name": "Edited",
+        "rate_limit": 6,
+    }
+
+
+@pytest.mark.asyncio
+async def test_update_my_api_key_adapter_passes_rate_limit_and_name(monkeypatch: Any) -> None:
+    captured: dict[str, object] = {}
+
+    def _sync(user_id: str, key_id: str, request: object) -> dict[str, object]:
+        captured["user_id"] = user_id
+        captured["key_id"] = key_id
+        captured["name"] = getattr(request, "name", None)
+        captured["rate_limit"] = getattr(request, "rate_limit", None)
+        return {"id": key_id, "name": captured["name"], "rate_limit": captured["rate_limit"]}
+
+    monkeypatch.setattr("src.api.user_me.routes._update_my_api_key_sync", _sync)
+
+    adapter = UpdateMyApiKeyAdapter(key_id="key-2")
+    context = SimpleNamespace(
+        db=MagicMock(),
+        user=SimpleNamespace(id="user-1"),
+        request=SimpleNamespace(state=SimpleNamespace()),
+        ensure_json_body=lambda: {"name": "Edited Again", "rate_limit": 15},
+        add_audit_metadata=lambda **_: None,
+    )
+
+    result = await adapter.handle(context)
+
+    assert result["id"] == "key-2"
+    assert captured == {
+        "user_id": "user-1",
+        "key_id": "key-2",
+        "name": "Edited Again",
+        "rate_limit": 15,
+    }
@@ -0,0 +1,148 @@
+from __future__ import annotations
+
+from collections.abc import AsyncGenerator
+from unittest.mock import AsyncMock, MagicMock
+
+import pytest
+
+from src.services.rate_limit.user_rpm_limiter import RpmCheckResult, UserRpmLimiter
+
+
+@pytest.fixture
+async def limiter(monkeypatch: pytest.MonkeyPatch) -> AsyncGenerator[UserRpmLimiter]:
+    limiter = UserRpmLimiter()
+    limiter._redis = None
+    limiter._memory_counts.clear()
+    await limiter.close()
+    # 阻止 initialize() 连接真实 Redis，确保纯内存模式
+    monkeypatch.setattr(limiter, "initialize", AsyncMock())
+    yield limiter
+    limiter._memory_counts.clear()
+    limiter._redis = None
+    await limiter.close()
+
+
+@pytest.mark.asyncio
+async def test_check_and_consume_enforces_user_scope_in_memory(
+    limiter: UserRpmLimiter,
+) -> None:
+    user_key = limiter.get_user_rpm_key("user-1")
+    key_key = limiter.get_key_rpm_key("key-1")
+
+    first = await limiter.check_and_consume(
+        user_rpm_key=user_key,
+        user_rpm_limit=1,
+        key_rpm_key=key_key,
+        key_rpm_limit=0,
+    )
+    second = await limiter.check_and_consume(
+        user_rpm_key=user_key,
+        user_rpm_limit=1,
+        key_rpm_key=key_key,
+        key_rpm_limit=0,
+    )
+
+    assert first == RpmCheckResult(allowed=True, remaining=0)
+    assert second.allowed is False
+    assert second.scope == "user"
+    assert second.limit == 1
+    assert second.remaining == 0
+
+
+@pytest.mark.asyncio
+async def test_check_and_consume_enforces_key_scope_in_memory(
+    limiter: UserRpmLimiter,
+) -> None:
+    user_key = limiter.get_user_rpm_key("user-1")
+    key_key = limiter.get_key_rpm_key("key-1")
+
+    first = await limiter.check_and_consume(
+        user_rpm_key=user_key,
+        user_rpm_limit=3,
+        key_rpm_key=key_key,
+        key_rpm_limit=1,
+    )
+    second = await limiter.check_and_consume(
+        user_rpm_key=user_key,
+        user_rpm_limit=3,
+        key_rpm_key=key_key,
+        key_rpm_limit=1,
+    )
+
+    assert first.allowed is True
+    assert second.allowed is False
+    assert second.scope == "key"
+    assert second.limit == 1
+
+
+@pytest.mark.asyncio
+async def test_check_and_consume_redis_failure_falls_back_to_memory_when_fail_close(
+    monkeypatch: pytest.MonkeyPatch,
+    limiter: UserRpmLimiter,
+) -> None:
+    fake_redis = MagicMock()
+    fake_redis.eval = AsyncMock(side_effect=RuntimeError("redis down"))
+    limiter._redis = fake_redis
+
+    monkeypatch.setattr(
+        "src.services.rate_limit.user_rpm_limiter.config.rate_limit_fail_open", False
+    )
+
+    user_key = limiter.get_user_rpm_key("user-1")
+    key_key = limiter.get_key_rpm_key("key-1")
+
+    first = await limiter.check_and_consume(
+        user_rpm_key=user_key,
+        user_rpm_limit=1,
+        key_rpm_key=key_key,
+        key_rpm_limit=0,
+    )
+    second = await limiter.check_and_consume(
+        user_rpm_key=user_key,
+        user_rpm_limit=1,
+        key_rpm_key=key_key,
+        key_rpm_limit=0,
+    )
+
+    assert first.allowed is True
+    assert second.allowed is False
+    assert second.scope == "user"
+
+
+@pytest.mark.asyncio
+async def test_check_and_consume_redis_failure_fail_open_allows_request(
+    monkeypatch: pytest.MonkeyPatch,
+    limiter: UserRpmLimiter,
+) -> None:
+    fake_redis = MagicMock()
+    fake_redis.eval = AsyncMock(side_effect=RuntimeError("redis down"))
+    limiter._redis = fake_redis
+
+    monkeypatch.setattr(
+        "src.services.rate_limit.user_rpm_limiter.config.rate_limit_fail_open", True
+    )
+
+    result = await limiter.check_and_consume(
+        user_rpm_key=limiter.get_user_rpm_key("user-1"),
+        user_rpm_limit=1,
+        key_rpm_key=limiter.get_key_rpm_key("key-1"),
+        key_rpm_limit=0,
+    )
+
+    assert result.allowed is True
+
+
+@pytest.mark.asyncio
+async def test_check_and_consume_skips_when_all_limits_are_zero(
+    limiter: UserRpmLimiter,
+) -> None:
+    result = await limiter.check_and_consume(
+        user_rpm_key=limiter.get_user_rpm_key("user-1"),
+        user_rpm_limit=0,
+        key_rpm_key=limiter.get_key_rpm_key("key-1"),
+        key_rpm_limit=0,
+    )
+
+    assert result.allowed is True
+    assert result.scope is None
+    assert limiter._memory_counts == {}
@@ -53,3 +53,29 @@ def test_import_user_api_key_material_keeps_legacy_encrypted_payload() -> None:

    assert key_hash == legacy_hash
    assert key_encrypted == legacy_encrypted
+
+
+def test_import_user_rate_limit_defaults_to_none_when_missing() -> None:
+    assert AdminImportUsersAdapter._normalize_imported_user_rate_limit({}) is None
+
+
+def test_import_legacy_standalone_key_null_rate_limit_becomes_unlimited() -> None:
+    assert (
+        AdminImportUsersAdapter._normalize_imported_api_key_rate_limit(
+            {"rate_limit": None},
+            is_standalone=True,
+            legacy_export=True,
+        )
+        == 0
+    )
+
+
+def test_import_new_standalone_key_null_rate_limit_keeps_inherit_semantics() -> None:
+    assert (
+        AdminImportUsersAdapter._normalize_imported_api_key_rate_limit(
+            {"rate_limit": None},
+            is_standalone=True,
+            legacy_export=False,
+        )
+        is None
+    )