Merge pull request #418 from Avilianb/feat/codex-responses-websocket

feat: support Codex Responses WebSocket transport

.gitignore

@@ -204,6 +204,7 @@ logs/
 
 # Git backup
 .git.backup/
+.worktrees/
 
 # Database backups
 backups/

Dockerfile.app

@@ -90,6 +90,11 @@ RUN printf '%s\n' \
     '    ""      $remote_addr;' \
     '}' \
     '' \
+    'map $http_upgrade $connection_upgrade {' \
+    '    default upgrade;' \
+    '    ""      "";' \
+    '}' \
+    '' \
     'server {' \
     '    listen 80;' \
     '    server_name _;' \
@@ -160,7 +165,8 @@ RUN printf '%s\n' \
     '        proxy_set_header X-Real-IP $real_ip;' \
     '        proxy_set_header X-Forwarded-For $forwarded_for;' \
     '        proxy_set_header X-Forwarded-Proto $scheme;' \
-    '        proxy_set_header Connection "";' \
+    '        proxy_set_header Upgrade $http_upgrade;' \
+    '        proxy_set_header Connection $connection_upgrade;' \
     '        proxy_set_header Accept $http_accept;' \
     '        proxy_set_header Content-Type $content_type;' \
     '        proxy_set_header Authorization $http_authorization;' \

Dockerfile.app.local

@@ -112,6 +112,11 @@ RUN printf '%s\n' \
     '    ""      $remote_addr;' \
     '}' \
     '' \
+    'map $http_upgrade $connection_upgrade {' \
+    '    default upgrade;' \
+    '    ""      "";' \
+    '}' \
+    '' \
     'server {' \
     '    listen 80;' \
     '    server_name _;' \
@@ -182,7 +187,8 @@ RUN printf '%s\n' \
     '        proxy_set_header X-Real-IP $real_ip;' \
     '        proxy_set_header X-Forwarded-For $forwarded_for;' \
     '        proxy_set_header X-Forwarded-Proto $scheme;' \
-    '        proxy_set_header Connection "";' \
+    '        proxy_set_header Upgrade $http_upgrade;' \
+    '        proxy_set_header Connection $connection_upgrade;' \
     '        proxy_set_header Accept $http_accept;' \
     '        proxy_set_header Content-Type $content_type;' \
     '        proxy_set_header Authorization $http_authorization;' \

alembic/versions/20260324_1400_c3d4e5f6a7b8_usage_token_semantics_v2.py

@@ -0,0 +1,317 @@
+"""usage token semantics v2
+
+Revision ID: c3d4e5f6a7b8
+Revises: c9d8e7f6a5b4
+Create Date: 2026-03-24 14:00:00.000000+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 = "c3d4e5f6a7b8"
+down_revision: str | None = "c9d8e7f6a5b4"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+BACKFILL_BATCH_SIZE = 2000
+
+# 最大批次数，防止因数据异常导致死循环（2000 * 500000 = 10亿行上限）
+_MAX_BATCHES = 500000
+
+# 使用子查询中间层展开 input_output_total_tokens 的计算，
+# 确保 total_tokens 引用的是本次 SET 后的新值而非旧值。
+_UPGRADE_BACKFILL_SQL = sa.text(
+    """
+    UPDATE usage
+    SET
+        input_output_total_tokens = src.new_iot,
+        input_context_tokens      = src.new_ict,
+        total_tokens              = src.new_total,
+        cache_creation_cost_usd_5m        = src.new_cc5m,
+        cache_creation_cost_usd_1h        = src.new_cc1h,
+        actual_cache_creation_cost_usd_5m = src.new_acc5m,
+        actual_cache_creation_cost_usd_1h = src.new_acc1h,
+        actual_cache_cost_usd             = src.new_accu,
+        cache_creation_price_per_1m_5m    = src.new_cp5m,
+        cache_creation_price_per_1m_1h    = src.new_cp1h,
+        cache_cost_usd                    = src.new_ccu
+    FROM (
+        SELECT
+            id,
+            COALESCE(input_output_total_tokens,
+                     COALESCE(input_tokens, 0) + COALESCE(output_tokens, 0))
+                AS new_iot,
+            COALESCE(input_tokens, 0) + COALESCE(cache_read_input_tokens, 0)
+                AS new_ict,
+            /* total_tokens 引用本行计算出的 new_iot，避免依赖 SET 顺序 */
+            COALESCE(input_output_total_tokens,
+                     COALESCE(input_tokens, 0) + COALESCE(output_tokens, 0))
+                + COALESCE(cache_creation_input_tokens, 0)
+                + COALESCE(cache_read_input_tokens, 0)
+                AS new_total,
+            CASE
+                WHEN COALESCE(cache_creation_input_tokens_5m, 0) > 0
+                     AND COALESCE(cache_creation_input_tokens_1h, 0) = 0
+                THEN COALESCE(cache_creation_cost_usd, 0)
+                WHEN COALESCE(cache_creation_input_tokens_5m, 0) > 0
+                     AND COALESCE(cache_creation_input_tokens, 0) > 0
+                THEN COALESCE(cache_creation_cost_usd, 0)
+                     * (COALESCE(cache_creation_input_tokens_5m, 0) * 1.0
+                        / GREATEST(COALESCE(cache_creation_input_tokens, 0), 1))
+                ELSE 0
+            END AS new_cc5m,
+            CASE
+                WHEN COALESCE(cache_creation_input_tokens_1h, 0) > 0
+                     AND COALESCE(cache_creation_input_tokens_5m, 0) = 0
+                THEN COALESCE(cache_creation_cost_usd, 0)
+                WHEN COALESCE(cache_creation_input_tokens_1h, 0) > 0
+                     AND COALESCE(cache_creation_input_tokens, 0) > 0
+                THEN COALESCE(cache_creation_cost_usd, 0)
+                     * (COALESCE(cache_creation_input_tokens_1h, 0) * 1.0
+                        / GREATEST(COALESCE(cache_creation_input_tokens, 0), 1))
+                ELSE 0
+            END AS new_cc1h,
+            CASE
+                WHEN COALESCE(cache_creation_input_tokens_5m, 0) > 0
+                     AND COALESCE(cache_creation_input_tokens_1h, 0) = 0
+                THEN COALESCE(actual_cache_creation_cost_usd, 0)
+                WHEN COALESCE(cache_creation_input_tokens_5m, 0) > 0
+                     AND COALESCE(cache_creation_input_tokens, 0) > 0
+                THEN COALESCE(actual_cache_creation_cost_usd, 0)
+                     * (COALESCE(cache_creation_input_tokens_5m, 0) * 1.0
+                        / GREATEST(COALESCE(cache_creation_input_tokens, 0), 1))
+                ELSE 0
+            END AS new_acc5m,
+            CASE
+                WHEN COALESCE(cache_creation_input_tokens_1h, 0) > 0
+                     AND COALESCE(cache_creation_input_tokens_5m, 0) = 0
+                THEN COALESCE(actual_cache_creation_cost_usd, 0)
+                WHEN COALESCE(cache_creation_input_tokens_1h, 0) > 0
+                     AND COALESCE(cache_creation_input_tokens, 0) > 0
+                THEN COALESCE(actual_cache_creation_cost_usd, 0)
+                     * (COALESCE(cache_creation_input_tokens_1h, 0) * 1.0
+                        / GREATEST(COALESCE(cache_creation_input_tokens, 0), 1))
+                ELSE 0
+            END AS new_acc1h,
+            COALESCE(actual_cache_creation_cost_usd, 0)
+                + COALESCE(actual_cache_read_cost_usd, 0) AS new_accu,
+            CASE
+                WHEN COALESCE(cache_creation_input_tokens_5m, 0) > 0
+                     AND COALESCE(cache_creation_input_tokens_1h, 0) = 0
+                THEN cache_creation_price_per_1m
+                ELSE NULL
+            END AS new_cp5m,
+            CASE
+                WHEN COALESCE(cache_creation_input_tokens_1h, 0) > 0
+                     AND COALESCE(cache_creation_input_tokens_5m, 0) = 0
+                THEN cache_creation_price_per_1m
+                ELSE NULL
+            END AS new_cp1h,
+            COALESCE(cache_creation_cost_usd, 0)
+                + COALESCE(cache_read_cost_usd, 0) AS new_ccu
+        FROM usage
+        WHERE id IN (
+            SELECT id FROM usage
+            WHERE
+                input_output_total_tokens IS DISTINCT FROM
+                    COALESCE(input_output_total_tokens,
+                             COALESCE(input_tokens, 0) + COALESCE(output_tokens, 0))
+                OR input_context_tokens IS DISTINCT FROM
+                    COALESCE(input_tokens, 0) + COALESCE(cache_read_input_tokens, 0)
+                OR total_tokens IS DISTINCT FROM (
+                    COALESCE(input_output_total_tokens,
+                             COALESCE(input_tokens, 0) + COALESCE(output_tokens, 0))
+                    + COALESCE(cache_creation_input_tokens, 0)
+                    + COALESCE(cache_read_input_tokens, 0)
+                )
+                OR cache_creation_cost_usd_5m IS DISTINCT FROM (
+                    CASE
+                        WHEN COALESCE(cache_creation_input_tokens_5m, 0) > 0
+                             AND COALESCE(cache_creation_input_tokens_1h, 0) = 0
+                        THEN COALESCE(cache_creation_cost_usd, 0)
+                        WHEN COALESCE(cache_creation_input_tokens_5m, 0) > 0
+                             AND COALESCE(cache_creation_input_tokens, 0) > 0
+                        THEN COALESCE(cache_creation_cost_usd, 0)
+                             * (COALESCE(cache_creation_input_tokens_5m, 0) * 1.0
+                                / GREATEST(COALESCE(cache_creation_input_tokens, 0), 1))
+                        ELSE 0
+                    END
+                )
+                OR cache_creation_cost_usd_1h IS DISTINCT FROM (
+                    CASE
+                        WHEN COALESCE(cache_creation_input_tokens_1h, 0) > 0
+                             AND COALESCE(cache_creation_input_tokens_5m, 0) = 0
+                        THEN COALESCE(cache_creation_cost_usd, 0)
+                        WHEN COALESCE(cache_creation_input_tokens_1h, 0) > 0
+                             AND COALESCE(cache_creation_input_tokens, 0) > 0
+                        THEN COALESCE(cache_creation_cost_usd, 0)
+                             * (COALESCE(cache_creation_input_tokens_1h, 0) * 1.0
+                                / GREATEST(COALESCE(cache_creation_input_tokens, 0), 1))
+                        ELSE 0
+                    END
+                )
+                OR actual_cache_creation_cost_usd_5m IS DISTINCT FROM (
+                    CASE
+                        WHEN COALESCE(cache_creation_input_tokens_5m, 0) > 0
+                             AND COALESCE(cache_creation_input_tokens_1h, 0) = 0
+                        THEN COALESCE(actual_cache_creation_cost_usd, 0)
+                        WHEN COALESCE(cache_creation_input_tokens_5m, 0) > 0
+                             AND COALESCE(cache_creation_input_tokens, 0) > 0
+                        THEN COALESCE(actual_cache_creation_cost_usd, 0)
+                             * (COALESCE(cache_creation_input_tokens_5m, 0) * 1.0
+                                / GREATEST(COALESCE(cache_creation_input_tokens, 0), 1))
+                        ELSE 0
+                    END
+                )
+                OR actual_cache_creation_cost_usd_1h IS DISTINCT FROM (
+                    CASE
+                        WHEN COALESCE(cache_creation_input_tokens_1h, 0) > 0
+                             AND COALESCE(cache_creation_input_tokens_5m, 0) = 0
+                        THEN COALESCE(actual_cache_creation_cost_usd, 0)
+                        WHEN COALESCE(cache_creation_input_tokens_1h, 0) > 0
+                             AND COALESCE(cache_creation_input_tokens, 0) > 0
+                        THEN COALESCE(actual_cache_creation_cost_usd, 0)
+                             * (COALESCE(cache_creation_input_tokens_1h, 0) * 1.0
+                                / GREATEST(COALESCE(cache_creation_input_tokens, 0), 1))
+                        ELSE 0
+                    END
+                )
+                OR actual_cache_cost_usd IS DISTINCT FROM (
+                    COALESCE(actual_cache_creation_cost_usd, 0)
+                    + COALESCE(actual_cache_read_cost_usd, 0)
+                )
+                OR cache_creation_price_per_1m_5m IS DISTINCT FROM (
+                    CASE
+                        WHEN COALESCE(cache_creation_input_tokens_5m, 0) > 0
+                             AND COALESCE(cache_creation_input_tokens_1h, 0) = 0
+                        THEN cache_creation_price_per_1m
+                        ELSE NULL
+                    END
+                )
+                OR cache_creation_price_per_1m_1h IS DISTINCT FROM (
+                    CASE
+                        WHEN COALESCE(cache_creation_input_tokens_1h, 0) > 0
+                             AND COALESCE(cache_creation_input_tokens_5m, 0) = 0
+                        THEN cache_creation_price_per_1m
+                        ELSE NULL
+                    END
+                )
+                OR cache_cost_usd IS DISTINCT FROM (
+                    COALESCE(cache_creation_cost_usd, 0) + COALESCE(cache_read_cost_usd, 0)
+                )
+            ORDER BY id
+            LIMIT :batch_size
+        )
+    ) AS src
+    WHERE usage.id = src.id
+    """
+)
+
+_DOWNGRADE_BACKFILL_SQL = sa.text(
+    """
+    UPDATE usage
+    SET total_tokens = COALESCE(input_output_total_tokens, COALESCE(input_tokens, 0) + COALESCE(output_tokens, 0))
+    WHERE id IN (
+        SELECT id
+        FROM usage
+        WHERE total_tokens IS DISTINCT FROM
+            COALESCE(input_output_total_tokens, COALESCE(input_tokens, 0) + COALESCE(output_tokens, 0))
+        ORDER BY id
+        LIMIT :batch_size
+    )
+    """
+)
+
+
+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 run_backfill_in_batches(sql: sa.TextClause, batch_size: int = BACKFILL_BATCH_SIZE) -> None:
+    context = op.get_context()
+    for _ in range(_MAX_BATCHES):
+        # Commit the preceding schema transaction before each batch so PostgreSQL
+        # does not keep ALTER TABLE locks for the entire data backfill.
+        with context.autocommit_block():
+            rowcount = op.get_bind().execute(sql, {"batch_size": batch_size}).rowcount
+        if rowcount == 0:
+            break
+    else:
+        raise RuntimeError(
+            f"Backfill did not converge after {_MAX_BATCHES} batches "
+            f"(batch_size={batch_size}). Possible infinite loop due to data anomaly."
+        )
+
+
+def upgrade() -> None:
+    if column_exists("usage", "total_tokens") and not column_exists("usage", "input_output_total_tokens"):
+        with op.batch_alter_table("usage") as batch_op:
+            batch_op.alter_column(
+                "total_tokens",
+                new_column_name="input_output_total_tokens",
+                existing_type=sa.Integer(),
+                existing_nullable=True,
+            )
+
+    with op.batch_alter_table("usage") as batch_op:
+        if not column_exists("usage", "input_context_tokens"):
+            batch_op.add_column(sa.Column("input_context_tokens", sa.Integer(), nullable=False, server_default="0"))
+        if not column_exists("usage", "total_tokens"):
+            batch_op.add_column(sa.Column("total_tokens", sa.Integer(), nullable=False, server_default="0"))
+        if not column_exists("usage", "cache_creation_cost_usd_5m"):
+            batch_op.add_column(sa.Column("cache_creation_cost_usd_5m", sa.Numeric(20, 8), nullable=False, server_default="0"))
+        if not column_exists("usage", "cache_creation_cost_usd_1h"):
+            batch_op.add_column(sa.Column("cache_creation_cost_usd_1h", sa.Numeric(20, 8), nullable=False, server_default="0"))
+        if not column_exists("usage", "actual_cache_creation_cost_usd_5m"):
+            batch_op.add_column(sa.Column("actual_cache_creation_cost_usd_5m", sa.Numeric(20, 8), nullable=False, server_default="0"))
+        if not column_exists("usage", "actual_cache_creation_cost_usd_1h"):
+            batch_op.add_column(sa.Column("actual_cache_creation_cost_usd_1h", sa.Numeric(20, 8), nullable=False, server_default="0"))
+        if not column_exists("usage", "actual_cache_cost_usd"):
+            batch_op.add_column(sa.Column("actual_cache_cost_usd", sa.Numeric(20, 8), nullable=False, server_default="0"))
+        if not column_exists("usage", "cache_creation_price_per_1m_5m"):
+            batch_op.add_column(sa.Column("cache_creation_price_per_1m_5m", sa.Numeric(20, 8), nullable=True))
+        if not column_exists("usage", "cache_creation_price_per_1m_1h"):
+            batch_op.add_column(sa.Column("cache_creation_price_per_1m_1h", sa.Numeric(20, 8), nullable=True))
+
+    run_backfill_in_batches(_UPGRADE_BACKFILL_SQL)
+
+
+def downgrade() -> None:
+    run_backfill_in_batches(_DOWNGRADE_BACKFILL_SQL)
+
+    with op.batch_alter_table("usage") as batch_op:
+        if column_exists("usage", "cache_creation_price_per_1m_1h"):
+            batch_op.drop_column("cache_creation_price_per_1m_1h")
+        if column_exists("usage", "cache_creation_price_per_1m_5m"):
+            batch_op.drop_column("cache_creation_price_per_1m_5m")
+        if column_exists("usage", "actual_cache_cost_usd"):
+            batch_op.drop_column("actual_cache_cost_usd")
+        if column_exists("usage", "actual_cache_creation_cost_usd_1h"):
+            batch_op.drop_column("actual_cache_creation_cost_usd_1h")
+        if column_exists("usage", "actual_cache_creation_cost_usd_5m"):
+            batch_op.drop_column("actual_cache_creation_cost_usd_5m")
+        if column_exists("usage", "cache_creation_cost_usd_1h"):
+            batch_op.drop_column("cache_creation_cost_usd_1h")
+        if column_exists("usage", "cache_creation_cost_usd_5m"):
+            batch_op.drop_column("cache_creation_cost_usd_5m")
+        if column_exists("usage", "input_context_tokens"):
+            batch_op.drop_column("input_context_tokens")
+        if column_exists("usage", "total_tokens"):
+            batch_op.drop_column("total_tokens")
+        if column_exists("usage", "input_output_total_tokens"):
+            batch_op.alter_column(
+                "input_output_total_tokens",
+                new_column_name="total_tokens",
+                existing_type=sa.Integer(),
+                existing_nullable=True,
+            )

frontend/src/api/admin.ts

@@ -380,102 +380,6 @@ export interface AdminApiKeysResponse {
   skip: number
 }
 
-export interface LeaderboardItem {
-  rank: number
-  id: string
-  name: string
-  value: number
-  requests: number
-  tokens: number
-  cost: number
-}
-
-export interface LeaderboardResponse {
-  items: LeaderboardItem[]
-  total: number
-  metric: string
-  start_date?: string | null
-  end_date?: string | null
-}
-
-export interface CostForecastResponse {
-  history: Array<{ date: string; total_cost: number }>
-  forecast: Array<{ date: string; total_cost: number }>
-  slope: number
-  intercept: number
-  start_date: string
-  end_date: string
-}
-
-export interface CostSavingsResponse {
-  cache_read_tokens: number
-  cache_read_cost: number
-  cache_creation_cost: number
-  estimated_full_cost: number
-  cache_savings: number
-}
-
-export interface QuotaUsageProvider {
-  id: string
-  name: string
-  quota_usd: number
-  used_usd: number
-  remaining_usd: number
-  usage_percent: number
-  quota_expires_at?: string | null
-  estimated_exhaust_at?: string | null
-}
-
-export interface QuotaUsageResponse {
-  providers: QuotaUsageProvider[]
-}
-
-export interface PercentileItem {
-  date: string
-  p50_response_time_ms?: number | null
-  p90_response_time_ms?: number | null
-  p99_response_time_ms?: number | null
-  p50_first_byte_time_ms?: number | null
-  p90_first_byte_time_ms?: number | null
-  p99_first_byte_time_ms?: number | null
-}
-
-export interface ErrorDistributionItem {
-  category: string
-  count: number
-}
-
-export interface ErrorTrendItem {
-  date: string
-  total: number
-  categories: Record<string, number>
-}
-
-export interface ErrorDistributionResponse {
-  distribution: ErrorDistributionItem[]
-  trend: ErrorTrendItem[]
-}
-
-export interface ComparisonMetric {
-  total_requests: number
-  total_tokens: number
-  total_cost: number
-  actual_total_cost: number
-  avg_response_time_ms: number
-  error_requests: number
-}
-
-export interface ComparisonResponse {
-  current: ComparisonMetric
-  comparison: ComparisonMetric
-  change_percent: Record<string, number | null>
-  current_start: string
-  current_end: string
-  comparison_start: string
-  comparison_end: string
-}
-
-
 export interface ApiKeyToggleResponse {
   id: string // UUID
   is_active: boolean
@@ -753,197 +657,6 @@ export const adminApi = {
     return response.data
   },
 
-  // Stats / Leaderboards
-  async getLeaderboardUsers(params?: {
-    start_date?: string
-    end_date?: string
-    preset?: string
-    timezone?: string
-    tz_offset_minutes?: number
-    metric?: 'requests' | 'tokens' | 'cost'
-    order?: 'asc' | 'desc'
-    limit?: number
-    offset?: number
-    provider_name?: string
-    model?: string
-    include_inactive?: boolean
-    exclude_admin?: boolean
-  }): Promise<LeaderboardResponse> {
-    const cacheKey = buildCacheKey('admin:stats:leaderboard:users', params)
-    return cachedRequest(
-      cacheKey,
-      async () => {
-        const response = await apiClient.get<LeaderboardResponse>('/api/admin/stats/leaderboard/users', {
-          params
-        })
-        return response.data
-      },
-      20 * 1000
-    )
-  },
-
-  async getLeaderboardApiKeys(params?: {
-    start_date?: string
-    end_date?: string
-    preset?: string
-    timezone?: string
-    tz_offset_minutes?: number
-    metric?: 'requests' | 'tokens' | 'cost'
-    order?: 'asc' | 'desc'
-    limit?: number
-    offset?: number
-    provider_name?: string
-    model?: string
-    include_inactive?: boolean
-    exclude_admin?: boolean
-  }): Promise<LeaderboardResponse> {
-    const cacheKey = buildCacheKey('admin:stats:leaderboard:api-keys', params)
-    return cachedRequest(
-      cacheKey,
-      async () => {
-        const response = await apiClient.get<LeaderboardResponse>('/api/admin/stats/leaderboard/api-keys', {
-          params
-        })
-        return response.data
-      },
-      20 * 1000
-    )
-  },
-
-  async getLeaderboardModels(params?: {
-    start_date?: string
-    end_date?: string
-    preset?: string
-    timezone?: string
-    tz_offset_minutes?: number
-    metric?: 'requests' | 'tokens' | 'cost'
-    order?: 'asc' | 'desc'
-    limit?: number
-    offset?: number
-    provider_name?: string
-    model?: string
-  }): Promise<LeaderboardResponse> {
-    const cacheKey = buildCacheKey('admin:stats:leaderboard:models', params)
-    return cachedRequest(
-      cacheKey,
-      async () => {
-        const response = await apiClient.get<LeaderboardResponse>('/api/admin/stats/leaderboard/models', {
-          params
-        })
-        return response.data
-      },
-      20 * 1000
-    )
-  },
-
-  async getCostForecast(params?: {
-    start_date?: string
-    end_date?: string
-    preset?: string
-    timezone?: string
-    tz_offset_minutes?: number
-    days?: number
-    forecast_days?: number
-  }): Promise<CostForecastResponse> {
-    const cacheKey = buildCacheKey('admin:stats:cost:forecast', params)
-    return cachedRequest(
-      cacheKey,
-      async () => {
-        const response = await apiClient.get<CostForecastResponse>('/api/admin/stats/cost/forecast', {
-          params
-        })
-        return response.data
-      },
-      30 * 1000
-    )
-  },
-
-  async getCostSavings(params?: {
-    start_date?: string
-    end_date?: string
-    preset?: string
-    timezone?: string
-    tz_offset_minutes?: number
-    provider_name?: string
-    model?: string
-  }): Promise<CostSavingsResponse> {
-    const cacheKey = buildCacheKey('admin:stats:cost:savings', params)
-    return cachedRequest(
-      cacheKey,
-      async () => {
-        const response = await apiClient.get<CostSavingsResponse>('/api/admin/stats/cost/savings', {
-          params
-        })
-        return response.data
-      },
-      30 * 1000
-    )
-  },
-
-  async getQuotaUsage(): Promise<QuotaUsageResponse> {
-    return cachedRequest(
-      'admin:stats:providers:quota-usage',
-      async () => {
-        const response = await apiClient.get<QuotaUsageResponse>('/api/admin/stats/providers/quota-usage')
-        return response.data
-      },
-      30 * 1000
-    )
-  },
-
-  async getPercentiles(params?: {
-    start_date?: string
-    end_date?: string
-    preset?: string
-    timezone?: string
-    tz_offset_minutes?: number
-  }): Promise<PercentileItem[]> {
-    const cacheKey = buildCacheKey('admin:stats:performance:percentiles', params)
-    return cachedRequest(
-      cacheKey,
-      async () => {
-        const response = await apiClient.get<PercentileItem[]>('/api/admin/stats/performance/percentiles', {
-          params
-        })
-        return response.data
-      },
-      20 * 1000
-    )
-  },
-
-  async getErrorDistribution(params?: {
-    start_date?: string
-    end_date?: string
-    preset?: string
-    timezone?: string
-    tz_offset_minutes?: number
-  }): Promise<ErrorDistributionResponse> {
-    const cacheKey = buildCacheKey('admin:stats:errors:distribution', params)
-    return cachedRequest(
-      cacheKey,
-      async () => {
-        const response = await apiClient.get<ErrorDistributionResponse>('/api/admin/stats/errors/distribution', {
-          params
-        })
-        return response.data
-      },
-      20 * 1000
-    )
-  },
-
-  async getComparison(params: {
-    current_start: string
-    current_end: string
-    comparison_type?: 'period' | 'year'
-    timezone?: string
-    tz_offset_minutes?: number
-  }): Promise<ComparisonResponse> {
-    const response = await apiClient.get<ComparisonResponse>('/api/admin/stats/comparison', {
-      params
-    })
-    return response.data
-  },
-
   // 数据清空
   purgeConfig: () => purge<{ message: string; deleted: Record<string, number> }>('config'),
   purgeUsers: () => purge<{ message: string; deleted: Record<string, number> }>('users'),
@@ -952,26 +665,4 @@ export const adminApi = {
   purgeRequestBodies: () => purge<{ message: string; cleaned: Record<string, number> }>('request-bodies'),
   purgeStats: () => purge<{ message: string }>('stats'),
 
-  async getTimeSeries(params?: {
-    start_date?: string
-    end_date?: string
-    preset?: string
-    granularity?: 'hour' | 'day' | 'week' | 'month'
-    timezone?: string
-    tz_offset_minutes?: number
-    user_id?: string
-    model?: string
-    provider_name?: string
-  }): Promise<Array<Record<string, unknown>>> {
-    const cacheKey = buildCacheKey('admin:stats:time-series', params)
-    return cachedRequest(
-      cacheKey,
-      async () => {
-        const response = await apiClient.get<Array<Record<string, unknown>>>('/api/admin/stats/time-series', { params })
-        return response.data
-      },
-      20 * 1000
-    )
-  },
-
 }

frontend/src/api/analytics.ts

@@ -0,0 +1,396 @@
+import apiClient from './client'
+import { cachedRequest, buildCacheKey } from '@/utils/cache'
+import type { ActivityHeatmap } from '@/types/activity'
+
+export type AnalyticsScopeKind = 'global' | 'me' | 'user' | 'api_key'
+export type AnalyticsGranularity = 'hour' | 'day' | 'week' | 'month'
+export type AnalyticsBreakdownDimension = 'model' | 'provider' | 'api_format' | 'api_key' | 'user'
+export type AnalyticsBreakdownMetric = 'requests_total' | 'total_tokens' | 'total_cost_usd' | 'actual_total_cost_usd'
+export type AnalyticsLeaderboardEntity = 'user' | 'api_key'
+export type AnalyticsLeaderboardMetric = 'requests_total' | 'total_tokens' | 'total_cost_usd'
+
+export interface AnalyticsScope {
+  kind: AnalyticsScopeKind
+  user_id?: string | null
+  api_key_id?: string | null
+}
+
+export interface AnalyticsTimeRange {
+  start_date?: string
+  end_date?: string
+  preset?: string
+  granularity?: AnalyticsGranularity
+  timezone?: string | null
+  tz_offset_minutes?: number
+}
+
+export interface AnalyticsFilters {
+  user_ids?: string[]
+  provider_names?: string[]
+  models?: string[]
+  target_models?: string[]
+  api_key_ids?: string[]
+  api_formats?: string[]
+  request_types?: string[]
+  statuses?: string[]
+  error_categories?: string[]
+  is_stream?: boolean | null
+  has_format_conversion?: boolean | null
+}
+
+export interface AnalyticsBaseRequest {
+  scope: AnalyticsScope
+  time_range: AnalyticsTimeRange
+  filters?: AnalyticsFilters
+}
+
+export interface AnalyticsSummary {
+  requests_total: number
+  requests_success: number
+  requests_error: number
+  requests_stream: number
+  success_rate: number
+  input_tokens: number
+  output_tokens: number
+  input_output_total_tokens: number
+  cache_creation_input_tokens: number
+  cache_creation_input_tokens_5m: number
+  cache_creation_input_tokens_1h: number
+  cache_read_input_tokens: number
+  input_context_tokens: number
+  total_tokens: number
+  cache_hit_rate: number
+  input_cost_usd: number
+  output_cost_usd: number
+  cache_creation_cost_usd: number
+  cache_creation_cost_usd_5m: number
+  cache_creation_cost_usd_1h: number
+  cache_read_cost_usd: number
+  cache_cost_usd: number
+  request_cost_usd: number
+  total_cost_usd: number
+  actual_total_cost_usd: number
+  actual_cache_cost_usd: number
+  avg_response_time_ms: number
+  avg_first_byte_time_ms: number
+  format_conversion_count: number
+  models_used_count: number
+}
+
+export interface AnalyticsCompositionSegment {
+  key: string
+  value: number
+  percentage: number
+}
+
+export interface AnalyticsOverviewResponse {
+  query_context: {
+    scope: AnalyticsScope
+    time_range: AnalyticsTimeRange
+  }
+  summary: AnalyticsSummary
+  composition: {
+    token_segments: AnalyticsCompositionSegment[]
+    cost_segments: AnalyticsCompositionSegment[]
+  }
+}
+
+export interface AnalyticsTimeseriesBucket extends AnalyticsSummary {
+  bucket_start: string
+  bucket_end: string
+}
+
+export interface AnalyticsTimeseriesResponse {
+  buckets: AnalyticsTimeseriesBucket[]
+}
+
+export interface AnalyticsBreakdownRequest extends AnalyticsBaseRequest {
+  dimension: AnalyticsBreakdownDimension
+  metric?: AnalyticsBreakdownMetric
+  limit?: number
+}
+
+export interface AnalyticsBreakdownRow extends AnalyticsSummary {
+  key: string
+  label: string
+  share_of_total_cost: number
+  share_of_total_tokens: number
+}
+
+export interface AnalyticsBreakdownResponse {
+  dimension: AnalyticsBreakdownDimension
+  metric: AnalyticsBreakdownMetric
+  rows: AnalyticsBreakdownRow[]
+}
+
+export interface AnalyticsRecordsRequest extends AnalyticsBaseRequest {
+  search?: {
+    text?: string | null
+    request_id?: string | null
+  }
+  pagination?: {
+    limit?: number
+    offset?: number
+  }
+}
+
+export interface AnalyticsRecord {
+  id: string
+  request_id: string
+  created_at: string | null
+  user_id: string | null
+  username: string | null
+  api_key_id: string | null
+  api_key_name: string | null
+  provider_api_key_name: string | null
+  provider_name: string | null
+  model: string
+  target_model: string | null
+  api_format: string | null
+  request_type: string | null
+  status: string
+  billing_status: string
+  is_stream: boolean
+  has_format_conversion: boolean | null
+  has_fallback?: boolean
+  has_retry?: boolean
+  status_code: number | null
+  error_message: string | null
+  error_category: string | null
+  response_time_ms: number | null
+  first_byte_time_ms: number | null
+  input_tokens: number
+  output_tokens: number
+  input_output_total_tokens: number
+  cache_creation_input_tokens: number
+  cache_creation_input_tokens_5m: number
+  cache_creation_input_tokens_1h: number
+  cache_read_input_tokens: number
+  input_context_tokens: number
+  total_tokens: number
+  input_cost_usd: number
+  output_cost_usd: number
+  cache_creation_cost_usd: number
+  cache_creation_cost_usd_5m: number
+  cache_creation_cost_usd_1h: number
+  cache_read_cost_usd: number
+  cache_cost_usd: number
+  request_cost_usd: number
+  total_cost_usd: number
+  actual_total_cost_usd: number
+  actual_cache_cost_usd: number
+  rate_multiplier: number
+}
+
+export interface AnalyticsRecordsResponse {
+  total: number
+  limit: number
+  offset: number
+  records: AnalyticsRecord[]
+}
+
+export interface AnalyticsFilterOption {
+  value: string
+  label: string
+}
+
+export interface AnalyticsFilterOptionsResponse {
+  providers: AnalyticsFilterOption[]
+  models: AnalyticsFilterOption[]
+  target_models: AnalyticsFilterOption[]
+  api_formats: AnalyticsFilterOption[]
+  request_types: AnalyticsFilterOption[]
+  error_categories: AnalyticsFilterOption[]
+  statuses: AnalyticsFilterOption[]
+  users?: AnalyticsFilterOption[]
+  api_keys?: AnalyticsFilterOption[]
+}
+
+export interface AnalyticsLeaderboardRequest extends AnalyticsBaseRequest {
+  entity: AnalyticsLeaderboardEntity
+  metric: AnalyticsLeaderboardMetric
+  limit?: number
+}
+
+export interface AnalyticsLeaderboardItem {
+  rank: number
+  id: string
+  label: string
+  requests_total: number
+  total_tokens: number
+  total_cost_usd: number
+  actual_total_cost_usd: number
+  metric_value: number
+}
+
+export interface AnalyticsLeaderboardResponse {
+  entity: AnalyticsLeaderboardEntity
+  metric: AnalyticsLeaderboardMetric
+  items: AnalyticsLeaderboardItem[]
+}
+
+export interface AnalyticsPercentilePoint {
+  date: string
+  p50_response_time_ms: number | null
+  p90_response_time_ms: number | null
+  p99_response_time_ms: number | null
+  p50_first_byte_time_ms: number | null
+  p90_first_byte_time_ms: number | null
+  p99_first_byte_time_ms: number | null
+}
+
+export interface AnalyticsErrorCategory {
+  category: string
+  label: string
+  count: number
+}
+
+export interface AnalyticsErrorTrendItem {
+  date: string
+  total: number
+}
+
+export interface AnalyticsProviderHealthItem {
+  provider_name: string
+  requests_total: number
+  success_rate: number
+  error_rate: number
+  avg_response_time_ms: number
+  avg_first_byte_time_ms: number
+}
+
+export interface AnalyticsPerformanceResponse {
+  latency: {
+    response_time_ms: {
+      avg: number
+      p50: number | null
+      p90: number | null
+      p99: number | null
+    }
+    first_byte_time_ms: {
+      avg: number
+      p50: number | null
+      p90: number | null
+      p99: number | null
+    }
+  }
+  percentiles: AnalyticsPercentilePoint[]
+  errors: {
+    total: number
+    rate: number
+    categories: AnalyticsErrorCategory[]
+    trend: AnalyticsErrorTrendItem[]
+  }
+  provider_health: AnalyticsProviderHealthItem[]
+}
+
+export interface AnalyticsActiveRequest {
+  id: string
+  status: 'pending' | 'streaming' | 'completed' | 'failed' | 'cancelled'
+  input_tokens: number
+  output_tokens: number
+  cache_creation_input_tokens?: number | null
+  cache_read_input_tokens?: number | null
+  cost?: number
+  total_cost_usd?: number
+  actual_cost?: number | null
+  actual_total_cost_usd?: number | null
+  rate_multiplier?: number | null
+  response_time_ms: number | null
+  first_byte_time_ms: number | null
+  provider?: string | null
+  provider_name?: string | null
+  api_key_name?: string | null
+  api_format?: string | null
+  endpoint_api_format?: string | null
+  has_format_conversion?: boolean | null
+  target_model?: string | null
+}
+
+export interface AnalyticsActiveRequestsResponse {
+  requests: AnalyticsActiveRequest[]
+}
+
+export const analyticsApi = {
+  async getOverview(payload: AnalyticsBaseRequest): Promise<AnalyticsOverviewResponse> {
+    const response = await apiClient.post<AnalyticsOverviewResponse>('/api/analytics/overview', payload)
+    return response.data
+  },
+
+  async getTimeseries(payload: AnalyticsBaseRequest): Promise<AnalyticsTimeseriesResponse> {
+    const response = await apiClient.post<AnalyticsTimeseriesResponse>('/api/analytics/timeseries', payload)
+    return response.data
+  },
+
+  async getBreakdown(payload: AnalyticsBreakdownRequest): Promise<AnalyticsBreakdownResponse> {
+    const response = await apiClient.post<AnalyticsBreakdownResponse>('/api/analytics/breakdown', payload)
+    return response.data
+  },
+
+  async getRecords(payload: AnalyticsRecordsRequest): Promise<AnalyticsRecordsResponse> {
+    const response = await apiClient.post<AnalyticsRecordsResponse>('/api/analytics/records', payload)
+    return response.data
+  },
+
+  async getFilterOptions(payload: AnalyticsBaseRequest): Promise<AnalyticsFilterOptionsResponse> {
+    const response = await apiClient.post<AnalyticsFilterOptionsResponse>('/api/analytics/filter-options', payload)
+    return response.data
+  },
+
+  async getLeaderboard(payload: AnalyticsLeaderboardRequest): Promise<AnalyticsLeaderboardResponse> {
+    const response = await apiClient.post<AnalyticsLeaderboardResponse>('/api/analytics/leaderboard', payload)
+    return response.data
+  },
+
+  async getPerformance(payload: AnalyticsBaseRequest): Promise<AnalyticsPerformanceResponse> {
+    const response = await apiClient.post<AnalyticsPerformanceResponse>('/api/analytics/performance', payload)
+    return response.data
+  },
+
+  async getHeatmap(payload: {
+    scope: AnalyticsScope
+    user_id?: string | null
+    api_key_id?: string | null
+  }): Promise<ActivityHeatmap> {
+    const cacheKey = buildCacheKey('analytics:heatmap', payload as Record<string, unknown>)
+    return cachedRequest(
+      cacheKey,
+      async () => {
+        const response = await apiClient.post<ActivityHeatmap>('/api/analytics/heatmap', payload)
+        return response.data
+      },
+      60000,
+    )
+  },
+
+  async getActiveRequests(payload: {
+    scope: AnalyticsScope
+    ids?: string[]
+  }): Promise<AnalyticsActiveRequestsResponse> {
+    const response = await apiClient.post<AnalyticsActiveRequestsResponse>('/api/analytics/active-requests', {
+      ...payload,
+      ids: payload.ids ?? [],
+    })
+    return response.data
+  },
+
+  async analyzeCacheAffinityTTL(payload: {
+    scope: AnalyticsScope
+    user_id?: string | null
+    api_key_id?: string | null
+    hours?: number
+  }): Promise<Record<string, unknown>> {
+    const response = await apiClient.post('/api/analytics/cache-affinity/ttl-analysis', payload)
+    return response.data
+  },
+
+  async analyzeCacheAffinityHit(payload: {
+    scope: AnalyticsScope
+    user_id?: string | null
+    api_key_id?: string | null
+    hours?: number
+  }): Promise<Record<string, unknown>> {
+    const response = await apiClient.post('/api/analytics/cache-affinity/hit-analysis', payload)
+    return response.data
+  },
+}

frontend/src/api/audit.ts

@@ -33,6 +33,203 @@ export interface AuditFilters {
   offset?: number
 }
 
+export type MonitoringMetricStatus = 'ok' | 'warning' | 'danger' | 'degraded' | 'error' | 'unknown'
+
+export interface MonitoringCpuMetric {
+  status: MonitoringMetricStatus
+  label: string
+  usage_percent: number | null
+  load_percent: number | null
+  core_count: number
+  message?: string | null
+}
+
+export interface MonitoringMemoryMetric {
+  status: MonitoringMetricStatus
+  label: string
+  used_percent: number | null
+  used_bytes: number | null
+  available_bytes: number | null
+  total_bytes: number | null
+  message?: string | null
+}
+
+export interface MonitoringServiceMetric {
+  status: MonitoringMetricStatus
+  label: string
+  latency_ms: number | null
+  memory_status?: MonitoringMetricStatus
+  memory_label?: string | null
+  used_memory_bytes?: number | null
+  peak_memory_bytes?: number | null
+  maxmemory_bytes?: number | null
+  memory_ceiling_bytes?: number | null
+  memory_source?: 'configured' | 'maxmemory' | 'system' | 'unknown'
+  available_memory_bytes?: number | null
+  memory_percent?: number | null
+  message?: string | null
+}
+
+export interface MonitoringPostgresMetric {
+  status: MonitoringMetricStatus
+  label: string
+  usage_percent: number | null
+  pool_usage_percent?: number | null
+  checked_out: number
+  pool_size: number
+  overflow: number
+  max_capacity: number
+  pool_timeout: number
+  server_connections?: number | null
+  server_max_connections?: number | null
+  server_usage_percent?: number | null
+  storage_status?: MonitoringMetricStatus
+  storage_label?: string | null
+  storage_total_bytes?: number | null
+  storage_free_bytes?: number | null
+  storage_free_percent?: number | null
+  database_size_bytes?: number | null
+  storage_message?: string | null
+  message?: string | null
+}
+
+export interface MonitoringSystemStatus {
+  timestamp: string
+  users: {
+    total: number
+    active: number
+  }
+  providers: {
+    total: number
+    active: number
+  }
+  api_keys: {
+    total: number
+    active: number
+  }
+  today_stats: {
+    requests: number
+    tokens: number
+    cost_usd: number
+  }
+  recent_errors: number
+  system_metrics?: {
+    cpu: MonitoringCpuMetric
+    memory: MonitoringMemoryMetric
+    redis: MonitoringServiceMetric
+    postgres: MonitoringPostgresMetric
+  }
+}
+
+function parseNumericLike(value: unknown): number {
+  if (typeof value === 'number') {
+    return Number.isFinite(value) ? value : 0
+  }
+  if (typeof value === 'string') {
+    const normalized = value.trim().replace(/[$,\s]/g, '')
+    if (!normalized) return 0
+    const parsed = Number(normalized)
+    return Number.isFinite(parsed) ? parsed : 0
+  }
+  return 0
+}
+
+function parseNullableNumericLike(value: unknown): number | null {
+  if (value == null) {
+    return null
+  }
+  if (typeof value === 'number') {
+    return Number.isFinite(value) ? value : null
+  }
+  if (typeof value === 'string') {
+    const normalized = value.trim().replace(/[$,\s]/g, '')
+    if (!normalized) return null
+    const parsed = Number(normalized)
+    return Number.isFinite(parsed) ? parsed : null
+  }
+  return null
+}
+
+function normalizeMonitoringSystemStatus(data: MonitoringSystemStatus): MonitoringSystemStatus {
+  return {
+    ...data,
+    users: {
+      total: parseNumericLike(data.users?.total),
+      active: parseNumericLike(data.users?.active),
+    },
+    providers: {
+      total: parseNumericLike(data.providers?.total),
+      active: parseNumericLike(data.providers?.active),
+    },
+    api_keys: {
+      total: parseNumericLike(data.api_keys?.total),
+      active: parseNumericLike(data.api_keys?.active),
+    },
+    today_stats: {
+      requests: parseNumericLike(data.today_stats?.requests),
+      tokens: parseNumericLike(data.today_stats?.tokens),
+      cost_usd: parseNumericLike(data.today_stats?.cost_usd),
+    },
+    recent_errors: parseNumericLike(data.recent_errors),
+    system_metrics: data.system_metrics ? {
+      cpu: {
+        status: data.system_metrics.cpu?.status ?? 'unknown',
+        label: data.system_metrics.cpu?.label ?? '未知',
+        usage_percent: parseNullableNumericLike(data.system_metrics.cpu?.usage_percent),
+        load_percent: parseNullableNumericLike(data.system_metrics.cpu?.load_percent),
+        core_count: parseNumericLike(data.system_metrics.cpu?.core_count),
+        message: data.system_metrics.cpu?.message ?? null,
+      },
+      memory: {
+        status: data.system_metrics.memory?.status ?? 'unknown',
+        label: data.system_metrics.memory?.label ?? '未知',
+        used_percent: parseNullableNumericLike(data.system_metrics.memory?.used_percent),
+        used_bytes: parseNullableNumericLike(data.system_metrics.memory?.used_bytes),
+        available_bytes: parseNullableNumericLike(data.system_metrics.memory?.available_bytes),
+        total_bytes: parseNullableNumericLike(data.system_metrics.memory?.total_bytes),
+        message: data.system_metrics.memory?.message ?? null,
+      },
+      redis: {
+        status: data.system_metrics.redis?.status ?? 'unknown',
+        label: data.system_metrics.redis?.label ?? '未知',
+        latency_ms: parseNullableNumericLike(data.system_metrics.redis?.latency_ms),
+        memory_status: data.system_metrics.redis?.memory_status ?? 'unknown',
+        memory_label: data.system_metrics.redis?.memory_label ?? '未知',
+        used_memory_bytes: parseNullableNumericLike(data.system_metrics.redis?.used_memory_bytes),
+        peak_memory_bytes: parseNullableNumericLike(data.system_metrics.redis?.peak_memory_bytes),
+        maxmemory_bytes: parseNullableNumericLike(data.system_metrics.redis?.maxmemory_bytes),
+        memory_ceiling_bytes: parseNullableNumericLike(data.system_metrics.redis?.memory_ceiling_bytes),
+        memory_source: (data.system_metrics.redis?.memory_source as MonitoringServiceMetric['memory_source']) ?? 'unknown',
+        available_memory_bytes: parseNullableNumericLike(data.system_metrics.redis?.available_memory_bytes),
+        memory_percent: parseNullableNumericLike(data.system_metrics.redis?.memory_percent),
+        message: data.system_metrics.redis?.message ?? null,
+      },
+      postgres: {
+        status: data.system_metrics.postgres?.status ?? 'unknown',
+        label: data.system_metrics.postgres?.label ?? '未知',
+        usage_percent: parseNullableNumericLike(data.system_metrics.postgres?.usage_percent),
+        pool_usage_percent: parseNullableNumericLike(data.system_metrics.postgres?.pool_usage_percent),
+        checked_out: parseNumericLike(data.system_metrics.postgres?.checked_out),
+        pool_size: parseNumericLike(data.system_metrics.postgres?.pool_size),
+        overflow: parseNumericLike(data.system_metrics.postgres?.overflow),
+        max_capacity: parseNumericLike(data.system_metrics.postgres?.max_capacity),
+        pool_timeout: parseNumericLike(data.system_metrics.postgres?.pool_timeout),
+        server_connections: parseNullableNumericLike(data.system_metrics.postgres?.server_connections),
+        server_max_connections: parseNullableNumericLike(data.system_metrics.postgres?.server_max_connections),
+        server_usage_percent: parseNullableNumericLike(data.system_metrics.postgres?.server_usage_percent),
+        storage_status: data.system_metrics.postgres?.storage_status ?? 'unknown',
+        storage_label: data.system_metrics.postgres?.storage_label ?? '未知',
+        storage_total_bytes: parseNullableNumericLike(data.system_metrics.postgres?.storage_total_bytes),
+        storage_free_bytes: parseNullableNumericLike(data.system_metrics.postgres?.storage_free_bytes),
+        storage_free_percent: parseNullableNumericLike(data.system_metrics.postgres?.storage_free_percent),
+        database_size_bytes: parseNullableNumericLike(data.system_metrics.postgres?.database_size_bytes),
+        storage_message: data.system_metrics.postgres?.storage_message ?? null,
+        message: data.system_metrics.postgres?.message ?? null,
+      },
+    } : undefined,
+  }
+}
+
 function normalizeAuditResponse(data: Record<string, unknown>): AuditLogsResponse {
   const items: AuditLog[] = (data.items ?? data.logs ?? []) as AuditLog[]
   const meta: PaginationMeta = (data.meta as PaginationMeta) ?? {
@@ -78,6 +275,11 @@ export const auditApi = {
     return response.data
   },
 
+  async getSystemStatus(): Promise<MonitoringSystemStatus> {
+    const response = await apiClient.get<MonitoringSystemStatus>('/api/admin/monitoring/system-status')
+    return normalizeMonitoringSystemStatus(response.data)
+  },
+
   // 分析用户行为 (管理员)
   async analyzeUserBehavior(userId: number, days: number = 7): Promise<{
     analysis: Record<string, unknown>

frontend/src/api/auth.ts

@@ -24,13 +24,6 @@ export interface UserPreferences {
   [key: string]: unknown // 允许扩展其他偏好设置
 }
 
-export interface UserStats {
-  total_requests?: number
-  total_cost?: number
-  last_request_at?: string
-  [key: string]: unknown // 允许扩展其他统计数据
-}
-
 export interface SendVerificationCodeRequest {
   email: string
 }
@@ -113,13 +106,9 @@ export interface User {
   role: string  // 'admin' or 'user'
   is_active: boolean
   billing?: BillingSummary
-  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
   created_at: string
   last_login_at?: string
   preferences?: UserPreferences
-  stats?: UserStats
 }
 
 export const authApi = {

frontend/src/api/cache.ts

@@ -288,7 +288,12 @@ export const cacheAnalysisApi = {
     api_key_id?: string
     hours?: number
   }): Promise<TTLAnalysisResponse> {
-    const response = await api.get('/api/admin/usage/cache-affinity/ttl-analysis', { params })
+    const response = await api.post('/api/analytics/cache-affinity/ttl-analysis', {
+      scope: { kind: 'global' },
+      user_id: params?.user_id ?? null,
+      api_key_id: params?.api_key_id ?? null,
+      hours: params?.hours,
+    })
     return response.data
   },
 
@@ -300,7 +305,12 @@ export const cacheAnalysisApi = {
     api_key_id?: string
     hours?: number
   }): Promise<CacheHitAnalysisResponse> {
-    const response = await api.get('/api/admin/usage/cache-affinity/hit-analysis', { params })
+    const response = await api.post('/api/analytics/cache-affinity/hit-analysis', {
+      scope: { kind: 'global' },
+      user_id: params?.user_id ?? null,
+      api_key_id: params?.api_key_id ?? null,
+      hours: params?.hours,
+    })
     return response.data
   },
 
@@ -319,7 +329,13 @@ export const cacheAnalysisApi = {
     return cachedRequest(
       cacheKey,
       async () => {
-        const response = await api.get('/api/admin/usage/cache-affinity/interval-timeline', { params })
+        const response = await api.post('/api/analytics/interval-timeline', {
+          scope: params?.user_id ? { kind: 'user', user_id: params.user_id } : { kind: 'global' },
+          user_id: params?.user_id ?? null,
+          hours: params?.hours,
+          limit: params?.limit,
+          include_user_info: params?.include_user_info ?? false,
+        })
         return response.data
       },
       30000

frontend/src/api/dashboard.ts

@@ -1,382 +0,0 @@
-import apiClient from './client'
-import { cachedRequest, buildCacheKey } from '@/utils/cache'
-
-export interface DashboardStat {
-  name: string
-  value: string
-  subValue?: string
-  change?: string
-  changeType?: 'increase' | 'decrease' | 'neutral'
-  extraBadge?: string
-  icon: string
-}
-
-export interface RecentRequest {
-  id: string // UUID
-  user: string
-  model: string
-  tokens: number
-  time: string
-}
-
-export interface ProviderStatus {
-  name: string
-  status: 'active' | 'inactive'
-  requests: number
-}
-
-// 系统健康指标（管理员专用）
-export interface SystemHealth {
-  avg_response_time: number
-  error_rate: number
-  error_requests: number
-  fallback_count: number
-  total_requests: number
-}
-
-// 成本统计（管理员专用）
-export interface CostStats {
-  total_cost: number
-  total_actual_cost: number
-  cost_savings: number
-}
-
-// 缓存统计
-export interface CacheStats {
-  cache_creation_tokens: number
-  cache_read_tokens: number
-  cache_creation_cost?: number
-  cache_read_cost?: number
-  cache_hit_rate?: number
-  total_cache_tokens: number
-}
-
-// 用户统计（管理员专用）
-export interface UserStats {
-  total: number
-  active: number
-}
-
-// Token 详细分类
-export interface TokenBreakdown {
-  input: number
-  output: number
-  cache_creation: number
-  cache_read: number
-}
-
-export interface DashboardStatsResponse {
-  stats: DashboardStat[]
-  today?: {
-    requests: number
-    tokens: number
-    cost: number
-    actual_cost?: number
-    cache_creation_tokens?: number
-    cache_read_tokens?: number
-  }
-  api_keys?: {
-    total: number
-    active: number
-  }
-  tokens?: {
-    month: number
-  }
-  // 管理员专用字段
-  system_health?: SystemHealth
-  cost_stats?: CostStats
-  cache_stats?: CacheStats
-  users?: UserStats
-  token_breakdown?: TokenBreakdown
-  // 普通用户专用字段
-  monthly_cost?: number
-}
-
-export interface RecentRequestsResponse {
-  requests: RecentRequest[]
-}
-
-export interface ProviderStatusResponse {
-  providers: ProviderStatus[]
-}
-
-// 视频/图像/音频计费信息
-export interface VideoBilling {
-  task_type: 'video' | 'image' | 'audio'
-  duration_seconds?: number  // 视频时长（秒）
-  resolution?: string        // 分辨率
-  video_price_per_second?: number  // 每秒单价
-  video_cost?: number        // 视频费用
-  cost?: number              // 总费用
-  rule_name?: string         // 计费规则名称
-  expression?: string        // 计费公式
-  status?: string            // 计费状态
-}
-
-export interface RequestDetail {
-  id: string // UUID
-  request_id: string
-  user: {
-    id: string // UUID
-    username: string
-    email: string
-  }
-  api_key: {
-    id: string // UUID
-    name: string
-    display: string
-  }
-  provider: string
-  api_format?: string
-  model: string
-  target_model?: string | null  // 映射后的目标模型名
-  tokens: {
-    input: number
-    output: number
-    total: number
-  }
-  cost: {
-    input: number
-    output: number
-    total: number
-  }
-  // Additional token fields
-  input_tokens?: number
-  output_tokens?: number
-  total_tokens?: number
-  cache_creation_input_tokens?: number
-  cache_creation_input_tokens_5m?: number
-  cache_creation_input_tokens_1h?: number
-  cache_read_input_tokens?: number
-  // Additional cost fields
-  input_cost?: number
-  output_cost?: number
-  total_cost?: number
-  cache_creation_cost?: number
-  cache_read_cost?: number
-  request_cost?: number  // 按次计费费用
-  // Historical pricing fields (per 1M tokens)
-  input_price_per_1m?: number
-  output_price_per_1m?: number
-  cache_creation_price_per_1m?: number
-  cache_read_price_per_1m?: number
-  price_per_request?: number  // 按次计费价格
-  request_type: string
-  is_stream: boolean
-  status_code: number
-  status?: string  // pending, streaming, completed, failed, cancelled
-  error_message?: string
-  response_time_ms: number
-  created_at: string
-  request_headers?: Record<string, unknown>
-  request_body?: Record<string, unknown>
-  provider_request_headers?: Record<string, unknown>
-  provider_request_body?: Record<string, unknown>
-  response_headers?: Record<string, unknown>
-  client_response_headers?: Record<string, unknown>
-  response_body?: Record<string, unknown>
-  client_response_body?: Record<string, unknown>
-  has_request_body?: boolean
-  has_provider_request_body?: boolean
-  has_response_body?: boolean
-  has_client_response_body?: boolean
-  metadata?: Record<string, unknown>
-  // 阶梯计费信息
-  tiered_pricing?: {
-    total_input_context: number  // 总输入上下文 (input + cache_read)
-    tier_index: number  // 命中的阶梯索引 (0-based)
-    tier_count: number  // 阶梯总数
-    source?: 'provider' | 'global'  // 定价来源: 提供商或全局
-    current_tier: {  // 当前命中的阶梯配置
-      up_to?: number | null
-      input_price_per_1m: number
-      output_price_per_1m: number
-      cache_creation_price_per_1m?: number
-      cache_read_price_per_1m?: number
-      cache_ttl_pricing?: Array<{
-        ttl_minutes: number
-        cache_creation_price_per_1m?: number
-        cache_read_price_per_1m?: number
-      }>
-    }
-    tiers: Array<{  // 完整阶梯配置列表
-      up_to?: number | null
-      input_price_per_1m: number
-      output_price_per_1m: number
-      cache_creation_price_per_1m?: number
-      cache_read_price_per_1m?: number
-      cache_ttl_pricing?: Array<{
-        ttl_minutes: number
-        cache_creation_price_per_1m?: number
-        cache_read_price_per_1m?: number
-      }>
-    }>
-  } | null
-  // 视频/图像/音频计费信息
-  video_billing?: VideoBilling | null
-}
-
-export interface CurlData {
-  url: string
-  method: string
-  headers: Record<string, string>
-  body: Record<string, unknown>
-  curl: string
-}
-
-export interface ReplayRequest {
-  provider_id?: string
-  endpoint_id?: string
-  api_key_id?: string
-  body_override?: Record<string, unknown>
-}
-
-export interface ReplayResponse {
-  url: string
-  provider: string
-  status_code: number
-  response_headers: Record<string, string>
-  response_body: Record<string, unknown>
-  response_time_ms: number
-  mapping?: {
-    source_model: string
-    original_target_model?: string | null
-    resolved_model: string
-    target_provider_id: string
-    target_provider: string
-    target_endpoint_id: string
-    target_api_format: string
-    replay_mode: string
-    mapping_applied: boolean
-    mapping_source: string
-  }
-}
-
-export interface ModelBreakdown {
-  model: string
-  requests: number
-  tokens: number
-  cost: number
-}
-
-export interface ModelSummary {
-  model: string
-  requests: number
-  tokens: number
-  cost: number
-  avg_response_time: number
-  cost_per_request: number
-  tokens_per_request: number
-}
-
-export interface ProviderSummary {
-  provider: string
-  requests: number
-  tokens: number
-  cost: number
-}
-
-export interface DailyStat {
-  date: string // ISO date string
-  requests: number
-  tokens: number
-  cost: number
-  avg_response_time: number // in seconds
-  unique_models: number
-  unique_providers?: number // 仅管理员返回
-  model_breakdown: ModelBreakdown[]
-}
-
-export interface DailyStatsResponse {
-  daily_stats: DailyStat[]
-  model_summary: ModelSummary[]
-  provider_summary?: ProviderSummary[] // 仅管理员返回
-  period: {
-    start_date: string
-    end_date: string
-    days: number
-  }
-}
-
-export interface TimeRangeParams {
-  start_date?: string
-  end_date?: string
-  preset?: string
-  granularity?: 'hour' | 'day' | 'week' | 'month'
-  timezone?: string
-  tz_offset_minutes?: number
-}
-
-export const dashboardApi = {
-  // 获取仪表盘统计数据
-  async getStats(params?: TimeRangeParams): Promise<DashboardStatsResponse> {
-    const cacheKey = buildCacheKey('dashboard:stats', params)
-    return cachedRequest(
-      cacheKey,
-      async () => {
-        const response = await apiClient.get<DashboardStatsResponse>('/api/dashboard/stats', { params })
-        return response.data
-      },
-      10 * 1000
-    )
-  },
-
-  // 获取最近的请求记录
-  async getRecentRequests(limit: number = 10): Promise<RecentRequest[]> {
-    const response = await apiClient.get<RecentRequestsResponse>('/api/dashboard/recent-requests', {
-      params: { limit }
-    })
-    return response.data.requests
-  },
-
-  // 获取提供商状态
-  async getProviderStatus(): Promise<ProviderStatus[]> {
-    return cachedRequest(
-      'dashboard:provider-status',
-      async () => {
-        const response = await apiClient.get<ProviderStatusResponse>('/api/dashboard/provider-status')
-        return response.data.providers
-      },
-      20 * 1000
-    )
-  },
-
-  // 获取请求详情
-  // NOTE: This method now calls the new RESTful API at /api/admin/usage/{id}
-  async getRequestDetail(requestId: string, options: { includeBodies?: boolean } = {}): Promise<RequestDetail> {
-    const response = await apiClient.get<RequestDetail>(`/api/admin/usage/${requestId}`, {
-      params: { include_bodies: options.includeBodies ?? true },
-    })
-    return response.data
-  },
-
-  // 获取每日统计数据
-  async getDailyStats(params?: TimeRangeParams & { days?: number }): Promise<DailyStatsResponse> {
-    const cacheKey = buildCacheKey('dashboard:daily-stats', params)
-    return cachedRequest(
-      cacheKey,
-      async () => {
-        const response = await apiClient.get<DailyStatsResponse>('/api/dashboard/daily-stats', {
-          params
-        })
-        return response.data
-      },
-      20 * 1000
-    )
-  },
-
-  // 获取 cURL 命令数据（含明文 API Key）
-  async getCurlData(requestId: string): Promise<CurlData> {
-    const response = await apiClient.get<CurlData>(`/api/admin/usage/${requestId}/curl`)
-    return response.data
-  },
-
-  // 回放请求到提供商
-  async replayRequest(requestId: string, params?: ReplayRequest): Promise<ReplayResponse> {
-    const response = await apiClient.post<ReplayResponse>(
-      `/api/admin/usage/${requestId}/replay`,
-      params || {}
-    )
-    return response.data
-  }
-}

frontend/src/api/endpoints/keys.ts

@@ -4,6 +4,34 @@ import type { EndpointAPIKey, AllowedModels } from './types'
 // Re-export types for convenience
 export type { EndpointAPIKey, AllowedModels }
 
+export interface GroupedFormatKey {
+  id: string
+  provider_id: string
+  name: string
+  auth_type?: string
+  api_key_masked: string
+  internal_priority: number
+  global_priority_by_format: Record<string, number> | null
+  format_priority: number | null
+  rate_multipliers: Record<string, number> | null
+  is_active: boolean
+  provider_active: boolean
+  pool_enabled: boolean
+  circuit_breaker_open: boolean
+  provider_name: string
+  api_format: string
+  api_formats: string[]
+  capabilities: string[]
+  health_score: number | null
+  success_rate: number | null
+  avg_response_time_ms: number | null
+  request_count: number
+}
+
+function toNumberOrNull(value: unknown): number | null {
+  return typeof value === 'number' && Number.isFinite(value) ? value : null
+}
+
 /**
  * 能力定义类型
  */
@@ -125,6 +153,51 @@ export async function getProviderKeys(providerId: string): Promise<EndpointAPIKe
   return allKeys
 }
 
+/**
+ * 获取按 API 格式分组的 Key 列表
+ */
+export async function getKeysGroupedByFormat(): Promise<Record<string, GroupedFormatKey[]>> {
+  const response = await client.get('/api/admin/endpoints/keys/grouped-by-format')
+  const grouped = response.data as Record<string, Array<Record<string, unknown>>>
+  const result: Record<string, GroupedFormatKey[]> = {}
+
+  for (const [apiFormat, keys] of Object.entries(grouped || {})) {
+    if (!Array.isArray(keys)) continue
+
+    result[apiFormat] = keys.map((key) => ({
+      id: String(key.id || ''),
+      provider_id: String(key.provider_id || ''),
+      name: String(key.name || 'Unnamed Key'),
+      auth_type: typeof key.auth_type === 'string' ? key.auth_type : undefined,
+      api_key_masked: String(key.api_key_masked || '***'),
+      internal_priority: toNumberOrNull(key.internal_priority) ?? 0,
+      global_priority_by_format:
+        key.global_priority_by_format && typeof key.global_priority_by_format === 'object'
+          ? (key.global_priority_by_format as Record<string, number>)
+          : null,
+      format_priority: toNumberOrNull(key.format_priority),
+      rate_multipliers:
+        key.rate_multipliers && typeof key.rate_multipliers === 'object'
+          ? (key.rate_multipliers as Record<string, number>)
+          : null,
+      is_active: key.is_active !== false,
+      provider_active: key.provider_active !== false,
+      pool_enabled: key.pool_enabled === true,
+      circuit_breaker_open: key.circuit_breaker_open === true,
+      provider_name: String(key.provider_name || 'Unknown Provider'),
+      api_format: typeof key.api_format === 'string' ? key.api_format : apiFormat,
+      api_formats: Array.isArray(key.api_formats) ? key.api_formats.map(item => String(item)) : [apiFormat],
+      capabilities: Array.isArray(key.capabilities) ? key.capabilities.map(item => String(item)) : [],
+      health_score: toNumberOrNull(key.health_score),
+      success_rate: toNumberOrNull(key.success_rate),
+      avg_response_time_ms: toNumberOrNull(key.avg_response_time_ms),
+      request_count: toNumberOrNull(key.request_count) ?? 0,
+    }))
+  }
+
+  return result
+}
+
 /**
  * 为 Provider 添加 Key
  */

frontend/src/api/endpoints/provider_oauth.ts

@@ -94,7 +94,7 @@ export async function completeProviderLevelOAuth(
 
 export async function importProviderRefreshToken(
   providerId: string,
-  data: { refresh_token: string; name?: string; proxy_node_id?: string }
+  data: { refresh_token?: string; access_token?: string; name?: string; proxy_node_id?: string }
 ): Promise<ProviderOAuthCompleteResponseWithKey> {
   const resp = await client.post(`/api/admin/provider-oauth/providers/${providerId}/import-refresh-token`, data)
   return resp.data

frontend/src/api/endpoints/types/provider.ts

@@ -292,6 +292,7 @@ export interface EndpointAPIKey {
   oauth_account_user_id?: string | null  // Codex ChatGPT account-user 联合 ID
   oauth_account_name?: string | null
   oauth_organizations?: OAuthOrganizationInfo[] | null  // OAuth 关联组织/工作区摘要
+  oauth_temporary?: boolean | null  // 是否为仅 Access Token 导入的临时 OAuth 账号
   oauth_invalid_at?: number | null  // 兼容字段；优先使用 status_snapshot.oauth
   oauth_invalid_reason?: string | null  // 兼容字段；优先使用 status_snapshot.oauth
   status_snapshot?: ProviderKeyStatusSnapshot | null
@@ -594,6 +595,7 @@ export interface HealthSummary {
     total: number
     active: number
     unhealthy: number
+    circuit_open?: number
   }
 }
 

frontend/src/api/me.ts

@@ -1,7 +1,5 @@
 import apiClient from './client'
-import type { ActivityHeatmap } from '@/types/activity'
 import type { TieredPricingConfig } from './endpoints/types'
-import { cachedRequest, buildCacheKey } from '@/utils/cache'
 import type { BillingSummary } from './auth'
 import type { UserSession } from '@/types/session'
 
@@ -25,8 +23,6 @@ export interface Profile {
 export interface UserPreferences {
   avatar_url?: string
   bio?: string
-  default_provider_id?: string // UUID
-  default_provider?: Record<string, unknown>
   theme: string
   language: string
   timezone?: string
@@ -37,103 +33,6 @@ export interface UserPreferences {
   }
 }
 
-// 提供商配置接口
-export interface ProviderConfig {
-  provider_id: string
-  priority: number  // 优先级（越高越优先）
-  weight: number    // 负载均衡权重
-  enabled: boolean  // 是否启用
-}
-
-// 使用记录接口
-export interface UsageRecordDetail {
-  id: string
-  provider: string
-  model: string
-  input_tokens: number
-  output_tokens: number
-  total_tokens: number
-  cost: number  // 官方费率
-  actual_cost?: number  // 倍率消耗（仅管理员可见）
-  rate_multiplier?: number  // 成本倍率（仅管理员可见）
-  response_time_ms?: number
-  is_stream: boolean
-  created_at: string
-  cache_creation_input_tokens?: number
-  cache_read_input_tokens?: number
-  status_code: number
-  error_message?: string
-  input_price_per_1m: number
-  output_price_per_1m: number
-  cache_creation_price_per_1m?: number
-  cache_read_price_per_1m?: number
-  price_per_request?: number  // 按次计费价格
-  api_key?: {
-    id: string
-    name: string
-    display: string
-  }
-}
-
-// 模型统计接口
-export interface ModelSummary {
-  model: string
-  requests: number
-  input_tokens: number
-  output_tokens: number
-  total_tokens: number
-  cache_read_tokens?: number
-  cache_hit_rate?: number
-  total_cost_usd: number
-  actual_total_cost_usd?: number  // 倍率消耗（仅管理员可见）
-}
-
-// 提供商统计接口
-export interface ProviderSummary {
-  provider: string
-  requests: number
-  total_tokens: number
-  cache_read_tokens?: number
-  cache_hit_rate?: number
-  total_cost_usd: number
-  success_rate: number | null
-  avg_response_time_ms: number | null
-}
-
-// API 格式统计接口
-export interface ApiFormatSummary {
-  api_format: string
-  request_count: number
-  total_tokens: number
-  cache_read_tokens: number
-  cache_hit_rate: number
-  total_cost_usd: number
-  avg_response_time_ms: number
-}
-
-// 使用统计响应接口
-export interface UsageResponse {
-  total_requests: number
-  total_input_tokens: number
-  total_output_tokens: number
-  total_tokens: number
-  total_cost: number  // 官方费率
-  total_actual_cost?: number  // 倍率消耗（仅管理员可见）
-  avg_response_time: number
-  billing: BillingSummary
-  summary_by_model: ModelSummary[]
-  summary_by_provider?: ProviderSummary[]
-  summary_by_api_format?: ApiFormatSummary[]
-  pagination?: {
-    total: number
-    limit: number
-    offset: number
-    has_more: boolean
-  }
-  records: UsageRecordDetail[]
-  activity_heatmap?: ActivityHeatmap | null
-}
-
 export interface ApiKey {
   id: string // UUID
   name: string
@@ -146,7 +45,6 @@ export interface ApiKey {
   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
 }
 
@@ -249,51 +147,6 @@ export const meApi = {
     return response.data
   },
 
-  // 使用统计
-  async getUsage(params?: {
-    start_date?: string
-    end_date?: string
-    preset?: string
-    timezone?: string
-    tz_offset_minutes?: number
-    search?: string  // 通用搜索：密钥名、模型名
-    limit?: number
-    offset?: number
-  }): Promise<UsageResponse> {
-    const response = await apiClient.get<UsageResponse>('/api/users/me/usage', { params })
-    return response.data
-  },
-
-  // 获取活跃请求状态（用于轮询更新）
-  async getActiveRequests(ids?: string): Promise<{
-    requests: Array<{
-      id: string
-      status: 'pending' | 'streaming' | 'completed' | 'failed' | 'cancelled'
-      input_tokens: number
-      output_tokens: number
-      cache_creation_input_tokens?: number | null
-      cache_read_input_tokens?: number | null
-      cost: number
-      actual_cost?: number | null
-      rate_multiplier?: number | null
-      response_time_ms: number | null
-      first_byte_time_ms: number | null
-      api_format?: string | null
-      endpoint_api_format?: string | null
-      has_format_conversion?: boolean | null
-    }>
-  }> {
-    const params = ids ? { ids } : {}
-    const response = await apiClient.get('/api/users/me/usage/active', { params })
-    return response.data
-  },
-
-  // 获取可用的提供商
-  async getAvailableProviders(): Promise<Array<Record<string, unknown>>> {
-    const response = await apiClient.get('/api/users/me/providers')
-    return response.data
-  },
-
   // 获取用户可用的模型列表
   async getAvailableModels(params?: {
     skip?: number
@@ -334,16 +187,6 @@ export const meApi = {
     return response.data
   },
 
-  // 提供商绑定管理相关方法已移除，改为直接从可用提供商中选择
-
-  // API密钥提供商关联
-  async updateApiKeyProviders(keyId: string, data: {
-    allowed_providers?: ProviderConfig[]
-  }): Promise<{ message: string }> {
-    const response = await apiClient.put(`/api/users/me/api-keys/${keyId}/providers`, data)
-    return response.data
-  },
-
   // API密钥能力配置
   async updateApiKeyCapabilities(keyId: string, data: {
     force_capabilities?: Record<string, boolean> | null
@@ -370,39 +213,4 @@ export const meApi = {
     return response.data
   },
 
-  // 获取请求间隔时间线（用于散点图）
-  async getIntervalTimeline(params?: {
-    hours?: number
-    limit?: number
-  }): Promise<{
-    analysis_period_hours: number
-    total_points: number
-    points: Array<{ x: string; y: number; model?: string }>
-    models?: string[]
-  }> {
-    const cacheKey = buildCacheKey('me:interval-timeline', params as Record<string, unknown> | undefined)
-    return cachedRequest(
-      cacheKey,
-      async () => {
-        const response = await apiClient.get('/api/users/me/usage/interval-timeline', { params })
-        return response.data
-      },
-      30000
-    )
-  },
-
-  /**
-   * 获取活跃度热力图数据（用户）
-   * 后端已缓存5分钟
-   */
-  async getActivityHeatmap(): Promise<ActivityHeatmap> {
-    return cachedRequest(
-      'me-activity-heatmap',
-      async () => {
-        const response = await apiClient.get<ActivityHeatmap>('/api/users/me/usage/heatmap')
-        return response.data
-      },
-      60000
-    )
-  }
 }

frontend/src/api/request-details.ts

@@ -0,0 +1,161 @@
+import apiClient from './client'
+
+export interface VideoBilling {
+  task_type: 'video' | 'image' | 'audio'
+  duration_seconds?: number
+  resolution?: string
+  video_price_per_second?: number
+  video_cost?: number
+  cost?: number
+  rule_name?: string
+  expression?: string
+  status?: string
+}
+
+export interface RequestDetail {
+  id: string
+  request_id: string
+  user: {
+    id: string | null
+    username: string | null
+    email: string | null
+  }
+  api_key: {
+    id: string | null
+    name: string | null
+    display: string | null
+  }
+  provider_api_key: {
+    id: string | null
+    name: string | null
+  }
+  provider: string
+  api_format?: string
+  model: string
+  target_model?: string | null
+  tokens: {
+    input: number
+    output: number
+    total: number
+  }
+  cost: {
+    input: number
+    output: number
+    total: number
+  }
+  input_tokens?: number
+  output_tokens?: number
+  total_tokens?: number
+  cache_creation_input_tokens?: number
+  cache_creation_input_tokens_5m?: number
+  cache_creation_input_tokens_1h?: number
+  cache_read_input_tokens?: number
+  input_cost?: number
+  output_cost?: number
+  total_cost?: number
+  cache_creation_cost?: number
+  cache_read_cost?: number
+  request_cost?: number
+  input_price_per_1m?: number
+  output_price_per_1m?: number
+  cache_creation_price_per_1m?: number
+  cache_read_price_per_1m?: number
+  price_per_request?: number
+  request_type: string
+  is_stream: boolean
+  status_code: number
+  status?: string
+  error_message?: string
+  response_time_ms: number
+  created_at: string
+  request_headers?: Record<string, unknown>
+  request_body?: Record<string, unknown>
+  provider_request_headers?: Record<string, unknown>
+  provider_request_body?: Record<string, unknown>
+  response_headers?: Record<string, unknown>
+  client_response_headers?: Record<string, unknown>
+  response_body?: Record<string, unknown>
+  client_response_body?: Record<string, unknown>
+  has_request_body?: boolean
+  has_provider_request_body?: boolean
+  has_response_body?: boolean
+  has_client_response_body?: boolean
+  metadata?: Record<string, unknown>
+  tiered_pricing?: {
+    total_input_context: number
+    tier_index: number
+    tier_count: number
+    source?: 'provider' | 'global'
+    current_tier: {
+      up_to?: number | null
+      input_price_per_1m: number
+      output_price_per_1m: number
+      cache_creation_price_per_1m?: number
+      cache_read_price_per_1m?: number
+      cache_ttl_pricing?: Array<{
+        ttl_minutes: number
+        cache_creation_price_per_1m?: number
+        cache_read_price_per_1m?: number
+      }>
+    }
+    tiers: Array<{
+      up_to?: number | null
+      input_price_per_1m: number
+      output_price_per_1m: number
+      cache_creation_price_per_1m?: number
+      cache_read_price_per_1m?: number
+      cache_ttl_pricing?: Array<{
+        ttl_minutes: number
+        cache_creation_price_per_1m?: number
+        cache_read_price_per_1m?: number
+      }>
+    }>
+  } | null
+  video_billing?: VideoBilling | null
+}
+
+export interface CurlData {
+  url: string
+  method: string
+  headers: Record<string, string>
+  body: Record<string, unknown>
+  curl: string
+}
+
+export interface ReplayRequest {
+  provider_id?: string
+  endpoint_id?: string
+  api_key_id?: string
+  body_override?: Record<string, unknown>
+}
+
+export interface ReplayResponse {
+  url: string
+  provider: string
+  status_code: number
+  response_headers: Record<string, string>
+  response_body: Record<string, unknown>
+  response_time_ms: number
+}
+
+export const requestDetailsApi = {
+  async getRequestDetail(requestId: string, options: { includeBodies?: boolean } = {}): Promise<RequestDetail> {
+    const response = await apiClient.get<RequestDetail>(`/api/admin/usage/${requestId}`, {
+      params: { include_bodies: options.includeBodies ?? true },
+    })
+    return response.data
+  },
+
+  async getCurlData(requestId: string): Promise<CurlData> {
+    const response = await apiClient.get<CurlData>(`/api/admin/usage/${requestId}/curl`)
+    return response.data
+  },
+
+  async replayRequest(requestId: string, params?: ReplayRequest): Promise<ReplayResponse> {
+    const response = await apiClient.post<ReplayResponse>(
+      `/api/admin/usage/${requestId}/replay`,
+      params || {},
+    )
+    return response.data
+  },
+}

frontend/src/api/usage.ts

@@ -1,247 +0,0 @@
-import apiClient from './client'
-import { cachedRequest, dedupedRequest, buildCacheKey } from '@/utils/cache'
-import type { ActivityHeatmap } from '@/types/activity'
-
-export interface UsageRecord {
-  id: string // UUID
-  user_id: string // UUID
-  username?: string
-  provider_id?: string // UUID
-  provider_name?: string
-  model: string
-  input_tokens: number
-  output_tokens: number
-  cache_creation_input_tokens?: number
-  cache_read_input_tokens?: number
-  total_tokens: number
-  cost?: number
-  response_time?: number
-  created_at: string
-  has_fallback?: boolean // 🆕 是否发生了 fallback
-}
-
-export interface UsageStats {
-  total_requests: number
-  total_tokens: number
-  total_cost: number
-  total_actual_cost?: number
-  avg_response_time: number
-  today?: {
-    requests: number
-    tokens: number
-    cost: number
-  }
-  activity_heatmap?: ActivityHeatmap | null
-}
-
-export interface UsageByModel {
-  model: string
-  request_count: number
-  total_tokens: number
-  total_cost: number
-  avg_response_time?: number
-  cache_read_tokens?: number
-  cache_hit_rate?: number
-}
-
-export interface UsageByUser {
-  user_id: string // UUID
-  email: string
-  username: string
-  request_count: number
-  total_tokens: number
-  total_cost: number
-}
-
-export interface UsageByProvider {
-  provider_id: string
-  provider: string
-  request_count: number
-  total_tokens: number
-  total_cost: number
-  actual_cost: number
-  avg_response_time_ms: number
-  success_rate: number
-  error_count: number
-  cache_read_tokens?: number
-  cache_hit_rate?: number
-}
-
-export interface UsageByApiFormat {
-  api_format: string
-  request_count: number
-  total_tokens: number
-  total_cost: number
-  actual_cost: number
-  avg_response_time_ms: number
-  cache_read_tokens?: number
-  cache_hit_rate?: number
-}
-
-export interface UsageFilters {
-  user_id?: string // UUID
-  provider_id?: string // UUID
-  model?: string
-  start_date?: string
-  end_date?: string
-  preset?: string
-  granularity?: 'hour' | 'day' | 'week' | 'month'
-  timezone?: string
-  tz_offset_minutes?: number
-  page?: number
-  page_size?: number
-}
-
-export const usageApi = {
-  async getUsageRecords(filters?: UsageFilters): Promise<{
-    records: UsageRecord[]
-    total: number
-    page: number
-    page_size: number
-  }> {
-    const response = await apiClient.get('/api/usage', { params: filters })
-    return response.data
-  },
-
-  async getUsageStats(filters?: UsageFilters): Promise<UsageStats> {
-    // 为统计数据添加30秒缓存
-    const cacheKey = `usage-stats-${JSON.stringify(filters || {})}`
-    return cachedRequest(
-      cacheKey,
-      async () => {
-        const response = await apiClient.get<UsageStats>('/api/admin/usage/stats', { params: filters })
-        return response.data
-      },
-      30000 // 30秒缓存
-    )
-  },
-
-  /**
-   * Get usage aggregation by dimension (RESTful API)
-   * @param groupBy Aggregation dimension: 'model', 'user', 'provider', or 'api_format'
-   * @param filters Optional filters
-   */
-  async getUsageAggregation<T = UsageByModel[] | UsageByUser[] | UsageByProvider[] | UsageByApiFormat[]>(
-    groupBy: 'model' | 'user' | 'provider' | 'api_format',
-    filters?: UsageFilters & { limit?: number }
-  ): Promise<T> {
-    const cacheKey = `usage-aggregation-${groupBy}-${JSON.stringify(filters || {})}`
-    return cachedRequest(
-      cacheKey,
-      async () => {
-        const response = await apiClient.get<T>('/api/admin/usage/aggregation/stats', {
-          params: { group_by: groupBy, ...filters }
-        })
-        return response.data
-      },
-      30000 // 30秒缓存
-    )
-  },
-
-  // Shorthand methods using getUsageAggregation
-  async getUsageByModel(filters?: UsageFilters & { limit?: number }): Promise<UsageByModel[]> {
-    return this.getUsageAggregation<UsageByModel[]>('model', filters)
-  },
-
-  async getUsageByUser(filters?: UsageFilters & { limit?: number }): Promise<UsageByUser[]> {
-    return this.getUsageAggregation<UsageByUser[]>('user', filters)
-  },
-
-  async getUsageByProvider(filters?: UsageFilters & { limit?: number }): Promise<UsageByProvider[]> {
-    return this.getUsageAggregation<UsageByProvider[]>('provider', filters)
-  },
-
-  async getUsageByApiFormat(filters?: UsageFilters & { limit?: number }): Promise<UsageByApiFormat[]> {
-    return this.getUsageAggregation<UsageByApiFormat[]>('api_format', filters)
-  },
-
-  async getUserUsage(userId: string, filters?: UsageFilters): Promise<{
-    records: UsageRecord[]
-    stats: UsageStats
-  }> {
-    const response = await apiClient.get(`/api/users/${userId}/usage`, { params: filters })
-    return response.data
-  },
-
-  async exportUsage(format: 'csv' | 'json', filters?: UsageFilters): Promise<Blob> {
-    const response = await apiClient.get('/api/usage/export', {
-      params: { ...filters, format },
-      responseType: 'blob'
-    })
-    return response.data
-  },
-
-  async getAllUsageRecords(params?: {
-    start_date?: string
-    end_date?: string
-    preset?: string
-    granularity?: 'hour' | 'day' | 'week' | 'month'
-    timezone?: string
-    tz_offset_minutes?: number
-    search?: string  // 通用搜索：用户名、密钥名、模型名、提供商名
-    user_id?: string // UUID
-    username?: string
-    model?: string
-    provider?: string
-    api_format?: string  // API 格式筛选（如 openai:chat, claude:chat）
-    status?: string // 'stream' | 'standard' | 'error'
-    limit?: number
-    offset?: number
-  }): Promise<{
-    records: Array<Record<string, unknown>>
-    total: number
-    limit: number
-    offset: number
-  }> {
-    const key = buildCacheKey('usage:records', params as Record<string, unknown> | undefined)
-    return dedupedRequest(key, async () => {
-      const response = await apiClient.get('/api/admin/usage/records', { params })
-      return response.data
-    })
-  },
-
-  /**
-   * 获取活跃请求的状态（轻量级接口，用于轮询更新）
-   * @param ids 可选，逗号分隔的请求 ID 列表
-   */
-  async getActiveRequests(ids?: string[]): Promise<{
-    requests: Array<{
-      id: string
-      status: 'pending' | 'streaming' | 'completed' | 'failed' | 'cancelled'
-      input_tokens: number
-      output_tokens: number
-      cache_creation_input_tokens?: number | null
-      cache_read_input_tokens?: number | null
-      cost: number
-      actual_cost?: number | null
-      rate_multiplier?: number | null
-      response_time_ms: number | null
-      first_byte_time_ms: number | null
-      provider?: string | null
-      api_key_name?: string | null
-      api_format?: string | null
-      endpoint_api_format?: string | null
-      has_format_conversion?: boolean | null
-      target_model?: string | null
-    }>
-  }> {
-    const params = ids?.length ? { ids: ids.join(',') } : {}
-    const response = await apiClient.get('/api/admin/usage/active', { params })
-    return response.data
-  },
-
-  /**
-   * 获取活跃度热力图数据（管理员）
-   * 后端已缓存5分钟
-   */
-  async getActivityHeatmap(): Promise<ActivityHeatmap> {
-    return cachedRequest(
-      'admin-usage-activity-heatmap',
-      async () => {
-        const response = await apiClient.get<ActivityHeatmap>('/api/admin/usage/heatmap')
-        return response.data
-      },
-      60000
-    )
-  }
-}

frontend/src/api/users.ts

@@ -141,10 +141,5 @@ export const usersApi = {
       `/api/admin/users/${userId}/api-keys/${keyId}/full-key`
     )
     return response.data
-  },
-  // 管理员统计
-  async getUsageStats(): Promise<Record<string, unknown>> {
-    const response = await apiClient.get('/api/admin/usage/stats')
-    return response.data
   }
 }

frontend/src/components/charts/BarChart.vue

@@ -1,5 +1,8 @@
 <template>
-  <div class="w-full h-full">
+  <div
+    class="w-full"
+    :style="{ height: `${props.height}px` }"
+  >
     <canvas ref="chartRef" />
   </div>
 </template>
@@ -18,6 +21,7 @@ import {
   type ChartData,
   type ChartOptions
 } from 'chart.js'
+import { observeChartThemeChanges, resolveChartTheme } from '@/utils/chartTheme'
 
 const props = withDefaults(defineProps<Props>(), {
   height: 300,
@@ -44,6 +48,7 @@ interface Props {
 
 const chartRef = ref<HTMLCanvasElement>()
 let chart: ChartJS<'bar'> | null = null
+let stopThemeObserver: (() => void) | null = null
 
 const defaultOptions: ChartOptions<'bar'> = {
   responsive: true,
@@ -91,9 +96,7 @@ const defaultOptions: ChartOptions<'bar'> = {
   }
 }
 
-function createChart() {
-  if (!chartRef.value) return
-
+function buildChartOptions(): ChartOptions<'bar'> {
   const stackedOptions = props.stacked ? {
     scales: {
       x: { ...defaultOptions.scales?.x, stacked: true },
@@ -106,20 +109,26 @@ function createChart() {
     }
   }
 
-  chart = new ChartJS(chartRef.value, {
-    type: 'bar',
-    data: props.data,
-    options: {
+  return resolveChartTheme({
     ...defaultOptions,
     ...stackedOptions,
     ...props.options
+  })
 }
+
+function createChart() {
+  if (!chartRef.value) return
+
+  chart = new ChartJS(chartRef.value, {
+    type: 'bar',
+    data: resolveChartTheme(props.data),
+    options: buildChartOptions()
   })
 }
 
 function updateChart() {
   if (chart) {
-    chart.data = props.data
+    chart.data = resolveChartTheme(props.data)
     chart.update('none')
   }
 }
@@ -127,9 +136,17 @@ function updateChart() {
 onMounted(async () => {
   await nextTick()
   createChart()
+  stopThemeObserver = observeChartThemeChanges(() => {
+    if (!chart) return
+    chart.data = resolveChartTheme(props.data)
+    chart.options = buildChartOptions()
+    chart.update('none')
+  })
 })
 
 onUnmounted(() => {
+  stopThemeObserver?.()
+  stopThemeObserver = null
   if (chart) {
     chart.destroy()
     chart = null
@@ -139,11 +156,8 @@ onUnmounted(() => {
 watch(() => props.data, updateChart, { deep: true })
 watch(() => props.options, () => {
   if (chart) {
-    chart.options = {
-      ...defaultOptions,
-      ...props.options
-    }
-    chart.update()
+    chart.options = buildChartOptions()
+    chart.update('none')
   }
 }, { deep: true })
 </script>

frontend/src/components/charts/DoughnutChart.vue

@@ -1,5 +1,8 @@
 <template>
-  <div class="w-full h-full">
+  <div
+    class="w-full"
+    :style="{ height: `${props.height}px` }"
+  >
     <canvas ref="chartRef" />
   </div>
 </template>
@@ -16,10 +19,12 @@ import {
   type ChartData,
   type ChartOptions
 } from 'chart.js'
+import { observeChartThemeChanges, resolveChartTheme } from '@/utils/chartTheme'
 
 const props = withDefaults(defineProps<Props>(), {
   height: 300,
-  options: undefined
+  options: undefined,
+  showLegend: true,
 })
 
 ChartJS.register(
@@ -34,15 +39,22 @@ interface Props {
   data: ChartData<'doughnut'>
   options?: ChartOptions<'doughnut'>
   height?: number
+  showLegend?: boolean
 }
 
 const chartRef = ref<HTMLCanvasElement>()
 let chart: ChartJS<'doughnut'> | null = null
+let stopThemeObserver: (() => void) | null = null
 
 const defaultOptions: ChartOptions<'doughnut'> = {
   responsive: true,
   maintainAspectRatio: false,
   cutout: '60%',
+  events: ['mousemove', 'mouseout', 'click', 'touchstart', 'touchmove'],
+  interaction: {
+    mode: 'nearest',
+    intersect: true,
+  },
   plugins: {
     legend: {
       position: 'right',
@@ -71,22 +83,59 @@ const defaultOptions: ChartOptions<'doughnut'> = {
   }
 }
 
+function resolvedOptions(): ChartOptions<'doughnut'> {
+  const incomingOptions = props.options ?? {}
+  const incomingPlugins = (incomingOptions.plugins ?? {}) as Record<string, unknown>
+  const incomingLegend = (incomingPlugins.legend ?? {}) as Record<string, unknown>
+  const incomingTooltip = (incomingPlugins.tooltip ?? {}) as Record<string, unknown>
+  const defaultPlugins = (defaultOptions.plugins ?? {}) as Record<string, unknown>
+  const defaultLegend = (defaultPlugins.legend ?? {}) as Record<string, unknown>
+  const defaultTooltip = (defaultPlugins.tooltip ?? {}) as Record<string, unknown>
+
+  return resolveChartTheme({
+    ...defaultOptions,
+    ...incomingOptions,
+    interaction: {
+      ...defaultOptions.interaction,
+      ...incomingOptions.interaction,
+    },
+    plugins: {
+      ...defaultPlugins,
+      ...incomingPlugins,
+      legend: {
+        ...defaultLegend,
+        ...incomingLegend,
+        display: incomingLegend.display ?? props.showLegend,
+        labels: {
+          ...(defaultLegend.labels as Record<string, unknown> | undefined),
+          ...(incomingLegend.labels as Record<string, unknown> | undefined),
+        },
+      },
+      tooltip: {
+        ...defaultTooltip,
+        ...incomingTooltip,
+        callbacks: {
+          ...(defaultTooltip.callbacks as Record<string, unknown> | undefined),
+          ...(incomingTooltip.callbacks as Record<string, unknown> | undefined),
+        },
+      },
+    },
+  })
+}
+
 function createChart() {
   if (!chartRef.value) return
 
   chart = new ChartJS(chartRef.value, {
     type: 'doughnut',
-    data: props.data,
-    options: {
-      ...defaultOptions,
-      ...props.options
-    }
+    data: resolveChartTheme(props.data),
+    options: resolvedOptions(),
   })
 }
 
 function updateChart() {
   if (chart) {
-    chart.data = props.data
+    chart.data = resolveChartTheme(props.data)
     chart.update('none')
   }
 }
@@ -94,9 +143,17 @@ function updateChart() {
 onMounted(async () => {
   await nextTick()
   createChart()
+  stopThemeObserver = observeChartThemeChanges(() => {
+    if (!chart) return
+    chart.data = resolveChartTheme(props.data)
+    chart.options = resolvedOptions()
+    chart.update('none')
+  })
 })
 
 onUnmounted(() => {
+  stopThemeObserver?.()
+  stopThemeObserver = null
   if (chart) {
     chart.destroy()
     chart = null
@@ -106,8 +163,15 @@ onUnmounted(() => {
 watch(() => props.data, updateChart, { deep: true })
 watch(() => props.options, () => {
   if (chart) {
-    chart.options = { ...defaultOptions, ...props.options }
-    chart.update()
+    chart.options = resolvedOptions()
+    chart.update('none')
   }
 }, { deep: true })
+
+watch(() => props.showLegend, () => {
+  if (chart) {
+    chart.options = resolvedOptions()
+    chart.update('none')
+  }
+})
 </script>

frontend/src/components/charts/LineChart.vue

@@ -1,5 +1,8 @@
 <template>
-  <div class="w-full h-full">
+  <div
+    class="w-full"
+    :style="{ height: `${props.height}px` }"
+  >
     <canvas ref="chartRef" />
   </div>
 </template>
@@ -16,9 +19,11 @@ import {
   Title,
   Tooltip,
   Legend,
+  Filler,
   type ChartData,
   type ChartOptions
 } from 'chart.js'
+import { observeChartThemeChanges, resolveChartTheme } from '@/utils/chartTheme'
 
 const props = withDefaults(defineProps<Props>(), {
   height: 300,
@@ -34,7 +39,8 @@ ChartJS.register(
   LineController,
   Title,
   Tooltip,
-  Legend
+  Legend,
+  Filler
 )
 
 interface Props {
@@ -45,12 +51,17 @@ interface Props {
 
 const chartRef = ref<HTMLCanvasElement>()
 let chart: ChartJS<'line'> | null = null
+let stopThemeObserver: (() => void) | null = null
 
 function buildChartOptions(): ChartOptions<'line'> {
-  return {
+  return resolveChartTheme({
     ...defaultOptions,
     ...props.options
+  })
 }
+
+function buildChartData(): ChartData<'line'> {
+  return resolveChartTheme(props.data)
 }
 
 const defaultOptions: ChartOptions<'line'> = {
@@ -95,14 +106,14 @@ function createChart() {
 
   chart = new ChartJS(chartRef.value, {
     type: 'line',
-    data: props.data,
+    data: buildChartData(),
     options: buildChartOptions()
   })
 }
 
 function updateChart() {
   if (chart) {
-    chart.data = props.data
+    chart.data = buildChartData()
     chart.update('none') // 禁用动画以提高性能
   }
 }
@@ -110,9 +121,17 @@ function updateChart() {
 onMounted(async () => {
   await nextTick()
   createChart()
+  stopThemeObserver = observeChartThemeChanges(() => {
+    if (!chart) return
+    chart.data = buildChartData()
+    chart.options = buildChartOptions()
+    chart.update('none')
+  })
 })
 
 onUnmounted(() => {
+  stopThemeObserver?.()
+  stopThemeObserver = null
   if (chart) {
     chart.destroy()
     chart = null

frontend/src/components/common/TimeRangePicker.vue

@@ -3,39 +3,24 @@
     <Select
       v-model="selectedPreset"
     >
-      <SelectTrigger class="h-8 w-32 text-xs border-border/60">
+      <SelectTrigger :class="presetTriggerClass">
         <SelectValue placeholder="选择时间段" />
       </SelectTrigger>
       <SelectContent>
         <SelectItem value="today">
           今天
         </SelectItem>
-        <SelectItem value="yesterday">
-          昨天
-        </SelectItem>
         <SelectItem value="last7days">
           最近7天
         </SelectItem>
         <SelectItem value="last30days">
           最近30天
         </SelectItem>
-        <SelectItem value="last90days">
-          最近90天
+        <SelectItem value="last180days">
+          最近180天
         </SelectItem>
-        <SelectItem value="this_week">
-          本周
-        </SelectItem>
-        <SelectItem value="last_week">
-          上周
-        </SelectItem>
-        <SelectItem value="this_month">
-          本月
-        </SelectItem>
-        <SelectItem value="last_month">
-          上月
-        </SelectItem>
-        <SelectItem value="this_year">
-          今年
+        <SelectItem value="last1year">
+          最近一年
         </SelectItem>
         <SelectItem value="custom">
           自定义
@@ -50,13 +35,13 @@
       <Input
         v-model="startDate"
         type="date"
-        class="h-8 w-36 text-xs border-border/60"
+        :class="dateInputClass"
       />
       <span class="text-xs text-muted-foreground">至</span>
       <Input
         v-model="endDate"
         type="date"
-        class="h-8 w-36 text-xs border-border/60"
+        :class="dateInputClass"
       />
     </div>
 
@@ -64,10 +49,16 @@
       v-if="showGranularity"
       v-model="selectedGranularity"
     >
-      <SelectTrigger class="h-8 w-24 text-xs border-border/60">
+      <SelectTrigger :class="granularityTriggerClass">
         <SelectValue placeholder="粒度" />
       </SelectTrigger>
       <SelectContent>
+        <SelectItem
+          v-if="includeAutoGranularity"
+          value="auto"
+        >
+          自动
+        </SelectItem>
         <SelectItem
           v-if="allowHourly && canUseHourly"
           value="hour"
@@ -100,26 +91,51 @@ import {
 } from '@/components/ui'
 import type { DateRangeParams } from '@/features/usage/types'
 
+type PickerGranularity = NonNullable<DateRangeParams['granularity']>
+
 const props = defineProps<{
   modelValue: DateRangeParams
   showGranularity?: boolean
   allowHourly?: boolean
+  includeAutoGranularity?: boolean
+  compact?: boolean
 }>()
 
 const emit = defineEmits<{
   'update:modelValue': [value: DateRangeParams]
 }>()
 
-const selectedPreset = ref(props.modelValue.preset || 'last7days')
+function getDefaultGranularity(): PickerGranularity {
+  return props.includeAutoGranularity ? 'auto' : 'day'
+}
+
+const selectedPreset = ref(props.modelValue.preset || 'today')
 const startDate = ref(props.modelValue.start_date || '')
 const endDate = ref(props.modelValue.end_date || '')
-const selectedGranularity = ref(props.modelValue.granularity || 'day')
+const selectedGranularity = ref<PickerGranularity>(props.modelValue.granularity || getDefaultGranularity())
 
 const showGranularity = computed(() => props.showGranularity !== false)
 const allowHourly = computed(() => props.allowHourly === true)
+const includeAutoGranularity = computed(() => props.includeAutoGranularity === true)
+const compact = computed(() => props.compact === true)
+const presetTriggerClass = computed(() =>
+  compact.value
+    ? 'h-8 w-28 text-xs border-border/60'
+    : 'h-8 w-32 text-xs border-border/60',
+)
+const dateInputClass = computed(() =>
+  compact.value
+    ? 'h-8 w-32 text-xs border-border/60'
+    : 'h-8 w-36 text-xs border-border/60',
+)
+const granularityTriggerClass = computed(() =>
+  compact.value
+    ? 'h-8 w-20 text-xs border-border/60'
+    : 'h-8 w-24 text-xs border-border/60',
+)
 
 const canUseHourly = computed(() => {
-  if (selectedPreset.value === 'today' || selectedPreset.value === 'yesterday') return true
+  if (selectedPreset.value === 'today') return true
   if (selectedPreset.value === 'custom' && startDate.value && endDate.value) {
     return startDate.value === endDate.value
   }
@@ -165,7 +181,7 @@ watch(() => props.modelValue, (value) => {
   if (value.preset) selectedPreset.value = value.preset
   if (value.start_date !== undefined) startDate.value = value.start_date || ''
   if (value.end_date !== undefined) endDate.value = value.end_date || ''
-  if (value.granularity) selectedGranularity.value = value.granularity
+  selectedGranularity.value = value.granularity || getDefaultGranularity()
   // 同步更新 lastEmittedValue，避免外部设置值后触发重复 emit
   lastEmittedValue = getValueKey(value)
 }, { deep: true })
@@ -173,7 +189,7 @@ watch(() => props.modelValue, (value) => {
 watch([selectedPreset, startDate, endDate, selectedGranularity], () => {
   if (!allowHourly.value || !canUseHourly.value) {
     if (selectedGranularity.value === 'hour') {
-      selectedGranularity.value = 'day'
+      selectedGranularity.value = includeAutoGranularity.value ? 'auto' : 'day'
     }
   }
 

frontend/src/components/stats/ActivityHeatmap.vue

@@ -13,8 +13,11 @@
         <p class="mt-0.5">
           {{ tooltip.day.requests }} 次请求 · {{ formatTokens(tooltip.day.total_tokens) }}
         </p>
-        <p class="text-[11px] text-muted-foreground">
-          成本 {{ formatCurrency(tooltip.day.total_cost) }}
+        <p class="text-[11px] text-muted-foreground whitespace-nowrap">
+          {{ valueLabel }} {{ formatCurrency(tooltip.day.total_cost) }}
+          <template v-if="actualValueLabel && tooltip.day.actual_total_cost !== undefined">
+            · {{ actualValueLabel }} {{ formatCurrency(tooltip.day.actual_total_cost) }}
+          </template>
         </p>
       </div>
     </Teleport>
@@ -90,7 +93,7 @@
           class="flex items-center invisible"
         >周六</span>
       </div>
-      <div class="flex-1 min-w-[200px]">
+      <div class="flex-1 min-w-0">
         <div
           ref="heatmapWrapper"
           class="relative block w-full"
@@ -100,12 +103,12 @@
             :style="horizontalGapStyle"
           >
             <div
-              v-for="(week, weekIndex) in weekColumns"
+              v-for="(week, weekIndex) in visibleWeekColumns"
               :key="`month-${weekIndex}`"
               :style="monthCellStyle"
               class="text-center"
             >
-              <span v-if="monthMarkers[weekIndex]">{{ monthMarkers[weekIndex] }}</span>
+              <span v-if="visibleMonthMarkers[weekIndex]">{{ visibleMonthMarkers[weekIndex] }}</span>
             </div>
           </div>
           <div
@@ -113,7 +116,7 @@
             :style="horizontalGapStyle"
           >
             <div
-              v-for="(week, weekIndex) in weekColumns"
+              v-for="(week, weekIndex) in visibleWeekColumns"
               :key="weekIndex"
               class="flex flex-col"
               :style="verticalGapStyle"
@@ -161,11 +164,15 @@ const props = withDefaults(defineProps<{
   title?: string
   subtitle?: string
   showHeader?: boolean
+  valueLabel?: string
+  actualValueLabel?: string
 }>(), {
   data: undefined,
   title: undefined,
   subtitle: undefined,
-  showHeader: true
+  showHeader: true,
+  valueLabel: '费用',
+  actualValueLabel: undefined,
 })
 
 const legendLevels = [0.08, 0.25, 0.45, 0.65, 0.85]
@@ -270,7 +277,47 @@ const monthMarkers = computed(() => {
     if (month === lastMonth) {
       return
     }
-    markers[index] = `${month + 1}月`
+    markers[index] = String(month + 1)
+    lastMonth = month
+  })
+
+  return markers
+})
+
+const MIN_CELL_SIZE = 6
+
+// 根据容器宽度计算最多能放多少周，然后从后面截取（保留最新数据）
+const maxVisibleWeeks = computed(() => {
+  const total = weekColumns.value.length
+  if (!heatmapWidth.value || !total) return total
+  // 先尝试用理想尺寸放全部
+  const totalGapAll = Math.max(total - 1, 0) * cellGap.value
+  const rawSizeAll = (heatmapWidth.value - totalGapAll) / total
+  if (rawSizeAll >= MIN_CELL_SIZE) return total
+  // 放不下，算容器最多能容纳几周（用最小尺寸）
+  // n * MIN_CELL_SIZE + (n-1) * gap <= width
+  const maxN = Math.floor((heatmapWidth.value + cellGap.value) / (MIN_CELL_SIZE + cellGap.value))
+  return Math.max(1, Math.min(maxN, total))
+})
+
+const visibleWeekColumns = computed(() => {
+  const all = weekColumns.value
+  const n = maxVisibleWeeks.value
+  if (n >= all.length) return all
+  return all.slice(all.length - n)
+})
+
+const visibleMonthMarkers = computed(() => {
+  const markers: Record<number, string> = {}
+  const columns = visibleWeekColumns.value
+  let lastMonth: number | null = null
+
+  columns.forEach((week, index) => {
+    const firstValid = week.find((day): day is DayWithMeta => day !== null)
+    if (!firstValid) return
+    const month = firstValid.dateObj.getUTCMonth()
+    if (month === lastMonth) return
+    markers[index] = String(month + 1)
     lastMonth = month
   })
 
@@ -282,7 +329,7 @@ let mediaQuery: MediaQueryList | null = null
 let mediaQueryHandler: ((event?: MediaQueryListEvent) => void) | null = null
 
 const recalcCellSize = () => {
-  const columnCount = weekColumns.value.length
+  const columnCount = visibleWeekColumns.value.length
   if (!columnCount || !heatmapWidth.value) {
     return
   }
@@ -290,12 +337,12 @@ const recalcCellSize = () => {
   const totalGap = Math.max(columnCount - 1, 0) * cellGap.value
   const availableSpace = Math.max(heatmapWidth.value - totalGap, 0)
   const rawSize = availableSpace / columnCount
-  // 自适应尺寸，最小 6px
-  cellSize.value = Math.max(6, rawSize)
+  // 当所有列都能放下时，让格子继续随容器放大，避免宽屏下出现大块留白。
+  cellSize.value = Math.max(MIN_CELL_SIZE, rawSize)
 }
 
 watch(
-  [() => heatmapWidth.value, () => weekColumns.value.length, () => cellGap.value],
+  [() => heatmapWidth.value, () => visibleWeekColumns.value.length, () => cellGap.value],
   () => {
     recalcCellSize()
   },
@@ -345,7 +392,7 @@ onBeforeUnmount(() => {
 
 function handleHover(day: ActivityHeatmapDay, event: MouseEvent) {
   const cellRect = (event.currentTarget as HTMLElement).getBoundingClientRect()
-  const tooltipWidth = 200
+  const tooltipWidth = 280
   const tooltipHeight = 72
 
   // Calculate horizontal position (centered on cell)
@@ -402,10 +449,10 @@ function getCellStyle(requests: number) {
 
 function buildTooltip(day: ActivityHeatmapDay): string {
   const dateLabel = day.date
-  const costLabel = formatCurrency(day.total_cost || 0)
+  const costLabel = `${props.valueLabel} ${formatCurrency(day.total_cost || 0)}`
   const parts = [`${dateLabel}`, `${day.requests} 次请求`, `${formatTokens(day.total_tokens)} tokens`, costLabel]
-  if (day.actual_total_cost !== undefined) {
-    parts.push(`倍率: ${formatCurrency(day.actual_total_cost)}`)
+  if (props.actualValueLabel && day.actual_total_cost !== undefined) {
+    parts.push(`${props.actualValueLabel} ${formatCurrency(day.actual_total_cost)}`)
   }
   return parts.join(' · ')
 }

frontend/src/components/stats/ComparisonCard.vue

@@ -1,38 +0,0 @@
-<template>
-  <Card class="p-4 space-y-2">
-    <div class="text-xs text-muted-foreground">
-      {{ label }}
-    </div>
-    <div class="text-lg font-semibold">
-      {{ value }}
-    </div>
-    <div
-      class="text-xs"
-      :class="changeClass"
-    >
-      <span v-if="changePercent !== null">{{ changePercent }}%</span>
-      <span v-else>--</span>
-      <span class="ml-1 text-muted-foreground">vs 对比期</span>
-    </div>
-  </Card>
-</template>
-
-<script setup lang="ts">
-import { computed } from 'vue'
-import { Card } from '@/components/ui'
-
-interface Props {
-  label: string
-  value: string
-  changePercent: number | null
-}
-
-const props = defineProps<Props>()
-
-const changeClass = computed(() => {
-  if (props.changePercent === null) return 'text-muted-foreground'
-  if (props.changePercent > 0) return 'text-emerald-500'
-  if (props.changePercent < 0) return 'text-rose-500'
-  return 'text-muted-foreground'
-})
-</script>

frontend/src/components/stats/CostForecastChart.vue

@@ -1,93 +0,0 @@
-<template>
-  <div class="space-y-3">
-    <div class="flex items-center justify-between">
-      <h3 class="text-sm font-semibold">
-        {{ title }}
-      </h3>
-      <span
-        v-if="subtitle"
-        class="text-xs text-muted-foreground"
-      >{{ subtitle }}</span>
-    </div>
-    <div
-      v-if="loading"
-      class="p-6"
-    >
-      <LoadingState />
-    </div>
-    <div
-      v-else
-      class="h-[280px]"
-    >
-      <LineChart
-        :data="chartData"
-        :options="chartOptions"
-      />
-    </div>
-  </div>
-</template>
-
-<script setup lang="ts">
-import { computed } from 'vue'
-import LineChart from '@/components/charts/LineChart.vue'
-import { LoadingState } from '@/components/common'
-import { formatCurrency } from '@/utils/format'
-
-interface Props {
-  title: string
-  subtitle?: string
-  history: Array<{ date: string; total_cost: number }>
-  forecast: Array<{ date: string; total_cost: number }>
-  loading?: boolean
-}
-
-const props = withDefaults(defineProps<Props>(), {
-  subtitle: undefined,
-  loading: false
-})
-
-const labels = computed(() => [
-  ...props.history.map(item => item.date),
-  ...props.forecast.map(item => item.date)
-])
-
-const chartData = computed(() => {
-  const historyValues = props.history.map(item => item.total_cost)
-  const forecastValues = props.forecast.map(item => item.total_cost)
-  return {
-    labels: labels.value,
-    datasets: [
-      {
-        label: '实际成本',
-        data: historyValues.concat(new Array(forecastValues.length).fill(null)),
-        borderColor: 'rgb(59, 130, 246)',
-        backgroundColor: 'rgba(59, 130, 246, 0.15)',
-        tension: 0.25,
-        pointRadius: 2
-      },
-      {
-        label: '预测成本',
-        data: new Array(historyValues.length).fill(null).concat(forecastValues),
-        borderColor: 'rgb(234, 179, 8)',
-        backgroundColor: 'rgba(234, 179, 8, 0.15)',
-        borderDash: [6, 4],
-        tension: 0.25,
-        pointRadius: 2
-      }
-    ]
-  }
-})
-
-const chartOptions = computed(() => ({
-  plugins: {
-    tooltip: {
-      callbacks: {
-        label: (context: { parsed?: { y?: number }; dataset: { label?: string } }) => {
-          const value = context.parsed?.y ?? 0
-          return `${context.dataset.label}: ${formatCurrency(value)}`
-        }
-      }
-    }
-  }
-}))
-</script>

frontend/src/components/stats/ErrorDistributionChart.vue

@@ -1,14 +1,34 @@
 <template>
-  <div class="space-y-3">
-    <div class="flex items-center justify-between">
+  <div class="space-y-4">
+    <div class="flex flex-col gap-3 sm:flex-row sm:items-start sm:justify-between">
+      <div>
         <h3 class="text-sm font-semibold">
           {{ title }}
         </h3>
-      <span
+        <p
           v-if="subtitle"
-        class="text-xs text-muted-foreground"
-      >{{ subtitle }}</span>
+          class="mt-1 text-[11px] text-muted-foreground"
+        >
+          {{ subtitle }}
+        </p>
       </div>
+
+      <div
+        v-if="!unavailable"
+        class="flex flex-wrap gap-2"
+      >
+        <div class="rounded-full border border-rose-500/20 bg-rose-500/8 px-2.5 py-1 text-[10px] font-medium text-rose-700 dark:text-rose-300">
+          总错误 {{ totalErrors }}
+        </div>
+        <div
+          v-if="topCategory"
+          class="rounded-full border border-border/70 bg-muted/20 px-2.5 py-1 text-[10px] font-medium text-muted-foreground"
+        >
+          TOP {{ topCategory.label }} · {{ topCategory.count }}
+        </div>
+      </div>
+    </div>
+
     <div
       v-if="loading"
       class="p-6"
@@ -16,57 +36,116 @@
       <LoadingState />
     </div>
     <div
-      v-else
-      class="h-[260px]"
+      v-else-if="unavailable"
+      class="flex h-[272px] items-center justify-center rounded-2xl border border-dashed border-border/70 bg-muted/10 text-xs text-muted-foreground"
     >
+      {{ unavailableText }}
+    </div>
+    <div
+      v-else-if="!hasData"
+      class="flex h-[272px] items-center justify-center rounded-2xl border border-dashed border-border/70 bg-muted/10 text-xs text-muted-foreground"
+    >
+      暂无错误分布数据
+    </div>
+    <div
+      v-else
+      class="rounded-2xl border border-border/60 bg-background p-3"
+    >
+      <div class="mx-auto w-full max-w-[320px]">
         <DoughnutChart
           :data="chartData"
           :options="chartOptions"
+          :show-legend="true"
+          :height="248"
         />
       </div>
     </div>
+  </div>
 </template>
 
 <script setup lang="ts">
 import { computed } from 'vue'
+import type { ChartData, ChartOptions } from 'chart.js'
 import DoughnutChart from '@/components/charts/DoughnutChart.vue'
 import { LoadingState } from '@/components/common'
-import type { ErrorDistributionItem } from '@/api/admin'
+
+interface ErrorDistributionItem {
+  category: string
+  label: string
+  count: number
+}
 
 interface Props {
   title: string
   subtitle?: string
   distribution: ErrorDistributionItem[]
   loading?: boolean
+  unavailable?: boolean
+  unavailableText?: string
 }
 
 const props = withDefaults(defineProps<Props>(), {
   subtitle: undefined,
-  loading: false
+  loading: false,
+  unavailable: false,
+  unavailableText: '数据暂不可用',
 })
 
-const chartData = computed(() => ({
-  labels: props.distribution.map(item => item.category),
+const chartColors = [
+  'rgba(226, 85, 118, 0.88)',
+  'rgba(84, 135, 237, 0.86)',
+  'rgba(225, 165, 52, 0.84)',
+  'rgba(68, 187, 132, 0.82)',
+  'rgba(148, 163, 184, 0.8)',
+  'rgba(145, 118, 214, 0.82)',
+]
+
+const totalErrors = computed(() => props.distribution.reduce((sum, item) => sum + item.count, 0))
+const hasData = computed(() => totalErrors.value > 0)
+const distributionWithShare = computed(() => props.distribution.map(item => ({
+  ...item,
+  label: item.label?.trim() || item.category,
+  share: totalErrors.value > 0 ? `${((item.count / totalErrors.value) * 100).toFixed(1)}%` : '0.0%',
+})))
+const topCategory = computed(() => (
+  distributionWithShare.value.slice().sort((left, right) => right.count - left.count)[0] ?? null
+))
+
+const chartData = computed<ChartData<'doughnut'>>(() => ({
+  labels: distributionWithShare.value.map(item => item.label),
   datasets: [
     {
-      data: props.distribution.map(item => item.count),
-      backgroundColor: [
-        'rgba(239, 68, 68, 0.7)',
-        'rgba(59, 130, 246, 0.7)',
-        'rgba(234, 179, 8, 0.7)',
-        'rgba(34, 197, 94, 0.7)',
-        'rgba(148, 163, 184, 0.7)'
+      data: distributionWithShare.value.map(item => item.count),
+      backgroundColor: distributionWithShare.value.map((_, index) => chartColors[index % chartColors.length]),
+      borderColor: 'rgba(255,255,255,0.92)',
+      borderWidth: 2,
+      hoverOffset: 6,
+    },
   ],
-      borderWidth: 0
-    }
-  ]
 }))
 
-const chartOptions = computed(() => ({
+const chartOptions = computed<ChartOptions<'doughnut'>>(() => ({
+  cutout: '68%',
   plugins: {
     legend: {
-      position: 'bottom' as const
-    }
-  }
+      position: 'bottom',
+      align: 'center',
+      labels: {
+        usePointStyle: true,
+        pointStyle: 'circle',
+        boxWidth: 10,
+        padding: 14,
+      },
+    },
+    tooltip: {
+      callbacks: {
+        label: (context) => {
+          const value = Number(context.raw ?? 0)
+          const percentage = totalErrors.value > 0 ? ((value / totalErrors.value) * 100).toFixed(1) : '0.0'
+          return `${context.label ?? '错误'}: ${value} (${percentage}%)`
+        },
+      },
+    },
+  },
 }))
 </script>

frontend/src/components/stats/LeaderboardTable.vue

@@ -1,34 +1,30 @@
 <template>
   <TableCard :title="title">
     <template #actions>
-      <Select
-        v-if="showMetricSelect"
-        :model-value="metric"
-        @update:model-value="emitMetric"
+      <div
+        v-if="isRefreshing"
+        class="inline-flex h-7 items-center rounded-lg border border-border/70 bg-background px-2.5 text-[10px] text-muted-foreground"
       >
-        <SelectTrigger class="h-8 text-xs w-28">
-          <SelectValue placeholder="指标" />
-        </SelectTrigger>
-        <SelectContent>
-          <SelectItem value="requests">
-            请求数
-          </SelectItem>
-          <SelectItem value="tokens">
-            Tokens
-          </SelectItem>
-          <SelectItem value="cost">
-            成本
-          </SelectItem>
-        </SelectContent>
-      </Select>
+        更新中
+      </div>
     </template>
 
     <div
-      v-if="loading"
+      v-if="showLoadingState"
       class="p-6"
     >
       <LoadingState />
     </div>
+    <div
+      v-else-if="showUnavailableState"
+      class="p-6"
+    >
+      <EmptyState
+        type="error"
+        :title="unavailableTitle"
+        :description="unavailableDescription"
+      />
+    </div>
     <div
       v-else-if="items.length === 0"
       class="p-6"
@@ -38,20 +34,29 @@
         description="当前时间范围内没有统计结果"
       />
     </div>
-    <Table v-else>
+    <Table
+      v-else
+      class="table-fixed"
+      :class="{ 'opacity-60 transition-opacity': isRefreshing }"
+    >
       <TableHeader>
         <TableRow>
-          <TableHead class="w-16">
+          <TableHead class="w-14 px-2.5">
             排名
           </TableHead>
-          <TableHead>名称</TableHead>
-          <TableHead class="text-right">
+          <TableHead class="w-[34%] px-2.5">
+            名称
+          </TableHead>
+          <TableHead class="w-[15%] px-2.5 text-right">
             请求数
           </TableHead>
-          <TableHead class="text-right">
+          <TableHead class="w-[18%] px-2.5 text-right">
             Tokens
           </TableHead>
-          <TableHead class="text-right">
+          <TableHead class="w-[16%] px-2.5 text-right">
+            费用
+          </TableHead>
+          <TableHead class="w-[17%] px-2.5 text-right">
             成本
           </TableHead>
         </TableRow>
@@ -60,68 +65,118 @@
         <TableRow
           v-for="item in items"
           :key="item.id"
+          class="cursor-pointer"
+          :class="item.id === selectedId
+            ? 'bg-[linear-gradient(90deg,rgba(232,145,89,0.16),rgba(232,145,89,0.05),transparent)] shadow-[inset_4px_0_0_0_rgba(232,145,89,0.9)] hover:bg-[linear-gradient(90deg,rgba(232,145,89,0.2),rgba(232,145,89,0.06),transparent)]'
+            : 'hover:bg-muted/25'"
+          @click="emit('select', item.id)"
+        >
+          <TableCell class="px-2.5 py-2.5">
+            <div
+              class="inline-flex h-7 min-w-7 items-center justify-center rounded-full border px-2 text-[11px] font-semibold tabular-nums transition-colors"
+              :class="item.id === selectedId
+                ? 'border-amber-500/70 bg-amber-500/15 text-amber-700 dark:text-amber-300'
+                : 'border-border/70 bg-muted/35 text-muted-foreground'"
             >
-          <TableCell class="font-medium">
               {{ item.rank }}
+            </div>
           </TableCell>
-          <TableCell>{{ item.name }}</TableCell>
-          <TableCell class="text-right">
-            {{ item.requests }}
+          <TableCell class="px-2.5 py-2.5">
+            <div class="min-w-0">
+              <div
+                class="truncate text-sm font-medium"
+                :title="item.name"
+              >
+                {{ item.name }}
+              </div>
+              <div class="mt-0.5 truncate text-[10px] text-muted-foreground">
+                费用优先
+              </div>
+            </div>
           </TableCell>
-          <TableCell class="text-right">
+          <TableCell class="px-2.5 py-2.5 text-right text-[11px] tabular-nums">
+            {{ item.requests.toLocaleString() }}
+          </TableCell>
+          <TableCell class="px-2.5 py-2.5 text-right text-[11px] tabular-nums">
             {{ formatTokens(item.tokens) }}
           </TableCell>
-          <TableCell class="text-right">
+          <TableCell class="px-2.5 py-2.5 text-right text-[11px] font-medium tabular-nums text-amber-700 dark:text-amber-300">
             {{ formatCurrency(item.cost) }}
           </TableCell>
+          <TableCell class="px-2.5 py-2.5 text-right text-[11px] tabular-nums text-muted-foreground">
+            {{ formatCurrency(item.actualCost) }}
+          </TableCell>
         </TableRow>
       </TableBody>
     </Table>
+
+    <template v-if="showPagination" #pagination>
+      <Pagination
+        :current="currentPage"
+        :total="totalItems"
+        :page-size="pageSize"
+        :show-page-size-selector="false"
+        @update:current="emit('update:currentPage', $event)"
+      />
+    </template>
   </TableCard>
 </template>
 
 <script setup lang="ts">
 import { computed } from 'vue'
 import { EmptyState, LoadingState } from '@/components/common'
-import { TableCard } from '@/components/ui'
-import {
-  Select,
-  SelectContent,
-  SelectItem,
-  SelectTrigger,
-  SelectValue,
-  Table,
-  TableBody,
-  TableCell,
-  TableHead,
-  TableHeader,
-  TableRow
-} from '@/components/ui'
+import { Pagination, Table, TableBody, TableCell, TableHead, TableHeader, TableRow, TableCard } from '@/components/ui'
 import { formatCurrency, formatTokens } from '@/utils/format'
-import type { LeaderboardItem } from '@/api/admin'
+
+interface LeaderboardItem {
+  rank: number
+  id: string
+  name: string
+  requests: number
+  tokens: number
+  cost: number
+  actualCost: number
+}
 
 interface Props {
   title: string
   items: LeaderboardItem[]
-  metric: 'requests' | 'tokens' | 'cost'
+  selectedId?: string
   loading?: boolean
-  showMetricSelect?: boolean
+  hasLoaded?: boolean
+  currentPage?: number
+  totalItems?: number
+  pageSize?: number
+  showPagination?: boolean
+  unavailable?: boolean
+  unavailableTitle?: string
+  unavailableDescription?: string
 }
 
 const props = withDefaults(defineProps<Props>(), {
+  selectedId: undefined,
   loading: false,
-  showMetricSelect: true
+  hasLoaded: false,
+  currentPage: 1,
+  totalItems: 0,
+  pageSize: 5,
+  showPagination: false,
+  unavailable: false,
+  unavailableTitle: '排行榜暂不可用',
+  unavailableDescription: '接口未返回结果，请稍后重试',
 })
 
 const emit = defineEmits<{
-  (e: 'update:metric', value: 'requests' | 'tokens' | 'cost'): void
+  (e: 'select', value: string): void
+  (e: 'update:currentPage', value: number): void
 }>()
 
-const metric = computed(() => props.metric)
-
-function emitMetric(value: string) {
-  if (value === 'requests' || value === 'tokens' || value === 'cost') {
-    emit('update:metric', value)
-  }
-}
+const selectedId = computed(() => props.selectedId)
+const showLoadingState = computed(() => props.loading && (!props.hasLoaded || props.items.length === 0))
+const showUnavailableState = computed(() => props.unavailable && !props.loading && props.items.length === 0)
+const isRefreshing = computed(() => props.loading && props.hasLoaded && props.items.length > 0)
+const currentPage = computed(() => props.currentPage)
+const totalItems = computed(() => props.totalItems)
+const pageSize = computed(() => props.pageSize)
+const showPagination = computed(() => props.showPagination && props.totalItems > props.pageSize)
 </script>

frontend/src/components/stats/PercentileChart.vue

@@ -1,27 +1,59 @@
 <template>
-  <div class="space-y-3">
-    <div class="flex items-center justify-between">
+  <div class="space-y-4">
+    <div class="flex flex-col gap-3 sm:flex-row sm:items-start sm:justify-between">
+      <div>
         <h3 class="text-sm font-semibold">
           {{ title }}
         </h3>
-      <span
+        <p
           v-if="subtitle"
-        class="text-xs text-muted-foreground"
-      >{{ subtitle }}</span>
+          class="mt-1 text-[11px] text-muted-foreground"
+        >
+          {{ subtitle }}
+        </p>
       </div>
+
+      <div
+        v-if="summaryMetrics.length > 0 && !unavailable"
+        class="flex flex-wrap gap-2"
+      >
+        <div
+          v-for="metric in summaryMetrics"
+          :key="metric.label"
+          class="rounded-full border px-2.5 py-1 text-[10px] font-medium"
+          :class="metric.className"
+        >
+          {{ metric.label }} {{ metric.value }}
+        </div>
+      </div>
+    </div>
+
     <div
       v-if="loading"
       class="p-6"
     >
       <LoadingState />
     </div>
+    <div
+      v-else-if="unavailable"
+      class="flex h-[272px] items-center justify-center rounded-2xl border border-dashed border-border/70 bg-muted/10 text-xs text-muted-foreground"
+    >
+      {{ unavailableText }}
+    </div>
+    <div
+      v-else-if="!hasData"
+      class="flex h-[272px] items-center justify-center rounded-2xl border border-dashed border-border/70 bg-muted/10 text-xs text-muted-foreground"
+    >
+      暂无延迟数据
+    </div>
     <div
       v-else
-      class="h-[260px]"
+      class="rounded-2xl border border-border/60 bg-background p-3"
     >
       <LineChart
         :data="chartData"
         :options="chartOptions"
+        :height="248"
       />
     </div>
   </div>
@@ -29,9 +61,19 @@
 
 <script setup lang="ts">
 import { computed } from 'vue'
+import type { ChartData, ChartOptions } from 'chart.js'
 import LineChart from '@/components/charts/LineChart.vue'
 import { LoadingState } from '@/components/common'
-import type { PercentileItem } from '@/api/admin'
+
+interface PercentileItem {
+  date: string
+  p50_response_time_ms?: number | null
+  p90_response_time_ms?: number | null
+  p99_response_time_ms?: number | null
+  p50_first_byte_time_ms?: number | null
+  p90_first_byte_time_ms?: number | null
+  p99_first_byte_time_ms?: number | null
+}
 
 interface Props {
   title: string
@@ -39,61 +81,159 @@ interface Props {
   series: PercentileItem[]
   mode: 'response' | 'ttfb'
   loading?: boolean
+  unavailable?: boolean
+  unavailableText?: string
 }
 
 const props = withDefaults(defineProps<Props>(), {
   subtitle: undefined,
-  loading: false
+  loading: false,
+  unavailable: false,
+  unavailableText: '数据暂不可用',
 })
 
-const labels = computed(() => props.series.map(item => item.date))
-
-// 毫秒转秒
 function msToSeconds(ms: number | null | undefined): number | null {
   if (ms == null) return null
-  return ms / 1000
+  return Number((ms / 1000).toFixed(3))
 }
 
-const chartData = computed(() => {
-  const p50Key = props.mode === 'response' ? 'p50_response_time_ms' : 'p50_first_byte_time_ms'
-  const p90Key = props.mode === 'response' ? 'p90_response_time_ms' : 'p90_first_byte_time_ms'
-  const p99Key = props.mode === 'response' ? 'p99_response_time_ms' : 'p99_first_byte_time_ms'
+function getPercentileValue(item: PercentileItem, percentile: '50' | '90' | '99') {
+  if (props.mode === 'response') {
+    if (percentile === '50') return msToSeconds(item.p50_response_time_ms)
+    if (percentile === '90') return msToSeconds(item.p90_response_time_ms)
+    return msToSeconds(item.p99_response_time_ms)
+  }
 
-  return {
+  if (percentile === '50') return msToSeconds(item.p50_first_byte_time_ms)
+  if (percentile === '90') return msToSeconds(item.p90_first_byte_time_ms)
+  return msToSeconds(item.p99_first_byte_time_ms)
+}
+
+function getLatestMetricValue(percentile: '50' | '90' | '99') {
+  for (let index = props.series.length - 1; index >= 0; index -= 1) {
+    const value = getPercentileValue(props.series[index], percentile)
+    if (value != null) return value
+  }
+  return null
+}
+
+function formatSeconds(value: number | null) {
+  if (value == null) return '--'
+  if (value < 1) return `${Math.round(value * 1000)}ms`
+  return `${value.toFixed(2)}s`
+}
+
+const labels = computed(() => props.series.map(item => item.date?.slice(5) ?? ''))
+const hasData = computed(() => props.series.some(item => (
+  getPercentileValue(item, '50') != null
+  || getPercentileValue(item, '90') != null
+  || getPercentileValue(item, '99') != null
+)))
+
+const summaryMetrics = computed(() => ([
+  {
+    label: 'P50',
+    value: formatSeconds(getLatestMetricValue('50')),
+    className: 'border-sky-500/20 bg-sky-500/8 text-sky-700 dark:text-sky-300',
+  },
+  {
+    label: 'P90',
+    value: formatSeconds(getLatestMetricValue('90')),
+    className: 'border-amber-500/20 bg-amber-500/8 text-amber-700 dark:text-amber-300',
+  },
+  {
+    label: 'P99',
+    value: formatSeconds(getLatestMetricValue('99')),
+    className: 'border-rose-500/20 bg-rose-500/8 text-rose-700 dark:text-rose-300',
+  },
+]))
+
+const chartData = computed<ChartData<'line'>>(() => ({
   labels: labels.value,
   datasets: [
     {
       label: 'P50',
-        data: props.series.map(item => msToSeconds(item[p50Key])),
-        borderColor: 'rgb(59, 130, 246)',
-        tension: 0.25,
-        pointRadius: 2
+      data: props.series.map(item => getPercentileValue(item, '50')),
+      borderColor: 'rgb(70, 136, 240)',
+      backgroundColor: 'rgba(70, 136, 240, 0.10)',
+      borderWidth: 2.4,
+      fill: false,
+      tension: 0.32,
+      pointRadius: 0,
+      pointHoverRadius: 4,
+      pointHitRadius: 14,
     },
     {
       label: 'P90',
-        data: props.series.map(item => msToSeconds(item[p90Key])),
-        borderColor: 'rgb(234, 179, 8)',
-        tension: 0.25,
-        pointRadius: 2
+      data: props.series.map(item => getPercentileValue(item, '90')),
+      borderColor: 'rgb(223, 166, 55)',
+      backgroundColor: 'rgba(223, 166, 55, 0.08)',
+      borderWidth: 2.1,
+      fill: false,
+      tension: 0.3,
+      pointRadius: 0,
+      pointHoverRadius: 4,
+      pointHitRadius: 14,
     },
     {
       label: 'P99',
-        data: props.series.map(item => msToSeconds(item[p99Key])),
-        borderColor: 'rgb(239, 68, 68)',
-        tension: 0.25,
-        pointRadius: 2
-      }
-    ]
-  }
-})
+      data: props.series.map(item => getPercentileValue(item, '99')),
+      borderColor: 'rgb(214, 92, 111)',
+      backgroundColor: 'rgba(214, 92, 111, 0.08)',
+      borderWidth: 2,
+      borderDash: [6, 4],
+      fill: false,
+      tension: 0.28,
+      pointRadius: 0,
+      pointHoverRadius: 4,
+      pointHitRadius: 14,
+    },
+  ],
+}))
 
-const chartOptions = computed(() => ({
+const chartOptions = computed<ChartOptions<'line'>>(() => ({
+  responsive: true,
+  maintainAspectRatio: false,
+  interaction: {
+    mode: 'index',
+    intersect: false,
+  },
+  plugins: {
+    legend: {
+      position: 'bottom',
+      labels: {
+        usePointStyle: true,
+        boxWidth: 10,
+        padding: 14,
+      },
+    },
+    tooltip: {
+      callbacks: {
+        title: items => props.series[items[0]?.dataIndex ?? -1]?.date ?? '',
+        label: context => `${context.dataset.label}: ${formatSeconds(Number(context.raw ?? 0))}`,
+      },
+    },
+  },
   scales: {
-    y: {
+    x: {
+      grid: {
+        color: 'rgba(148, 163, 184, 0.06)',
+      },
       ticks: {
-        callback: (value: number) => `${value}s`
-      }
-    }
-  }
+        color: 'rgb(100, 116, 139)',
+        maxRotation: 0,
+      },
+    },
+    y: {
+      beginAtZero: true,
+      grid: {
+        color: 'rgba(148, 163, 184, 0.07)',
+      },
+      ticks: {
+        color: 'rgb(100, 116, 139)',
+        callback: value => formatSeconds(Number(value)),
+      },
+    },
+  },
 }))
 </script>

frontend/src/components/stats/QuotaProgressCard.vue

@@ -1,81 +0,0 @@
-<template>
-  <Card class="p-4 space-y-4">
-    <div class="flex items-center justify-between">
-      <h3 class="text-sm font-semibold">
-        {{ title }}
-      </h3>
-      <span
-        v-if="subtitle"
-        class="text-xs text-muted-foreground"
-      >{{ subtitle }}</span>
-    </div>
-
-    <div
-      v-if="loading"
-      class="p-4"
-    >
-      <LoadingState />
-    </div>
-    <div
-      v-else-if="providers.length === 0"
-      class="p-4"
-    >
-      <EmptyState
-        title="暂无数据"
-        description="暂无月卡配额数据"
-      />
-    </div>
-    <div
-      v-else
-      class="space-y-4"
-    >
-      <div
-        v-for="provider in providers"
-        :key="provider.id"
-        class="space-y-2"
-      >
-        <div class="flex items-center justify-between text-xs">
-          <span class="font-medium">{{ provider.name }}</span>
-          <span class="text-muted-foreground">
-            {{ formatCurrency(provider.used_usd) }} / {{ formatCurrency(provider.quota_usd) }}
-          </span>
-        </div>
-        <div class="h-2 rounded-full bg-muted">
-          <div
-            class="h-2 rounded-full bg-primary"
-            :style="{ width: `${Math.min(provider.usage_percent, 100)}%` }"
-          />
-        </div>
-        <div class="flex items-center justify-between text-[11px] text-muted-foreground">
-          <span>剩余 {{ formatCurrency(provider.remaining_usd) }}</span>
-          <span v-if="provider.estimated_exhaust_at">
-            预计耗尽 {{ formatDate(provider.estimated_exhaust_at) }}
-          </span>
-        </div>
-      </div>
-    </div>
-  </Card>
-</template>
-
-<script setup lang="ts">
-import { Card } from '@/components/ui'
-import { EmptyState, LoadingState } from '@/components/common'
-import { formatCurrency } from '@/utils/format'
-import type { QuotaUsageProvider } from '@/api/admin'
-
-interface Props {
-  title: string
-  subtitle?: string
-  providers: QuotaUsageProvider[]
-  loading?: boolean
-}
-
-withDefaults(defineProps<Props>(), {
-  subtitle: undefined,
-  loading: false
-})
-
-function formatDate(value: string) {
-  return new Date(value).toLocaleDateString()
-}
-</script>

frontend/src/components/stats/index.ts

@@ -1,7 +1,4 @@
 export { default as ActivityHeatmap } from './ActivityHeatmap.vue'
 export { default as LeaderboardTable } from './LeaderboardTable.vue'
-export { default as CostForecastChart } from './CostForecastChart.vue'
-export { default as QuotaProgressCard } from './QuotaProgressCard.vue'
 export { default as PercentileChart } from './PercentileChart.vue'
 export { default as ErrorDistributionChart } from './ErrorDistributionChart.vue'
-export { default as ComparisonCard } from './ComparisonCard.vue'

frontend/src/components/ui/select-item.vue

@@ -59,7 +59,7 @@ const isHidden = computed(
 )
 const itemClass = computed(() =>
   cn(
-    'relative flex w-full cursor-pointer select-none items-center rounded-lg py-1.5 pl-8 pr-2 text-sm outline-none',
+    'relative flex min-w-0 w-full cursor-pointer select-none items-center rounded-lg py-1.5 pl-8 pr-2 text-sm outline-none',
     'data-[highlighted]:bg-accent focus:bg-accent text-foreground',
     'transition-colors data-[disabled]:pointer-events-none data-[disabled]:opacity-50',
     isHidden.value && 'hidden',
@@ -101,8 +101,16 @@ onBeforeUnmount(() => {
         <Check class="h-4 w-4" />
       </SelectItemIndicator>
     </span>
-    <SelectItemText>
+    <SelectItemText
+      as="div"
+      class="min-w-0 flex-1 overflow-hidden"
+    >
+      <span
+        class="block w-full min-w-0 truncate"
+        :title="normalizedText"
+      >
         <slot />
+      </span>
     </SelectItemText>
   </SelectItemPrimitive>
 </template>

frontend/src/components/ui/select-trigger.vue

@@ -25,7 +25,7 @@ const triggerClass = computed(() =>
     :class="triggerClass"
     :disabled="disabled"
   >
-    <span class="truncate">
+    <span class="min-w-0 flex-1 truncate">
       <slot />
     </span>
     <ChevronDown class="h-4 w-4 opacity-50 pointer-events-none flex-shrink-0" />

frontend/src/composables/analytics/index.ts

@@ -0,0 +1,3 @@
+export { useLeaderboardData } from './useLeaderboardData'
+export type { LeaderboardDimension } from './useLeaderboardData'
+export { usePerformanceData } from './usePerformanceData'

frontend/src/composables/analytics/mappers.ts

@@ -0,0 +1,23 @@
+import type { AnalyticsLeaderboardItem } from '@/api/analytics'
+
+export interface LeaderboardTableRow {
+  rank: number
+  id: string
+  name: string
+  requests: number
+  tokens: number
+  cost: number
+  actualCost: number
+}
+
+export function mapAnalyticsLeaderboardItem(item: AnalyticsLeaderboardItem): LeaderboardTableRow {
+  return {
+    rank: item.rank,
+    id: item.id,
+    name: item.label,
+    requests: item.requests_total,
+    tokens: item.total_tokens,
+    cost: item.total_cost_usd,
+    actualCost: item.actual_total_cost_usd,
+  }
+}

frontend/src/composables/analytics/useLeaderboardData.ts

@@ -0,0 +1,272 @@
+import { computed, onMounted, onUnmounted, ref, watch, type Ref } from 'vue'
+import { analyticsApi, type AnalyticsFilterOption, type AnalyticsLeaderboardItem, type AnalyticsTimeseriesBucket } from '@/api/analytics'
+import { buildTimeRangeParams, createLoader, createRequestGuard, useInjectedAnalyticsFilters } from '@/composables/useAnalyticsFilters'
+import { fillMissingTimeseriesBuckets } from '@/utils/analyticsTimeseries'
+import { isAnalyticsSingleDayRange, resolveAnalyticsAutoGranularity, type AnalyticsGranularityOption } from '@/utils/analyticsGranularity'
+import { mapAnalyticsLeaderboardItem, type LeaderboardTableRow } from './mappers'
+
+export type LeaderboardDimension = 'users' | 'api_keys'
+
+export interface LeaderboardSummary {
+  total_requests: number
+  total_tokens: number
+  total_cost: number
+  total_actual_cost: number
+  avg_response_time: number
+}
+
+function dimensionToEntity(dimension: LeaderboardDimension): 'user' | 'api_key' {
+  return dimension === 'users' ? 'user' : 'api_key'
+}
+
+export function useLeaderboardData(dimension: Ref<LeaderboardDimension>) {
+  const filters = useInjectedAnalyticsFilters()
+
+  const leaderboardItems = ref<AnalyticsLeaderboardItem[]>([])
+  const latestUserOptions = ref<AnalyticsFilterOption[]>([])
+  const latestApiKeyOptions = ref<AnalyticsFilterOption[]>([])
+  const leaderboardLoading = ref(false)
+  const leaderboardError = ref(false)
+  const leaderboardHasLoaded = ref(false)
+  const selectedId = ref<string>('')
+
+  const selectedSummary = ref<LeaderboardSummary | null>(null)
+  const selectedTrendBuckets = ref<AnalyticsTimeseriesBucket[]>([])
+  const panelLoading = ref(false)
+  const panelError = ref(false)
+  const panelHasLoaded = ref(false)
+
+  const leaderboardGuard = createRequestGuard()
+  const filterOptionsGuard = createRequestGuard()
+  const panelGuard = createRequestGuard()
+
+  const entityOptions = computed(() => (
+    dimension.value === 'users' ? latestUserOptions.value : latestApiKeyOptions.value
+  ))
+  const entityLabelMap = computed(() => (
+    new Map(entityOptions.value.map(option => [option.value, option.label]))
+  ))
+
+  const leaderboard = computed<LeaderboardTableRow[]>(() => (
+    leaderboardItems.value.map(item => ({
+      ...mapAnalyticsLeaderboardItem(item),
+      name: entityLabelMap.value.get(item.id) ?? item.label,
+    }))
+  ))
+
+  const selectedItem = computed(() => (
+    leaderboard.value.find(item => item.id === selectedId.value) ?? null
+  ))
+
+  const selectedLabel = computed(() => (
+    selectedItem.value?.name ?? entityLabelMap.value.get(selectedId.value) ?? null
+  ))
+
+  const resolvedTrendGranularity = computed(() => resolveAnalyticsAutoGranularity(filters.timeRange.value))
+  const baseTimeRangePayload = computed(() => ({
+    start_date: filters.timeRange.value.start_date,
+    end_date: filters.timeRange.value.end_date,
+    preset: filters.timeRange.value.preset,
+    timezone: filters.timeRange.value.timezone,
+    tz_offset_minutes: filters.timeRange.value.tz_offset_minutes,
+  }))
+  const trendTimeRangePayload = computed(() => ({
+    ...baseTimeRangePayload.value,
+    granularity: resolvedTrendGranularity.value,
+  }))
+
+  const currentTrendGranularitySelection = computed<AnalyticsGranularityOption>({
+    get: () => filters.timeRange.value.granularity || 'auto',
+    set: (value) => {
+      if (value === 'hour' && !isAnalyticsSingleDayRange(filters.timeRange.value)) {
+        value = 'auto'
+      }
+      filters.timeRange.value = {
+        ...filters.timeRange.value,
+        granularity: value,
+      }
+    },
+  })
+
+  function resetPanelData(invalidate = false) {
+    if (invalidate) panelGuard.invalidate()
+    panelLoading.value = false
+    panelError.value = false
+    selectedSummary.value = null
+    selectedTrendBuckets.value = []
+  }
+
+  async function loadFilterOptions() {
+    const requestId = filterOptionsGuard.next()
+    const response = await analyticsApi.getFilterOptions({
+      scope: { kind: 'global' },
+      time_range: baseTimeRangePayload.value,
+      filters: {},
+    }).catch(() => null)
+
+    if (filterOptionsGuard.isStale(requestId) || !response) return
+    latestUserOptions.value = response.users ?? []
+    latestApiKeyOptions.value = response.api_keys ?? []
+  }
+
+  async function loadLeaderboard() {
+    const requestId = leaderboardGuard.next()
+    leaderboardLoading.value = true
+    leaderboardError.value = false
+    try {
+      const response = await analyticsApi.getLeaderboard({
+        scope: { kind: 'global' },
+        time_range: baseTimeRangePayload.value,
+        filters: {
+          statuses: ['completed', 'failed', 'cancelled'],
+        },
+        entity: dimensionToEntity(dimension.value),
+        metric: 'total_cost_usd',
+        limit: 20,
+      })
+
+      if (leaderboardGuard.isStale(requestId)) return
+      leaderboardItems.value = response.items
+
+      const nextSelectedId = response.items.some(item => item.id === selectedId.value)
+        ? selectedId.value
+        : (response.items[0]?.id ?? '')
+
+      if (selectedId.value !== nextSelectedId) {
+        selectedId.value = nextSelectedId
+      }
+      leaderboardHasLoaded.value = true
+    } catch {
+      if (leaderboardGuard.isStale(requestId)) return
+      leaderboardError.value = true
+    } finally {
+      if (leaderboardGuard.isCurrent(requestId)) {
+        leaderboardLoading.value = false
+      }
+    }
+  }
+
+  async function loadPanel() {
+    if (!selectedId.value) {
+      resetPanelData(true)
+      return
+    }
+
+    const requestId = panelGuard.next()
+    panelLoading.value = true
+    panelError.value = false
+    try {
+      const scope = dimension.value === 'users'
+        ? { kind: 'user' as const, user_id: selectedId.value }
+        : { kind: 'api_key' as const, api_key_id: selectedId.value }
+
+      const basePayload = {
+        scope,
+        time_range: baseTimeRangePayload.value,
+        filters: {
+          statuses: ['completed', 'failed', 'cancelled'],
+        },
+      }
+
+      const [overview, timeseries] = await Promise.all([
+        analyticsApi.getOverview(basePayload),
+        analyticsApi.getTimeseries({
+          ...basePayload,
+          time_range: trendTimeRangePayload.value,
+        }),
+      ])
+
+      if (panelGuard.isStale(requestId)) return
+      selectedSummary.value = {
+        total_requests: overview.summary.requests_total,
+        total_tokens: overview.summary.total_tokens,
+        total_cost: overview.summary.total_cost_usd,
+        total_actual_cost: overview.summary.actual_total_cost_usd,
+        avg_response_time: overview.summary.avg_response_time_ms / 1000,
+      }
+      selectedTrendBuckets.value = fillMissingTimeseriesBuckets(
+        timeseries.buckets,
+        trendTimeRangePayload.value,
+      )
+      panelHasLoaded.value = true
+    } catch {
+      if (panelGuard.isStale(requestId)) return
+      panelError.value = true
+    } finally {
+      if (panelGuard.isCurrent(requestId)) {
+        panelLoading.value = false
+      }
+    }
+  }
+
+  const filterOptionsLoader = createLoader(loadFilterOptions)
+  const leaderboardLoader = createLoader(loadLeaderboard)
+  const panelLoader = createLoader(loadPanel)
+
+  watch(
+    () => [filters.timeRange.value.preset, filters.timeRange.value.start_date, filters.timeRange.value.end_date],
+    () => {
+      if (filters.timeRange.value.granularity === 'hour' && !isAnalyticsSingleDayRange(filters.timeRange.value)) {
+        filters.timeRange.value = {
+          ...filters.timeRange.value,
+          granularity: 'auto',
+        }
+      }
+    },
+  )
+
+  watch(
+    () => [baseTimeRangePayload.value, dimension.value],
+    () => {
+      filterOptionsLoader.schedule()
+      leaderboardLoader.schedule()
+    },
+    { deep: true },
+  )
+
+  watch(
+    () => [
+      selectedId.value,
+      dimension.value,
+      baseTimeRangePayload.value,
+      trendTimeRangePayload.value,
+    ],
+    () => panelLoader.schedule(),
+    { deep: true },
+  )
+
+  onMounted(() => {
+    void (async () => {
+      await filterOptionsLoader.execute()
+      await leaderboardLoader.execute()
+      await panelLoader.execute()
+    })()
+  })
+
+  onUnmounted(() => {
+    filterOptionsLoader.cleanup()
+    leaderboardLoader.cleanup()
+    panelLoader.cleanup()
+    filterOptionsGuard.invalidate()
+    leaderboardGuard.invalidate()
+    panelGuard.invalidate()
+  })
+
+  return {
+    leaderboard,
+    leaderboardLoading,
+    leaderboardError,
+    leaderboardHasLoaded,
+    selectedId,
+    selectedItem,
+    selectedLabel,
+    selectedSummary,
+    selectedTrendBuckets,
+    panelLoading,
+    panelError,
+    panelHasLoaded,
+    resolvedTrendGranularity,
+    currentTrendGranularitySelection,
+    reloadLeaderboard: leaderboardLoader.schedule,
+  }
+}

frontend/src/composables/analytics/usePerformanceData.ts

@@ -0,0 +1,98 @@
+import { ref, watch, onMounted, onUnmounted } from 'vue'
+import { analyticsApi, type AnalyticsErrorCategory, type AnalyticsErrorTrendItem, type AnalyticsPercentilePoint, type AnalyticsPerformanceResponse, type AnalyticsProviderHealthItem } from '@/api/analytics'
+import { auditApi, type MonitoringSystemStatus } from '@/api/audit'
+import { getHealthSummary, type HealthSummary } from '@/api/endpoints'
+import { createLoader, createRequestGuard, useInjectedAnalyticsFilters } from '@/composables/useAnalyticsFilters'
+
+const EMPTY_LATENCY: AnalyticsPerformanceResponse['latency'] = {
+  response_time_ms: { avg: 0, p50: null, p90: null, p99: null },
+  first_byte_time_ms: { avg: 0, p50: null, p90: null, p99: null },
+}
+
+export function usePerformanceData() {
+  const filters = useInjectedAnalyticsFilters()
+
+  const latency = ref<AnalyticsPerformanceResponse['latency']>(EMPTY_LATENCY)
+  const percentiles = ref<AnalyticsPercentilePoint[]>([])
+  const errorDistribution = ref<AnalyticsErrorCategory[]>([])
+  const errorTrend = ref<AnalyticsErrorTrendItem[]>([])
+  const errorTotal = ref(0)
+  const errorRate = ref(0)
+  const providerStatus = ref<AnalyticsProviderHealthItem[]>([])
+  const healthSummary = ref<HealthSummary | null>(null)
+  const systemStatus = ref<MonitoringSystemStatus | null>(null)
+  const loading = ref(false)
+  const loadError = ref(false)
+  const hasLoaded = ref(false)
+  const guard = createRequestGuard()
+
+  async function load() {
+    const requestId = guard.next()
+    loading.value = true
+    loadError.value = false
+    try {
+      const [response, summaryResponse, systemResponse] = await Promise.all([
+        analyticsApi.getPerformance({
+          scope: { kind: 'global' },
+          time_range: filters.getTimeRangeParams(),
+          filters: {
+            statuses: ['completed', 'failed', 'cancelled'],
+          },
+        }),
+        getHealthSummary().catch(() => null),
+        auditApi.getSystemStatus().catch(() => null),
+      ])
+
+      if (guard.isStale(requestId)) return
+      latency.value = response.latency
+      percentiles.value = response.percentiles
+      errorTotal.value = response.errors.total
+      errorRate.value = response.errors.rate
+      errorDistribution.value = response.errors.categories
+      errorTrend.value = response.errors.trend
+      providerStatus.value = response.provider_health
+      if (summaryResponse) {
+        healthSummary.value = summaryResponse
+      }
+      if (systemResponse) {
+        systemStatus.value = systemResponse
+      }
+      hasLoaded.value = true
+    } catch {
+      if (guard.isStale(requestId)) return
+      loadError.value = true
+    } finally {
+      if (guard.isCurrent(requestId)) {
+        loading.value = false
+      }
+    }
+  }
+
+  const loader = createLoader(load)
+
+  watch(
+    () => [
+      filters.getTimeRangeParams(),
+    ],
+    () => loader.schedule(),
+    { deep: true },
+  )
+
+  onMounted(() => { void loader.execute() })
+  onUnmounted(() => { loader.cleanup(); guard.invalidate() })
+
+  return {
+    latency,
+    percentiles,
+    errorDistribution,
+    errorTrend,
+    errorTotal,
+    errorRate,
+    providerStatus,
+    healthSummary,
+    systemStatus,
+    loading,
+    loadError,
+    hasLoaded,
+  }
+}

frontend/src/composables/useAnalyticsFilters.ts

@@ -0,0 +1,144 @@
+import { inject, ref, watch } from 'vue'
+import { useRoute, useRouter } from 'vue-router'
+import type { DateRangeParams, PeriodValue } from '@/features/usage/types'
+import { getDateRangeFromPeriod } from '@/features/usage/composables'
+
+/**
+ * 构建时间范围 API 参数（从各页面提取的公共逻辑）
+ */
+export function buildTimeRangeParams(timeRange: DateRangeParams) {
+  return {
+    start_date: timeRange.start_date,
+    end_date: timeRange.end_date,
+    preset: timeRange.preset,
+    timezone: timeRange.timezone,
+    tz_offset_minutes: timeRange.tz_offset_minutes,
+    granularity: timeRange.granularity && timeRange.granularity !== 'auto'
+      ? timeRange.granularity
+      : 'day',
+  }
+}
+
+/**
+ * 通用防抖加载器工厂
+ * 封装 120ms debounce + in-flight promise 去重 + requestId 过期检查
+ */
+export function createLoader(loadFn: () => Promise<void>) {
+  let loadPromise: Promise<void> | null = null
+  let hasPending = false
+  let debounceTimer: ReturnType<typeof setTimeout> | null = null
+
+  function schedule() {
+    if (debounceTimer) clearTimeout(debounceTimer)
+    debounceTimer = setTimeout(() => {
+      debounceTimer = null
+      void execute()
+    }, 120)
+  }
+
+  async function execute() {
+    if (loadPromise) {
+      hasPending = true
+      return loadPromise
+    }
+    loadPromise = loadFn()
+      .finally(() => {
+        loadPromise = null
+        if (hasPending) {
+          hasPending = false
+          void execute()
+        }
+      })
+    return loadPromise
+  }
+
+  function cleanup() {
+    if (debounceTimer) {
+      clearTimeout(debounceTimer)
+      debounceTimer = null
+    }
+    hasPending = false
+    loadPromise = null
+  }
+
+  return { schedule, execute, cleanup }
+}
+
+/**
+ * 请求 ID 守卫：用于确保异步回调中的数据不被过期请求覆盖
+ */
+export function createRequestGuard() {
+  let id = 0
+  function next() { return ++id }
+  function isStale(requestId: number) { return requestId !== id }
+  function isCurrent(requestId: number) { return requestId === id }
+  function invalidate() { id++ }
+  return { next, isStale, isCurrent, invalidate }
+}
+
+export interface AnalyticsFiltersOptions {
+  defaultPreset?: PeriodValue
+  syncToUrl?: boolean
+}
+
+/**
+ * 统一的分析页面筛选状态
+ * - timeRange: 时间范围
+ * - modelFilter / providerFilter / apiKeyFilter / userFilter: 多维度筛选
+ * - activeTab: 当前 Tab
+ * - URL query 双向同步
+ */
+export function useAnalyticsFilters(options: AnalyticsFiltersOptions = {}) {
+  const { defaultPreset = 'last30days', syncToUrl = true } = options
+  const route = useRoute()
+  const router = useRouter()
+
+  // 从 URL query 初始化，无则用默认值
+  const initialPreset = (syncToUrl && route.query.preset as PeriodValue) || defaultPreset
+  const initialTab = (syncToUrl && route.query.tab as string) || 'detail'
+
+  const timeRange = ref<DateRangeParams>(getDateRangeFromPeriod(initialPreset))
+  const activeTab = ref(initialTab)
+  const modelFilter = ref<string[]>([])
+  const providerFilter = ref<string[]>([])
+  const apiKeyFilter = ref<string[]>([])
+  const userFilter = ref<string[]>([])
+
+  // URL 同步：仅同步 preset 和 tab（各类多选筛选太冗长不入 URL）
+  if (syncToUrl) {
+    watch([() => timeRange.value.preset, activeTab], ([preset, tab]) => {
+      const query: Record<string, string> = {}
+      if (preset && preset !== defaultPreset) query.preset = preset
+      if (tab && tab !== 'detail') query.tab = tab
+      // 保留其他 query 参数
+      const currentQuery = { ...route.query }
+      delete currentQuery.preset
+      delete currentQuery.tab
+      router.replace({ query: { ...currentQuery, ...query } })
+    }, { flush: 'post' })
+  }
+
+  function getTimeRangeParams() {
+    return buildTimeRangeParams(timeRange.value)
+  }
+
+  return {
+    timeRange,
+    activeTab,
+    modelFilter,
+    providerFilter,
+    apiKeyFilter,
+    userFilter,
+    getTimeRangeParams,
+  }
+}
+
+export type AnalyticsFilters = ReturnType<typeof useAnalyticsFilters>
+
+export function useInjectedAnalyticsFilters(): AnalyticsFilters {
+  const filters = inject<AnalyticsFilters>('analyticsFilters', null)
+  if (!filters) {
+    throw new Error('analyticsFilters provider is missing')
+  }
+  return filters
+}

frontend/src/composables/useReportsData.ts

@@ -0,0 +1,287 @@
+import { computed, ref, watch, onMounted, onUnmounted, type Ref } from 'vue'
+import {
+  analyticsApi,
+  type AnalyticsBreakdownRow,
+  type AnalyticsFilterOption,
+  type AnalyticsScope,
+  type AnalyticsSummary,
+  type AnalyticsTimeseriesBucket,
+} from '@/api/analytics'
+import type { DateRangeParams, PeriodValue } from '@/features/usage/types'
+import { getDateRangeFromPeriod } from '@/features/usage/composables'
+import { createLoader, createRequestGuard, buildTimeRangeParams } from '@/composables/useAnalyticsFilters'
+import { fillMissingTimeseriesBuckets } from '@/utils/analyticsTimeseries'
+import { getAnalyticsRangeDaysInclusive, resolveAnalyticsAutoGranularity, type ResolvedAnalyticsGranularity } from '@/utils/analyticsGranularity'
+import { type DailyUsageBreakdown } from '@/utils/usageBreakdown'
+
+export interface ModelSummary {
+  model: string
+  requests: number
+  tokens: number
+  cost: number
+  inputTokens: number
+  outputTokens: number
+  cacheCreationTokens: number
+  cacheReadTokens: number
+  cacheHitRate: number
+  avg_response_time?: number
+  avg_first_byte_time?: number
+}
+
+export interface ReportBucketStat {
+  bucket_start: string
+  bucket_end: string
+  requests: number
+  tokens: number
+  cost: number
+  avg_response_time: number
+  avg_first_byte_time: number
+  models_used_count: number
+}
+
+function mapBucketToReportStat(bucket: AnalyticsTimeseriesBucket): ReportBucketStat {
+  return {
+    bucket_start: bucket.bucket_start,
+    bucket_end: bucket.bucket_end,
+    requests: bucket.requests_total,
+    tokens: bucket.total_tokens,
+    cost: bucket.total_cost_usd,
+    avg_response_time: bucket.avg_response_time_ms / 1000,
+    avg_first_byte_time: bucket.avg_first_byte_time_ms / 1000,
+    models_used_count: bucket.models_used_count,
+  }
+}
+
+function mapBucketToBreakdown(bucket: AnalyticsTimeseriesBucket): DailyUsageBreakdown {
+  return {
+    date: bucket.bucket_start,
+    inputTokens: bucket.input_tokens,
+    outputTokens: bucket.output_tokens,
+    cacheCreationTokens: bucket.cache_creation_input_tokens,
+    cacheReadTokens: bucket.cache_read_input_tokens,
+    cacheCreationCost: bucket.cache_creation_cost_usd,
+    cacheReadCost: bucket.cache_read_cost_usd,
+    cacheHitRate: bucket.cache_hit_rate,
+    totalCacheCost: bucket.cache_cost_usd,
+    totalCacheTokens: bucket.cache_creation_input_tokens + bucket.cache_read_input_tokens,
+    totalTrackedTokens: bucket.total_tokens,
+    baseTokens: bucket.input_output_total_tokens,
+  }
+}
+
+function mapBreakdownRowToModelSummary(row: AnalyticsBreakdownRow): ModelSummary {
+  const requests = row.requests_total
+  const tokens = row.total_tokens
+  const cost = row.total_cost_usd
+  return {
+    model: row.label,
+    requests,
+    tokens,
+    cost,
+    inputTokens: row.input_tokens,
+    outputTokens: row.output_tokens,
+    cacheCreationTokens: row.cache_creation_input_tokens,
+    cacheReadTokens: row.cache_read_input_tokens,
+    cacheHitRate: row.cache_hit_rate,
+    avg_response_time: row.avg_response_time_ms / 1000,
+    avg_first_byte_time: row.avg_first_byte_time_ms / 1000,
+  }
+}
+
+type ReportGranularity = ResolvedAnalyticsGranularity
+
+function resolveReportGranularity(range: DateRangeParams): ReportGranularity {
+  return resolveAnalyticsAutoGranularity(range)
+}
+
+function areStringArraysEqual(left: string[], right: string[]): boolean {
+  if (left.length !== right.length) return false
+  return left.every((value, index) => value === right[index])
+}
+
+export interface UseReportsDataOptions {
+  scope?: AnalyticsScope
+  defaultPreset?: PeriodValue
+  timeRange?: Ref<DateRangeParams>
+  userFilter?: Ref<string[]>
+  apiKeyFilter?: Ref<string[]>
+  loadApiKeyOptions?: boolean
+}
+
+export function useReportsData(options: UseReportsDataOptions = {}) {
+  const {
+    scope = { kind: 'me' as const },
+    defaultPreset = 'today',
+    timeRange: externalTimeRange,
+    userFilter: externalUserFilter,
+    apiKeyFilter: externalApiKeyFilter,
+    loadApiKeyOptions = scope.kind === 'me',
+  } = options
+
+  const timeRange = externalTimeRange ?? ref<DateRangeParams>({
+    ...getDateRangeFromPeriod(defaultPreset),
+    granularity: 'auto',
+  })
+  const userFilter = externalUserFilter ?? ref<string[]>([])
+  const apiKeyFilter = externalApiKeyFilter ?? ref<string[]>([])
+  const apiKeyOptions = ref<AnalyticsFilterOption[]>([])
+
+  const summary = ref<AnalyticsSummary | null>(null)
+  const bucketStats = ref<ReportBucketStat[]>([])
+  const bucketBreakdowns = ref<Record<string, DailyUsageBreakdown>>({})
+  const modelSummary = ref<ModelSummary[]>([])
+  const loading = ref(false)
+  const loadError = ref(false)
+  const hasLoaded = ref(false)
+  const guard = createRequestGuard()
+  const apiKeysGuard = createRequestGuard()
+  let suppressNextApiKeyFilterLoad = false
+  const resolvedGranularity = computed(() => resolveReportGranularity(timeRange.value))
+  const timeseriesTimeRange = computed(() => ({
+    ...buildTimeRangeParams(timeRange.value),
+    granularity: resolvedGranularity.value,
+  }))
+
+  watch(
+    () => [timeRange.value.preset, timeRange.value.start_date, timeRange.value.end_date],
+    () => {
+      if (timeRange.value.granularity === 'hour' && getAnalyticsRangeDaysInclusive(timeRange.value) !== 1) {
+        timeRange.value = {
+          ...timeRange.value,
+          granularity: 'auto',
+        }
+      }
+    },
+  )
+
+  async function load() {
+    const requestId = guard.next()
+    loading.value = true
+    loadError.value = false
+    try {
+      const params = buildTimeRangeParams(timeRange.value)
+      const timeseriesParams = timeseriesTimeRange.value
+      const basePayload = {
+        scope,
+        time_range: params,
+        filters: {
+          user_ids: userFilter.value,
+          api_key_ids: apiKeyFilter.value,
+        },
+      }
+
+      const [overviewData, timeseriesData, modelBreakdownData] = await Promise.all([
+        analyticsApi.getOverview(basePayload),
+        analyticsApi.getTimeseries({
+          ...basePayload,
+          time_range: timeseriesParams,
+        }),
+        analyticsApi.getBreakdown({
+          ...basePayload,
+          dimension: 'model',
+          limit: 50,
+        }),
+      ])
+
+      if (guard.isStale(requestId)) return
+
+      const filledBuckets = fillMissingTimeseriesBuckets(timeseriesData.buckets, timeseriesParams)
+
+      summary.value = overviewData.summary
+      bucketStats.value = filledBuckets.map(mapBucketToReportStat)
+      bucketBreakdowns.value = Object.fromEntries(
+        filledBuckets.map(bucket => {
+          const breakdown = mapBucketToBreakdown(bucket)
+          return [breakdown.date, breakdown]
+        }),
+      )
+      modelSummary.value = modelBreakdownData.rows.map(mapBreakdownRowToModelSummary)
+      hasLoaded.value = true
+    } catch {
+      if (guard.isStale(requestId)) return
+      loadError.value = true
+    } finally {
+      if (guard.isCurrent(requestId)) {
+        loading.value = false
+      }
+    }
+  }
+
+  async function loadApiKeys() {
+    if (!loadApiKeyOptions) return
+    const requestId = apiKeysGuard.next()
+    const filterOptions = await analyticsApi.getFilterOptions({
+      scope,
+      time_range: buildTimeRangeParams(timeRange.value),
+      filters: {},
+    }).catch(() => null)
+    if (apiKeysGuard.isStale(requestId)) return
+
+    const nextOptions = filterOptions?.api_keys ?? []
+
+    apiKeyOptions.value = nextOptions
+    const validOptionValues = new Set(nextOptions.map(option => option.value))
+    const nextFilter = apiKeyFilter.value.filter(value => validOptionValues.has(value))
+    if (!areStringArraysEqual(nextFilter, apiKeyFilter.value)) {
+      suppressNextApiKeyFilterLoad = true
+      apiKeyFilter.value = nextFilter
+    }
+  }
+
+  const loader = createLoader(load)
+  const apiKeysLoader = createLoader(loadApiKeys)
+
+  watch(
+    timeRange,
+    () => {
+      void (async () => {
+        if (loadApiKeyOptions) {
+          await apiKeysLoader.execute()
+        }
+        loader.schedule()
+      })()
+    },
+    { deep: true },
+  )
+
+  watch(
+    [userFilter, apiKeyFilter],
+    () => {
+      if (suppressNextApiKeyFilterLoad) {
+        suppressNextApiKeyFilterLoad = false
+        return
+      }
+      loader.schedule()
+    },
+    { deep: true },
+  )
+
+  onMounted(() => {
+    void (async () => {
+      if (loadApiKeyOptions) {
+        await apiKeysLoader.execute()
+      }
+      await loader.execute()
+    })()
+  })
+  onUnmounted(() => {
+    loader.cleanup()
+    apiKeysLoader.cleanup()
+    guard.invalidate()
+    apiKeysGuard.invalidate()
+  })
+
+  return {
+    timeRange,
+    apiKeyFilter,
+    apiKeyOptions,
+    summary,
+    bucketStats,
+    bucketBreakdowns,
+    modelSummary,
+    loading,
+    loadError,
+    hasLoaded,
+    resolvedGranularity,
+  }
+}

frontend/src/composables/useRouteQuery.ts

@@ -0,0 +1,44 @@
+import { useRoute, useRouter, type LocationQueryValue, type LocationQuery } from 'vue-router'
+
+type QueryValue = LocationQueryValue | LocationQueryValue[]
+
+function normalizeQueryValue(value: QueryValue): string | undefined {
+  if (Array.isArray(value)) {
+    return value.length > 0 ? (value[value.length - 1] ?? undefined) : undefined
+  }
+  return typeof value === 'string' ? value : undefined
+}
+
+function queriesEqual(left: LocationQuery, right: LocationQuery): boolean {
+  const keys = new Set([...Object.keys(left), ...Object.keys(right)])
+  for (const key of keys) {
+    if (normalizeQueryValue(left[key]) !== normalizeQueryValue(right[key])) {
+      return false
+    }
+  }
+  return true
+}
+
+export function useRouteQuery() {
+  const route = useRoute()
+  const router = useRouter()
+
+  function getQueryValue(key: string): string | undefined {
+    return normalizeQueryValue(route.query[key])
+  }
+
+  function patchQuery(patch: Record<string, string | undefined | null>) {
+    const next: LocationQuery = { ...route.query }
+    for (const [key, value] of Object.entries(patch)) {
+      if (value == null || value.trim() === '') {
+        delete next[key]
+      } else {
+        next[key] = value
+      }
+    }
+    if (queriesEqual(route.query, next)) return
+    void router.replace({ query: next }).catch(() => {})
+  }
+
+  return { route, router, getQueryValue, patchQuery }
+}

frontend/src/features/providers/components/HealthMonitorCard.vue

@@ -3,18 +3,98 @@
     variant="default"
     class="overflow-hidden"
   >
-    <!-- 标题和筛选器 -->
-    <div class="px-6 py-3.5 border-b border-border/60">
-      <div class="flex items-center justify-between gap-4">
+    <div class="border-b border-border/60 bg-muted/20 px-6 py-5">
+      <div class="flex flex-col gap-4">
+        <div class="flex flex-col gap-3 xl:flex-row xl:items-start xl:justify-between">
+          <div class="space-y-1">
+            <div class="flex items-center gap-2 flex-wrap">
               <h3 class="text-base font-semibold">
                 {{ title }}
               </h3>
-        <div class="flex items-center gap-3">
-          <Label class="text-xs text-muted-foreground">回溯时间：</Label>
-          <Select
-            v-model="lookbackHours"
+              <Badge
+                variant="outline"
+                class="text-[10px] uppercase tracking-[0.25em]"
               >
-            <SelectTrigger class="w-28 h-8 text-xs border-border/60">
+                {{ isAdmin ? 'Admin' : 'Public' }}
+              </Badge>
+            </div>
+            <p class="text-xs text-muted-foreground">
+              异常优先排序，支持状态筛选、详情诊断与自动刷新，首屏先看出问题的格式。
+            </p>
+          </div>
+
+          <div class="flex flex-wrap items-center gap-3">
+            <div class="text-xs text-muted-foreground">
+              最后刷新 {{ generatedAt ? formatRelativeTime(generatedAt) : '未加载' }}
+            </div>
+            <div class="flex items-center gap-2 rounded-xl border border-border/60 bg-card/70 px-3 py-1.5">
+              <Switch v-model="autoRefreshEnabled" />
+              <span class="text-xs font-medium">
+                {{ autoRefreshEnabled ? '自动刷新 15s' : '自动刷新已关闭' }}
+              </span>
+            </div>
+            <Button
+              v-if="isAdmin"
+              variant="outline"
+              size="sm"
+              class="h-9 px-3 text-xs"
+              :disabled="recoveringAllKeys || (healthSummary?.keys.circuit_open ?? 0) <= 0"
+              @click="recoverAllConfirmOpen = true"
+            >
+              <Loader2
+                v-if="recoveringAllKeys"
+                class="mr-2 h-3.5 w-3.5 animate-spin"
+              />
+              恢复全部熔断 Key
+            </Button>
+            <RefreshButton
+              :loading="loadingMonitors"
+              @click="refreshData({ silent: false })"
+            />
+          </div>
+        </div>
+
+        <div class="grid grid-cols-2 gap-3 xl:grid-cols-6">
+          <Card
+            v-for="item in summaryCards"
+            :key="item.label"
+            variant="inset"
+            class="p-3"
+          >
+            <div class="flex items-start justify-between gap-3">
+              <div class="min-w-0">
+                <div class="text-[11px] font-semibold uppercase tracking-[0.18em] text-muted-foreground">
+                  {{ item.label }}
+                </div>
+                <div
+                  class="mt-2 text-xl font-semibold"
+                  :class="item.valueClass"
+                >
+                  {{ item.value }}
+                </div>
+                <div class="mt-1 text-[11px] text-muted-foreground">
+                  {{ item.help }}
+                </div>
+              </div>
+              <div
+                class="rounded-xl border border-border/60 p-2"
+                :class="item.iconClass"
+              >
+                <component
+                  :is="item.icon"
+                  class="h-4 w-4"
+                />
+              </div>
+            </div>
+          </Card>
+        </div>
+
+        <div class="flex flex-col gap-3 xl:flex-row xl:items-center xl:justify-between">
+          <div class="grid flex-1 grid-cols-1 gap-3 sm:grid-cols-2 xl:grid-cols-4">
+            <div>
+              <Label class="mb-1.5 block text-xs text-muted-foreground">回溯时间</Label>
+              <Select v-model="lookbackHours">
+                <SelectTrigger class="h-9 border-border/60 text-xs">
                   <SelectValue />
                 </SelectTrigger>
                 <SelectContent>
@@ -35,18 +115,119 @@
                   </SelectItem>
                 </SelectContent>
               </Select>
-          <RefreshButton
-            :loading="loading"
-            @click="refreshData"
-          />
+            </div>
+
+            <div>
+              <Label class="mb-1.5 block text-xs text-muted-foreground">状态筛选</Label>
+              <Select v-model="statusFilter">
+                <SelectTrigger class="h-9 border-border/60 text-xs">
+                  <SelectValue />
+                </SelectTrigger>
+                <SelectContent>
+                  <SelectItem value="all">
+                    全部状态
+                  </SelectItem>
+                  <SelectItem value="unhealthy">
+                    仅看异常
+                  </SelectItem>
+                  <SelectItem value="warning">
+                    仅看预警
+                  </SelectItem>
+                  <SelectItem value="healthy">
+                    仅看健康
+                  </SelectItem>
+                  <SelectItem value="unknown">
+                    仅看无流量
+                  </SelectItem>
+                </SelectContent>
+              </Select>
+            </div>
+
+            <div>
+              <Label class="mb-1.5 block text-xs text-muted-foreground">格式族</Label>
+              <Select v-model="familyFilter">
+                <SelectTrigger class="h-9 border-border/60 text-xs">
+                  <SelectValue />
+                </SelectTrigger>
+                <SelectContent>
+                  <SelectItem value="all">
+                    全部格式
+                  </SelectItem>
+                  <SelectItem value="claude">
+                    Claude
+                  </SelectItem>
+                  <SelectItem value="openai">
+                    OpenAI
+                  </SelectItem>
+                  <SelectItem value="gemini">
+                    Gemini
+                  </SelectItem>
+                  <SelectItem value="other">
+                    其他
+                  </SelectItem>
+                </SelectContent>
+              </Select>
+            </div>
+
+            <div>
+              <Label class="mb-1.5 block text-xs text-muted-foreground">显示范围</Label>
+              <div class="flex h-9 items-center justify-between rounded-xl border border-border/60 bg-card/70 px-3">
+                <span class="text-xs font-medium">仅看有流量</span>
+                <Switch v-model="onlyTraffic" />
+              </div>
+            </div>
+          </div>
+
+          <div class="flex items-center gap-2 text-xs text-muted-foreground">
+            <span>显示 {{ filteredMonitors.length }} / {{ monitors.length }} 个格式</span>
+            <Button
+              v-if="hasActiveFilters"
+              variant="ghost"
+              size="sm"
+              class="h-8 px-2 text-xs"
+              @click="resetFilters"
+            >
+              重置筛选
+            </Button>
+          </div>
+        </div>
+
+        <div class="flex flex-wrap items-center gap-2 text-xs">
+          <Badge
+            variant="success"
+            class="gap-1"
+          >
+            健康
+          </Badge>
+          <span class="text-muted-foreground">成功率 ≥ 95%</span>
+          <Badge
+            variant="warning"
+            class="gap-1"
+          >
+            预警
+          </Badge>
+          <span class="text-muted-foreground">70% - 95%</span>
+          <Badge
+            variant="destructive"
+            class="gap-1"
+          >
+            异常
+          </Badge>
+          <span class="text-muted-foreground">成功率 &lt; 70%</span>
+          <Badge
+            variant="outline"
+            class="gap-1"
+          >
+            无流量
+          </Badge>
+          <span class="text-muted-foreground">窗口内没有请求</span>
         </div>
       </div>
     </div>
 
-    <!-- 内容区域 -->
     <div class="p-6">
       <div
-        v-if="loadingMonitors"
+        v-if="showInitialLoading"
         class="flex items-center justify-center py-12"
       >
         <Loader2 class="w-6 h-6 animate-spin text-muted-foreground" />
@@ -57,62 +238,104 @@
         v-else-if="monitors.length === 0"
         class="flex flex-col items-center justify-center py-12 text-muted-foreground"
       >
-        <Activity class="w-12 h-12 mb-3 opacity-30" />
+        <Activity class="mb-3 h-12 w-12 opacity-30" />
         <p>暂无健康监控数据</p>
-        <p class="text-xs mt-1">
+        <p class="mt-1 text-xs">
           端点尚未产生请求记录
         </p>
       </div>
 
+      <div
+        v-else-if="filteredMonitors.length === 0"
+        class="flex flex-col items-center justify-center py-12 text-muted-foreground"
+      >
+        <AlertTriangle class="mb-3 h-12 w-12 opacity-30" />
+        <p>当前筛选条件下没有匹配项</p>
+        <p class="mt-1 text-xs">
+          可以放宽状态、格式族或流量条件
+        </p>
+        <Button
+          variant="outline"
+          size="sm"
+          class="mt-4"
+          @click="resetFilters"
+        >
+          重置筛选
+        </Button>
+      </div>
+
       <div
         v-else
         class="space-y-3"
       >
         <div
-          v-for="monitor in monitors"
+          v-for="monitor in filteredMonitors"
           :key="monitor.api_format"
-          class="border border-border/60 rounded-lg p-4 hover:border-primary/50 transition-colors"
+          class="rounded-xl border p-4 transition-colors"
+          :class="getMonitorCardClass(monitor)"
         >
-          <!-- 响应式布局：窄屏上下两行，宽屏左右结构 -->
-          <div class="flex flex-col sm:flex-row sm:gap-6 sm:items-center">
-            <!-- 第一行/左侧：信息区域 -->
-            <div class="sm:w-52 flex-shrink-0 space-y-1.5 mb-3 sm:mb-0">
-              <!-- API 格式标签和成功率 -->
+          <div class="flex flex-col gap-3 xl:flex-row xl:items-center xl:gap-6">
+            <div class="space-y-1.5 xl:w-64 xl:flex-shrink-0">
               <div class="flex items-center gap-2 flex-wrap">
                 <Badge
                   variant="outline"
-                  class="font-mono text-xs whitespace-nowrap"
+                  class="whitespace-nowrap font-mono text-xs"
                 >
                   {{ formatApiFormat(monitor.api_format) }}
                 </Badge>
+                <Badge
+                  :variant="getMonitorStatusVariant(monitor)"
+                  class="text-xs whitespace-nowrap"
+                >
+                  {{ getMonitorStatusLabel(monitor) }}
+                </Badge>
                 <Badge
                   v-if="monitor.total_attempts > 0"
                   :variant="getSuccessRateVariant(monitor.success_rate)"
                   class="text-xs whitespace-nowrap"
                 >
-                  {{ (monitor.success_rate * 100).toFixed(0) }}%
+                  {{ formatPercent(monitor.success_rate) }}
                 </Badge>
-                <!-- 提供商信息（仅管理员可见）- 窄屏时显示在同一行 -->
                 <span
                   v-if="showProviderInfo && 'provider_count' in monitor"
-                  class="text-xs text-muted-foreground sm:hidden"
+                  class="text-xs text-muted-foreground xl:hidden"
                 >
                   {{ monitor.provider_count }} 个提供商 / {{ monitor.key_count }} 个密钥
                 </span>
               </div>
 
-              <!-- 提供商信息（仅管理员可见）- 宽屏时显示在下方 -->
               <div
                 v-if="showProviderInfo && 'provider_count' in monitor"
-                class="text-xs text-muted-foreground hidden sm:block"
+                class="hidden text-xs text-muted-foreground xl:block"
               >
                 {{ monitor.provider_count }} 个提供商 / {{ monitor.key_count }} 个密钥
               </div>
+
+              <div class="text-xs text-muted-foreground">
+                {{ monitor.total_attempts }} 次请求
+                <span class="mx-1">·</span>
+                失败 {{ monitor.failed_count }}
+                <span class="mx-1">·</span>
+                跳过 {{ monitor.skipped_count }}
               </div>
 
-            <!-- 第二行/右侧：时间线区域 -->
-            <div class="flex-1 min-w-0 sm:flex sm:justify-end">
-              <div class="w-full sm:max-w-5xl">
+              <div class="text-xs text-muted-foreground">
+                最近事件 {{ monitor.last_event_at ? formatRelativeTime(monitor.last_event_at) : '暂无' }}
+              </div>
+
+              <div class="pt-1">
+                <Button
+                  variant="outline"
+                  size="sm"
+                  class="h-8 px-3 text-xs"
+                  @click="openMonitorDetail(monitor)"
+                >
+                  查看详情
+                </Button>
+              </div>
+            </div>
+
+            <div class="min-w-0 flex-1">
               <EndpointHealthTimeline
                 :monitor="monitor"
                 :lookback-hours="parseInt(lookbackHours)"
@@ -122,16 +345,42 @@
         </div>
       </div>
     </div>
-    </div>
   </Card>
+
+  <HealthMonitorDetailDialog
+    v-model="detailDialogOpen"
+    :monitor="selectedMonitor"
+    :is-admin="isAdmin"
+    :format-keys="selectedFormatKeys"
+    :loading-format-keys="loadingFormatKeys"
+    :recovering-key-id="recoveringKeyId"
+    @navigate-provider="goToProviderManagement"
+    @navigate-pool="goToPoolManagement"
+    @recover-key="recoverFormatKey"
+  />
+
+  <AlertDialog
+    v-model="recoverAllConfirmOpen"
+    title="恢复全部熔断 Key"
+    :description="recoverAllDialogDescription"
+    type="warning"
+    confirm-text="立即恢复"
+    :loading="recoveringAllKeys"
+    @confirm="confirmRecoverAllKeys"
+  />
 </template>
 
 <script setup lang="ts">
-import { ref, onMounted, watch } from 'vue'
-import { Activity, Loader2 } from 'lucide-vue-next'
+import { computed, onMounted, onUnmounted, ref, watch, type Component } from 'vue'
+import { useRouter } from 'vue-router'
+import { Activity, AlertTriangle, Gauge, KeyRound, Loader2, Radio, ShieldAlert } from 'lucide-vue-next'
+import AlertDialog from '@/components/common/AlertDialog.vue'
+import HealthMonitorDetailDialog from './HealthMonitorDetailDialog.vue'
 import Card from '@/components/ui/card.vue'
 import Badge from '@/components/ui/badge.vue'
+import Button from '@/components/ui/button.vue'
 import Label from '@/components/ui/label.vue'
+import Switch from '@/components/ui/switch.vue'
 import Select from '@/components/ui/select.vue'
 import SelectTrigger from '@/components/ui/select-trigger.vue'
 import SelectValue from '@/components/ui/select-value.vue'
@@ -139,12 +388,24 @@ import SelectContent from '@/components/ui/select-content.vue'
 import SelectItem from '@/components/ui/select-item.vue'
 import RefreshButton from '@/components/ui/refresh-button.vue'
 import EndpointHealthTimeline from './EndpointHealthTimeline.vue'
-import { getEndpointStatusMonitor, getPublicEndpointStatusMonitor } from '@/api/endpoints/health'
-import type { EndpointStatusMonitor, PublicEndpointStatusMonitor } from '@/api/endpoints/types'
+import { getKeysGroupedByFormat, type GroupedFormatKey } from '@/api/endpoints/keys'
+import { getEndpointStatusMonitor, getHealthSummary, getPublicEndpointStatusMonitor, recoverAllKeysHealth, recoverKeyHealth } from '@/api/endpoints/health'
+import type { EndpointStatusMonitor, HealthSummary, PublicEndpointStatusMonitor } from '@/api/endpoints/types'
 import { useToast } from '@/composables/useToast'
 import { parseApiError } from '@/utils/errorParser'
 import { formatApiFormat } from '@/api/endpoints/types/api-format'
 
+type MonitorItem = EndpointStatusMonitor | PublicEndpointStatusMonitor
+type MonitorStatus = 'healthy' | 'warning' | 'unhealthy' | 'unknown'
+type SummaryCardItem = {
+  label: string
+  value: string
+  help: string
+  icon: Component
+  valueClass: string
+  iconClass: string
+}
+
 const props = withDefaults(defineProps<{
   title?: string
   isAdmin?: boolean
@@ -155,55 +416,510 @@ const props = withDefaults(defineProps<{
   showProviderInfo: false
 })
 
-const { error: showError } = useToast()
+const AUTO_REFRESH_INTERVAL_MS = 15_000
+const STATUS_SORT_ORDER: Record<MonitorStatus, number> = {
+  unhealthy: 0,
+  warning: 1,
+  unknown: 2,
+  healthy: 3
+}
+
+const router = useRouter()
+const { error: showError, success: showSuccess, warning: showWarning } = useToast()
 
-const loading = ref(false)
 const loadingMonitors = ref(false)
-const monitors = ref<(EndpointStatusMonitor | PublicEndpointStatusMonitor)[]>([])
+const monitors = ref<MonitorItem[]>([])
+const generatedAt = ref<string | null>(null)
+const healthSummary = ref<HealthSummary | null>(null)
 const lookbackHours = ref('6')
+const statusFilter = ref<'all' | MonitorStatus>('all')
+const familyFilter = ref<'all' | 'claude' | 'openai' | 'gemini' | 'other'>('all')
+const onlyTraffic = ref(false)
+const autoRefreshEnabled = ref(true)
+const detailDialogOpen = ref(false)
+const selectedMonitorFormat = ref<string | null>(null)
+const recoverAllConfirmOpen = ref(false)
+const recoveringAllKeys = ref(false)
+const keysByFormat = ref<Record<string, GroupedFormatKey[]>>({})
+const loadingFormatKeys = ref(false)
+const recoveringKeyId = ref<string | null>(null)
 
-async function loadMonitors() {
+let autoRefreshTimer: ReturnType<typeof setInterval> | null = null
+let loadRequestId = 0
+
+const isAdmin = computed(() => props.isAdmin)
+const showInitialLoading = computed(() => loadingMonitors.value && monitors.value.length === 0)
+
+const unhealthyCount = computed(() =>
+  monitors.value.filter(item => getMonitorStatus(item) === 'unhealthy').length
+)
+
+const warningCount = computed(() =>
+  monitors.value.filter(item => getMonitorStatus(item) === 'warning').length
+)
+
+const trafficCount = computed(() =>
+  monitors.value.filter(item => item.total_attempts > 0).length
+)
+
+const overallSuccessRate = computed(() => {
+  const totals = monitors.value.reduce((acc, item) => {
+    acc.success += item.success_count
+    acc.failed += item.failed_count
+    return acc
+  }, { success: 0, failed: 0 })
+
+  const denominator = totals.success + totals.failed
+  if (denominator <= 0) return null
+  return totals.success / denominator
+})
+
+const summaryCards = computed<SummaryCardItem[]>(() => {
+  const items: SummaryCardItem[] = [
+    {
+      label: '异常格式',
+      value: `${unhealthyCount.value}`,
+      help: `共 ${monitors.value.length} 个格式`,
+      icon: AlertTriangle,
+      valueClass: unhealthyCount.value > 0 ? 'text-destructive' : 'text-foreground',
+      iconClass: unhealthyCount.value > 0 ? 'bg-destructive/10 text-destructive' : 'bg-muted/60 text-muted-foreground'
+    },
+    {
+      label: '预警格式',
+      value: `${warningCount.value}`,
+      help: '建议优先观察',
+      icon: Radio,
+      valueClass: warningCount.value > 0 ? 'text-yellow-600 dark:text-yellow-400' : 'text-foreground',
+      iconClass: warningCount.value > 0 ? 'bg-yellow-500/10 text-yellow-600 dark:text-yellow-400' : 'bg-muted/60 text-muted-foreground'
+    },
+    {
+      label: '有流量格式',
+      value: `${trafficCount.value}`,
+      help: `最近 ${lookbackHours.value} 小时`,
+      icon: Activity,
+      valueClass: 'text-foreground',
+      iconClass: 'bg-primary/10 text-primary'
+    },
+    {
+      label: '整体成功率',
+      value: overallSuccessRate.value == null ? '--' : formatPercent(overallSuccessRate.value),
+      help: '按 success / (success + failed)',
+      icon: Gauge,
+      valueClass: overallSuccessRate.value != null && overallSuccessRate.value < 0.95
+        ? 'text-yellow-600 dark:text-yellow-400'
+        : 'text-foreground',
+      iconClass: 'bg-primary/10 text-primary'
+    }
+  ]
+
+  if (props.isAdmin) {
+    items.push(
+      {
+        label: '熔断 Key',
+        value: `${healthSummary.value?.keys.circuit_open ?? 0}`,
+        help: '已打开熔断器',
+        icon: ShieldAlert,
+        valueClass: (healthSummary.value?.keys.circuit_open ?? 0) > 0 ? 'text-destructive' : 'text-foreground',
+        iconClass: (healthSummary.value?.keys.circuit_open ?? 0) > 0 ? 'bg-destructive/10 text-destructive' : 'bg-muted/60 text-muted-foreground'
+      },
+      {
+        label: '活跃 Key',
+        value: `${healthSummary.value?.keys.active ?? 0}`,
+        help: '当前可调度密钥',
+        icon: KeyRound,
+        valueClass: 'text-foreground',
+        iconClass: 'bg-primary/10 text-primary'
+      }
+    )
+  }
+
+  return items
+})
+
+const filteredMonitors = computed(() => {
+  const filtered = monitors.value.filter((monitor) => {
+    const status = getMonitorStatus(monitor)
+    const family = getMonitorFamily(monitor.api_format)
+
+    if (statusFilter.value !== 'all' && status !== statusFilter.value) {
+      return false
+    }
+
+    if (familyFilter.value !== 'all' && family !== familyFilter.value) {
+      return false
+    }
+
+    if (onlyTraffic.value && monitor.total_attempts <= 0) {
+      return false
+    }
+
+    return true
+  })
+
+  return filtered.sort((left, right) => {
+    const statusGap = STATUS_SORT_ORDER[getMonitorStatus(left)] - STATUS_SORT_ORDER[getMonitorStatus(right)]
+    if (statusGap !== 0) return statusGap
+
+    const leftTime = left.last_event_at ? new Date(left.last_event_at).getTime() : 0
+    const rightTime = right.last_event_at ? new Date(right.last_event_at).getTime() : 0
+    if (leftTime !== rightTime) return rightTime - leftTime
+
+    if (left.total_attempts !== right.total_attempts) {
+      return right.total_attempts - left.total_attempts
+    }
+
+    return left.api_format.localeCompare(right.api_format)
+  })
+})
+
+const hasActiveFilters = computed(() =>
+  statusFilter.value !== 'all' || familyFilter.value !== 'all' || onlyTraffic.value
+)
+
+const selectedMonitor = computed<MonitorItem | null>(() => {
+  if (!selectedMonitorFormat.value) return null
+  return monitors.value.find(item => item.api_format === selectedMonitorFormat.value) ?? null
+})
+
+
+const selectedFormatKeys = computed<GroupedFormatKey[]>(() => {
+  if (!selectedMonitorFormat.value) return []
+  const keys = keysByFormat.value[selectedMonitorFormat.value] ?? []
+  return [...keys].sort((left, right) => {
+    const circuitGap = Number(right.circuit_breaker_open) - Number(left.circuit_breaker_open)
+    if (circuitGap !== 0) return circuitGap
+
+    const leftHealth = left.health_score ?? 1
+    const rightHealth = right.health_score ?? 1
+    if (leftHealth !== rightHealth) return leftHealth - rightHealth
+
+    if (left.request_count !== right.request_count) {
+      return right.request_count - left.request_count
+    }
+
+    return left.name.localeCompare(right.name)
+  })
+})
+
+
+const recoverAllDialogDescription = computed(() => {
+  const circuitOpen = healthSummary.value?.keys.circuit_open ?? 0
+  return [
+    '这会重置所有已熔断 Key 的健康状态并关闭熔断器。',
+    `当前检测到 **${circuitOpen}** 个熔断 Key。`,
+    '建议在确认上游服务恢复后执行。'
+  ].join('\n')
+})
+
+async function loadMonitors(silent = false) {
+  const currentRequestId = ++loadRequestId
   loadingMonitors.value = true
+
   try {
     const params = {
       lookback_hours: parseInt(lookbackHours.value),
-      per_format_limit: 100
+      per_format_limit: 60
+    }
+
+    const monitorPromise = props.isAdmin
+      ? getEndpointStatusMonitor(params)
+      : getPublicEndpointStatusMonitor(params)
+
+    const [monitorResult, summaryResult] = await Promise.allSettled([
+      monitorPromise,
+      props.isAdmin ? getHealthSummary() : Promise.resolve(null)
+    ])
+
+    if (currentRequestId !== loadRequestId) return
+
+    if (monitorResult.status === 'fulfilled') {
+      monitors.value = monitorResult.value.formats || []
+      generatedAt.value = monitorResult.value.generated_at ?? null
+    } else if (!silent) {
+      showError(parseApiError(monitorResult.reason, '加载健康监控数据失败'), '错误')
     }
 
     if (props.isAdmin) {
-      const data = await getEndpointStatusMonitor(params)
-      monitors.value = data.formats || []
+      if (summaryResult.status === 'fulfilled') {
+        healthSummary.value = summaryResult.value
+      }
     } else {
-      const data = await getPublicEndpointStatusMonitor(params)
-      monitors.value = data.formats || []
+      healthSummary.value = null
+    }
+
+    if (selectedMonitorFormat.value && !monitors.value.some(item => item.api_format === selectedMonitorFormat.value)) {
+      detailDialogOpen.value = false
+      selectedMonitorFormat.value = null
     }
   } catch (err: unknown) {
+    if (!silent) {
       showError(parseApiError(err, '加载健康监控数据失败'), '错误')
+    }
   } finally {
+    if (currentRequestId === loadRequestId) {
       loadingMonitors.value = false
     }
   }
+}
 
-async function refreshData() {
-  loading.value = true
+async function refreshData(options: { silent?: boolean } = {}) {
+  await loadMonitors(Boolean(options.silent))
+}
+
+async function loadKeysByFormat(force = false) {
+  if (!props.isAdmin) return
+  if (!force && Object.keys(keysByFormat.value).length > 0) return
+
+  loadingFormatKeys.value = true
   try {
-    await loadMonitors()
+    keysByFormat.value = await getKeysGroupedByFormat()
+  } catch (err: unknown) {
+    showError(parseApiError(err, '加载 Key 诊断数据失败'), '错误')
   } finally {
-    loading.value = false
+    loadingFormatKeys.value = false
   }
 }
 
-function getSuccessRateVariant(rate: number): 'default' | 'secondary' | 'destructive' | 'outline' {
-  if (rate >= 0.95) return 'default'
-  if (rate >= 0.8) return 'secondary'
-  return 'destructive'
+function resetFilters() {
+  statusFilter.value = 'all'
+  familyFilter.value = 'all'
+  onlyTraffic.value = false
 }
 
+
+function stopAutoRefresh() {
+  if (!autoRefreshTimer) return
+  clearInterval(autoRefreshTimer)
+  autoRefreshTimer = null
+}
+
+function startAutoRefresh() {
+  stopAutoRefresh()
+  if (!autoRefreshEnabled.value) return
+
+  autoRefreshTimer = setInterval(() => {
+    if (document.visibilityState === 'hidden') return
+    void refreshData({ silent: true })
+  }, AUTO_REFRESH_INTERVAL_MS)
+}
+
+function handleVisibilityChange() {
+  if (document.visibilityState === 'visible' && autoRefreshEnabled.value) {
+    void refreshData({ silent: true })
+  }
+}
+
+async function openMonitorDetail(monitor: MonitorItem) {
+  selectedMonitorFormat.value = monitor.api_format
+  detailDialogOpen.value = true
+  if (props.isAdmin) {
+    await loadKeysByFormat()
+  }
+}
+
+async function confirmRecoverAllKeys() {
+  if (recoveringAllKeys.value) return
+
+  recoveringAllKeys.value = true
+  try {
+    const result = await recoverAllKeysHealth()
+    recoverAllConfirmOpen.value = false
+
+    if (result.recovered_count > 0) {
+      showSuccess(result.message || `已恢复 ${result.recovered_count} 个熔断 Key`)
+    } else {
+      showWarning(result.message || '当前没有需要恢复的熔断 Key')
+    }
+
+    await refreshData({ silent: false })
+  } catch (err: unknown) {
+    showError(parseApiError(err, '批量恢复熔断 Key 失败'), '错误')
+  } finally {
+    recoveringAllKeys.value = false
+  }
+}
+
+async function recoverFormatKey(key: GroupedFormatKey) {
+  if (recoveringKeyId.value || !selectedMonitorFormat.value) return
+
+  recoveringKeyId.value = key.id
+  try {
+    const result = await recoverKeyHealth(key.id, selectedMonitorFormat.value)
+    showSuccess(result.message || '当前格式已恢复')
+    await Promise.all([
+      refreshData({ silent: true }),
+      loadKeysByFormat(true)
+    ])
+  } catch (err: unknown) {
+    showError(parseApiError(err, '恢复 Key 健康状态失败'), '错误')
+  } finally {
+    recoveringKeyId.value = null
+  }
+}
+
+
+function goToProviderManagement(providerId: string) {
+  void router.push({
+    name: 'ProviderManagement',
+    query: { providerId }
+  })
+}
+
+function goToPoolManagement(providerId: string, search?: string) {
+  const query: Record<string, string> = { providerId }
+  const normalizedSearch = search?.trim()
+  if (normalizedSearch) {
+    query.search = normalizedSearch
+  }
+
+  void router.push({
+    name: 'PoolManagement',
+    query
+  })
+}
+
+function getMonitorFamily(apiFormat: string): 'claude' | 'openai' | 'gemini' | 'other' {
+  if (apiFormat.startsWith('claude:')) return 'claude'
+  if (apiFormat.startsWith('openai:')) return 'openai'
+  if (apiFormat.startsWith('gemini:')) return 'gemini'
+  return 'other'
+}
+
+function getLatestTimelineStatus(monitor: MonitorItem): MonitorStatus | null {
+  const timeline = monitor.timeline ?? []
+  for (let index = timeline.length - 1; index >= 0; index -= 1) {
+    const status = timeline[index]
+    if (status === 'healthy' || status === 'warning' || status === 'unhealthy') {
+      return status
+    }
+  }
+  return null
+}
+
+function getMonitorStatus(monitor: MonitorItem): MonitorStatus {
+  const timelineStatus = getLatestTimelineStatus(monitor)
+  if (timelineStatus === 'unhealthy') return 'unhealthy'
+  if (timelineStatus === 'warning') return 'warning'
+
+  if (monitor.total_attempts <= 0) {
+    return 'unknown'
+  }
+
+  if (monitor.success_rate < 0.7) return 'unhealthy'
+  if (monitor.success_rate < 0.95) return 'warning'
+  if (timelineStatus === 'healthy' || monitor.success_rate >= 0.95) return 'healthy'
+  return 'unknown'
+}
+
+function getMonitorStatusLabel(monitor: MonitorItem): string {
+  switch (getMonitorStatus(monitor)) {
+    case 'unhealthy':
+      return '异常'
+    case 'warning':
+      return '预警'
+    case 'healthy':
+      return '健康'
+    default:
+      return '无流量'
+  }
+}
+
+function getMonitorStatusVariant(monitor: MonitorItem): 'success' | 'warning' | 'destructive' | 'outline' {
+  switch (getMonitorStatus(monitor)) {
+    case 'healthy':
+      return 'success'
+    case 'warning':
+      return 'warning'
+    case 'unhealthy':
+      return 'destructive'
+    default:
+      return 'outline'
+  }
+}
+
+function getMonitorCardClass(monitor: MonitorItem): string {
+  switch (getMonitorStatus(monitor)) {
+    case 'unhealthy':
+      return 'border-red-300/70 bg-red-500/5 hover:border-red-400/80'
+    case 'warning':
+      return 'border-yellow-300/70 bg-yellow-500/5 hover:border-yellow-400/80'
+    case 'healthy':
+      return 'border-border/60 hover:border-primary/50'
+    default:
+      return 'border-border/40 bg-muted/20 hover:border-border/70'
+  }
+}
+
+function getSuccessRateVariant(rate: number): 'success' | 'warning' | 'destructive' | 'outline' {
+  if (rate >= 0.95) return 'success'
+  if (rate >= 0.8) return 'warning'
+  if (rate >= 0) return 'destructive'
+  return 'outline'
+}
+
+
+function formatPercent(rate: number): string {
+  return `${(rate * 100).toFixed(0)}%`
+}
+
+
+
+function formatRelativeTime(dateString: string): string {
+  const diff = Date.now() - new Date(dateString).getTime()
+  if (Number.isNaN(diff)) return '未知'
+  if (diff < 60_000) return '刚刚'
+
+  const minutes = Math.floor(diff / 60_000)
+  if (minutes < 60) return `${minutes} 分钟前`
+
+  const hours = Math.floor(minutes / 60)
+  if (hours < 24) return `${hours} 小时前`
+
+  const days = Math.floor(hours / 24)
+  if (days < 7) return `${days} 天前`
+
+  return formatAbsoluteTime(dateString)
+}
+
+function formatAbsoluteTime(dateString: string): string {
+  return new Date(dateString).toLocaleString('zh-CN', {
+    month: '2-digit',
+    day: '2-digit',
+    hour: '2-digit',
+    minute: '2-digit',
+    second: '2-digit'
+  })
+}
+
+
 watch(lookbackHours, () => {
-  loadMonitors()
+  void refreshData({ silent: false })
 })
 
+watch(() => props.isAdmin, () => {
+  keysByFormat.value = {}
+  void refreshData({ silent: false })
+})
+
+watch(autoRefreshEnabled, () => {
+  startAutoRefresh()
+})
+
+watch(detailDialogOpen, (open) => {
+  if (!open) {
+    selectedMonitorFormat.value = null
+  }
+})
+
+
 onMounted(() => {
-  refreshData()
+  document.addEventListener('visibilitychange', handleVisibilityChange)
+  startAutoRefresh()
+  void refreshData({ silent: false })
+})
+
+onUnmounted(() => {
+  document.removeEventListener('visibilitychange', handleVisibilityChange)
+  stopAutoRefresh()
 })
 </script>

frontend/src/features/providers/components/HealthMonitorDetailDialog.vue

@@ -0,0 +1,936 @@
+<template>
+  <Dialog
+    v-model="open"
+    size="5xl"
+    close-on-backdrop
+    :title="monitor ? `${formatApiFormat(monitor.api_format)} 详情` : '监控详情'"
+    :description="dialogDescription"
+  >
+    <template #header-actions>
+      <Button
+        variant="ghost"
+        size="icon"
+        class="h-8 w-8"
+        aria-label="关闭详情"
+        @click="open = false"
+      >
+        <X class="h-4 w-4" />
+      </Button>
+    </template>
+
+    <div
+      v-if="monitor"
+      class="space-y-4"
+    >
+      <div class="grid grid-cols-2 gap-3 xl:grid-cols-5">
+        <Card
+          variant="inset"
+          class="p-3"
+        >
+          <div class="text-[11px] uppercase tracking-[0.18em] text-muted-foreground">
+            当前状态
+          </div>
+          <div class="mt-2">
+            <Badge :variant="getMonitorStatusVariant(monitor)">
+              {{ getMonitorStatusLabel(monitor) }}
+            </Badge>
+          </div>
+        </Card>
+
+        <Card
+          variant="inset"
+          class="p-3"
+        >
+          <div class="text-[11px] uppercase tracking-[0.18em] text-muted-foreground">
+            成功率
+          </div>
+          <div class="mt-2 text-xl font-semibold">
+            {{ formatPercent(monitor.success_rate) }}
+          </div>
+        </Card>
+
+        <Card
+          variant="inset"
+          class="p-3"
+        >
+          <div class="text-[11px] uppercase tracking-[0.18em] text-muted-foreground">
+            请求总数
+          </div>
+          <div class="mt-2 text-xl font-semibold">
+            {{ monitor.total_attempts }}
+          </div>
+        </Card>
+
+        <Card
+          variant="inset"
+          class="p-3"
+        >
+          <div class="text-[11px] uppercase tracking-[0.18em] text-muted-foreground">
+            最近失败
+          </div>
+          <div class="mt-2 text-xl font-semibold">
+            {{ failureEvents.length }}
+          </div>
+        </Card>
+
+        <Card
+          variant="inset"
+          class="p-3"
+        >
+          <div class="text-[11px] uppercase tracking-[0.18em] text-muted-foreground">
+            最近事件
+          </div>
+          <div class="mt-2 text-sm font-medium">
+            {{ monitor.last_event_at ? formatAbsoluteTime(monitor.last_event_at) : '暂无' }}
+          </div>
+        </Card>
+      </div>
+
+      <div
+        v-if="monitorApiPath"
+        class="rounded-xl border border-border/60 bg-muted/20 px-4 py-3 text-sm"
+      >
+        <div class="text-[11px] uppercase tracking-[0.18em] text-muted-foreground">
+          请求入口
+        </div>
+        <div class="mt-2 font-mono text-foreground">
+          {{ monitorApiPath }}
+        </div>
+      </div>
+
+      <div
+        v-if="isAdmin"
+        class="grid grid-cols-1 gap-4 xl:grid-cols-[minmax(0,0.85fr)_minmax(0,1.15fr)]"
+      >
+        <Card
+          variant="inset"
+          class="p-4"
+        >
+          <div class="mb-3 flex items-center justify-between">
+            <h4 class="text-sm font-semibold">
+              受影响 Provider
+            </h4>
+            <span class="text-xs text-muted-foreground">
+              {{ providerSummaries.length }} 个
+            </span>
+          </div>
+
+          <div
+            v-if="loadingFormatKeys"
+            class="flex items-center gap-2 py-6 text-sm text-muted-foreground"
+          >
+            <Loader2 class="h-4 w-4 animate-spin" />
+            正在加载 Key 诊断数据...
+          </div>
+
+          <div
+            v-else-if="providerSummaries.length === 0"
+            class="py-6 text-sm text-muted-foreground"
+          >
+            当前格式下没有可诊断的 provider/key 数据。
+          </div>
+
+          <div
+            v-else
+            class="max-h-[420px] space-y-2 overflow-y-auto pr-1"
+          >
+            <div
+              v-for="provider in providerSummaries"
+              :key="provider.provider_id"
+              class="rounded-xl border p-3 transition-colors"
+              :class="providerFilter === provider.provider_id
+                ? 'border-primary/60 bg-primary/5'
+                : 'border-border/50 bg-card/70'"
+            >
+              <div class="flex items-center justify-between gap-3">
+                <div class="min-w-0">
+                  <div class="flex items-center gap-2 flex-wrap">
+                    <span class="text-sm font-medium">
+                      {{ provider.provider_name }}
+                    </span>
+                    <Badge :variant="provider.circuitOpenKeys > 0 ? 'destructive' : provider.providerActive ? 'success' : 'outline'">
+                      {{ provider.circuitOpenKeys > 0 ? `${provider.circuitOpenKeys} 熔断` : provider.providerActive ? '活跃' : '停用' }}
+                    </Badge>
+                  </div>
+                  <div class="mt-1 text-xs text-muted-foreground">
+                    {{ provider.totalKeys }} 个 Key
+                    <span class="mx-1">·</span>
+                    活跃 {{ provider.activeKeys }}
+                    <span class="mx-1">·</span>
+                    平均健康 {{ formatScore(provider.avgHealthScore) }}
+                  </div>
+                </div>
+                <div class="text-right text-xs text-muted-foreground">
+                  <div>{{ provider.requestCount }} 请求</div>
+                </div>
+              </div>
+
+              <div class="mt-3 flex flex-wrap gap-2">
+                <Button
+                  variant="outline"
+                  size="sm"
+                  :class="FILTER_BUTTON_CLASS"
+                  @click="toggleProviderFilter(provider.provider_id)"
+                >
+                  {{ providerFilter === provider.provider_id ? '取消筛选' : '只看该 Provider' }}
+                </Button>
+                <Button
+                  variant="outline"
+                  size="sm"
+                  :class="FILTER_BUTTON_CLASS"
+                  @click="$emit('navigateProvider', provider.provider_id)"
+                >
+                  前往 Provider
+                </Button>
+                <Button
+                  v-if="provider.poolEnabled"
+                  variant="outline"
+                  size="sm"
+                  :class="FILTER_BUTTON_CLASS"
+                  @click="$emit('navigatePool', provider.provider_id)"
+                >
+                  前往号池
+                </Button>
+              </div>
+            </div>
+          </div>
+        </Card>
+
+        <Card
+          variant="inset"
+          class="p-4"
+        >
+          <div class="mb-3 flex flex-col gap-3">
+            <div class="flex items-center justify-between gap-3">
+              <h4 class="text-sm font-semibold">
+                Key 诊断
+              </h4>
+              <span class="text-xs text-muted-foreground">
+                {{ filteredFormatKeys.length }} / {{ sortedFormatKeys.length }} 个 Key
+              </span>
+            </div>
+
+            <div class="flex flex-wrap gap-2">
+              <Button
+                variant="outline"
+                size="sm"
+                :class="[FILTER_BUTTON_CLASS, keyScopeFilter === 'all' ? 'border-primary/60 bg-primary/5 text-primary' : '']"
+                @click="keyScopeFilter = 'all'"
+              >
+                全部
+              </Button>
+              <Button
+                variant="outline"
+                size="sm"
+                :class="[FILTER_BUTTON_CLASS, keyScopeFilter === 'circuit-open' ? 'border-primary/60 bg-primary/5 text-primary' : '']"
+                @click="keyScopeFilter = 'circuit-open'"
+              >
+                仅看熔断
+              </Button>
+              <Button
+                variant="outline"
+                size="sm"
+                :class="[FILTER_BUTTON_CLASS, keyScopeFilter === 'recoverable' ? 'border-primary/60 bg-primary/5 text-primary' : '']"
+                @click="keyScopeFilter = 'recoverable'"
+              >
+                仅看可恢复项
+              </Button>
+              <Button
+                v-if="providerFilter !== 'all' || keyScopeFilter !== 'all'"
+                variant="outline"
+                size="sm"
+                :class="FILTER_BUTTON_CLASS"
+                @click="resetFilters"
+              >
+                清空局部筛选
+              </Button>
+            </div>
+
+            <div
+              v-if="providerFilter !== 'all' || keyScopeFilter !== 'all'"
+              class="text-xs text-muted-foreground"
+            >
+              <span v-if="providerFilter !== 'all'">
+                Provider：{{ providerFilterName }}
+              </span>
+              <span v-if="providerFilter !== 'all' && keyScopeFilter !== 'all'"> · </span>
+              <span v-if="keyScopeFilter === 'circuit-open'">范围：仅熔断项</span>
+              <span v-else-if="keyScopeFilter === 'recoverable'">范围：仅可恢复项</span>
+            </div>
+          </div>
+
+          <div
+            v-if="loadingFormatKeys"
+            class="flex items-center gap-2 py-6 text-sm text-muted-foreground"
+          >
+            <Loader2 class="h-4 w-4 animate-spin" />
+            正在加载 Key 诊断数据...
+          </div>
+
+          <div
+            v-else-if="sortedFormatKeys.length === 0"
+            class="py-6 text-sm text-muted-foreground"
+          >
+            当前格式下没有 Key 样本。
+          </div>
+
+          <div
+            v-else-if="filteredFormatKeys.length === 0"
+            class="py-6 text-sm text-muted-foreground"
+          >
+            当前局部筛选下没有匹配的 Key。
+          </div>
+
+          <div
+            v-else
+            class="max-h-[420px] space-y-2 overflow-y-auto pr-1"
+          >
+            <div
+              v-for="key in filteredFormatKeys"
+              :key="key.id"
+              class="rounded-xl border border-border/50 bg-card/70 p-3"
+            >
+              <div class="flex flex-col gap-3 sm:flex-row sm:items-start sm:justify-between">
+                <div class="min-w-0">
+                  <div class="flex items-center gap-2 flex-wrap">
+                    <span class="text-sm font-medium">
+                      {{ key.name }}
+                    </span>
+                    <Badge variant="outline">
+                      {{ key.provider_name }}
+                    </Badge>
+                    <Badge :variant="key.circuit_breaker_open ? 'destructive' : key.is_active && key.provider_active ? 'success' : 'outline'">
+                      {{ key.circuit_breaker_open ? '已熔断' : key.is_active && key.provider_active ? '可用' : '停用' }}
+                    </Badge>
+                    <Badge
+                      v-if="shouldShowRecoverAction(key)"
+                      variant="warning"
+                    >
+                      建议恢复
+                    </Badge>
+                  </div>
+                  <div class="mt-1 font-mono text-xs text-muted-foreground">
+                    {{ key.api_key_masked }}
+                  </div>
+                  <div class="mt-2 flex flex-wrap gap-x-3 gap-y-1 text-xs text-muted-foreground">
+                    <span>健康 {{ formatScore(key.health_score) }}</span>
+                    <span v-if="key.success_rate != null">{{ formatPercent(key.success_rate) }}</span>
+                    <span>{{ key.request_count }} 请求</span>
+                    <span v-if="key.avg_response_time_ms != null">均延迟 {{ formatLatency(key.avg_response_time_ms) }}</span>
+                  </div>
+                </div>
+
+                <div class="flex flex-wrap gap-2 sm:justify-end">
+                  <Button
+                    v-if="key.pool_enabled"
+                    variant="outline"
+                    size="sm"
+                    :class="ACTION_BUTTON_CLASS"
+                    @click="$emit('navigatePool', key.provider_id, key.name)"
+                  >
+                    号池内查看
+                  </Button>
+                  <Button
+                    v-if="shouldShowRecoverAction(key)"
+                    variant="outline"
+                    size="sm"
+                    :class="ACTION_BUTTON_CLASS"
+                    :disabled="recoveringKeyId !== null"
+                    @click="$emit('recoverKey', key)"
+                  >
+                    <Loader2
+                      v-if="recoveringKeyId === key.id"
+                      class="mr-2 h-3.5 w-3.5 animate-spin"
+                    />
+                    恢复健康度
+                  </Button>
+                </div>
+              </div>
+            </div>
+          </div>
+        </Card>
+      </div>
+
+      <div class="grid grid-cols-1 gap-4 xl:grid-cols-2">
+        <Card
+          variant="inset"
+          class="p-4"
+        >
+          <div class="mb-3 flex items-center justify-between">
+            <h4 class="text-sm font-semibold">
+              最近失败
+            </h4>
+            <span class="text-xs text-muted-foreground">
+              最近 {{ Math.min(failureEvents.length, 8) }} 条
+            </span>
+          </div>
+
+          <div
+            v-if="failureEvents.length === 0"
+            class="py-6 text-sm text-muted-foreground"
+          >
+            当前窗口内没有失败事件。
+          </div>
+
+          <div
+            v-else
+            class="space-y-2"
+          >
+            <div
+              v-for="event in failureEvents.slice(0, 8)"
+              :key="buildEventKey(event)"
+              class="rounded-xl border border-border/50 bg-card/70 p-3"
+            >
+              <div class="flex items-center justify-between gap-3">
+                <div class="flex items-center gap-2">
+                  <Badge variant="destructive">
+                    失败
+                  </Badge>
+                  <span
+                    v-if="event.status_code"
+                    class="text-xs font-medium text-foreground"
+                  >
+                    HTTP {{ event.status_code }}
+                  </span>
+                </div>
+                <span class="text-xs text-muted-foreground">
+                  {{ formatAbsoluteTime(event.timestamp) }}
+                </span>
+              </div>
+              <div class="mt-2 text-sm text-foreground">
+                {{ event.error_type || event.error_message || '未提供错误类型' }}
+              </div>
+              <div class="mt-1 text-xs text-muted-foreground">
+                {{ event.latency_ms != null ? `延迟 ${formatLatency(event.latency_ms)}` : '延迟未知' }}
+              </div>
+            </div>
+          </div>
+        </Card>
+
+        <Card
+          variant="inset"
+          class="p-4"
+        >
+          <div class="mb-3">
+            <h4 class="text-sm font-semibold">
+              错误分布
+            </h4>
+          </div>
+
+          <div class="space-y-4">
+            <div>
+              <div class="mb-2 text-xs uppercase tracking-[0.18em] text-muted-foreground">
+                Error Type
+              </div>
+              <div
+                v-if="errorTypes.length === 0"
+                class="text-sm text-muted-foreground"
+              >
+                当前没有可归类的错误类型。
+              </div>
+              <div
+                v-else
+                class="flex flex-wrap gap-2"
+              >
+                <Badge
+                  v-for="item in errorTypes"
+                  :key="item.label"
+                  variant="outline"
+                >
+                  {{ item.label }} × {{ item.count }}
+                </Badge>
+              </div>
+            </div>
+
+            <div>
+              <div class="mb-2 text-xs uppercase tracking-[0.18em] text-muted-foreground">
+                Status Code
+              </div>
+              <div
+                v-if="statusCodes.length === 0"
+                class="text-sm text-muted-foreground"
+              >
+                当前没有状态码样本。
+              </div>
+              <div
+                v-else
+                class="flex flex-wrap gap-2"
+              >
+                <Badge
+                  v-for="item in statusCodes"
+                  :key="item.label"
+                  variant="outline"
+                >
+                  {{ item.label }} × {{ item.count }}
+                </Badge>
+              </div>
+            </div>
+
+            <div>
+              <div class="mb-2 text-xs uppercase tracking-[0.18em] text-muted-foreground">
+                延迟样本
+              </div>
+              <div class="grid grid-cols-3 gap-2 text-sm">
+                <div class="rounded-lg border border-border/50 bg-card/70 px-3 py-2">
+                  <div class="text-[11px] text-muted-foreground">
+                    P50
+                  </div>
+                  <div class="mt-1 font-semibold">
+                    {{ latencyStats.p50 }}
+                  </div>
+                </div>
+                <div class="rounded-lg border border-border/50 bg-card/70 px-3 py-2">
+                  <div class="text-[11px] text-muted-foreground">
+                    P95
+                  </div>
+                  <div class="mt-1 font-semibold">
+                    {{ latencyStats.p95 }}
+                  </div>
+                </div>
+                <div class="rounded-lg border border-border/50 bg-card/70 px-3 py-2">
+                  <div class="text-[11px] text-muted-foreground">
+                    Max
+                  </div>
+                  <div class="mt-1 font-semibold">
+                    {{ latencyStats.max }}
+                  </div>
+                </div>
+              </div>
+            </div>
+          </div>
+        </Card>
+      </div>
+
+      <Card
+        variant="inset"
+        class="p-4"
+      >
+        <div class="mb-3 flex items-center justify-between">
+          <h4 class="text-sm font-semibold">
+            最近活动
+          </h4>
+          <span class="text-xs text-muted-foreground">
+            最近 {{ Math.min(sortedEvents.length, 12) }} 条事件
+          </span>
+        </div>
+
+        <div
+          v-if="sortedEvents.length === 0"
+          class="py-6 text-sm text-muted-foreground"
+        >
+          当前没有事件样本。
+        </div>
+
+        <div
+          v-else
+          class="space-y-2"
+        >
+          <div
+            v-for="event in sortedEvents.slice(0, 12)"
+            :key="buildEventKey(event)"
+            class="flex flex-col gap-2 rounded-xl border border-border/50 bg-card/70 p-3 sm:flex-row sm:items-center sm:justify-between"
+          >
+            <div class="flex min-w-0 items-center gap-2 flex-wrap">
+              <Badge :variant="getEventStatusVariant(event.status)">
+                {{ getEventStatusLabel(event.status) }}
+              </Badge>
+              <span
+                v-if="event.status_code"
+                class="text-xs text-foreground"
+              >
+                HTTP {{ event.status_code }}
+              </span>
+              <span
+                v-if="event.latency_ms != null"
+                class="text-xs text-muted-foreground"
+              >
+                {{ formatLatency(event.latency_ms) }}
+              </span>
+              <span
+                v-if="event.error_type"
+                class="truncate text-xs text-muted-foreground"
+              >
+                {{ event.error_type }}
+              </span>
+            </div>
+            <div class="text-xs text-muted-foreground">
+              {{ formatAbsoluteTime(event.timestamp) }}
+            </div>
+          </div>
+        </div>
+      </Card>
+    </div>
+  </Dialog>
+</template>
+
+<script setup lang="ts">
+import { computed, ref, watch } from 'vue'
+import { Loader2, X } from 'lucide-vue-next'
+import Card from '@/components/ui/card.vue'
+import Badge from '@/components/ui/badge.vue'
+import Button from '@/components/ui/button.vue'
+import Dialog from '@/components/ui/dialog/Dialog.vue'
+import type { GroupedFormatKey } from '@/api/endpoints/keys'
+import type { EndpointHealthEvent, EndpointStatusMonitor, PublicEndpointStatusMonitor, PublicHealthEvent } from '@/api/endpoints/types'
+import { formatApiFormat } from '@/api/endpoints/types/api-format'
+
+type MonitorItem = EndpointStatusMonitor | PublicEndpointStatusMonitor
+type MonitorEvent = EndpointHealthEvent | PublicHealthEvent
+type MonitorStatus = 'healthy' | 'warning' | 'unhealthy' | 'unknown'
+type CountItem = { label: string; count: number }
+type KeyScopeFilter = 'all' | 'circuit-open' | 'recoverable'
+type ProviderImpactSummary = {
+  provider_id: string
+  provider_name: string
+  providerActive: boolean
+  poolEnabled: boolean
+  totalKeys: number
+  activeKeys: number
+  circuitOpenKeys: number
+  avgHealthScore: number | null
+  requestCount: number
+}
+
+const props = defineProps<{
+  monitor: MonitorItem | null
+  isAdmin: boolean
+  formatKeys: GroupedFormatKey[]
+  loadingFormatKeys: boolean
+  recoveringKeyId: string | null
+}>()
+
+defineEmits<{
+  navigateProvider: [providerId: string]
+  navigatePool: [providerId: string, search?: string]
+  recoverKey: [key: GroupedFormatKey]
+}>()
+
+const open = defineModel<boolean>({ default: false })
+
+const FILTER_BUTTON_CLASS = 'h-8 px-3 text-xs'
+const ACTION_BUTTON_CLASS = 'h-8 px-3 text-xs'
+
+const providerFilter = ref<string>('all')
+const keyScopeFilter = ref<KeyScopeFilter>('all')
+
+watch(open, (value) => {
+  if (!value) {
+    providerFilter.value = 'all'
+    keyScopeFilter.value = 'all'
+  }
+})
+
+const monitorApiPath = computed(() =>
+  props.monitor && 'api_path' in props.monitor ? props.monitor.api_path : null
+)
+
+const sortedEvents = computed<MonitorEvent[]>(() => {
+  const events = props.monitor?.events ?? []
+  return [...events].sort((left, right) =>
+    new Date(right.timestamp).getTime() - new Date(left.timestamp).getTime()
+  )
+})
+
+const failureEvents = computed<MonitorEvent[]>(() =>
+  sortedEvents.value.filter(event => event.status === 'failed')
+)
+
+const sortedFormatKeys = computed<GroupedFormatKey[]>(() =>
+  [...props.formatKeys].sort((left, right) => {
+    const circuitGap = Number(right.circuit_breaker_open) - Number(left.circuit_breaker_open)
+    if (circuitGap !== 0) return circuitGap
+
+    const leftHealth = left.health_score ?? 1
+    const rightHealth = right.health_score ?? 1
+    if (leftHealth !== rightHealth) return leftHealth - rightHealth
+
+    if (left.request_count !== right.request_count) {
+      return right.request_count - left.request_count
+    }
+
+    return left.name.localeCompare(right.name)
+  })
+)
+
+const providerSummaries = computed<ProviderImpactSummary[]>(() => {
+  const grouped = new Map<string, ProviderImpactSummary & { healthValues: number[] }>()
+
+  for (const key of sortedFormatKeys.value) {
+    const current = grouped.get(key.provider_id) ?? {
+      provider_id: key.provider_id,
+      provider_name: key.provider_name,
+      providerActive: key.provider_active,
+      poolEnabled: key.pool_enabled,
+      totalKeys: 0,
+      activeKeys: 0,
+      circuitOpenKeys: 0,
+      avgHealthScore: null,
+      requestCount: 0,
+      healthValues: []
+    }
+
+    current.totalKeys += 1
+    current.requestCount += key.request_count
+    current.providerActive = current.providerActive || key.provider_active
+    current.poolEnabled = current.poolEnabled || key.pool_enabled
+
+    if (key.is_active) current.activeKeys += 1
+    if (key.circuit_breaker_open) current.circuitOpenKeys += 1
+    if (key.health_score != null) current.healthValues.push(key.health_score)
+
+    grouped.set(key.provider_id, current)
+  }
+
+  return Array.from(grouped.values())
+    .map((provider) => ({
+      provider_id: provider.provider_id,
+      provider_name: provider.provider_name,
+      providerActive: provider.providerActive,
+      poolEnabled: provider.poolEnabled,
+      totalKeys: provider.totalKeys,
+      activeKeys: provider.activeKeys,
+      circuitOpenKeys: provider.circuitOpenKeys,
+      avgHealthScore: provider.healthValues.length > 0
+        ? provider.healthValues.reduce((sum, value) => sum + value, 0) / provider.healthValues.length
+        : null,
+      requestCount: provider.requestCount,
+    }))
+    .sort((left, right) => {
+      if (left.circuitOpenKeys !== right.circuitOpenKeys) {
+        return right.circuitOpenKeys - left.circuitOpenKeys
+      }
+
+      const leftHealth = left.avgHealthScore ?? 1
+      const rightHealth = right.avgHealthScore ?? 1
+      if (leftHealth !== rightHealth) return leftHealth - rightHealth
+
+      return right.requestCount - left.requestCount
+    })
+})
+
+const providerFilterName = computed(() =>
+  providerSummaries.value.find(p => p.provider_id === providerFilter.value)?.provider_name ?? '当前 Provider'
+)
+
+const filteredFormatKeys = computed<GroupedFormatKey[]>(() =>
+  sortedFormatKeys.value.filter((key) => {
+    if (providerFilter.value !== 'all' && key.provider_id !== providerFilter.value) {
+      return false
+    }
+
+    if (keyScopeFilter.value === 'circuit-open') {
+      return key.circuit_breaker_open
+    }
+
+    if (keyScopeFilter.value === 'recoverable') {
+      return shouldShowRecoverAction(key)
+    }
+
+    return true
+  })
+)
+
+const dialogDescription = computed(() => {
+  if (!props.monitor) return '查看最近失败、错误分布与事件样本'
+  const timeText = props.monitor.last_event_at
+    ? `最近事件 ${formatRelativeTime(props.monitor.last_event_at)}`
+    : '当前窗口没有事件'
+  return `${getMonitorStatusLabel(props.monitor)} · ${props.monitor.total_attempts} 次请求 · ${timeText}`
+})
+
+const errorTypes = computed<CountItem[]>(() =>
+  buildCountItems(
+    failureEvents.value
+      .map(event => event.error_type?.trim())
+      .filter((value): value is string => Boolean(value))
+  )
+)
+
+const statusCodes = computed<CountItem[]>(() =>
+  buildCountItems(
+    failureEvents.value
+      .map(event => event.status_code != null ? String(event.status_code) : null)
+      .filter((value): value is string => Boolean(value))
+  )
+)
+
+const latencyStats = computed(() => {
+  const values = sortedEvents.value
+    .map(event => event.latency_ms)
+    .filter((value): value is number => value != null)
+    .sort((left, right) => left - right)
+
+  if (values.length === 0) {
+    return { p50: '--', p95: '--', max: '--' }
+  }
+
+  return {
+    p50: formatLatency(pickPercentile(values, 0.5)),
+    p95: formatLatency(pickPercentile(values, 0.95)),
+    max: formatLatency(values[values.length - 1])
+  }
+})
+
+watch(sortedFormatKeys, (keys) => {
+  if (providerFilter.value === 'all') return
+  if (!keys.some(key => key.provider_id === providerFilter.value)) {
+    providerFilter.value = 'all'
+  }
+})
+
+function resetFilters() {
+  providerFilter.value = 'all'
+  keyScopeFilter.value = 'all'
+}
+
+function toggleProviderFilter(providerId: string) {
+  providerFilter.value = providerFilter.value === providerId ? 'all' : providerId
+}
+
+function shouldShowRecoverAction(key: GroupedFormatKey): boolean {
+  if (key.health_score == null) return false
+  return key.health_score < 0.999
+}
+
+function getLatestTimelineStatus(monitor: MonitorItem): MonitorStatus | null {
+  const timeline = monitor.timeline ?? []
+  for (let index = timeline.length - 1; index >= 0; index -= 1) {
+    const status = timeline[index]
+    if (status === 'healthy' || status === 'warning' || status === 'unhealthy') {
+      return status
+    }
+  }
+  return null
+}
+
+function getMonitorStatus(monitor: MonitorItem): MonitorStatus {
+  const timelineStatus = getLatestTimelineStatus(monitor)
+  if (timelineStatus === 'unhealthy') return 'unhealthy'
+  if (timelineStatus === 'warning') return 'warning'
+
+  if (monitor.total_attempts <= 0) {
+    return 'unknown'
+  }
+
+  if (monitor.success_rate < 0.7) return 'unhealthy'
+  if (monitor.success_rate < 0.95) return 'warning'
+  if (timelineStatus === 'healthy' || monitor.success_rate >= 0.95) return 'healthy'
+  return 'unknown'
+}
+
+function getMonitorStatusLabel(monitor: MonitorItem): string {
+  switch (getMonitorStatus(monitor)) {
+    case 'unhealthy':
+      return '异常'
+    case 'warning':
+      return '预警'
+    case 'healthy':
+      return '健康'
+    default:
+      return '无流量'
+  }
+}
+
+function getMonitorStatusVariant(monitor: MonitorItem): 'success' | 'warning' | 'destructive' | 'outline' {
+  switch (getMonitorStatus(monitor)) {
+    case 'healthy':
+      return 'success'
+    case 'warning':
+      return 'warning'
+    case 'unhealthy':
+      return 'destructive'
+    default:
+      return 'outline'
+  }
+}
+
+function getEventStatusVariant(status: string): 'success' | 'warning' | 'destructive' | 'outline' {
+  switch (status) {
+    case 'success':
+      return 'success'
+    case 'failed':
+      return 'destructive'
+    case 'skipped':
+      return 'warning'
+    default:
+      return 'outline'
+  }
+}
+
+function getEventStatusLabel(status: string): string {
+  switch (status) {
+    case 'success':
+      return '成功'
+    case 'failed':
+      return '失败'
+    case 'skipped':
+      return '跳过'
+    case 'started':
+      return '执行中'
+    default:
+      return '未知'
+  }
+}
+
+function buildCountItems(values: string[]): CountItem[] {
+  const counter = new Map<string, number>()
+  for (const value of values) {
+    counter.set(value, (counter.get(value) ?? 0) + 1)
+  }
+  return Array.from(counter.entries())
+    .map(([label, count]) => ({ label, count }))
+    .sort((left, right) => right.count - left.count)
+    .slice(0, 8)
+}
+
+function pickPercentile(sortedValues: number[], percentile: number): number {
+  if (sortedValues.length === 0) return 0
+  const index = Math.min(sortedValues.length - 1, Math.max(0, Math.ceil(sortedValues.length * percentile) - 1))
+  return sortedValues[index]
+}
+
+function formatPercent(rate: number): string {
+  return `${(rate * 100).toFixed(0)}%`
+}
+
+function formatScore(score: number | null | undefined): string {
+  if (score == null) return '--'
+  return `${(score * 100).toFixed(0)}%`
+}
+
+function formatLatency(latencyMs: number): string {
+  if (latencyMs < 1000) return `${latencyMs}ms`
+  return `${(latencyMs / 1000).toFixed(2)}s`
+}
+
+function formatRelativeTime(dateString: string): string {
+  const diff = Date.now() - new Date(dateString).getTime()
+  if (Number.isNaN(diff)) return '未知'
+  if (diff < 60_000) return '刚刚'
+
+  const minutes = Math.floor(diff / 60_000)
+  if (minutes < 60) return `${minutes} 分钟前`
+
+  const hours = Math.floor(minutes / 60)
+  if (hours < 24) return `${hours} 小时前`
+
+  const days = Math.floor(hours / 24)
+  if (days < 7) return `${days} 天前`
+
+  return formatAbsoluteTime(dateString)
+}
+
+function formatAbsoluteTime(dateString: string): string {
+  return new Date(dateString).toLocaleString('zh-CN', {
+    month: '2-digit',
+    day: '2-digit',
+    hour: '2-digit',
+    minute: '2-digit',
+    second: '2-digit'
+  })
+}
+
+function buildEventKey(event: MonitorEvent): string {
+  return `${event.timestamp}-${event.status}-${event.status_code ?? 'na'}-${event.error_type ?? 'none'}-${event.latency_ms ?? 'na'}`
+}
+</script>

frontend/src/features/providers/components/ModelMappingDialog.vue

@@ -282,8 +282,6 @@ export interface AliasGroup {
 const props = defineProps<{
   open: boolean
   providerId: string
-  /** @deprecated */
-  providerApiFormats?: string[]
   models: Model[]
   editingGroup?: AliasGroup | null
   preselectedModelId?: string | null

frontend/src/features/providers/components/OAuthAccountDialog.vue

@@ -410,8 +410,8 @@
             :reset-key="importInputResetKey"
             drop-title="拖入授权文件或点击选择"
             drop-hint="支持 .json / .txt，可多选"
-            manual-placeholder="粘贴 Refresh Token 或 JSON 内容"
-            paste-toggle-text="或手动粘贴 Refresh Token"
+            manual-placeholder="粘贴 Refresh Token / Access Token 或 JSON 内容"
+            paste-toggle-text="或手动粘贴 Token"
             file-toggle-text="或选择 JSON 文件导入"
             textarea-class="min-h-[200px] text-xs font-mono break-all !rounded-xl"
             @error="handleImportInputError"
@@ -893,7 +893,7 @@ function isBatchImport(text: string): boolean {
   return lines.length > 1
 }
 
-function parseImportText(text: string): { refresh_token: string; name?: string } | null {
+function parseImportText(text: string): { refresh_token?: string; access_token?: string; name?: string } | null {
   const trimmed = text.trim()
   if (!trimmed) return null
 
@@ -907,9 +907,19 @@ function parseImportText(text: string): { refresh_token: string; name?: string }
     if (typeof parsed === 'object' && parsed !== null) {
       const obj = parsed as Record<string, unknown>
       const refreshToken = obj.refresh_token
-      if (typeof refreshToken === 'string' && refreshToken.trim()) {
+      const refreshTokenCamel = obj.refreshToken
+      const accessToken = obj.access_token
+      const accessTokenCamel = obj.accessToken
+      const normalizedRefreshToken = typeof refreshToken === 'string' && refreshToken.trim()
+        ? refreshToken.trim()
+        : (typeof refreshTokenCamel === 'string' && refreshTokenCamel.trim() ? refreshTokenCamel.trim() : undefined)
+      const normalizedAccessToken = typeof accessToken === 'string' && accessToken.trim()
+        ? accessToken.trim()
+        : (typeof accessTokenCamel === 'string' && accessTokenCamel.trim() ? accessTokenCamel.trim() : undefined)
+      if (normalizedRefreshToken || normalizedAccessToken) {
         return {
-          refresh_token: refreshToken.trim(),
+          refresh_token: normalizedRefreshToken,
+          access_token: normalizedAccessToken,
           name: (typeof obj.name === 'string' ? obj.name : undefined) || (typeof obj.oauth_email === 'string' ? obj.oauth_email : undefined),
         }
       }
@@ -919,9 +929,34 @@ function parseImportText(text: string): { refresh_token: string; name?: string }
     // Not JSON: treat as raw token.
   }
 
+  if (isLikelyJwtToken(trimmed)) {
+    return { access_token: trimmed }
+  }
+
   return { refresh_token: trimmed }
 }
 
+function isLikelyJwtToken(token: string): boolean {
+  const parts = token.trim().split('.')
+  if (parts.length !== 3 || parts.some(part => !part)) return false
+
+  try {
+    const header = JSON.parse(decodeBase64Url(parts[0])) as Record<string, unknown>
+    const payload = JSON.parse(decodeBase64Url(parts[1])) as Record<string, unknown>
+    const tokenType = typeof header.typ === 'string' ? header.typ.toLowerCase() : ''
+    if (tokenType && tokenType !== 'jwt' && tokenType !== 'at+jwt') return false
+    return ['exp', 'aud', 'iss', 'scope', 'scp'].some(key => key in payload)
+  } catch {
+    return false
+  }
+}
+
+function decodeBase64Url(value: string): string {
+  const normalized = value.replace(/-/g, '+').replace(/_/g, '/')
+  const padded = normalized.padEnd(normalized.length + ((4 - (normalized.length % 4)) % 4), '=')
+  return atob(padded)
+}
+
 function handleImportInputError(payload: { message: string; title?: string }) {
   showError(payload.message, payload.title)
 }

frontend/src/features/providers/components/ProviderDetailDrawer.vue

@@ -370,7 +370,16 @@
                                 >
                                   {{ getKeyOAuthExpires(key)?.text }}
                                 </span>
+                                <Badge
+                                  v-if="key.oauth_temporary"
+                                  variant="outline"
+                                  class="text-[10px] px-1.5 py-0 shrink-0"
+                                  title="仅通过 Access Token 导入，无法自动刷新，到期后需要重新导入"
+                                >
+                                  临时
+                                </Badge>
                                 <Button
+                                  v-else
                                   variant="ghost"
                                   size="icon"
                                   class="h-4 w-4 shrink-0"
@@ -548,7 +557,7 @@
                               v-if="key.upstream_metadata.codex?.updated_at"
                               class="text-[9px] text-muted-foreground/70"
                             >
-                            {{ formatCodexUpdatedAt(key.upstream_metadata.codex.updated_at) }}
+                              {{ formatUpdatedAt(key.upstream_metadata.codex.updated_at) }}
                             </span>
                         </div>
                       </div>
@@ -674,7 +683,7 @@
                               v-if="key.upstream_metadata.antigravity?.updated_at"
                               class="text-[9px] text-muted-foreground/70"
                             >
-                              {{ formatAntigravityUpdatedAt(key.upstream_metadata.antigravity.updated_at) }}
+                              {{ formatUpdatedAt(key.upstream_metadata.antigravity.updated_at) }}
                             </span>
                           </div>
                         </div>
@@ -2471,10 +2480,6 @@ function formatUpdatedAt(updatedAt: number): string {
   return `${days}天前更新`
 }
 
-// 兼容旧函数名
-const formatCodexUpdatedAt = formatUpdatedAt
-const formatAntigravityUpdatedAt = formatUpdatedAt
-
 function secondsUntilReset(resetTime: string): number | null {
   if (!resetTime) return null
   const ts = Date.parse(resetTime)

frontend/src/features/providers/components/provider-tabs/ModelAliasesTab.vue

@@ -165,7 +165,6 @@
   <ModelMappingDialog
     v-model:open="dialogOpen"
     :provider-id="provider.id"
-    :provider-api-formats="providerApiFormats"
     :models="models"
     :editing-group="editingGroup"
     :preselected-model-id="preselectedModelId"
@@ -195,7 +194,6 @@ import { useToast } from '@/composables/useToast'
 import {
   getProviderModels,
   testModel,
-  API_FORMAT_LABELS,
   type Model,
   type ProviderModelAlias
 } from '@/api/endpoints'
@@ -227,16 +225,6 @@ const preselectedModelId = ref<string | null>(null)
 // 列表展开状态
 const expandedAliasGroups = ref<Set<string>>(new Set())
 
-// 获取 Provider 支持的 API 格式
-const providerApiFormats = computed(() => {
-  const formats = props.provider?.api_formats
-  if (Array.isArray(formats) && formats.length > 0) {
-    const order = Object.keys(API_FORMAT_LABELS)
-    return [...formats].sort((a, b) => order.indexOf(a) - order.indexOf(b))
-  }
-  return []
-})
-
 // 生成作用域唯一键
 function getApiFormatsKey(formats: string[] | undefined): string {
   if (!formats || formats.length === 0) return ''

frontend/src/features/usage/components/ActivityHeatmapCard.vue

@@ -20,26 +20,31 @@
     </div>
     <div
       v-if="isLoading"
-      class="h-full min-h-[160px] flex items-center justify-center text-sm text-muted-foreground"
+      class="flex items-center justify-center text-sm text-muted-foreground"
+      :style="{ minHeight: `${minHeight}px` }"
     >
       <Loader2 class="h-5 w-5 animate-spin mr-2" />
       加载中...
     </div>
     <div
       v-else-if="hasError"
-      class="h-full min-h-[160px] flex items-center justify-center text-sm text-destructive"
+      class="flex items-center justify-center text-sm text-destructive"
+      :style="{ minHeight: `${minHeight}px` }"
     >
       <AlertCircle class="h-4 w-4 mr-1.5" />
-      加载失败
+      数据暂不可用
     </div>
     <ActivityHeatmap
       v-else-if="hasData"
       :data="data"
       :show-header="false"
+      :value-label="valueLabel"
+      :actual-value-label="actualValueLabel"
     />
     <div
       v-else
-      class="h-full min-h-[160px] flex items-center justify-center text-sm text-muted-foreground"
+      class="flex items-center justify-center text-sm text-muted-foreground"
+      :style="{ minHeight: `${minHeight}px` }"
     >
       暂无活跃数据
     </div>
@@ -58,6 +63,9 @@ const props = defineProps<{
   title: string
   isLoading?: boolean
   hasError?: boolean
+  minHeight?: number
+  valueLabel?: string
+  actualValueLabel?: string
 }>()
 
 const legendLevels = [0.08, 0.25, 0.45, 0.65, 0.85]
@@ -65,4 +73,6 @@ const legendLevels = [0.08, 0.25, 0.45, 0.65, 0.85]
 const hasData = computed(() =>
   props.data && props.data.days && props.data.days.length > 0
 )
+
+const minHeight = computed(() => props.minHeight ?? 160)
 </script>

frontend/src/features/usage/components/HorizontalRequestTimeline.vue

@@ -780,11 +780,11 @@ const poolAttemptCandidates = computed<CandidateRecord[]>(() => {
           }
 
       merged.status = normalizeTimelineStatus(raw.status ?? merged.status)
-      if (typeof raw.provider_id === 'string') merged.provider_id = raw.provider_id
-      if (typeof raw.provider_name === 'string') merged.provider_name = raw.provider_name
-      if (typeof raw.endpoint_id === 'string') merged.endpoint_id = raw.endpoint_id
-      if (typeof raw.key_id === 'string') merged.key_id = raw.key_id
-      if (typeof raw.key_name === 'string') merged.key_name = raw.key_name
+      if (!merged.provider_id && typeof raw.provider_id === 'string') merged.provider_id = raw.provider_id
+      if (!merged.provider_name && typeof raw.provider_name === 'string') merged.provider_name = raw.provider_name
+      if (!merged.endpoint_id && typeof raw.endpoint_id === 'string') merged.endpoint_id = raw.endpoint_id
+      if (!merged.key_id && typeof raw.key_id === 'string') merged.key_id = raw.key_id
+      if (!merged.key_name && typeof raw.key_name === 'string') merged.key_name = raw.key_name
       if (typeof raw.status_code === 'number') merged.status_code = raw.status_code
       if (typeof raw.error_type === 'string') merged.error_type = raw.error_type
       const rawPoolGroupId = typeof raw.pool_group_id === 'string' ? raw.pool_group_id.trim() : ''

frontend/src/features/usage/components/IntervalTimelineCard.vue

@@ -1,325 +0,0 @@
-<template>
-  <Card class="p-4">
-    <div class="flex items-center justify-between mb-3">
-      <p class="text-sm font-semibold">
-        {{ title }}
-      </p>
-      <div
-        v-if="displayLegendItems.length > 0"
-        class="flex items-center gap-2 flex-wrap justify-end text-[11px]"
-      >
-        <div
-          v-for="item in displayLegendItems"
-          :key="item.id"
-          class="flex items-center gap-1"
-        >
-          <div
-            class="w-2.5 h-2.5 rounded-full"
-            :style="{ backgroundColor: item.color }"
-          />
-          <span class="text-muted-foreground">{{ item.name }}</span>
-        </div>
-        <span
-          v-if="hiddenLegendCount > 0"
-          class="text-muted-foreground"
-        >
-          +{{ hiddenLegendCount }} 更多
-        </span>
-      </div>
-    </div>
-    <div
-      v-if="loading"
-      class="h-[160px] flex items-center justify-center"
-    >
-      <div class="text-sm text-muted-foreground">
-        Loading...
-      </div>
-    </div>
-    <div
-      v-else-if="hasData"
-      class="h-[160px]"
-    >
-      <ScatterChart
-        :data="chartData"
-        :options="chartOptions"
-      />
-    </div>
-    <div
-      v-else
-      class="h-[160px] flex items-center justify-center text-sm text-muted-foreground"
-    >
-      暂无请求间隔数据
-    </div>
-  </Card>
-</template>
-
-<script setup lang="ts">
-import { computed, ref, onMounted, onBeforeUnmount, watch } from 'vue'
-import Card from '@/components/ui/card.vue'
-import ScatterChart from '@/components/charts/ScatterChart.vue'
-import { cacheAnalysisApi, type IntervalTimelineResponse } from '@/api/cache'
-import { meApi } from '@/api/me'
-import type { ChartData, ChartOptions } from 'chart.js'
-import { log } from '@/utils/logger'
-
-const props = withDefaults(defineProps<{
-  title: string
-  isAdmin: boolean
-  hours?: number
-}>(), {
-  hours: 24  // 默认当天
-})
-
-const loading = ref(false)
-const timelineData = ref<IntervalTimelineResponse | null>(null)
-const primaryColor = ref('201, 100, 66')  // 默认主题色
-let loadRequestId = 0
-
-const ADMIN_TIMELINE_LIMIT = 1500
-const USER_TIMELINE_LIMIT = 1200
-
-// 获取主题色
-function getPrimaryColor(): string {
-  if (typeof window === 'undefined') return '201, 100, 66'
-  // CSS 变量定义在 body 上，不是 documentElement
-  const body = document.body
-  const style = getComputedStyle(body)
-  const rgb = style.getPropertyValue('--color-primary-rgb').trim()
-  return rgb || '201, 100, 66'
-}
-
-onMounted(() => {
-  primaryColor.value = getPrimaryColor()
-  void loadData()
-})
-
-// 预定义的颜色列表（用于区分不同用户/模型）
-const COLORS = [
-  'rgba(59, 130, 246, 0.7)',   // blue
-  'rgba(236, 72, 153, 0.7)',   // pink
-  'rgba(34, 197, 94, 0.7)',    // green
-  'rgba(249, 115, 22, 0.7)',   // orange
-  'rgba(168, 85, 247, 0.7)',   // purple
-  'rgba(234, 179, 8, 0.7)',    // yellow
-  'rgba(14, 165, 233, 0.7)',   // sky
-  'rgba(239, 68, 68, 0.7)',    // red
-  'rgba(20, 184, 166, 0.7)',   // teal
-  'rgba(99, 102, 241, 0.7)',   // indigo
-]
-
-const hasData = computed(() =>
-  timelineData.value && timelineData.value.points && timelineData.value.points.length > 0
-)
-
-// 判断是否有多个分组（管理员按用户分组，普通用户按模型分组）
-const hasMultipleGroups = computed(() => {
-  if (props.isAdmin) {
-    // 管理员视图：按用户分组
-    return timelineData.value?.users && Object.keys(timelineData.value.users).length > 1
-  } else {
-    // 用户视图：按模型分组
-    return timelineData.value?.models && timelineData.value.models.length > 1
-  }
-})
-
-// 图例最多显示数量
-const MAX_LEGEND_ITEMS = 6
-
-// 图例项（管理员显示用户，普通用户显示模型）
-const legendItems = computed(() => {
-  if (props.isAdmin && timelineData.value?.users) {
-    // 管理员视图：显示用户图例
-    const users = Object.entries(timelineData.value.users)
-    if (users.length <= 1) return []
-    return users.map(([userId, username], index) => ({
-      id: userId,
-      name: username || userId.slice(0, 8),
-      color: COLORS[index % COLORS.length]
-    }))
-  } else if (timelineData.value?.models && timelineData.value.models.length > 1) {
-    // 用户视图：显示模型图例
-    return timelineData.value.models.map((model, index) => ({
-      id: model,
-      name: formatModelName(model),
-      color: COLORS[index % COLORS.length]
-    }))
-  }
-  return []
-})
-
-// 显示的图例项（限制数量）
-const displayLegendItems = computed(() => {
-  return legendItems.value.slice(0, MAX_LEGEND_ITEMS)
-})
-
-// 隐藏的图例数量
-const hiddenLegendCount = computed(() => {
-  return Math.max(0, legendItems.value.length - MAX_LEGEND_ITEMS)
-})
-
-// 格式化模型名称，使其更简洁
-function formatModelName(model: string): string {
-  // 常见的简化规则
-  if (model.includes('claude')) {
-    // claude-3-5-sonnet-20241022 -> Claude 3.5 Sonnet
-    const match = model.match(/claude-(\d+)-(\d+)?-?(\w+)?/i)
-    if (match) {
-      const major = match[1]
-      const minor = match[2]
-      const variant = match[3]
-      let name = `Claude ${major}`
-      if (minor) name += `.${minor}`
-      if (variant) name += ` ${variant.charAt(0).toUpperCase() + variant.slice(1)}`
-      return name
-    }
-  }
-  // 其他模型保持原样但截断
-  return model.length > 20 ? `${model.slice(0, 17)  }...` : model
-}
-
-// 构建图表数据
-const chartData = computed<ChartData<'scatter'>>(() => {
-  if (!timelineData.value?.points) {
-    return { datasets: [] }
-  }
-
-  const points = timelineData.value.points
-
-  // 管理员视图且有多个用户：按用户分组
-  if (props.isAdmin && timelineData.value.users && Object.keys(timelineData.value.users).length > 1) {
-    const userIds = Object.keys(timelineData.value.users)
-    const userColorMap: Record<string, string> = {}
-    userIds.forEach((userId, index) => {
-      userColorMap[userId] = COLORS[index % COLORS.length]
-    })
-
-    // 按用户分组数据
-    const groupedData: Record<string, Array<{ x: string; y: number }>> = {}
-    for (const point of points) {
-      const userId = point.user_id || 'unknown'
-      if (!groupedData[userId]) {
-        groupedData[userId] = []
-      }
-      groupedData[userId].push({ x: point.x, y: point.y })
-    }
-
-    // 创建每个用户的 dataset
-    const datasets = Object.entries(groupedData).map(([userId, data]) => ({
-      label: timelineData.value?.users?.[userId] || userId.slice(0, 8),
-      data,
-      backgroundColor: userColorMap[userId] || 'rgba(59, 130, 246, 0.6)',
-      borderColor: userColorMap[userId] || 'rgba(59, 130, 246, 0.8)',
-      pointRadius: 3,
-      pointHoverRadius: 5,
-    }))
-
-    return { datasets }
-  }
-
-  // 用户视图且有多个模型：按模型分组
-  if (!props.isAdmin && timelineData.value.models && timelineData.value.models.length > 1) {
-    const models = timelineData.value.models
-    const modelColorMap: Record<string, string> = {}
-    models.forEach((model, index) => {
-      modelColorMap[model] = COLORS[index % COLORS.length]
-    })
-
-    // 按模型分组数据
-    const groupedData: Record<string, Array<{ x: string; y: number }>> = {}
-    for (const point of points) {
-      const model = point.model || 'unknown'
-      if (!groupedData[model]) {
-        groupedData[model] = []
-      }
-      groupedData[model].push({ x: point.x, y: point.y })
-    }
-
-    // 创建每个模型的 dataset
-    const datasets = Object.entries(groupedData).map(([model, data]) => ({
-      label: formatModelName(model),
-      data,
-      backgroundColor: modelColorMap[model] || 'rgba(59, 130, 246, 0.6)',
-      borderColor: modelColorMap[model] || 'rgba(59, 130, 246, 0.8)',
-      pointRadius: 3,
-      pointHoverRadius: 5,
-    }))
-
-    return { datasets }
-  }
-
-  // 单用户或单模型：使用主题色
-  return {
-    datasets: [{
-      label: '请求间隔',
-      data: points.map(p => ({ x: p.x, y: p.y })),
-      backgroundColor: `rgba(${primaryColor.value}, 0.6)`,
-      borderColor: `rgba(${primaryColor.value}, 0.8)`,
-      pointRadius: 3,
-      pointHoverRadius: 5,
-    }]
-  }
-})
-
-const chartOptions = computed<ChartOptions<'scatter'>>(() => ({
-  plugins: {
-    legend: {
-      display: false  // 使用自定义图例
-    },
-    tooltip: {
-      callbacks: {
-        label: (context) => {
-          const point = context.raw as { x: string; y: number; _originalY?: number }
-          const realY = point._originalY ?? point.y
-          const datasetLabel = context.dataset.label || ''
-          if (hasMultipleGroups.value) {
-            return `${datasetLabel}: ${realY.toFixed(1)} 分钟`
-          }
-          return `间隔: ${realY.toFixed(1)} 分钟`
-        }
-      }
-    }
-  }
-}))
-
-async function loadData() {
-  const requestId = ++loadRequestId
-  loading.value = true
-  try {
-    const limit = props.isAdmin ? ADMIN_TIMELINE_LIMIT : USER_TIMELINE_LIMIT
-    if (props.isAdmin) {
-      // 管理员：获取所有用户数据（按比例采样）
-      const data = await cacheAnalysisApi.getIntervalTimeline({
-        hours: props.hours,
-        include_user_info: true,
-        limit,
-      })
-      if (requestId !== loadRequestId) return
-      timelineData.value = data
-    } else {
-      // 普通用户：获取自己的数据
-      const data = await meApi.getIntervalTimeline({
-        hours: props.hours,
-        limit,
-      })
-      if (requestId !== loadRequestId) return
-      timelineData.value = data
-    }
-  } catch (error) {
-    if (requestId !== loadRequestId) return
-    log.error('加载请求间隔时间线失败:', error)
-    timelineData.value = null
-  } finally {
-    if (requestId === loadRequestId) {
-      loading.value = false
-    }
-  }
-}
-
-watch([() => props.hours, () => props.isAdmin], () => {
-  void loadData()
-})
-
-onBeforeUnmount(() => {
-  loadRequestId++
-})
-</script>

frontend/src/features/usage/components/ReplayDialog.vue

@@ -322,7 +322,7 @@
 
 <script setup lang="ts">
 import { ref, computed, watch } from 'vue'
-import { dashboardApi, type ReplayResponse, type RequestDetail } from '@/api/dashboard'
+import { requestDetailsApi, type ReplayResponse, type RequestDetail } from '@/api/request-details'
 import { getProvidersSummary } from '@/api/endpoints/providers'
 import { getProviderKeys } from '@/api/endpoints/keys'
 import type { EndpointAPIKey } from '@/api/endpoints/types'
@@ -482,7 +482,7 @@ async function doReplay() {
     if (selectedProviderId.value) params.provider_id = selectedProviderId.value
     if (selectedKeyId.value) params.api_key_id = selectedKeyId.value
 
-    replayResult.value = await dashboardApi.replayRequest(
+    replayResult.value = await requestDetailsApi.replayRequest(
       props.requestId,
       Object.keys(params).length > 0 ? params : undefined,
     )

frontend/src/features/usage/components/RequestDetailDrawer.vue

@@ -696,7 +696,7 @@ import Skeleton from '@/components/ui/skeleton.vue'
 import Tabs from '@/components/ui/tabs.vue'
 import TabsContent from '@/components/ui/tabs-content.vue'
 import { Copy, Check, Maximize2, Minimize2, Columns2, RefreshCw, X, Monitor, Server, MessageSquareText, Code2, Terminal, Play } from 'lucide-vue-next'
-import { dashboardApi, type RequestDetail } from '@/api/dashboard'
+import { requestDetailsApi, type RequestDetail } from '@/api/request-details'
 import { formatApiFormat } from '@/api/endpoints/types/api-format'
 import { log } from '@/utils/logger'
 
@@ -1319,7 +1319,7 @@ async function ensureBodyContentLoaded() {
   const requestId = ++bodyLoadRequestId
   bodyLoading.value = true
   try {
-    const response = await dashboardApi.getRequestDetail(props.requestId, { includeBodies: true })
+    const response = await requestDetailsApi.getRequestDetail(props.requestId, { includeBodies: true })
     if (requestId !== bodyLoadRequestId || !detail.value) return
     detail.value = {
       ...detail.value,
@@ -1359,7 +1359,7 @@ async function loadDetail(id: string, silent = false) {
   }
   error.value = null
   try {
-    const response = await dashboardApi.getRequestDetail(id, { includeBodies: false })
+    const response = await requestDetailsApi.getRequestDetail(id, { includeBodies: false })
     if (requestId !== loadDetailRequestId) return
 
     const previousDetail = detail.value
@@ -1737,7 +1737,7 @@ async function copyCurlCommand() {
   if (!props.requestId || curlCopying.value) return
   curlCopying.value = true
   try {
-    const data = await dashboardApi.getCurlData(props.requestId)
+    const data = await requestDetailsApi.getCurlData(props.requestId)
     if (data.curl) {
       copyToClipboard(data.curl, false)
       curlCopied.value = true

frontend/src/features/usage/components/RequestDetailDrawer/RequestHeadersContent.vue

@@ -159,7 +159,7 @@
 import { computed, ref } from 'vue'
 import Card from '@/components/ui/card.vue'
 import JsonContent from './JsonContent.vue'
-import type { RequestDetail } from '@/api/dashboard'
+import type { RequestDetail } from '@/api/request-details'
 
 const props = withDefaults(defineProps<{
   detail: RequestDetail

frontend/src/features/usage/components/UsageApiFormatTable.vue

@@ -1,100 +0,0 @@
-<template>
-  <Card class="overflow-hidden flex flex-col">
-    <div class="px-3 py-2 border-b flex-shrink-0">
-      <h3 class="text-sm font-medium">
-        按API格式分析
-      </h3>
-    </div>
-    <div class="overflow-auto max-h-[320px]">
-      <Table class="text-sm">
-        <TableHeader>
-          <TableRow>
-            <TableHead class="h-8 px-2">
-              API格式
-            </TableHead>
-            <TableHead class="h-8 px-2 text-right">
-              请求数
-            </TableHead>
-            <TableHead class="h-8 px-2 text-right">
-              <div class="flex flex-col text-xs gap-0.5 whitespace-nowrap">
-                <span>输入/输出</span>
-                <span class="text-muted-foreground font-normal">缓存</span>
-              </div>
-            </TableHead>
-            <TableHead class="h-8 px-2 text-right">
-              费用
-            </TableHead>
-            <TableHead class="h-8 px-2 text-right">
-              缓存命中率
-            </TableHead>
-            <TableHead class="h-8 px-2 text-right">
-              平均响应
-            </TableHead>
-          </TableRow>
-        </TableHeader>
-        <TableBody>
-          <TableRow v-if="data.length === 0">
-            <TableCell
-              :colspan="6"
-              class="text-center py-6 text-muted-foreground px-2"
-            >
-              暂无API格式统计数据
-            </TableCell>
-          </TableRow>
-          <TableRow
-            v-for="item in data"
-            :key="item.api_format"
-          >
-            <TableCell class="font-medium py-2 px-2">
-              {{ formatApiFormat(item.api_format) }}
-            </TableCell>
-            <TableCell class="text-right py-2 px-2">
-              {{ item.request_count }}
-            </TableCell>
-            <TableCell class="text-right py-2 px-2">
-              <div class="flex flex-col items-end text-xs gap-0.5 whitespace-nowrap">
-                <span>{{ formatTokens(item.total_input_context || 0) }} / {{ formatTokens(item.output_tokens || 0) }}</span>
-                <span class="text-muted-foreground">{{ formatTokens((item.cache_read_tokens || 0) + (item.cache_creation_tokens || 0)) }}</span>
-              </div>
-            </TableCell>
-            <TableCell class="text-right py-2 px-2">
-              <div class="flex flex-col items-end text-xs gap-0.5">
-                <span class="text-primary font-medium">{{ formatCurrency(item.total_cost) }}</span>
-                <span
-                  v-if="isAdmin && item.actual_cost !== undefined"
-                  class="text-muted-foreground text-[10px]"
-                >
-                  {{ formatCurrency(item.actual_cost) }}
-                </span>
-              </div>
-            </TableCell>
-            <TableCell class="text-right py-2 px-2">
-              <span>{{ formatHitRate(item.cache_hit_rate) }}</span>
-            </TableCell>
-            <TableCell class="text-right text-muted-foreground py-2 px-2">
-              {{ item.avgResponseTime }}
-            </TableCell>
-          </TableRow>
-        </TableBody>
-      </Table>
-    </div>
-  </Card>
-</template>
-
-<script setup lang="ts">
-import Card from '@/components/ui/card.vue'
-import Table from '@/components/ui/table.vue'
-import TableHeader from '@/components/ui/table-header.vue'
-import TableBody from '@/components/ui/table-body.vue'
-import TableRow from '@/components/ui/table-row.vue'
-import TableHead from '@/components/ui/table-head.vue'
-import TableCell from '@/components/ui/table-cell.vue'
-import { formatTokens, formatCurrency, formatHitRate } from '@/utils/format'
-import { formatApiFormat } from '@/api/endpoints/types/api-format'
-import type { ApiFormatStatsItem } from '../types'
-
-defineProps<{
-  data: ApiFormatStatsItem[]
-  isAdmin: boolean
-}>()
-</script>

frontend/src/features/usage/components/UsageModelTable.vue

@@ -1,99 +0,0 @@
-<template>
-  <Card class="overflow-hidden flex flex-col">
-    <div class="px-3 py-2 border-b flex-shrink-0">
-      <h3 class="text-sm font-medium">
-        按模型分析
-      </h3>
-    </div>
-    <div class="overflow-auto max-h-[320px]">
-      <Table class="text-sm">
-        <TableHeader>
-          <TableRow>
-            <TableHead class="h-8 px-2">
-              模型
-            </TableHead>
-            <TableHead class="h-8 px-2 text-right">
-              请求数
-            </TableHead>
-            <TableHead class="h-8 px-2 text-right">
-              <div class="flex flex-col text-xs gap-0.5 whitespace-nowrap">
-                <span>输入/输出</span>
-                <span class="text-muted-foreground font-normal">缓存</span>
-              </div>
-            </TableHead>
-            <TableHead class="h-8 px-2 text-right">
-              费用
-            </TableHead>
-            <TableHead class="h-8 px-2 text-right">
-              缓存命中率
-            </TableHead>
-            <TableHead class="h-8 px-2 text-right">
-              效率
-            </TableHead>
-          </TableRow>
-        </TableHeader>
-        <TableBody>
-          <TableRow v-if="data.length === 0">
-            <TableCell
-              :colspan="6"
-              class="text-center py-6 text-muted-foreground px-2"
-            >
-              暂无模型统计数据
-            </TableCell>
-          </TableRow>
-          <TableRow
-            v-for="model in data"
-            :key="model.model"
-          >
-            <TableCell class="font-medium py-2 px-2">
-              {{ model.model.replace('claude-', '') }}
-            </TableCell>
-            <TableCell class="text-right py-2 px-2">
-              {{ model.request_count }}
-            </TableCell>
-            <TableCell class="text-right py-2 px-2">
-              <div class="flex flex-col items-end text-xs gap-0.5 whitespace-nowrap">
-                <span>{{ formatTokens(model.total_input_context || 0) }} / {{ formatTokens(model.output_tokens || 0) }}</span>
-                <span class="text-muted-foreground">{{ formatTokens(model.cache_read_tokens || 0) }}</span>
-              </div>
-            </TableCell>
-            <TableCell class="text-right py-2 px-2">
-              <div class="flex flex-col items-end text-xs gap-0.5">
-                <span class="text-primary font-medium">{{ formatCurrency(model.total_cost) }}</span>
-                <span
-                  v-if="isAdmin && model.actual_cost !== undefined"
-                  class="text-muted-foreground text-[10px]"
-                >
-                  {{ formatCurrency(model.actual_cost) }}
-                </span>
-              </div>
-            </TableCell>
-            <TableCell class="text-right py-2 px-2">
-              <span>{{ formatHitRate(model.cache_hit_rate) }}</span>
-            </TableCell>
-            <TableCell class="text-right text-muted-foreground py-2 px-2">
-              {{ model.costPerToken }}
-            </TableCell>
-          </TableRow>
-        </TableBody>
-      </Table>
-    </div>
-  </Card>
-</template>
-
-<script setup lang="ts">
-import Card from '@/components/ui/card.vue'
-import Table from '@/components/ui/table.vue'
-import TableHeader from '@/components/ui/table-header.vue'
-import TableBody from '@/components/ui/table-body.vue'
-import TableRow from '@/components/ui/table-row.vue'
-import TableHead from '@/components/ui/table-head.vue'
-import TableCell from '@/components/ui/table-cell.vue'
-import { formatTokens, formatCurrency, formatHitRate } from '@/utils/format'
-import type { EnhancedModelStatsItem } from '../types'
-
-defineProps<{
-  data: EnhancedModelStatsItem[]
-  isAdmin: boolean
-}>()
-</script>

frontend/src/features/usage/components/UsageProviderTable.vue

@@ -1,111 +0,0 @@
-<template>
-  <Card class="overflow-hidden flex flex-col">
-    <div class="px-3 py-2 border-b flex-shrink-0">
-      <h3 class="text-sm font-medium">
-        按提供商分析
-      </h3>
-    </div>
-    <div class="overflow-auto max-h-[320px]">
-      <Table class="text-sm">
-        <TableHeader>
-          <TableRow>
-            <TableHead class="h-8 px-2">
-              提供商
-            </TableHead>
-            <TableHead class="h-8 px-2 text-right">
-              请求数
-            </TableHead>
-            <TableHead class="h-8 px-2 text-right">
-              <div class="flex flex-col text-xs gap-0.5 whitespace-nowrap">
-                <span>输入/输出</span>
-                <span class="text-muted-foreground font-normal">缓存</span>
-              </div>
-            </TableHead>
-            <TableHead class="h-8 px-2 text-right">
-              费用
-            </TableHead>
-            <TableHead class="h-8 px-2 text-right">
-              缓存命中率
-            </TableHead>
-            <TableHead class="h-8 px-2 text-right">
-              成功率
-            </TableHead>
-            <TableHead class="h-8 px-2 text-right">
-              平均响应
-            </TableHead>
-          </TableRow>
-        </TableHeader>
-        <TableBody>
-          <TableRow v-if="data.length === 0">
-            <TableCell
-              :colspan="7"
-              class="text-center py-6 text-muted-foreground px-2"
-            >
-              暂无提供商统计数据
-            </TableCell>
-          </TableRow>
-          <TableRow
-            v-for="provider in data"
-            :key="provider.provider"
-          >
-            <TableCell class="font-medium py-2 px-2">
-              {{ provider.provider }}
-            </TableCell>
-            <TableCell class="text-right py-2 px-2">
-              {{ provider.requests }}
-            </TableCell>
-            <TableCell class="text-right py-2 px-2">
-              <div class="flex flex-col items-end text-xs gap-0.5 whitespace-nowrap">
-                <span>{{ formatTokens(provider.totalInputContext || 0) }} / {{ formatTokens(provider.outputTokens || 0) }}</span>
-                <span class="text-muted-foreground">{{ formatTokens((provider.cacheReadTokens || 0) + (provider.cacheCreationTokens || 0)) }}</span>
-              </div>
-            </TableCell>
-            <TableCell class="text-right py-2 px-2">
-              <div class="flex flex-col items-end text-xs gap-0.5">
-                <span class="text-primary font-medium">{{ formatCurrency(provider.totalCost) }}</span>
-                <span
-                  v-if="isAdmin && provider.actualCost !== undefined"
-                  class="text-muted-foreground text-[10px]"
-                >
-                  {{ formatCurrency(provider.actualCost) }}
-                </span>
-              </div>
-            </TableCell>
-            <TableCell class="text-right py-2 px-2">
-              <span>{{ formatHitRate(provider.cacheHitRate) }}</span>
-            </TableCell>
-            <TableCell class="text-right py-2 px-2">
-              <span :class="getSuccessRateClass(provider.successRate)">{{ provider.successRate }}%</span>
-            </TableCell>
-            <TableCell class="text-right text-muted-foreground py-2 px-2">
-              {{ provider.avgResponseTime }}
-            </TableCell>
-          </TableRow>
-        </TableBody>
-      </Table>
-    </div>
-  </Card>
-</template>
-
-<script setup lang="ts">
-import Card from '@/components/ui/card.vue'
-import Table from '@/components/ui/table.vue'
-import TableHeader from '@/components/ui/table-header.vue'
-import TableBody from '@/components/ui/table-body.vue'
-import TableRow from '@/components/ui/table-row.vue'
-import TableHead from '@/components/ui/table-head.vue'
-import TableCell from '@/components/ui/table-cell.vue'
-import { formatTokens, formatCurrency, formatHitRate } from '@/utils/format'
-import type { ProviderStatsItem } from '../types'
-
-defineProps<{
-  data: ProviderStatsItem[]
-  isAdmin: boolean
-}>()
-
-// 成功率样式 - 简化为两种状态
-function getSuccessRateClass(rate: number): string {
-  if (rate < 90) return 'text-destructive'
-  return ''  // 默认颜色
-}
-</script>

frontend/src/features/usage/components/UsageRecordsTable.vue

@@ -36,10 +36,33 @@
           </SelectItem>
           <SelectItem
             v-for="user in availableUsers"
-            :key="user.id"
-            :value="user.id"
+            :key="user.value"
+            :value="user.value"
           >
-            {{ user.username || user.email }}
+            {{ user.label }}
+          </SelectItem>
+        </SelectContent>
+      </Select>
+
+      <!-- Key 筛选 -->
+      <Select
+        v-if="availableApiKeys.length > 0"
+        :model-value="filterApiKey"
+        @update:model-value="$emit('update:filterApiKey', $event)"
+      >
+        <SelectTrigger class="flex-1 min-w-0 sm:flex-none sm:w-36 h-8 text-xs border-border/60">
+          <SelectValue placeholder="Key" />
+        </SelectTrigger>
+        <SelectContent>
+          <SelectItem value="__all__">
+            全部 Key
+          </SelectItem>
+          <SelectItem
+            v-for="apiKey in availableApiKeys"
+            :key="apiKey.value"
+            :value="apiKey.value"
+          >
+            {{ apiKey.label }}
           </SelectItem>
         </SelectContent>
       </Select>
@@ -58,10 +81,10 @@
           </SelectItem>
           <SelectItem
             v-for="model in availableModels"
-            :key="model"
-            :value="model"
+            :key="model.value"
+            :value="model.value"
           >
-            {{ model.replace('claude-', '') }}
+            {{ model.label }}
           </SelectItem>
         </SelectContent>
       </Select>
@@ -81,10 +104,10 @@
           </SelectItem>
           <SelectItem
             v-for="provider in availableProviders"
-            :key="provider"
-            :value="provider"
+            :key="provider.value"
+            :value="provider.value"
           >
-            {{ provider }}
+            {{ provider.label }}
           </SelectItem>
         </SelectContent>
       </Select>
@@ -106,7 +129,7 @@
             :key="format.value"
             :value="format.value"
           >
-            {{ format.label }}
+            {{ getApiFormatOptionLabel(format) }}
           </SelectItem>
         </SelectContent>
       </Select>
@@ -123,26 +146,12 @@
           <SelectItem value="__all__">
             全部状态
           </SelectItem>
-          <SelectItem value="stream">
-            流式
-          </SelectItem>
-          <SelectItem value="standard">
-            标准
-          </SelectItem>
-          <SelectItem value="active">
-            活跃
-          </SelectItem>
-          <SelectItem value="failed">
-            失败
-          </SelectItem>
-          <SelectItem value="cancelled">
-            已取消
-          </SelectItem>
-          <SelectItem value="has_retry">
-            发生重试
-          </SelectItem>
-          <SelectItem value="has_fallback">
-            发生转移
+          <SelectItem
+            v-for="status in availableStatuses"
+            :key="status.value"
+            :value="status.value"
+          >
+            {{ getStatusOptionLabel(status) }}
           </SelectItem>
         </SelectContent>
       </Select>
@@ -350,18 +359,18 @@
           <TableCell
             v-if="isAdmin"
             class="py-4 w-[100px] truncate"
-            :title="record.username || record.user_email || (record.user_id ? `User ${record.user_id}` : '已删除用户')"
+            :title="getUserDisplayName(record)"
           >
             <div class="flex flex-col text-xs gap-0.5">
               <span class="truncate">
-                {{ record.username || record.user_email || (record.user_id ? `User ${record.user_id}` : '已删除用户') }}
+                {{ getUserDisplayName(record) }}
               </span>
               <span
-                v-if="record.api_key?.name"
+                v-if="getUserApiKeyLabel(record)"
                 class="text-muted-foreground truncate"
-                :title="record.api_key.name"
+                :title="getUserApiKeyLabel(record) || undefined"
               >
-                {{ record.api_key.name }}
+                {{ getUserApiKeyLabel(record) }}
               </span>
             </div>
           </TableCell>
@@ -369,10 +378,10 @@
           <TableCell
             v-if="!isAdmin"
             class="py-4 w-[100px]"
-            :title="record.api_key?.name || '-'"
+            :title="getUserApiKeyLabel(record) || '-'"
           >
             <div class="flex flex-col text-xs gap-0.5">
-              <span class="truncate">{{ record.api_key?.name || '-' }}</span>
+              <span class="truncate">{{ getUserApiKeyLabel(record) || '-' }}</span>
               <span
                 v-if="record.api_key?.display"
                 class="text-muted-foreground truncate"
@@ -417,7 +426,7 @@
           >
             <div class="flex items-center gap-1">
               <div class="flex flex-col text-xs gap-0.5">
-                <span>{{ record.provider }}</span>
+                <span>{{ formatProviderLabel(record.provider) }}</span>
                 <span
                   v-if="record.api_key_name"
                   class="text-muted-foreground truncate"
@@ -672,16 +681,11 @@ import { formatTokens, formatCurrency } from '@/utils/format'
 import { formatDateTime } from '../composables'
 import { useRowClick } from '@/composables/useRowClick'
 import { formatApiFormat } from '@/api/endpoints/types/api-format'
+import type { AnalyticsFilterOption } from '@/api/analytics'
 import type { DateRangeParams, UsageRecord } from '../types'
 import { TimeRangePicker } from '@/components/common'
 import ElapsedTimeText from './ElapsedTimeText.vue'
 
-export interface UserOption {
-  id: string
-  username: string
-  email: string
-}
-
 const props = defineProps<{
   records: UsageRecord[]
   isAdmin: boolean
@@ -692,13 +696,17 @@ const props = defineProps<{
   // 筛选
   filterSearch: string
   filterUser: string
+  filterApiKey: string
   filterModel: string
   filterProvider: string
   filterApiFormat: string
   filterStatus: string
-  availableUsers: UserOption[]
-  availableModels: string[]
-  availableProviders: string[]
+  availableUsers: AnalyticsFilterOption[]
+  availableApiKeys: AnalyticsFilterOption[]
+  availableModels: AnalyticsFilterOption[]
+  availableApiFormats: AnalyticsFilterOption[]
+  availableProviders: AnalyticsFilterOption[]
+  availableStatuses: AnalyticsFilterOption[]
   // 分页
   currentPage: number
   pageSize: number
@@ -712,6 +720,7 @@ const emit = defineEmits<{
   'update:timeRange': [value: DateRangeParams]
   'update:filterSearch': [value: string]
   'update:filterUser': [value: string]
+  'update:filterApiKey': [value: string]
   'update:filterModel': [value: string]
   'update:filterProvider': [value: string]
   'update:filterApiFormat': [value: string]
@@ -723,22 +732,6 @@ const emit = defineEmits<{
   'showDetail': [id: string]
 }>()
 
-// 静态常量（放在 defineProps/defineEmits 之后）
-const AVAILABLE_API_FORMATS = [
-  { value: 'openai:chat', label: 'OpenAI Chat' },
-  { value: 'openai:cli', label: 'OpenAI CLI' },
-  { value: 'openai:compact', label: 'OpenAI Compact' },
-  { value: 'openai:video', label: 'OpenAI Video' },
-  { value: 'claude:chat', label: 'Claude Chat' },
-  { value: 'claude:cli', label: 'Claude CLI' },
-  { value: 'gemini:chat', label: 'Gemini Chat' },
-  { value: 'gemini:cli', label: 'Gemini CLI' },
-  { value: 'gemini:video', label: 'Gemini Video' },
-] as const
-
-// 使用模块级常量
-const availableApiFormats = AVAILABLE_API_FORMATS
-
 const timeRangeModel = computed({
   get: () => props.timeRange,
   set: (value: DateRangeParams) => emit('update:timeRange', value)
@@ -760,6 +753,44 @@ watch(localSearch, (value) => {
   emitSearchDebounced(value)
 })
 
+function getApiFormatOptionLabel(option: AnalyticsFilterOption): string {
+  return formatApiFormat(option.label || option.value)
+}
+
+function getStatusOptionLabel(option: AnalyticsFilterOption): string {
+  const statusLabelMap: Record<string, string> = {
+    pending: '等待',
+    streaming: '传输中',
+    completed: '已完成',
+    failed: '失败',
+    cancelled: '已取消',
+    active: '活跃',
+    stream: '流式',
+    standard: '标准',
+    has_retry: '发生重试',
+    has_fallback: '发生转移',
+  }
+
+  return statusLabelMap[option.value] || option.label || option.value
+}
+
+function formatProviderLabel(value: string | null | undefined): string {
+  if (!value) return '-'
+  if (value === 'pending') return '待分配提供商'
+  if (value === 'unknown') return '未识别提供商'
+  return value
+}
+
+function getUserDisplayName(record: UsageRecord): string {
+  return record.username || record.user_email || (record.user_id ? `User ${record.user_id}` : '已删除用户')
+}
+
+function getUserApiKeyLabel(record: UsageRecord): string | null {
+  const name = record.api_key?.name?.trim()
+  if (name) return name
+  return record.api_key?.id ? null : '已删除Key'
+}
+
 // 使用复用的行点击逻辑
 const { handleMouseDown, shouldTriggerRowClick } = useRowClick()
 

frontend/src/features/usage/components/index.ts

@@ -1,8 +1,3 @@
-export { default as UsageModelTable } from './UsageModelTable.vue'
-export { default as UsageProviderTable } from './UsageProviderTable.vue'
-export { default as UsageApiFormatTable } from './UsageApiFormatTable.vue'
 export { default as UsageRecordsTable } from './UsageRecordsTable.vue'
 export { default as ActivityHeatmapCard } from './ActivityHeatmapCard.vue'
 export { default as RequestDetailDrawer } from './RequestDetailDrawer.vue'
-export { default as HorizontalRequestTimeline } from './HorizontalRequestTimeline.vue'
-export { default as IntervalTimelineCard } from './IntervalTimelineCard.vue'

frontend/src/features/usage/composables/useDateRange.ts

@@ -28,18 +28,17 @@ export function getDateRangeFromPeriod(period: PeriodValue): DateRangeParams {
     case 'today':
       startDate = new Date(now.getFullYear(), now.getMonth(), now.getDate())
       break
-    case 'yesterday':
-      startDate = new Date(now.getFullYear(), now.getMonth(), now.getDate() - 1)
-      endDate = new Date(now.getFullYear(), now.getMonth(), now.getDate())
-      break
     case 'last7days':
       startDate = new Date(now.getTime() - 7 * 24 * 60 * 60 * 1000)
       break
     case 'last30days':
       startDate = new Date(now.getTime() - 30 * 24 * 60 * 60 * 1000)
       break
-    case 'last90days':
-      startDate = new Date(now.getTime() - 90 * 24 * 60 * 60 * 1000)
+    case 'last180days':
+      startDate = new Date(now.getTime() - 180 * 24 * 60 * 60 * 1000)
+      break
+    case 'last1year':
+      startDate = new Date(now.getFullYear() - 1, now.getMonth(), now.getDate())
       break
     default:
       return {} // 返回空对象表示不过滤时间

frontend/src/features/usage/composables/useUsageData.ts

@@ -1,16 +1,10 @@
-import { ref, computed, type Ref } from 'vue'
-import { usageApi } from '@/api/usage'
-import { meApi } from '@/api/me'
+import { ref, type Ref } from 'vue'
+import { analyticsApi, type AnalyticsBaseRequest, type AnalyticsFilterOption, type AnalyticsRecord, type AnalyticsScope } from '@/api/analytics'
+import { buildTimeRangeParams } from '@/composables/useAnalyticsFilters'
 import type {
-  UsageStatsState,
-  ModelStatsItem,
-  ProviderStatsItem,
-  ApiFormatStatsItem,
   UsageRecord,
   DateRangeParams,
-  EnhancedModelStatsItem
 } from '../types'
-import { createDefaultStats } from '../types'
 import { log } from '@/utils/logger'
 import { getErrorStatus } from '@/types/api-error'
 
@@ -26,27 +20,21 @@ export interface PaginationParams {
 export interface FilterParams {
   search?: string
   user_id?: string
+  api_key_id?: string
   model?: string
   provider?: string
   api_format?: string
   status?: string
 }
 
+type FilterOptionDimension = 'user' | 'api_key' | 'model' | 'provider' | 'api_format' | 'status'
+
 export function useUsageData(options: UseUsageDataOptions) {
   const { isAdminPage } = options
 
-  // 加载状态
-  const isLoadingStats = ref(true)
   const isLoadingRecords = ref(false)
-  const loading = computed(() => isLoadingStats.value || isLoadingRecords.value)
 
-  // 统计数据
-  const stats = ref<UsageStatsState>(createDefaultStats())
-  const modelStats = ref<ModelStatsItem[]>([])
-  const providerStats = ref<ProviderStatsItem[]>([])
-  const apiFormatStats = ref<ApiFormatStatsItem[]>([])
-
-  // 记录数据 - 只存储当前页
+  // 记录数据：前后端统一按当前页维护
   const currentRecords = ref<UsageRecord[]>([])
   const totalRecords = ref(0)
 
@@ -56,195 +44,93 @@ export function useUsageData(options: UseUsageDataOptions) {
   let loadRecordsRequestId = 0
 
   // 可用的筛选选项（从统计数据获取，而不是从记录中）
-  const availableModels = ref<string[]>([])
-  const availableProviders = ref<string[]>([])
+  const availableUsers = ref<AnalyticsFilterOption[]>([])
+  const availableModels = ref<AnalyticsFilterOption[]>([])
+  const availableProviders = ref<AnalyticsFilterOption[]>([])
+  const availableApiKeys = ref<AnalyticsFilterOption[]>([])
+  const availableApiFormats = ref<AnalyticsFilterOption[]>([])
+  const availableStatuses = ref<AnalyticsFilterOption[]>([])
 
-  // 增强的模型统计（包含效率分析）
-  const enhancedModelStats = computed<EnhancedModelStatsItem[]>(() => {
-    return modelStats.value.map(model => ({
-      ...model,
-      costPerToken: model.total_tokens > 0
-        ? `$${(model.total_cost / model.total_tokens * 1000000).toFixed(2)}/M`
-        : '-'
-    }))
-  })
+  function mapAnalyticsRecord(record: AnalyticsRecord): UsageRecord {
+    return {
+      id: record.id,
+      user_id: record.user_id || undefined,
+      username: record.username || undefined,
+      api_key: {
+        id: record.api_key_id,
+        name: record.api_key_name,
+        display: null,
+      },
+      provider: record.provider_name,
+      api_key_name: record.provider_api_key_name || undefined,
+      rate_multiplier: record.rate_multiplier,
+      model: record.model,
+      target_model: record.target_model,
+      api_format: record.api_format || undefined,
+      endpoint_api_format: undefined,
+      has_format_conversion: record.has_format_conversion ?? undefined,
+      input_tokens: record.input_tokens,
+      output_tokens: record.output_tokens,
+      cache_creation_input_tokens: record.cache_creation_input_tokens,
+      cache_read_input_tokens: record.cache_read_input_tokens,
+      total_tokens: record.total_tokens,
+      cost: record.total_cost_usd,
+      actual_cost: record.actual_total_cost_usd,
+      response_time_ms: record.response_time_ms ?? undefined,
+      first_byte_time_ms: record.first_byte_time_ms ?? undefined,
+      is_stream: record.is_stream,
+      status_code: record.status_code ?? undefined,
+      error_message: record.error_message ?? undefined,
+      status: record.status as UsageRecord['status'],
+      created_at: record.created_at || '',
+      has_fallback: record.has_fallback ?? undefined,
+      has_retry: record.has_retry ?? undefined,
+    }
+  }
+
+  function buildScope(): AnalyticsScope {
+    if (isAdminPage.value) {
+      return { kind: 'global' }
+    }
+    return { kind: 'me' }
+  }
+
+  function buildAnalyticsFilters(
+    filters?: FilterParams,
+    excludeDimension?: FilterOptionDimension,
+  ): NonNullable<AnalyticsBaseRequest['filters']> {
+    return {
+      user_ids: filters?.user_id && excludeDimension !== 'user' ? [filters.user_id] : [],
+      models: filters?.model && excludeDimension !== 'model' ? [filters.model] : [],
+      provider_names: filters?.provider && excludeDimension !== 'provider' ? [filters.provider] : [],
+      api_key_ids: filters?.api_key_id && excludeDimension !== 'api_key' ? [filters.api_key_id] : [],
+      api_formats: filters?.api_format && excludeDimension !== 'api_format' ? [filters.api_format] : [],
+      statuses: filters?.status && excludeDimension !== 'status' ? [filters.status] : [],
+    }
+  }
 
   // 加载统计数据（不加载记录）
-  async function loadStats(dateRange?: DateRangeParams) {
+  async function loadStats(dateRange?: DateRangeParams, filters?: FilterParams) {
     const requestId = ++loadStatsRequestId
-    isLoadingStats.value = true
     currentDateRange.value = dateRange
 
     try {
-      if (isAdminPage.value) {
-        // 管理员页面顺序加载统计数据，避免刷新使用记录时瞬时打满后端 worker。
-        const statsData = await usageApi.getUsageStats(dateRange)
-
-        if (requestId !== loadStatsRequestId) {
-          return
-        }
-
-        // statsData may contain additional fields not declared in UsageStats
-        const statsRaw = statsData as Record<string, unknown>
-        stats.value = {
-          total_requests: statsData.total_requests || 0,
-          total_tokens: statsData.total_tokens || 0,
-          total_cost: statsData.total_cost || 0,
-          total_actual_cost: statsData.total_actual_cost,
-          avg_response_time: statsData.avg_response_time || 0,
-          error_count: typeof statsRaw.error_count === 'number' ? statsRaw.error_count : undefined,
-          error_rate: typeof statsRaw.error_rate === 'number' ? statsRaw.error_rate : undefined,
-          cache_stats: statsRaw.cache_stats as UsageStatsState['cache_stats'],
-          period_start: '',
-          period_end: '',
-        }
-
-        const modelData = await usageApi.getUsageByModel(dateRange)
-        if (requestId !== loadStatsRequestId) {
-          return
-        }
-
-        modelStats.value = modelData.map(item => {
-          const raw = item as Record<string, unknown>
-          return {
-            model: item.model,
-            request_count: item.request_count || 0,
-            total_tokens: item.total_tokens || 0,
-            total_input_context: typeof raw.total_input_context === 'number' ? raw.total_input_context : 0,
-            output_tokens: typeof raw.output_tokens === 'number' ? raw.output_tokens : 0,
-            cache_read_tokens: typeof raw.cache_read_tokens === 'number' ? raw.cache_read_tokens : 0,
-            cache_creation_tokens: typeof raw.cache_creation_tokens === 'number' ? raw.cache_creation_tokens : 0,
-            cache_hit_rate: typeof raw.cache_hit_rate === 'number' ? raw.cache_hit_rate : 0,
-            total_cost: item.total_cost || 0,
-            actual_cost: typeof raw.actual_cost === 'number' ? raw.actual_cost : undefined
-          }
+      const response = await analyticsApi.getFilterOptions({
+        scope: buildScope(),
+        time_range: buildTimeRangeParams(dateRange || currentDateRange.value || {}),
+        filters: buildAnalyticsFilters(filters),
       })
 
-        const providerData = await usageApi.getUsageByProvider(dateRange)
       if (requestId !== loadStatsRequestId) {
         return
       }
 
-        providerStats.value = providerData.map(item => ({
-          provider: item.provider,
-          requests: item.request_count,
-          totalTokens: item.total_tokens || 0,
-          totalInputContext: item.total_input_context || 0,
-          outputTokens: item.output_tokens || 0,
-          cacheReadTokens: item.cache_read_tokens || 0,
-          cacheCreationTokens: item.cache_creation_tokens || 0,
-          cacheHitRate: item.cache_hit_rate || 0,
-          totalCost: item.total_cost,
-          actualCost: item.actual_cost,
-          successRate: item.success_rate,
-          avgResponseTime: item.avg_response_time_ms > 0
-            ? `${(item.avg_response_time_ms / 1000).toFixed(2)}s`
-            : '-'
-        }))
-
-        const apiFormatData = await usageApi.getUsageByApiFormat(dateRange)
-        if (requestId !== loadStatsRequestId) {
-          return
-        }
-
-        apiFormatStats.value = apiFormatData.map(item => ({
-          api_format: item.api_format,
-          request_count: item.request_count || 0,
-          total_tokens: item.total_tokens || 0,
-          total_input_context: item.total_input_context || 0,
-          output_tokens: item.output_tokens || 0,
-          cache_read_tokens: item.cache_read_tokens || 0,
-          cache_creation_tokens: item.cache_creation_tokens || 0,
-          cache_hit_rate: item.cache_hit_rate || 0,
-          total_cost: item.total_cost || 0,
-          actual_cost: item.actual_cost,
-          avgResponseTime: item.avg_response_time_ms > 0
-            ? `${(item.avg_response_time_ms / 1000).toFixed(2)}s`
-            : '-'
-        }))
-
-        // 从统计数据中提取可用的筛选选项
-        availableModels.value = modelData.map(item => item.model).filter(Boolean).sort()
-        availableProviders.value = providerData.map(item => item.provider).filter(Boolean).sort()
-
-      } else {
-        // 用户页面
-        const userData = await meApi.getUsage(dateRange)
-        if (requestId !== loadStatsRequestId) {
-          return
-        }
-
-        stats.value = {
-          total_requests: userData.total_requests || 0,
-          total_tokens: userData.total_tokens || 0,
-          total_cost: userData.total_cost || 0,
-          total_actual_cost: userData.total_actual_cost,
-          avg_response_time: userData.avg_response_time || 0,
-          period_start: '',
-          period_end: '',
-        }
-
-        modelStats.value = (userData.summary_by_model || []).map((item) => ({
-          model: item.model,
-          request_count: item.requests || 0,
-          total_tokens: item.total_tokens || 0,
-          total_input_context: item.total_input_context || 0,
-          output_tokens: item.output_tokens || 0,
-          cache_read_tokens: item.cache_read_tokens || 0,
-          cache_creation_tokens: item.cache_creation_tokens || 0,
-          cache_hit_rate: item.cache_hit_rate || 0,
-          total_cost: item.total_cost_usd || 0,
-          actual_cost: item.actual_total_cost_usd
-        }))
-
-        providerStats.value = (userData.summary_by_provider || []).map((item) => ({
-          provider: item.provider,
-          requests: item.requests || 0,
-          totalTokens: item.total_tokens || 0,
-          totalInputContext: item.total_input_context || 0,
-          outputTokens: item.output_tokens || 0,
-          cacheReadTokens: item.cache_read_tokens || 0,
-          cacheCreationTokens: item.cache_creation_tokens || 0,
-          cacheHitRate: item.cache_hit_rate || 0,
-          totalCost: item.total_cost_usd || 0,
-          successRate: item.success_rate || 0,
-          avgResponseTime: (item.avg_response_time_ms ?? 0) > 0
-            ? `${((item.avg_response_time_ms ?? 0) / 1000).toFixed(2)}s`
-            : '-'
-        }))
-
-        // 用户页面：记录直接从 userData 获取（数量较少）
-        // 使用 mergeRecordStatus 保护已有的活跃状态，避免轮询更新被覆盖
-        const nextRecords = (userData.records || []) as UsageRecord[]
-        currentRecords.value = mergeRecordStatus(currentRecords.value, nextRecords)
-        totalRecords.value = userData.pagination?.total ?? currentRecords.value.length
-
-        // 从记录中提取筛选选项
-        const models = new Set<string>()
-        const providers = new Set<string>()
-        currentRecords.value.forEach(record => {
-          if (record.model) models.add(record.model)
-          if (record.provider) providers.add(record.provider)
-        })
-        availableModels.value = Array.from(models).sort()
-        availableProviders.value = Array.from(providers).sort()
-
-        // API 格式统计直接使用后端聚合数据
-        apiFormatStats.value = (userData.summary_by_api_format || []).map(item => ({
-          api_format: item.api_format,
-          request_count: item.request_count || 0,
-          total_tokens: item.total_tokens || 0,
-          total_input_context: item.total_input_context || 0,
-          output_tokens: item.output_tokens || 0,
-          cache_read_tokens: item.cache_read_tokens || 0,
-          cache_creation_tokens: item.cache_creation_tokens || 0,
-          cache_hit_rate: item.cache_hit_rate || 0,
-          total_cost: item.total_cost_usd || 0,
-          avgResponseTime: (item.avg_response_time_ms ?? 0) > 0
-            ? `${((item.avg_response_time_ms ?? 0) / 1000).toFixed(2)}s`
-            : '-'
-        }))
-      }
+      availableUsers.value = response.users ?? []
+      availableModels.value = response.models ?? []
+      availableProviders.value = response.providers ?? []
+      availableApiKeys.value = response.api_keys ?? []
+      availableApiFormats.value = response.api_formats ?? []
+      availableStatuses.value = response.statuses ?? []
     } catch (error: unknown) {
       if (requestId !== loadStatsRequestId) {
         return
@@ -252,17 +138,17 @@ export function useUsageData(options: UseUsageDataOptions) {
       if (getErrorStatus(error) !== 403) {
         log.error('加载统计数据失败:', error)
       }
-      stats.value = createDefaultStats()
-      modelStats.value = []
+      availableUsers.value = []
+      availableModels.value = []
+      availableProviders.value = []
       currentRecords.value = []
-    } finally {
-      if (requestId === loadStatsRequestId) {
-        isLoadingStats.value = false
-      }
+      availableApiKeys.value = []
+      availableApiFormats.value = []
+      availableStatuses.value = []
     }
   }
 
-  // 加载记录（真正的后端分页）
+  // 加载记录：前后端统一由后端分页和筛选
   async function loadRecords(
     pagination: PaginationParams,
     filters?: FilterParams,
@@ -278,53 +164,24 @@ export function useUsageData(options: UseUsageDataOptions) {
         currentDateRange.value = dateRange
       }
 
-      // 构建请求参数
-      const params: Record<string, unknown> = {
+      const response = await analyticsApi.getRecords({
+        scope: buildScope(),
+        time_range: buildTimeRangeParams(currentDateRange.value || {}),
+        filters: buildAnalyticsFilters(filters),
+        search: {
+          text: filters?.search || null,
+        },
+        pagination: {
           limit: pagination.pageSize,
           offset,
-        ...effectiveDateRange
-      }
-
-      // 添加筛选条件
-      if (filters?.search?.trim()) {
-        params.search = filters.search.trim()
-      }
-
-      if (isAdminPage.value) {
-        // 管理员页面：使用管理员 API
-        if (filters?.user_id) {
-          params.user_id = filters.user_id
-        }
-        if (filters?.model) {
-          params.model = filters.model
-        }
-        if (filters?.provider) {
-          params.provider = filters.provider
-        }
-        if (filters?.api_format) {
-          params.api_format = filters.api_format
-        }
-        if (filters?.status) {
-          params.status = filters.status
-        }
-
-        const response = await usageApi.getAllUsageRecords(params)
+        },
+      })
       if (requestId !== loadRecordsRequestId) {
         return
       }
-        const nextRecords = (response.records || []) as UsageRecord[]
+      const nextRecords = (response.records || []).map(mapAnalyticsRecord)
       currentRecords.value = mergeRecordStatus(currentRecords.value, nextRecords)
       totalRecords.value = response.total || 0
-      } else {
-        // 用户页面：使用用户 API
-        const userData = await meApi.getUsage(params)
-        if (requestId !== loadRecordsRequestId) {
-          return
-        }
-        const nextRecords = (userData.records || []) as UsageRecord[]
-        currentRecords.value = mergeRecordStatus(currentRecords.value, nextRecords)
-        totalRecords.value = userData.pagination?.total || currentRecords.value.length
-      }
     } catch (error) {
       if (requestId !== loadRecordsRequestId) {
         return
@@ -412,33 +269,19 @@ export function useUsageData(options: UseUsageDataOptions) {
     })
   }
 
-  // 刷新所有数据
-  async function refreshData(dateRange?: DateRangeParams) {
-    await loadStats(dateRange)
-  }
-
   return {
-    // 状态
-    loading,
-    isLoadingStats,
     isLoadingRecords,
-    stats,
-    modelStats,
-    providerStats,
-    apiFormatStats,
     currentRecords,
     totalRecords,
 
-    // 筛选选项
+    availableUsers,
     availableModels,
     availableProviders,
+    availableApiKeys,
+    availableApiFormats,
+    availableStatuses,
 
-    // 计算属性
-    enhancedModelStats,
-
-    // 方法
     loadStats,
     loadRecords,
-    refreshData
   }
 }

frontend/src/features/usage/types.ts

@@ -1,74 +1,3 @@
-// 统计数据状态
-export interface UsageStatsState {
-  total_requests: number
-  total_tokens: number
-  total_cost: number
-  total_actual_cost?: number  // 倍率消耗（仅管理员可见）
-  avg_response_time: number
-  error_count?: number
-  error_rate?: number
-  cache_stats?: {
-    cache_creation_tokens: number
-    cache_read_tokens: number
-    cache_creation_cost: number
-    cache_read_cost: number
-  }
-  period_start: string
-  period_end: string
-}
-
-// 模型统计
-export interface ModelStatsItem {
-  model: string
-  request_count: number
-  total_tokens: number
-  total_input_context?: number
-  output_tokens?: number
-  cache_read_tokens?: number
-  cache_creation_tokens?: number
-  cache_hit_rate?: number
-  total_cost: number
-  actual_cost?: number  // 倍率消耗
-}
-
-// 增强的模型统计（包含效率分析）
-export interface EnhancedModelStatsItem extends ModelStatsItem {
-  costPerToken: string
-}
-
-// 提供商统计
-export interface ProviderStatsItem {
-  provider: string
-  requests: number
-  totalTokens: number
-  totalInputContext?: number
-  outputTokens?: number
-  cacheReadTokens?: number
-  cacheCreationTokens?: number
-  cacheHitRate?: number
-  totalCost: number
-  actualCost?: number
-  successRate: number
-  avgResponseTime: string
-}
-
-// API格式统计
-export interface ApiFormatStatsItem {
-  api_format: string
-  request_count: number
-  total_tokens: number
-  total_input_context?: number
-  output_tokens?: number
-  cache_read_tokens?: number
-  cache_creation_tokens?: number
-  cache_hit_rate?: number
-  total_cost: number
-  actual_cost?: number
-  avgResponseTime: string
-}
-
-// 请求记录
-// 请求状态类型
 export type RequestStatus = 'pending' | 'streaming' | 'completed' | 'failed' | 'cancelled'
 
 export interface UsageRecord {
@@ -82,7 +11,7 @@ export interface UsageRecord {
     display: string | null
   } | null
   provider?: string  // 仅管理员可见
-  api_key_name?: string
+  api_key_name?: string  // 提供商 Key 名称（管理员列使用）
   rate_multiplier?: number
   model: string
   target_model?: string | null  // 映射后的目标模型名（若无映射则为空）
@@ -113,29 +42,13 @@ export interface DateRangeParams {
   start_date?: string
   end_date?: string
   preset?: string
-  granularity?: 'hour' | 'day' | 'week' | 'month'
+  granularity?: 'auto' | 'hour' | 'day' | 'week' | 'month'
   timezone?: string
   tz_offset_minutes?: number
 }
 
 // 时间段选项
-export type PeriodValue = 'today' | 'yesterday' | 'last7days' | 'last30days' | 'last90days'
+export type PeriodValue = 'today' | 'last7days' | 'last30days' | 'last180days' | 'last1year'
 
-// 筛选状态（简化为常用维度）
-export type FilterStatusValue = '__all__' | 'stream' | 'standard' | 'active' | 'failed' | 'cancelled'
-
-// 默认统计状态
-export function createDefaultStats(): UsageStatsState {
-  return {
-    total_requests: 0,
-    total_tokens: 0,
-    total_cost: 0,
-    total_actual_cost: undefined,
-    avg_response_time: 0,
-    error_count: undefined,
-    error_rate: undefined,
-    cache_stats: undefined,
-    period_start: '',
-    period_end: ''
-  }
-}
+// 筛选状态由后端 filter-options 决定
+export type FilterStatusValue = string

frontend/src/layouts/MainLayout.vue

@@ -516,6 +516,7 @@ const navigation = computed(() => {
       items: [
         { name: '仪表盘', href: '/dashboard', icon: Home },
         { name: '健康监控', href: '/dashboard/endpoint-status', icon: Activity },
+        { name: '综合报表', href: '/dashboard/reports', icon: BarChart3 },
       ]
     },
     {
@@ -529,7 +530,7 @@ const navigation = computed(() => {
       title: '账户',
       items: [
          { name: '钱包中心', href: '/dashboard/wallet', icon: Wallet },
-         { name: '使用统计', href: '/dashboard/usage', icon: BarChart3 },
+         { name: '使用记录', href: '/dashboard/usage', icon: Activity },
          { name: '异步任务', href: '/dashboard/async-tasks', icon: Zap },
       ]
     }
@@ -573,10 +574,8 @@ const navigation = computed(() => {
       title: '概览',
       items: [
         { name: '仪表盘', href: '/admin/dashboard', icon: Home },
+        { name: '数据分析', href: '/admin/analytics', icon: BarChart3 },
         { name: '健康监控', href: '/admin/health-monitor', icon: Activity },
-        { name: '用户统计', href: '/admin/user-stats', icon: BarChart3 },
-        { name: '成本分析', href: '/admin/cost-analysis', icon: Gauge },
-        { name: '性能分析', href: '/admin/performance-analysis', icon: Activity },
       ]
     },
     {

frontend/src/mocks/data.ts

@@ -4,10 +4,9 @@
  */
 
 import type { User, LoginResponse } from '@/api/auth'
-import type { DashboardStatsResponse, RecentRequest, ProviderStatus, DailyStatsResponse } from '@/api/dashboard'
 import type { User as AdminUser } from '@/api/users'
 import type { AdminApiKeysResponse } from '@/api/admin'
-import type { Profile, UsageResponse } from '@/api/me'
+import type { Profile } from '@/api/me'
 import type { ProviderWithEndpointsSummary, GlobalModelResponse } from '@/api/endpoints/types'
 
 // ========== 用户数据 ==========
@@ -128,163 +127,6 @@ export const MOCK_USER_PROFILE: Profile = {
   }
 }
 
-// ========== Dashboard 数据 ==========
-
-export const MOCK_DASHBOARD_STATS: DashboardStatsResponse = {
-  stats: [
-    {
-      name: '今日请求',
-      value: '1,234',
-      subValue: '成功率 99.2%',
-      change: '+12.5%',
-      changeType: 'increase',
-      icon: 'Activity'
-    },
-    {
-      name: '今日 Token',
-      value: '2.5M',
-      subValue: '输入 1.8M / 输出 0.7M',
-      change: '+8.3%',
-      changeType: 'increase',
-      icon: 'Zap'
-    },
-    {
-      name: '今日费用',
-      value: '$45.67',
-      subValue: '节省 $12.34 (21%)',
-      change: '-5.2%',
-      changeType: 'decrease',
-      icon: 'DollarSign'
-    },
-    {
-      name: '活跃用户',
-      value: '28',
-      subValue: '总用户 156',
-      change: '+3',
-      changeType: 'increase',
-      icon: 'Users'
-    }
-  ],
-  today: {
-    requests: 1234,
-    tokens: 2500000,
-    cost: 45.67,
-    actual_cost: 33.33,
-    cache_creation_tokens: 50000,
-    cache_read_tokens: 200000
-  },
-  api_keys: {
-    total: 45,
-    active: 38
-  },
-  tokens: {
-    month: 75000000
-  },
-  system_health: {
-    avg_response_time: 1.23,
-    error_rate: 0.8,
-    error_requests: 10,
-    fallback_count: 5,
-    total_requests: 1234
-  },
-  cost_stats: {
-    total_cost: 45.67,
-    total_actual_cost: 33.33,
-    cost_savings: 12.34
-  },
-  cache_stats: {
-    cache_creation_tokens: 50000,
-    cache_read_tokens: 200000,
-    cache_creation_cost: 0.25,
-    cache_read_cost: 0.10,
-    cache_hit_rate: 0.35,
-    total_cache_tokens: 250000
-  },
-  users: {
-    total: 156,
-    active: 28
-  },
-  token_breakdown: {
-    input: 1800000,
-    output: 700000,
-    cache_creation: 50000,
-    cache_read: 200000
-  },
-  // 普通用户专用字段
-  monthly_cost: 45.67
-}
-
-export const MOCK_RECENT_REQUESTS: RecentRequest[] = [
-  { id: 'req-001', user: 'alice', model: 'claude-sonnet-4-5-20250929', tokens: 15234, time: '2 分钟前' },
-  { id: 'req-002', user: 'bob', model: 'gpt-5.1', tokens: 8765, time: '5 分钟前' },
-  { id: 'req-003', user: 'charlie', model: 'claude-opus-4-5-20251101', tokens: 32100, time: '8 分钟前' },
-  { id: 'req-004', user: 'diana', model: 'gemini-3-pro-preview', tokens: 4521, time: '12 分钟前' },
-  { id: 'req-005', user: 'eve', model: 'claude-sonnet-4-5-20250929', tokens: 9876, time: '15 分钟前' },
-  { id: 'req-006', user: 'frank', model: 'gpt-5.1-codex-mini', tokens: 2345, time: '18 分钟前' },
-  { id: 'req-007', user: 'grace', model: 'claude-haiku-4-5-20251001', tokens: 6789, time: '22 分钟前' },
-  { id: 'req-008', user: 'henry', model: 'gemini-3-pro-preview', tokens: 12345, time: '25 分钟前' }
-]
-
-export const MOCK_PROVIDER_STATUS: ProviderStatus[] = [
-  { name: 'Anthropic Official', status: 'active', requests: 456 },
-  { name: 'OpenAI Official', status: 'active', requests: 389 },
-  { name: 'Google AI', status: 'active', requests: 234 },
-  { name: 'AWS Bedrock', status: 'active', requests: 89 },
-  { name: 'Azure OpenAI', status: 'inactive', requests: 0 },
-  { name: 'Vertex AI', status: 'active', requests: 66 }
-]
-
-// 生成过去7天的每日统计数据
-function generateDailyStats(): DailyStatsResponse {
-  const dailyStats = []
-  const now = new Date()
-
-  for (let i = 6; i >= 0; i--) {
-    const date = new Date(now)
-    date.setDate(date.getDate() - i)
-    const dateStr = date.toISOString().split('T')[0]
-
-    const baseRequests = 800 + Math.floor(Math.random() * 600)
-    const baseTokens = 1500000 + Math.floor(Math.random() * 1500000)
-    const baseCost = 30 + Math.random() * 30
-
-    dailyStats.push({
-      date: dateStr,
-      requests: baseRequests,
-      tokens: baseTokens,
-      cost: Number(baseCost.toFixed(2)),
-      avg_response_time: 0.8 + Math.random() * 0.8,
-      unique_models: 8 + Math.floor(Math.random() * 5),
-      unique_providers: 4 + Math.floor(Math.random() * 3),
-      model_breakdown: [
-        { model: 'claude-sonnet-4-5-20250929', requests: Math.floor(baseRequests * 0.35), tokens: Math.floor(baseTokens * 0.35), cost: Number((baseCost * 0.35).toFixed(2)) },
-        { model: 'gpt-5.1', requests: Math.floor(baseRequests * 0.25), tokens: Math.floor(baseTokens * 0.25), cost: Number((baseCost * 0.25).toFixed(2)) },
-        { model: 'claude-opus-4-5-20251101', requests: Math.floor(baseRequests * 0.15), tokens: Math.floor(baseTokens * 0.15), cost: Number((baseCost * 0.20).toFixed(2)) },
-        { model: 'gemini-3-pro-preview', requests: Math.floor(baseRequests * 0.15), tokens: Math.floor(baseTokens * 0.15), cost: Number((baseCost * 0.10).toFixed(2)) },
-        { model: 'claude-haiku-4-5-20251001', requests: Math.floor(baseRequests * 0.10), tokens: Math.floor(baseTokens * 0.10), cost: Number((baseCost * 0.10).toFixed(2)) }
-      ]
-    })
-  }
-
-  return {
-    daily_stats: dailyStats,
-    model_summary: [
-      { model: 'claude-sonnet-4-5-20250929', requests: 2456, tokens: 8500000, cost: 125.45, avg_response_time: 1.2, cost_per_request: 0.051, tokens_per_request: 3461 },
-      { model: 'gpt-5.1', requests: 1823, tokens: 6200000, cost: 98.32, avg_response_time: 0.9, cost_per_request: 0.054, tokens_per_request: 3401 },
-      { model: 'claude-opus-4-5-20251101', requests: 987, tokens: 4100000, cost: 156.78, avg_response_time: 2.1, cost_per_request: 0.159, tokens_per_request: 4154 },
-      { model: 'gemini-3-pro-preview', requests: 1234, tokens: 3800000, cost: 28.56, avg_response_time: 0.6, cost_per_request: 0.023, tokens_per_request: 3079 },
-      { model: 'claude-haiku-4-5-20251001', requests: 2100, tokens: 5200000, cost: 32.10, avg_response_time: 0.5, cost_per_request: 0.015, tokens_per_request: 2476 }
-    ],
-    period: {
-      start_date: dailyStats[0].date,
-      end_date: dailyStats[dailyStats.length - 1].date,
-      days: 7
-    }
-  }
-}
-
-export const MOCK_DAILY_STATS = generateDailyStats()
-
 // ========== 用户管理数据 ==========
 
 export const MOCK_ALL_USERS: AdminUser[] = [
@@ -808,57 +650,6 @@ export const MOCK_GLOBAL_MODELS: GlobalModelResponse[] = [
   }
 ]
 
-// ========== Usage 数据 ==========
-
-export const MOCK_USAGE_RESPONSE: UsageResponse = {
-  total_requests: 1234,
-  total_input_tokens: 1800000,
-  total_output_tokens: 700000,
-  total_tokens: 2500000,
-  total_cost: 45.67,
-  total_actual_cost: 33.33,
-  avg_response_time: 1.23,
-  billing: MOCK_USER_BILLING,
-  summary_by_model: [
-    { model: 'claude-sonnet-4-5-20250929', requests: 456, input_tokens: 650000, output_tokens: 250000, total_tokens: 900000, total_cost_usd: 18.50, actual_total_cost_usd: 13.50 },
-    { model: 'gpt-5.1', requests: 312, input_tokens: 480000, output_tokens: 180000, total_tokens: 660000, total_cost_usd: 12.30, actual_total_cost_usd: 9.20 },
-    { model: 'claude-haiku-4-5-20251001', requests: 289, input_tokens: 420000, output_tokens: 170000, total_tokens: 590000, total_cost_usd: 8.50, actual_total_cost_usd: 6.30 },
-    { model: 'gemini-3-pro-preview', requests: 177, input_tokens: 250000, output_tokens: 100000, total_tokens: 350000, total_cost_usd: 6.37, actual_total_cost_usd: 4.33 }
-  ],
-  records: [
-    {
-      id: 'usage-001',
-      provider: 'anthropic',
-      model: 'claude-sonnet-4-5-20250929',
-      input_tokens: 1500,
-      output_tokens: 800,
-      total_tokens: 2300,
-      cost: 0.0165,
-      response_time_ms: 1234,
-      is_stream: true,
-      created_at: new Date().toISOString(),
-      status_code: 200,
-      input_price_per_1m: 3,
-      output_price_per_1m: 15
-    },
-    {
-      id: 'usage-002',
-      provider: 'openai',
-      model: 'gpt-5.1',
-      input_tokens: 2000,
-      output_tokens: 500,
-      total_tokens: 2500,
-      cost: 0.01,
-      response_time_ms: 890,
-      is_stream: false,
-      created_at: new Date(Date.now() - 300000).toISOString(),
-      status_code: 200,
-      input_price_per_1m: 2.5,
-      output_price_per_1m: 10
-    }
-  ]
-}
-
 // ========== 系统配置 ==========
 
 export const MOCK_SYSTEM_CONFIGS = [

frontend/src/router/index.ts

@@ -112,6 +112,11 @@ const routes: RouteRecordRaw[] = [
         name: 'MyUsage',
         component: () => importWithRetry(() => import('@/views/shared/Usage.vue'))
       },
+      {
+        path: 'reports',
+        name: 'Reports',
+        component: () => importWithRetry(() => import('@/views/user/Reports.vue'))
+      },
       {
         path: 'endpoint-status',
         name: 'EndpointStatus',
@@ -196,19 +201,9 @@ const routes: RouteRecordRaw[] = [
         component: () => importWithRetry(() => import('@/views/shared/Usage.vue'))
       },
       {
-        path: 'user-stats',
-        name: 'UserStats',
-        component: () => importWithRetry(() => import('@/views/admin/UserStats.vue'))
-      },
-      {
-        path: 'cost-analysis',
-        name: 'CostAnalysis',
-        component: () => importWithRetry(() => import('@/views/admin/CostAnalysis.vue'))
-      },
-      {
-        path: 'performance-analysis',
-        name: 'PerformanceAnalysis',
-        component: () => importWithRetry(() => import('@/views/admin/PerformanceAnalysis.vue'))
+        path: 'analytics',
+        name: 'Analytics',
+        component: () => importWithRetry(() => import('@/views/admin/Analytics.vue'))
       },
       {
         path: 'system',
@@ -272,11 +267,6 @@ const routes: RouteRecordRaw[] = [
         path: 'gemini-files',
         name: 'GeminiFilesManagement',
         component: () => importWithRetry(() => import('@/views/admin/GeminiFilesManagement.vue'))
-      },
-      // 保留旧路由兼容性
-      {
-        path: 'video-tasks',
-        redirect: '/admin/async-tasks'
       }
     ]
   }

frontend/src/utils/analyticsGranularity.ts

@@ -0,0 +1,137 @@
+import type { DateRangeParams } from '@/features/usage/types'
+
+export type AnalyticsGranularityOption = NonNullable<DateRangeParams['granularity']>
+export type ResolvedAnalyticsGranularity = Exclude<AnalyticsGranularityOption, 'auto'>
+
+export const analyticsGranularityTabs: Array<{ value: AnalyticsGranularityOption; label: string }> = [
+  { value: 'auto', label: '自动' },
+  { value: 'hour', label: '小时' },
+  { value: 'day', label: '天' },
+  { value: 'week', label: '周' },
+  { value: 'month', label: '月' },
+]
+
+export function parseAnalyticsDateKey(dateKey?: string): Date | null {
+  if (!dateKey) return null
+  return new Date(`${dateKey}T00:00:00`)
+}
+
+export function resolveAnalyticsRangeBounds(range: DateRangeParams): { startDate: Date | null; endDate: Date | null } {
+  const today = new Date()
+
+  switch (range.preset) {
+    case 'today':
+      return {
+        startDate: new Date(today.getFullYear(), today.getMonth(), today.getDate()),
+        endDate: new Date(today.getFullYear(), today.getMonth(), today.getDate()),
+      }
+    case 'last7days':
+      return {
+        startDate: new Date(today.getFullYear(), today.getMonth(), today.getDate() - 6),
+        endDate: new Date(today.getFullYear(), today.getMonth(), today.getDate()),
+      }
+    case 'last30days':
+      return {
+        startDate: new Date(today.getFullYear(), today.getMonth(), today.getDate() - 29),
+        endDate: new Date(today.getFullYear(), today.getMonth(), today.getDate()),
+      }
+    case 'last180days':
+      return {
+        startDate: new Date(today.getFullYear(), today.getMonth(), today.getDate() - 179),
+        endDate: new Date(today.getFullYear(), today.getMonth(), today.getDate()),
+      }
+    case 'last1year':
+      return {
+        startDate: new Date(today.getFullYear() - 1, today.getMonth(), today.getDate()),
+        endDate: new Date(today.getFullYear(), today.getMonth(), today.getDate()),
+      }
+    default:
+      return {
+        startDate: parseAnalyticsDateKey(range.start_date),
+        endDate: parseAnalyticsDateKey(range.end_date),
+      }
+  }
+}
+
+export function getAnalyticsRangeDaysInclusive(range: DateRangeParams): number | null {
+  const { startDate, endDate } = resolveAnalyticsRangeBounds(range)
+  if (!startDate || !endDate) return null
+  const diff = endDate.getTime() - startDate.getTime()
+  if (diff < 0) return null
+  return Math.floor(diff / (24 * 60 * 60 * 1000)) + 1
+}
+
+export function isAnalyticsSingleDayRange(range: DateRangeParams): boolean {
+  return getAnalyticsRangeDaysInclusive(range) === 1
+}
+
+export function resolveAnalyticsAutoGranularity(range: DateRangeParams): ResolvedAnalyticsGranularity {
+  if (range.granularity && range.granularity !== 'auto') {
+    if (range.granularity === 'hour') {
+      return isAnalyticsSingleDayRange(range) ? 'hour' : 'day'
+    }
+    return range.granularity
+  }
+
+  const rangeDays = getAnalyticsRangeDaysInclusive(range)
+  if (rangeDays == null) return 'day'
+  if (rangeDays === 1) return 'hour'
+  if (rangeDays > 180) return 'month'
+  if (rangeDays > 31) return 'week'
+  return 'day'
+}
+
+export function formatAnalyticsTrendDescription(granularity: ResolvedAnalyticsGranularity, metricsLabel: string): string {
+  if (granularity === 'hour') {
+    return `按小时观察今天截至当前时刻的${metricsLabel}`
+  }
+  if (granularity === 'week') {
+    return `按周观察${metricsLabel}`
+  }
+  if (granularity === 'month') {
+    return `按月观察${metricsLabel}`
+  }
+  return `按日观察${metricsLabel}`
+}
+
+export function formatAnalyticsTrendBucketLabel(
+  bucketStart: string,
+  granularity: ResolvedAnalyticsGranularity,
+): string {
+  if (granularity === 'hour') {
+    return `${bucketStart.slice(11, 13)}:00`
+  }
+  if (granularity === 'month') {
+    return bucketStart.slice(0, 7)
+  }
+  return bucketStart.slice(5, 10)
+}
+
+function formatAnalyticsDateRangeLabel(start: Date, end: Date): string {
+  const startLabel = `${String(start.getMonth() + 1).padStart(2, '0')}/${String(start.getDate()).padStart(2, '0')}`
+  const endLabel = `${String(end.getMonth() + 1).padStart(2, '0')}/${String(end.getDate()).padStart(2, '0')}`
+  return `${startLabel} - ${endLabel}`
+}
+
+export function formatAnalyticsTrendTooltipTitle(
+  bucketStart: string,
+  granularity: ResolvedAnalyticsGranularity,
+): string {
+  if (granularity === 'hour') {
+    return `${bucketStart.slice(0, 10)} ${bucketStart.slice(11, 13)}:00`
+  }
+
+  if (granularity === 'month') {
+    return bucketStart.slice(0, 7)
+  }
+
+  if (granularity === 'week') {
+    const start = parseAnalyticsDateKey(bucketStart.slice(0, 10))
+    if (!start) return bucketStart.slice(0, 10)
+    const end = new Date(start)
+    end.setDate(end.getDate() + 6)
+    return formatAnalyticsDateRangeLabel(start, end)
+  }
+
+  return bucketStart.slice(0, 10)
+}

frontend/src/utils/analyticsTimeseries.ts

@@ -0,0 +1,369 @@
+import type { AnalyticsGranularity, AnalyticsSummary, AnalyticsTimeRange, AnalyticsTimeseriesBucket } from '@/api/analytics'
+
+type DailyPreset = 'today' | 'last7days' | 'last30days' | 'last180days' | 'last1year'
+
+const ZERO_ANALYTICS_SUMMARY: AnalyticsSummary = {
+  requests_total: 0,
+  requests_success: 0,
+  requests_error: 0,
+  requests_stream: 0,
+  success_rate: 0,
+  input_tokens: 0,
+  output_tokens: 0,
+  input_output_total_tokens: 0,
+  cache_creation_input_tokens: 0,
+  cache_creation_input_tokens_5m: 0,
+  cache_creation_input_tokens_1h: 0,
+  cache_read_input_tokens: 0,
+  input_context_tokens: 0,
+  total_tokens: 0,
+  cache_hit_rate: 0,
+  input_cost_usd: 0,
+  output_cost_usd: 0,
+  cache_creation_cost_usd: 0,
+  cache_creation_cost_usd_5m: 0,
+  cache_creation_cost_usd_1h: 0,
+  cache_read_cost_usd: 0,
+  cache_cost_usd: 0,
+  request_cost_usd: 0,
+  total_cost_usd: 0,
+  actual_total_cost_usd: 0,
+  actual_cache_cost_usd: 0,
+  avg_response_time_ms: 0,
+  avg_first_byte_time_ms: 0,
+  format_conversion_count: 0,
+  models_used_count: 0,
+}
+
+function parseDateKey(dateKey: string | undefined): Date | null {
+  if (!dateKey) return null
+  const match = /^(\d{4})-(\d{2})-(\d{2})$/.exec(dateKey.trim())
+  if (!match) return null
+  const [, year, month, day] = match
+  return new Date(Date.UTC(Number(year), Number(month) - 1, Number(day)))
+}
+
+function formatDateKey(date: Date): string {
+  return date.toISOString().slice(0, 10)
+}
+
+function formatLocalDateKey(date: Date): string {
+  const year = date.getFullYear()
+  const month = String(date.getMonth() + 1).padStart(2, '0')
+  const day = String(date.getDate()).padStart(2, '0')
+  return `${year}-${month}-${day}`
+}
+
+function formatHourKey(dateKey: string, hour: number): string {
+  return `${dateKey}T${String(hour).padStart(2, '0')}`
+}
+
+function getStartOfWeekDateKey(date: Date): string {
+  const weekday = date.getUTCDay()
+  const diff = weekday === 0 ? -6 : 1 - weekday
+  const result = new Date(date)
+  result.setUTCDate(result.getUTCDate() + diff)
+  return formatDateKey(result)
+}
+
+function getStartOfMonthDateKey(date: Date): string {
+  const result = new Date(date)
+  result.setUTCDate(1)
+  return formatDateKey(result)
+}
+
+function getNextDateKey(dateKey: string): string {
+  const date = parseDateKey(dateKey)
+  if (!date) return dateKey
+  date.setUTCDate(date.getUTCDate() + 1)
+  return formatDateKey(date)
+}
+
+function getNextWeekKey(dateKey: string): string {
+  const date = parseDateKey(dateKey)
+  if (!date) return dateKey
+  date.setUTCDate(date.getUTCDate() + 7)
+  return formatDateKey(date)
+}
+
+function getNextMonthKey(dateKey: string): string {
+  const date = parseDateKey(dateKey)
+  if (!date) return dateKey
+  date.setUTCMonth(date.getUTCMonth() + 1, 1)
+  return formatDateKey(date)
+}
+
+function resolvePresetDateRange(preset?: string): { start_date: string; end_date: string } | null {
+  const now = new Date()
+  let startDate: Date
+  const endDate = new Date(now)
+
+  switch (preset as DailyPreset | undefined) {
+    case 'today':
+      startDate = new Date(now.getFullYear(), now.getMonth(), now.getDate())
+      break
+    case 'last7days':
+      startDate = new Date(now.getFullYear(), now.getMonth(), now.getDate() - 6)
+      break
+    case 'last30days':
+      startDate = new Date(now.getFullYear(), now.getMonth(), now.getDate() - 29)
+      break
+    case 'last180days':
+      startDate = new Date(now.getFullYear(), now.getMonth(), now.getDate() - 179)
+      break
+    case 'last1year':
+      startDate = new Date(now.getFullYear() - 1, now.getMonth(), now.getDate())
+      break
+    default:
+      return null
+  }
+
+  return {
+    start_date: formatLocalDateKey(startDate),
+    end_date: formatLocalDateKey(endDate),
+  }
+}
+
+function resolveDateBounds(
+  timeRange: AnalyticsTimeRange,
+  fallbackBuckets: AnalyticsTimeseriesBucket[],
+): { startDate: Date | null; endDate: Date | null } {
+  const sortedBuckets = sortBuckets(fallbackBuckets)
+  const fallbackStart = sortedBuckets[0]?.bucket_start.slice(0, 10)
+  const fallbackEnd = sortedBuckets[sortedBuckets.length - 1]?.bucket_start.slice(0, 10)
+  const presetRange = resolvePresetDateRange(timeRange.preset)
+  const startDate = parseDateKey(timeRange.start_date ?? presetRange?.start_date ?? fallbackStart)
+  const endDate = parseDateKey(timeRange.end_date ?? presetRange?.end_date ?? fallbackEnd)
+  return { startDate, endDate }
+}
+
+function sortBuckets(buckets: AnalyticsTimeseriesBucket[]): AnalyticsTimeseriesBucket[] {
+  return buckets.slice().sort((left, right) => left.bucket_start.localeCompare(right.bucket_start))
+}
+
+function normalizeGranularity(granularity?: string): AnalyticsGranularity {
+  if (granularity === 'hour' || granularity === 'week' || granularity === 'month') {
+    return granularity
+  }
+  return 'day'
+}
+
+function resolveTodayDateKey(): string {
+  return formatLocalDateKey(new Date())
+}
+
+export function getDateKeysInRange(
+  timeRange: AnalyticsTimeRange,
+  fallbackBuckets: AnalyticsTimeseriesBucket[] = [],
+): string[] {
+  const sortedBuckets = sortBuckets(fallbackBuckets)
+  const { startDate, endDate } = resolveDateBounds(timeRange, sortedBuckets)
+
+  if (!startDate || !endDate || startDate.getTime() > endDate.getTime()) {
+    return sortedBuckets.map(bucket => bucket.bucket_start.slice(0, 10))
+  }
+
+  const dates: string[] = []
+  const cursor = new Date(startDate)
+
+  while (cursor.getTime() <= endDate.getTime()) {
+    dates.push(formatDateKey(cursor))
+    cursor.setUTCDate(cursor.getUTCDate() + 1)
+  }
+
+  return dates
+}
+
+export function getHourKeysInRange(
+  timeRange: AnalyticsTimeRange,
+  fallbackBuckets: AnalyticsTimeseriesBucket[] = [],
+): string[] {
+  const sortedBuckets = sortBuckets(fallbackBuckets)
+  const { startDate, endDate } = resolveDateBounds(timeRange, sortedBuckets)
+
+  if (!startDate || !endDate || startDate.getTime() !== endDate.getTime()) {
+    return sortedBuckets.map(bucket => bucket.bucket_start.slice(0, 13))
+  }
+
+  const dateKey = formatDateKey(startDate)
+  const isToday = dateKey === resolveTodayDateKey()
+  const maxHour = isToday ? new Date().getHours() : 23
+
+  return Array.from({ length: maxHour + 1 }, (_, hour) => formatHourKey(dateKey, hour))
+}
+
+export function getWeekKeysInRange(
+  timeRange: AnalyticsTimeRange,
+  fallbackBuckets: AnalyticsTimeseriesBucket[] = [],
+): string[] {
+  const sortedBuckets = sortBuckets(fallbackBuckets)
+  const { startDate, endDate } = resolveDateBounds(timeRange, sortedBuckets)
+
+  if (!startDate || !endDate || startDate.getTime() > endDate.getTime()) {
+    return sortedBuckets.map(bucket => bucket.bucket_start.slice(0, 10))
+  }
+
+  const keys: string[] = []
+  let cursorKey = getStartOfWeekDateKey(startDate)
+  const endKey = getStartOfWeekDateKey(endDate)
+
+  while (cursorKey <= endKey) {
+    keys.push(cursorKey)
+    cursorKey = getNextWeekKey(cursorKey)
+  }
+
+  return keys
+}
+
+export function getMonthKeysInRange(
+  timeRange: AnalyticsTimeRange,
+  fallbackBuckets: AnalyticsTimeseriesBucket[] = [],
+): string[] {
+  const sortedBuckets = sortBuckets(fallbackBuckets)
+  const { startDate, endDate } = resolveDateBounds(timeRange, sortedBuckets)
+
+  if (!startDate || !endDate || startDate.getTime() > endDate.getTime()) {
+    return sortedBuckets.map(bucket => bucket.bucket_start.slice(0, 10))
+  }
+
+  const keys: string[] = []
+  let cursorKey = getStartOfMonthDateKey(startDate)
+  const endKey = getStartOfMonthDateKey(endDate)
+
+  while (cursorKey <= endKey) {
+    keys.push(cursorKey)
+    cursorKey = getNextMonthKey(cursorKey)
+  }
+
+  return keys
+}
+
+export function createZeroDailyBucket(dateKey: string): AnalyticsTimeseriesBucket {
+  return {
+    ...ZERO_ANALYTICS_SUMMARY,
+    bucket_start: `${dateKey}T00:00:00`,
+    bucket_end: `${getNextDateKey(dateKey)}T00:00:00`,
+  }
+}
+
+export function createZeroHourlyBucket(hourKey: string): AnalyticsTimeseriesBucket {
+  const dateKey = hourKey.slice(0, 10)
+  const hour = Number(hourKey.slice(11, 13))
+  const nextDate = new Date(Date.UTC(
+    Number(dateKey.slice(0, 4)),
+    Number(dateKey.slice(5, 7)) - 1,
+    Number(dateKey.slice(8, 10)),
+    hour,
+  ))
+  nextDate.setUTCHours(nextDate.getUTCHours() + 1)
+  const nextDateKey = formatDateKey(nextDate)
+  const nextHour = String(nextDate.getUTCHours()).padStart(2, '0')
+
+  return {
+    ...ZERO_ANALYTICS_SUMMARY,
+    bucket_start: `${hourKey}:00:00`,
+    bucket_end: `${nextDateKey}T${nextHour}:00:00`,
+  }
+}
+
+export function createZeroWeeklyBucket(dateKey: string): AnalyticsTimeseriesBucket {
+  return {
+    ...ZERO_ANALYTICS_SUMMARY,
+    bucket_start: `${dateKey}T00:00:00`,
+    bucket_end: `${getNextWeekKey(dateKey)}T00:00:00`,
+  }
+}
+
+export function createZeroMonthlyBucket(dateKey: string): AnalyticsTimeseriesBucket {
+  return {
+    ...ZERO_ANALYTICS_SUMMARY,
+    bucket_start: `${dateKey}T00:00:00`,
+    bucket_end: `${getNextMonthKey(dateKey)}T00:00:00`,
+  }
+}
+
+export function fillMissingDailyTimeseriesBuckets(
+  buckets: AnalyticsTimeseriesBucket[],
+  timeRange: AnalyticsTimeRange,
+): AnalyticsTimeseriesBucket[] {
+  const sortedBuckets = sortBuckets(buckets)
+  const dateKeys = getDateKeysInRange(timeRange, sortedBuckets)
+
+  if (!dateKeys.length) {
+    return sortedBuckets
+  }
+
+  const bucketByDate = new Map(
+    sortedBuckets.map(bucket => [bucket.bucket_start.slice(0, 10), bucket]),
+  )
+
+  return dateKeys.map(dateKey => bucketByDate.get(dateKey) ?? createZeroDailyBucket(dateKey))
+}
+
+export function fillMissingWeeklyTimeseriesBuckets(
+  buckets: AnalyticsTimeseriesBucket[],
+  timeRange: AnalyticsTimeRange,
+): AnalyticsTimeseriesBucket[] {
+  const sortedBuckets = sortBuckets(buckets)
+  const weekKeys = getWeekKeysInRange(timeRange, sortedBuckets)
+
+  if (!weekKeys.length) {
+    return sortedBuckets
+  }
+
+  const bucketByWeek = new Map(
+    sortedBuckets.map(bucket => [bucket.bucket_start.slice(0, 10), bucket]),
+  )
+
+  return weekKeys.map(weekKey => bucketByWeek.get(weekKey) ?? createZeroWeeklyBucket(weekKey))
+}
+
+export function fillMissingMonthlyTimeseriesBuckets(
+  buckets: AnalyticsTimeseriesBucket[],
+  timeRange: AnalyticsTimeRange,
+): AnalyticsTimeseriesBucket[] {
+  const sortedBuckets = sortBuckets(buckets)
+  const monthKeys = getMonthKeysInRange(timeRange, sortedBuckets)
+
+  if (!monthKeys.length) {
+    return sortedBuckets
+  }
+
+  const bucketByMonth = new Map(
+    sortedBuckets.map(bucket => [bucket.bucket_start.slice(0, 10), bucket]),
+  )
+
+  return monthKeys.map(monthKey => bucketByMonth.get(monthKey) ?? createZeroMonthlyBucket(monthKey))
+}
+
+export function fillMissingTimeseriesBuckets(
+  buckets: AnalyticsTimeseriesBucket[],
+  timeRange: AnalyticsTimeRange,
+): AnalyticsTimeseriesBucket[] {
+  const granularity = normalizeGranularity(timeRange.granularity)
+  const sortedBuckets = sortBuckets(buckets)
+
+  if (granularity === 'week') {
+    return fillMissingWeeklyTimeseriesBuckets(sortedBuckets, timeRange)
+  }
+
+  if (granularity === 'month') {
+    return fillMissingMonthlyTimeseriesBuckets(sortedBuckets, timeRange)
+  }
+
+  if (granularity !== 'hour') {
+    return fillMissingDailyTimeseriesBuckets(sortedBuckets, timeRange)
+  }
+
+  const hourKeys = getHourKeysInRange(timeRange, sortedBuckets)
+
+  if (!hourKeys.length) {
+    return sortedBuckets
+  }
+
+  const bucketByHour = new Map(
+    sortedBuckets.map(bucket => [bucket.bucket_start.slice(0, 13), bucket]),
+  )
+
+  return hourKeys.map(hourKey => bucketByHour.get(hourKey) ?? createZeroHourlyBucket(hourKey))
+}

frontend/src/utils/chartTheme.ts

@@ -0,0 +1,135 @@
+const COLOR_KEYS = new Set([
+  'backgroundColor',
+  'borderColor',
+  'color',
+  'pointBackgroundColor',
+  'pointBorderColor',
+  'pointHoverBackgroundColor',
+  'pointHoverBorderColor',
+  'hoverBackgroundColor',
+  'hoverBorderColor',
+  'tickColor',
+])
+
+const CSS_COLOR_PATTERN = /(var\(|oklch\(|hsl\(|hsla\(|rgb\(|rgba\()/i
+const HSL_VAR_PATTERN = /^hsl\(var\((--[^)]+)\)(?:\s*\/\s*([^)]+))?\)$/i
+
+let resolverElement: HTMLSpanElement | null = null
+
+function getResolverElement(): HTMLSpanElement | null {
+  if (typeof document === 'undefined') return null
+  if (resolverElement && resolverElement.isConnected) return resolverElement
+
+  const element = document.createElement('span')
+  element.setAttribute('aria-hidden', 'true')
+  element.style.position = 'fixed'
+  element.style.pointerEvents = 'none'
+  element.style.opacity = '0'
+  element.style.visibility = 'hidden'
+  element.style.inset = '-9999px'
+  document.body.appendChild(element)
+  resolverElement = element
+  return resolverElement
+}
+
+export function resolveCssColor(value: string): string {
+  if (!CSS_COLOR_PATTERN.test(value)) return value
+
+  const hslVarMatch = value.trim().match(HSL_VAR_PATTERN)
+  if (hslVarMatch) {
+    const [, variableName, alphaValue] = hslVarMatch
+    const baseColor = resolveCssColor(`var(${variableName})`)
+    if (!alphaValue) return baseColor
+    return applyAlpha(baseColor, alphaValue)
+  }
+
+  const element = getResolverElement()
+  if (!element) return value
+
+  element.style.color = ''
+  element.style.color = value
+  if (!element.style.color) {
+    return value
+  }
+  const resolved = window.getComputedStyle(element).color.trim()
+
+  return resolved || value
+}
+
+function applyAlpha(color: string, alphaValue: string): string {
+  const alpha = parseAlpha(alphaValue)
+  if (alpha == null) return color
+
+  const rgbMatch = color.match(/^rgba?\(([^)]+)\)$/i)
+  if (!rgbMatch) return color
+
+  const parts = rgbMatch[1].split(',').map(part => part.trim())
+  if (parts.length < 3) return color
+
+  return `rgba(${parts[0]}, ${parts[1]}, ${parts[2]}, ${alpha})`
+}
+
+function parseAlpha(value: string): number | null {
+  const trimmed = value.trim()
+  if (trimmed.endsWith('%')) {
+    const percentage = Number.parseFloat(trimmed.slice(0, -1))
+    return Number.isFinite(percentage) ? percentage / 100 : null
+  }
+
+  const numeric = Number.parseFloat(trimmed)
+  return Number.isFinite(numeric) ? numeric : null
+}
+
+function isObject(value: unknown): value is Record<string, unknown> {
+  return typeof value === 'object' && value !== null
+}
+
+function isColorKey(key: string | undefined): boolean {
+  if (!key) return false
+  return COLOR_KEYS.has(key) || /color$/i.test(key)
+}
+
+function resolveValue(value: unknown, key?: string): unknown {
+  if (typeof value === 'string') {
+    return isColorKey(key) ? resolveCssColor(value) : value
+  }
+
+  if (Array.isArray(value)) {
+    return value.map(item => resolveValue(item, key))
+  }
+
+  if (!isObject(value)) {
+    return value
+  }
+
+  const output: Record<string, unknown> = {}
+  for (const [childKey, childValue] of Object.entries(value)) {
+    output[childKey] = resolveValue(childValue, childKey)
+  }
+  return output
+}
+
+export function resolveChartTheme<T>(value: T): T {
+  return resolveValue(value) as T
+}
+
+export function observeChartThemeChanges(callback: () => void): () => void {
+  if (typeof MutationObserver === 'undefined' || typeof document === 'undefined') {
+    return () => {}
+  }
+
+  const observer = new MutationObserver(() => callback())
+  observer.observe(document.documentElement, {
+    attributes: true,
+    attributeFilter: ['class', 'style'],
+  })
+
+  if (document.body) {
+    observer.observe(document.body, {
+      attributes: true,
+      attributeFilter: ['class', 'style', 'theme-mode'],
+    })
+  }
+
+  return () => observer.disconnect()
+}

frontend/src/utils/csvExport.ts

@@ -0,0 +1,58 @@
+/**
+ * 客户端 CSV 导出工具
+ * 将数据转为 CSV 并触发浏览器下载
+ */
+
+function escapeCsvField(value: string | number | null | undefined): string {
+  if (value == null) return ''
+  const str = String(value)
+  if (str.includes(',') || str.includes('"') || str.includes('\n')) {
+    return `"${str.replace(/"/g, '""')}"`
+  }
+  return str
+}
+
+/**
+ * 生成 CSV 内容并触发下载
+ * @param filename 文件名（不含扩展名）
+ * @param headers 列标题
+ * @param rows 数据行
+ */
+export function downloadCsv(
+  filename: string,
+  headers: string[],
+  rows: (string | number | null | undefined)[][],
+): void {
+  const bom = '\uFEFF' // UTF-8 BOM，确保 Excel 正确识别中文
+  const headerLine = headers.map(escapeCsvField).join(',')
+  const dataLines = rows.map(row => row.map(escapeCsvField).join(','))
+  const csv = bom + [headerLine, ...dataLines].join('\n')
+
+  const blob = new Blob([csv], { type: 'text/csv;charset=utf-8' })
+  const url = URL.createObjectURL(blob)
+  const a = document.createElement('a')
+  a.href = url
+  a.download = `${filename}.csv`
+  document.body.appendChild(a)
+  a.click()
+  document.body.removeChild(a)
+  URL.revokeObjectURL(url)
+}
+
+/**
+ * 从对象数组生成 CSV 下载
+ * @param filename 文件名
+ * @param data 对象数组
+ * @param columns 列定义 [{ key, label }]
+ */
+export function downloadCsvFromObjects<T extends Record<string, unknown>>(
+  filename: string,
+  data: T[],
+  columns: { key: string; label: string }[],
+): void {
+  const headers = columns.map(c => c.label)
+  const rows = data.map(item =>
+    columns.map(c => item[c.key] as string | number | null),
+  )
+  downloadCsv(filename, headers, rows)
+}

frontend/src/utils/format.ts

@@ -1,17 +1,34 @@
+function toFiniteNumber(value: number | string | undefined | null): number | null {
+  if (value === undefined || value === null) return null
+  if (typeof value === 'number') {
+    return Number.isFinite(value) ? value : null
+  }
+
+  const normalized = value
+    .trim()
+    .replace(/[$,\s]/g, '')
+
+  if (!normalized) return null
+
+  const parsed = Number(normalized)
+  return Number.isFinite(parsed) ? parsed : null
+}
+
 // Token formatting - intelligent display based on value size
-export function formatTokens(num: number | undefined | null): string {
-  if (num === undefined || num === null || num === 0) {
+export function formatTokens(num: number | string | undefined | null): string {
+  const normalized = toFiniteNumber(num)
+  if (normalized === null || normalized === 0) {
     return '0'
   }
 
   // For very small values (< 1000), show as is without unit
-  if (num < 1000) {
-    return num.toString()
+  if (normalized < 1000) {
+    return normalized.toString()
   }
 
   // For values 1K-999K, show in thousands
-  if (num < 1000000) {
-    const thousands = num / 1000
+  if (normalized < 1000000) {
+    const thousands = normalized / 1000
     if (thousands >= 100) {
       return `${Math.round(thousands)  }K`
     } else if (thousands >= 10) {
@@ -22,7 +39,7 @@ export function formatTokens(num: number | undefined | null): string {
   }
 
   // For values >= 1M, show in millions
-  const millions = num / 1000000
+  const millions = normalized / 1000000
   if (millions >= 100) {
     return `${Math.round(millions)  }M`
   } else if (millions >= 10) {
@@ -33,53 +50,54 @@ export function formatTokens(num: number | undefined | null): string {
 }
 
 // Currency formatting with high precision for small values
-export function formatCurrency(amount: number | undefined | null): string {
-  if (amount === undefined || amount === null || amount === 0) {
+export function formatCurrency(amount: number | string | undefined | null): string {
+  const normalized = toFiniteNumber(amount)
+  if (normalized === null || normalized === 0) {
     return '$0.00'
   }
 
   // For very small amounts (< $0.00001), show up to 8 decimal places
-  if (amount > 0 && amount < 0.00001) {
-    const formatted = amount.toFixed(8)
+  if (normalized > 0 && normalized < 0.00001) {
+    const formatted = normalized.toFixed(8)
     // Remove trailing zeros but keep at least 2 decimal places
     const trimmed = formatted.replace(/(\.\d\d)0+$/, '$1')
     return `$${  trimmed}`
   }
 
   // For small amounts (< $0.0001), show up to 6 decimal places
-  if (amount < 0.0001) {
-    const formatted = amount.toFixed(6)
+  if (normalized < 0.0001) {
+    const formatted = normalized.toFixed(6)
     // Remove trailing zeros but keep at least 2 decimal places
     const trimmed = formatted.replace(/(\.\d\d)0+$/, '$1')
     return `$${  trimmed}`
   }
 
   // For small amounts (< $0.01), show up to 5 decimal places
-  if (amount < 0.01) {
-    const formatted = amount.toFixed(5)
+  if (normalized < 0.01) {
+    const formatted = normalized.toFixed(5)
     // Remove trailing zeros but keep at least 2 decimal places
     const trimmed = formatted.replace(/(\.\d\d)0+$/, '$1')
     return `$${  trimmed}`
   }
 
   // For amounts less than $1, show 4 decimal places
-  if (amount < 1) {
-    const formatted = amount.toFixed(4)
+  if (normalized < 1) {
+    const formatted = normalized.toFixed(4)
     // Remove trailing zeros but keep at least 2 decimal places
     const trimmed = formatted.replace(/(\.\d\d)0+$/, '$1')
     return `$${  trimmed}`
   }
 
   // For amounts $1-$100, show 2-3 decimal places
-  if (amount < 100) {
-    const formatted = amount.toFixed(3)
+  if (normalized < 100) {
+    const formatted = normalized.toFixed(3)
     // Remove trailing zeros but keep at least 2 decimal places
     const trimmed = formatted.replace(/(\.\d\d)0+$/, '$1')
     return `$${  trimmed}`
   }
 
   // For larger amounts, show 2 decimal places
-  return `$${  amount.toFixed(2)}`
+  return `$${  normalized.toFixed(2)}`
 }
 
 // Number formatting with locale support
@@ -157,7 +175,8 @@ export function formatRemainingTime(expireAt: number | undefined, currentTime: n
 // Cache hit rate formatting
 export function formatHitRate(rate: number | undefined): string {
   if (typeof rate !== 'number' || Number.isNaN(rate)) return '-'
-  return `${rate.toFixed(2)}%`
+  const normalized = rate <= 1 ? rate * 100 : rate
+  return `${normalized.toFixed(2)}%`
 }
 
 // Rate limit formatting (supports "inherit" semantics: null = inherit system default)

frontend/src/utils/usageBreakdown.ts

@@ -0,0 +1,14 @@
+export interface DailyUsageBreakdown {
+  date: string
+  inputTokens: number
+  outputTokens: number
+  cacheCreationTokens: number
+  cacheReadTokens: number
+  cacheCreationCost: number
+  cacheReadCost: number
+  cacheHitRate: number | null
+  totalCacheCost: number
+  totalCacheTokens: number
+  totalTrackedTokens: number
+  baseTokens: number
+}

frontend/src/views/admin/Analytics.vue

@@ -0,0 +1,246 @@
+<template>
+  <div class="space-y-6 px-4 sm:px-6 lg:px-0">
+    <div class="flex flex-col gap-3 sm:flex-row sm:items-start sm:justify-between">
+      <div class="min-w-0">
+        <h1 class="text-lg font-semibold">
+          数据分析
+        </h1>
+        <p class="text-xs text-muted-foreground">
+          用量报表、模型报表与全局性能洞察
+        </p>
+      </div>
+    </div>
+
+    <!-- Tabs -->
+    <Tabs v-model="filters.activeTab.value">
+      <div class="flex flex-col gap-3 sm:flex-row sm:items-center sm:justify-between">
+        <TabsList class="tabs-button-list">
+          <TabsTrigger value="detail">
+            用量报表
+          </TabsTrigger>
+          <TabsTrigger value="models">
+            模型报表
+          </TabsTrigger>
+          <TabsTrigger value="leaderboard">
+            排行榜
+          </TabsTrigger>
+          <TabsTrigger value="performance">
+            性能
+          </TabsTrigger>
+        </TabsList>
+
+        <div class="flex flex-wrap items-center gap-2 sm:justify-end">
+          <TimeRangePicker
+            v-model="filters.timeRange.value"
+            :show-granularity="false"
+            :include-auto-granularity="true"
+            :compact="true"
+          />
+          <template v-if="filters.activeTab.value === 'leaderboard'">
+            <Select v-model="leaderboardDimension">
+              <SelectTrigger class="h-8 w-full text-xs sm:w-32">
+                <SelectValue placeholder="维度" />
+              </SelectTrigger>
+              <SelectContent>
+                <SelectItem value="users">
+                  用户
+                </SelectItem>
+                <SelectItem value="api_keys">
+                  API Key
+                </SelectItem>
+              </SelectContent>
+            </Select>
+          </template>
+          <template v-else-if="showEntityFilters">
+            <Select v-model="selectedUserValue">
+              <SelectTrigger class="h-8 w-full text-xs sm:w-36">
+                <SelectValue placeholder="全部用户">
+                  <span
+                    class="block min-w-0 truncate"
+                    :title="selectedUserLabel"
+                  >
+                    {{ selectedUserLabel }}
+                  </span>
+                </SelectValue>
+              </SelectTrigger>
+              <SelectContent class="w-[min(18rem,var(--radix-select-trigger-width))]">
+                <SelectItem value="__all__">
+                  全部用户
+                </SelectItem>
+                <SelectItem
+                  v-for="option in userOptions"
+                  :key="option.value"
+                  :value="option.value"
+                  :text-value="option.label"
+                >
+                  {{ option.label }}
+                </SelectItem>
+              </SelectContent>
+            </Select>
+            <Select v-model="selectedApiKeyValue">
+              <SelectTrigger class="h-8 w-full text-xs sm:w-36">
+                <SelectValue placeholder="全部 Key">
+                  <span
+                    class="block min-w-0 truncate"
+                    :title="selectedApiKeyLabel"
+                  >
+                    {{ selectedApiKeyLabel }}
+                  </span>
+                </SelectValue>
+              </SelectTrigger>
+              <SelectContent class="w-[min(18rem,var(--radix-select-trigger-width))]">
+                <SelectItem value="__all__">
+                  全部 Key
+                </SelectItem>
+                <SelectItem
+                  v-for="option in apiKeyOptions"
+                  :key="option.value"
+                  :value="option.value"
+                  :text-value="option.label"
+                >
+                  {{ option.label }}
+                </SelectItem>
+              </SelectContent>
+            </Select>
+          </template>
+        </div>
+      </div>
+
+      <div class="mt-4">
+        <ReportsTab
+          v-if="isReportsTab"
+          :mode="reportMode"
+          :selected-user-label="selectedUserExportLabel"
+          :selected-api-key-label="selectedApiKeyExportLabel"
+        />
+        <LeaderboardTab
+          v-else-if="filters.activeTab.value === 'leaderboard'"
+          :dimension="leaderboardDimension"
+        />
+        <PerformanceTab v-else-if="filters.activeTab.value === 'performance'" />
+      </div>
+    </Tabs>
+  </div>
+</template>
+
+<script setup lang="ts">
+import { computed, provide, ref, watch, onMounted } from 'vue'
+import { Tabs, TabsList, TabsTrigger, Select, SelectContent, SelectItem, SelectTrigger, SelectValue } from '@/components/ui'
+import { TimeRangePicker } from '@/components/common'
+import { analyticsApi, type AnalyticsFilterOption } from '@/api/analytics'
+import { useAnalyticsFilters } from '@/composables/useAnalyticsFilters'
+import type { LeaderboardDimension } from '@/composables/analytics/useLeaderboardData'
+import ReportsTab from './components/analytics/ReportsTab.vue'
+import LeaderboardTab from './components/analytics/LeaderboardTab.vue'
+import PerformanceTab from './components/analytics/PerformanceTab.vue'
+
+const filters = useAnalyticsFilters()
+provide('analyticsFilters', filters)
+
+const userOptions = ref<AnalyticsFilterOption[]>([])
+const apiKeyOptions = ref<AnalyticsFilterOption[]>([])
+const leaderboardDimension = ref<LeaderboardDimension>('users')
+const supportedTabs = new Set(['detail', 'models', 'leaderboard', 'performance'])
+const isReportsTab = computed(() => filters.activeTab.value === 'detail' || filters.activeTab.value === 'models')
+const showEntityFilters = computed(() => isReportsTab.value)
+const reportMode = computed<'detail' | 'models'>(() => (
+  filters.activeTab.value === 'models' ? 'models' : 'detail'
+))
+let suppressNextUserFilterLoad = false
+let suppressNextApiKeyFilterLoad = false
+const selectedUserValue = computed({
+  get: () => filters.userFilter.value[0] ?? '__all__',
+  set: (value: string) => {
+    filters.userFilter.value = value && value !== '__all__' ? [value] : []
+  },
+})
+const selectedUserLabel = computed(() => (
+  selectedUserValue.value === '__all__'
+    ? '全部用户'
+    : userOptions.value.find(option => option.value === selectedUserValue.value)?.label || '全部用户'
+))
+const selectedApiKeyValue = computed({
+  get: () => filters.apiKeyFilter.value[0] ?? '__all__',
+  set: (value: string) => {
+    filters.apiKeyFilter.value = value && value !== '__all__' ? [value] : []
+  },
+})
+const selectedApiKeyLabel = computed(() => (
+  selectedApiKeyValue.value === '__all__'
+    ? '全部 Key'
+    : apiKeyOptions.value.find(option => option.value === selectedApiKeyValue.value)?.label || '全部 Key'
+))
+const selectedUserExportLabel = computed(() => (
+  selectedUserValue.value === '__all__' ? '' : selectedUserLabel.value
+))
+const selectedApiKeyExportLabel = computed(() => (
+  selectedApiKeyValue.value === '__all__' ? '' : selectedApiKeyLabel.value
+))
+
+function hasOption(options: AnalyticsFilterOption[], value: string): boolean {
+  return options.some(option => option.value === value)
+}
+
+watch(
+  () => filters.activeTab.value,
+  (tab) => {
+    if (!supportedTabs.has(tab)) {
+      filters.activeTab.value = 'detail'
+    }
+  },
+  { immediate: true },
+)
+
+async function loadFilterOptions() {
+  if (!showEntityFilters.value) return
+  const response = await analyticsApi.getFilterOptions({
+    scope: { kind: 'global' },
+    time_range: filters.getTimeRangeParams(),
+    filters: {
+      user_ids: filters.userFilter.value,
+      api_key_ids: filters.apiKeyFilter.value,
+    },
+  }).catch(() => null)
+
+  if (!response) return
+  userOptions.value = response.users ?? []
+  apiKeyOptions.value = response.api_keys ?? []
+
+  if (filters.userFilter.value.length > 0 && !hasOption(userOptions.value, filters.userFilter.value[0])) {
+    suppressNextUserFilterLoad = true
+    filters.userFilter.value = []
+  }
+  if (filters.apiKeyFilter.value.length > 0 && !hasOption(apiKeyOptions.value, filters.apiKeyFilter.value[0])) {
+    suppressNextApiKeyFilterLoad = true
+    filters.apiKeyFilter.value = []
+  }
+}
+
+watch(
+  () => [
+    showEntityFilters.value,
+    filters.getTimeRangeParams(),
+    filters.userFilter.value.slice(),
+    filters.apiKeyFilter.value.slice(),
+  ],
+  ([shouldShow]) => {
+    if (!shouldShow) return
+    if (suppressNextUserFilterLoad) {
+      suppressNextUserFilterLoad = false
+      return
+    }
+    if (suppressNextApiKeyFilterLoad) {
+      suppressNextApiKeyFilterLoad = false
+      return
+    }
+    void loadFilterOptions()
+  },
+  { deep: true },
+)
+
+onMounted(() => {
+  if (showEntityFilters.value) {
+    void loadFilterOptions()
+  }
+})
+</script>

frontend/src/views/admin/CostAnalysis.vue

@@ -1,207 +0,0 @@
-<template>
-  <div class="space-y-6 px-4 sm:px-6 lg:px-0">
-    <div class="flex flex-col sm:flex-row sm:items-center sm:justify-between gap-3">
-      <div>
-        <h1 class="text-lg font-semibold">
-          成本分析
-        </h1>
-        <p class="text-xs text-muted-foreground">
-          成本趋势、预测与节省统计
-        </p>
-      </div>
-      <TimeRangePicker v-model="timeRange" />
-    </div>
-
-    <div class="grid grid-cols-1 lg:grid-cols-3 gap-4">
-      <Card class="p-4 space-y-2">
-        <div class="text-xs text-muted-foreground">
-          缓存节省
-        </div>
-        <div class="text-lg font-semibold">
-          {{ formatCurrency(costSavings?.cache_savings ?? 0) }}
-        </div>
-        <div class="text-xs text-muted-foreground">
-          读取成本 {{ formatCurrency(costSavings?.cache_read_cost ?? 0) }}
-        </div>
-      </Card>
-      <Card class="p-4 space-y-2">
-        <div class="text-xs text-muted-foreground">
-          缓存读取 Tokens
-        </div>
-        <div class="text-lg font-semibold">
-          {{ formatTokens(costSavings?.cache_read_tokens ?? 0) }}
-        </div>
-        <div class="text-xs text-muted-foreground">
-          预计全额成本 {{ formatCurrency(costSavings?.estimated_full_cost ?? 0) }}
-        </div>
-      </Card>
-      <Card class="p-4 space-y-2">
-        <div class="text-xs text-muted-foreground">
-          缓存创建成本
-        </div>
-        <div class="text-lg font-semibold">
-          {{ formatCurrency(costSavings?.cache_creation_cost ?? 0) }}
-        </div>
-        <div class="text-xs text-muted-foreground">
-          基于当前时间范围
-        </div>
-      </Card>
-    </div>
-
-    <div class="grid grid-cols-1 lg:grid-cols-2 gap-4">
-      <Card class="p-4">
-        <CostForecastChart
-          title="成本趋势预测"
-          :history="forecastHistory"
-          :forecast="forecastFuture"
-          :loading="forecastLoading"
-        />
-      </Card>
-      <QuotaProgressCard
-        title="月卡消耗进度"
-        :providers="quotaProviders"
-        :loading="quotaLoading"
-      />
-    </div>
-
-    <UsageProviderTable
-      :data="providerStats"
-      :is-admin="true"
-    />
-  </div>
-</template>
-
-<script setup lang="ts">
-import { ref, computed, onMounted, onUnmounted, watch } from 'vue'
-import Card from '@/components/ui/card.vue'
-import { TimeRangePicker } from '@/components/common'
-import { CostForecastChart, QuotaProgressCard } from '@/components/stats'
-import { UsageProviderTable } from '@/features/usage/components'
-import { adminApi, type CostForecastResponse, type CostSavingsResponse, type QuotaUsageProvider } from '@/api/admin'
-import { usageApi } from '@/api/usage'
-import { formatCurrency, formatTokens } from '@/utils/format'
-import { getDateRangeFromPeriod } from '@/features/usage/composables'
-import type { DateRangeParams } from '@/features/usage/types'
-import type { ProviderStatsItem } from '@/features/usage/types'
-
-const timeRange = ref<DateRangeParams>(getDateRangeFromPeriod('last30days'))
-
-const forecast = ref<CostForecastResponse | null>(null)
-const costSavings = ref<CostSavingsResponse | null>(null)
-const quotaProviders = ref<QuotaUsageProvider[]>([])
-const providerStats = ref<ProviderStatsItem[]>([])
-
-const forecastLoading = ref(false)
-const quotaLoading = ref(false)
-let forecastRequestId = 0
-let savingsRequestId = 0
-let quotaRequestId = 0
-let providerStatsRequestId = 0
-let loadAllPromise: Promise<void> | null = null
-let hasPendingLoadAll = false
-let loadAllDebounceTimer: ReturnType<typeof setTimeout> | null = null
-
-const forecastHistory = computed(() => forecast.value?.history || [])
-const forecastFuture = computed(() => forecast.value?.forecast || [])
-
-function buildTimeRangeParams() {
-  return {
-    start_date: timeRange.value.start_date,
-    end_date: timeRange.value.end_date,
-    preset: timeRange.value.preset,
-    timezone: timeRange.value.timezone,
-    tz_offset_minutes: timeRange.value.tz_offset_minutes
-  }
-}
-
-async function loadForecast() {
-  const requestId = ++forecastRequestId
-  forecastLoading.value = true
-  try {
-    const data = await adminApi.getCostForecast(buildTimeRangeParams())
-    if (requestId !== forecastRequestId) return
-    forecast.value = data
-  } finally {
-    if (requestId === forecastRequestId) {
-      forecastLoading.value = false
-    }
-  }
-}
-
-async function loadSavings() {
-  const requestId = ++savingsRequestId
-  const data = await adminApi.getCostSavings(buildTimeRangeParams())
-  if (requestId !== savingsRequestId) return
-  costSavings.value = data
-}
-
-async function loadQuotaUsage() {
-  const requestId = ++quotaRequestId
-  quotaLoading.value = true
-  try {
-    const response = await adminApi.getQuotaUsage()
-    if (requestId !== quotaRequestId) return
-    quotaProviders.value = response.providers
-  } finally {
-    if (requestId === quotaRequestId) {
-      quotaLoading.value = false
-    }
-  }
-}
-
-async function loadProviderStats() {
-  const requestId = ++providerStatsRequestId
-  const stats = await usageApi.getUsageByProvider({
-    ...buildTimeRangeParams(),
-    limit: 8
-  })
-  if (requestId !== providerStatsRequestId) return
-  providerStats.value = stats
-}
-
-async function loadAll() {
-  if (loadAllPromise) {
-    hasPendingLoadAll = true
-    return loadAllPromise
-  }
-  loadAllPromise = Promise.all([loadForecast(), loadSavings(), loadQuotaUsage(), loadProviderStats()])
-    .then(() => undefined)
-    .finally(() => {
-      loadAllPromise = null
-      if (hasPendingLoadAll) {
-        hasPendingLoadAll = false
-        void loadAll()
-      }
-    })
-  return loadAllPromise
-}
-
-function scheduleLoadAll() {
-  if (loadAllDebounceTimer) {
-    clearTimeout(loadAllDebounceTimer)
-  }
-  loadAllDebounceTimer = setTimeout(() => {
-    loadAllDebounceTimer = null
-    void loadAll()
-  }, 120)
-}
-
-watch(timeRange, scheduleLoadAll, { deep: true })
-
-onMounted(() => {
-  void loadAll()
-})
-
-onUnmounted(() => {
-  if (loadAllDebounceTimer) {
-    clearTimeout(loadAllDebounceTimer)
-    loadAllDebounceTimer = null
-  }
-  hasPendingLoadAll = false
-  loadAllPromise = null
-  forecastRequestId += 1
-  savingsRequestId += 1
-  quotaRequestId += 1
-  providerStatsRequestId += 1
-})
-</script>

frontend/src/views/admin/PerformanceAnalysis.vue

@@ -1,232 +0,0 @@
-<template>
-  <div class="space-y-6 px-4 sm:px-6 lg:px-0">
-    <div class="flex flex-col sm:flex-row sm:items-center sm:justify-between gap-3">
-      <div>
-        <h1 class="text-lg font-semibold">
-          性能分析
-        </h1>
-        <p class="text-xs text-muted-foreground">
-          延迟分布与错误统计
-        </p>
-      </div>
-      <TimeRangePicker v-model="timeRange" />
-    </div>
-
-    <div class="grid grid-cols-1 lg:grid-cols-2 gap-4">
-      <Card class="p-4">
-        <PercentileChart
-          title="响应延迟百分位"
-          :series="percentiles"
-          mode="response"
-          :loading="percentileLoading"
-        />
-      </Card>
-      <Card class="p-4">
-        <PercentileChart
-          title="首字节延迟百分位"
-          :series="percentiles"
-          mode="ttfb"
-          :loading="percentileLoading"
-        />
-      </Card>
-    </div>
-
-    <div class="grid grid-cols-1 lg:grid-cols-2 gap-4">
-      <Card class="p-4">
-        <ErrorDistributionChart
-          title="错误分布"
-          :distribution="errorDistribution"
-          :loading="errorLoading"
-        />
-      </Card>
-      <Card class="p-4 space-y-3">
-        <h3 class="text-sm font-semibold">
-          错误趋势
-        </h3>
-        <div
-          v-if="errorLoading"
-          class="p-6"
-        >
-          <LoadingState />
-        </div>
-        <div
-          v-else
-          class="h-[260px]"
-        >
-          <LineChart :data="errorTrendChartData" />
-        </div>
-      </Card>
-    </div>
-
-    <Card class="p-4 space-y-3">
-      <h3 class="text-sm font-semibold">
-        提供商健康度
-      </h3>
-      <div
-        v-if="providerLoading"
-        class="p-4"
-      >
-        <LoadingState />
-      </div>
-      <div
-        v-else
-        class="grid grid-cols-1 sm:grid-cols-2 lg:grid-cols-3 gap-3 text-sm"
-      >
-        <div
-          v-for="provider in providerStatus"
-          :key="provider.name"
-          class="p-3 border rounded-lg"
-        >
-          <div class="flex items-center justify-between">
-            <span class="font-medium">{{ provider.name }}</span>
-            <span class="text-xs text-muted-foreground">{{ provider.requests }} 请求</span>
-          </div>
-          <div class="text-xs text-muted-foreground mt-1">
-            状态: {{ provider.status }}
-          </div>
-        </div>
-      </div>
-    </Card>
-  </div>
-</template>
-
-<script setup lang="ts">
-import { ref, computed, onMounted, onUnmounted, watch } from 'vue'
-import Card from '@/components/ui/card.vue'
-import { LoadingState, TimeRangePicker } from '@/components/common'
-import { ErrorDistributionChart, PercentileChart } from '@/components/stats'
-import LineChart from '@/components/charts/LineChart.vue'
-import { adminApi, type ErrorDistributionResponse, type PercentileItem } from '@/api/admin'
-import { dashboardApi, type ProviderStatus } from '@/api/dashboard'
-import { getDateRangeFromPeriod } from '@/features/usage/composables'
-import type { DateRangeParams } from '@/features/usage/types'
-
-const timeRange = ref<DateRangeParams>(getDateRangeFromPeriod('last30days'))
-
-const percentiles = ref<PercentileItem[]>([])
-const percentileLoading = ref(false)
-
-const errorDistribution = ref<ErrorDistributionResponse['distribution']>([])
-const errorTrend = ref<ErrorDistributionResponse['trend']>([])
-const errorLoading = ref(false)
-
-const providerStatus = ref<ProviderStatus[]>([])
-const providerLoading = ref(false)
-let percentilesRequestId = 0
-let errorsRequestId = 0
-let providersRequestId = 0
-let loadAllPromise: Promise<void> | null = null
-let hasPendingLoadAll = false
-let loadAllDebounceTimer: ReturnType<typeof setTimeout> | null = null
-
-function buildTimeRangeParams() {
-  return {
-    start_date: timeRange.value.start_date,
-    end_date: timeRange.value.end_date,
-    preset: timeRange.value.preset,
-    timezone: timeRange.value.timezone,
-    tz_offset_minutes: timeRange.value.tz_offset_minutes
-  }
-}
-
-async function loadPercentiles() {
-  const requestId = ++percentilesRequestId
-  percentileLoading.value = true
-  try {
-    const data = await adminApi.getPercentiles(buildTimeRangeParams())
-    if (requestId !== percentilesRequestId) return
-    percentiles.value = data
-  } finally {
-    if (requestId === percentilesRequestId) {
-      percentileLoading.value = false
-    }
-  }
-}
-
-async function loadErrors() {
-  const requestId = ++errorsRequestId
-  errorLoading.value = true
-  try {
-    const response = await adminApi.getErrorDistribution(buildTimeRangeParams())
-    if (requestId !== errorsRequestId) return
-    errorDistribution.value = response.distribution
-    errorTrend.value = response.trend
-  } finally {
-    if (requestId === errorsRequestId) {
-      errorLoading.value = false
-    }
-  }
-}
-
-async function loadProviders() {
-  const requestId = ++providersRequestId
-  providerLoading.value = true
-  try {
-    const data = await dashboardApi.getProviderStatus()
-    if (requestId !== providersRequestId) return
-    providerStatus.value = data
-  } finally {
-    if (requestId === providersRequestId) {
-      providerLoading.value = false
-    }
-  }
-}
-
-const errorTrendChartData = computed(() => ({
-  labels: errorTrend.value.map(item => item.date),
-  datasets: [
-    {
-      label: '错误数',
-      data: errorTrend.value.map(item => item.total),
-      borderColor: 'rgb(239, 68, 68)',
-      tension: 0.25,
-      pointRadius: 2
-    }
-  ]
-}))
-
-async function loadAll() {
-  if (loadAllPromise) {
-    hasPendingLoadAll = true
-    return loadAllPromise
-  }
-  loadAllPromise = Promise.all([loadPercentiles(), loadErrors(), loadProviders()])
-    .then(() => undefined)
-    .finally(() => {
-      loadAllPromise = null
-      if (hasPendingLoadAll) {
-        hasPendingLoadAll = false
-        void loadAll()
-      }
-    })
-  return loadAllPromise
-}
-
-function scheduleLoadAll() {
-  if (loadAllDebounceTimer) {
-    clearTimeout(loadAllDebounceTimer)
-  }
-  loadAllDebounceTimer = setTimeout(() => {
-    loadAllDebounceTimer = null
-    void loadAll()
-  }, 120)
-}
-
-watch(timeRange, scheduleLoadAll, { deep: true })
-
-onMounted(() => {
-  void loadAll()
-})
-
-onUnmounted(() => {
-  if (loadAllDebounceTimer) {
-    clearTimeout(loadAllDebounceTimer)
-    loadAllDebounceTimer = null
-  }
-  hasPendingLoadAll = false
-  loadAllPromise = null
-  percentilesRequestId += 1
-  errorsRequestId += 1
-  providersRequestId += 1
-})
-</script>

frontend/src/views/admin/PoolManagement.vue

@@ -1013,7 +1013,7 @@
 
         <!-- Empty keys -->
         <div
-          v-if="keyPage.keys.length === 0 && !keysLoading"
+          v-if="keyPage.keys.length === 0 && !keysLoading && keysLoadedOnce"
           class="flex flex-col items-center justify-center py-16 text-center"
         >
           <div class="mx-auto flex h-16 w-16 items-center justify-center rounded-full bg-muted">
@@ -1155,6 +1155,7 @@ import { useToast } from '@/composables/useToast'
 import { useClipboard } from '@/composables/useClipboard'
 import { useCountdownTimer, getCodexResetCountdown } from '@/composables/useCountdownTimer'
 import { useConfirm } from '@/composables/useConfirm'
+import { useRouteQuery } from '@/composables/useRouteQuery'
 import { parseApiError } from '@/utils/errorParser'
 import {
   getPoolOverview,
@@ -1209,6 +1210,7 @@ const { confirm } = useConfirm()
 const { copyToClipboard } = useClipboard()
 const { tick: countdownTick, start: startCountdownTimer } = useCountdownTimer()
 const proxyNodesStore = useProxyNodesStore()
+const { getQueryValue, patchQuery } = useRouteQuery()
 
 // --- Overview ---
 const poolProviders = ref<PoolOverviewItem[]>([])
@@ -1230,23 +1232,28 @@ async function loadOverview() {
     const enabledProviders = allProviders.filter(item => item.pool_enabled)
     poolProviders.value = enabledProviders
 
-    // Keep selected provider aligned with dropdown options.
-    const selectedId = selectedProviderId.value
+    const queryProviderId = getQueryValue('providerId')
+    const queryProviderExists = Boolean(
+      queryProviderId && enabledProviders.some(item => item.provider_id === queryProviderId),
+    )
+    const selectedId = selectedProviderId.value || (queryProviderExists ? queryProviderId : null)
     const selectedStillExists = Boolean(
       selectedId && enabledProviders.some(item => item.provider_id === selectedId),
     )
 
     if (!selectedStillExists) {
       if (enabledProviders.length > 0) {
-        // Do not block overview loading on key list fetch; keys area has its own loader.
-        void selectProvider(enabledProviders[0].provider_id)
+        await selectProvider(enabledProviders[0].provider_id)
       } else {
         selectedProviderId.value = null
         selectedProviderData.value = null
+        keysLoadedOnce.value = false
         showAccountBatchDialog.value = false
         closeProviderProxyPopovers()
         resetKeyPage()
       }
+    } else if (selectedId && selectedId !== selectedProviderId.value) {
+      await selectProvider(selectedId, { preserveSearch: true })
     }
   } catch (err) {
     if (requestId !== overviewRequestId) return
@@ -1430,7 +1437,7 @@ const desktopColumnWidths = computed(() => {
   }
 })
 
-async function selectProvider(id: string) {
+async function selectProvider(id: string, options: { preserveSearch?: boolean } = {}) {
   const requestId = ++selectProviderRequestId
   selectedProviderId.value = id
   selectedProviderData.value = null
@@ -1444,13 +1451,16 @@ async function selectProvider(id: string) {
   proxyMobilePopoverOpenKeyId.value = null
   suppressFiltersWatch = true
   currentPage.value = 1
+  if (!options.preserveSearch) {
     searchQuery.value = ''
+  }
   statusFilter.value = 'all'
   suppressFiltersWatch = false
   if (keysSearchDebounceTimer !== null) {
     clearTimeout(keysSearchDebounceTimer)
     keysSearchDebounceTimer = null
   }
+  keysLoadedOnce.value = false
   resetKeyPage(1, pageSize.value)
   const keysTask = loadKeys()
   // Provider summary is non-blocking for key list rendering.
@@ -1482,6 +1492,7 @@ function createEmptyKeyPage(page = 1, pageSizeValue = 50): PoolKeysPageResponse
 
 const keyPage = ref<PoolKeysPageResponse>(createEmptyKeyPage())
 const keysLoading = ref(false)
+const keysLoadedOnce = ref(false)
 const refreshingCurrentPageQuota = ref(false)
 const searchQuery = ref('')
 const statusFilter = ref('all')
@@ -1504,6 +1515,33 @@ const keyFormDialogOpen = ref(false)
 const oauthKeyEditDialogOpen = ref(false)
 const editingKeyDetail = ref<PoolKeyDetail | null>(null)
 
+watch(
+  () => getQueryValue('search') ?? '',
+  (value) => {
+    if (searchQuery.value === value) return
+    searchQuery.value = value
+  },
+  { immediate: true },
+)
+
+watch(searchQuery, (value) => {
+  patchQuery({ search: value.trim() || undefined })
+})
+
+watch(
+  () => getQueryValue('providerId'),
+  (value) => {
+    if (overviewLoading.value) return
+    if (!value || value === selectedProviderId.value) return
+    if (!poolProviders.value.some(item => item.provider_id === value)) return
+    void selectProvider(value, { preserveSearch: true })
+  },
+)
+
+watch(selectedProviderId, (value) => {
+  patchQuery({ providerId: value || undefined })
+})
+
 interface QuotaProgressItem {
   label: string
   remainingPercent: number
@@ -1670,9 +1708,11 @@ async function loadKeys() {
     })
     if (requestId !== keysRequestId || selectedProviderId.value !== providerId) return
     keyPage.value = nextPage
+    keysLoadedOnce.value = true
   } catch (err) {
     if (requestId !== keysRequestId || selectedProviderId.value !== providerId) return
     resetKeyPage(page, pageSizeValue)
+    keysLoadedOnce.value = true
     showError(parseApiError(err))
   } finally {
     if (requestId === keysRequestId) {

frontend/src/views/admin/ProviderManagement.vue

@@ -286,6 +286,7 @@ import {
 } from '@/api/endpoints'
 import { adminApi } from '@/api/admin'
 import { parseApiError } from '@/utils/errorParser'
+import { useRouteQuery } from '@/composables/useRouteQuery'
 
 interface ProviderDeleteProgressState {
   providerId: string
@@ -302,6 +303,7 @@ interface ProviderDeleteProgressState {
 
 const { error: showError, success: showSuccess, info: showInfo } = useToast()
 const { confirmDanger } = useConfirm()
+const { getQueryValue, patchQuery } = useRouteQuery()
 
 // 状态
 const loading = ref(false)
@@ -460,6 +462,46 @@ const {
   stopTick,
 } = useProviderBalance()
 
+watch(
+  () => getQueryValue('search') ?? '',
+  (value) => {
+    if (searchQuery.value === value) return
+    searchQuery.value = value
+  },
+  { immediate: true },
+)
+
+watch(searchQuery, (value) => {
+  patchQuery({ search: value.trim() || undefined })
+})
+
+watch(
+  () => getQueryValue('providerId'),
+  (value) => {
+    if (value) {
+      if (selectedProviderId.value === value) {
+        if (!providerDrawerOpen.value) providerDrawerOpen.value = true
+        return
+      }
+      openProviderDrawer(value)
+      return
+    }
+    if (selectedProviderId.value) selectedProviderId.value = null
+    if (providerDrawerOpen.value) providerDrawerOpen.value = false
+  },
+  { immediate: true },
+)
+
+watch(selectedProviderId, (value) => {
+  patchQuery({ providerId: value || undefined })
+})
+
+watch(providerDrawerOpen, (open) => {
+  if (!open && selectedProviderId.value) {
+    selectedProviderId.value = null
+  }
+})
+
 // 扩展操作配置对话框
 const opsConfigDialogOpen = ref(false)
 const opsConfigProviderId = ref('')

frontend/src/views/admin/SystemSettings.vue

@@ -73,6 +73,17 @@
             @update:enable-format-conversion="systemConfig.enable_format_conversion = $event"
           />
 
+          <MonitoringCapacitySection
+            id="section-monitoring-capacity"
+            :redis-memory-total-g-b="redisMemoryTotalGB"
+            :postgres-storage-total-g-b="postgresStorageTotalGB"
+            :loading="monitoringConfigLoading"
+            :has-changes="hasMonitoringConfigChanges"
+            @save="saveMonitoringConfig"
+            @update:redis-memory-total-g-b="redisMemoryTotalGB = $event"
+            @update:postgres-storage-total-g-b="postgresStorageTotalGB = $event"
+          />
+
           <!-- 请求记录配置 -->
           <RequestLogSection
             id="section-request-log"
@@ -211,6 +222,7 @@ import UserDataSection from './system-settings/UserDataSection.vue'
 import DataManagementSection from './system-settings/DataManagementSection.vue'
 import ProxyConfigSection from './system-settings/ProxyConfigSection.vue'
 import BasicConfigSection from './system-settings/BasicConfigSection.vue'
+import MonitoringCapacitySection from './system-settings/MonitoringCapacitySection.vue'
 import RequestLogSection from './system-settings/RequestLogSection.vue'
 import CleanupPolicySection from './system-settings/CleanupPolicySection.vue'
 import ScheduledTasksSection from './system-settings/ScheduledTasksSection.vue'
@@ -230,6 +242,7 @@ const tocItems = [
   { id: 'section-data-mgmt', label: '数据管理' },
   { id: 'section-proxy', label: '网络代理' },
   { id: 'section-basic', label: '基础配置' },
+  { id: 'section-monitoring-capacity', label: '监控容量' },
   { id: 'section-request-log', label: '请求记录' },
   { id: 'section-cleanup', label: '记录清理策略' },
   { id: 'section-scheduled', label: '定时任务' },
@@ -290,21 +303,26 @@ const {
   siteInfoLoading,
   proxyConfigLoading,
   basicConfigLoading,
+  monitoringConfigLoading,
   logConfigLoading,
   cleanupConfigLoading,
   hasSiteInfoChanges,
   hasProxyConfigChanges,
   hasBasicConfigChanges,
+  hasMonitoringConfigChanges,
   hasLogConfigChanges,
   hasCleanupConfigChanges,
   maxRequestBodySizeKB,
   maxResponseBodySizeKB,
+  redisMemoryTotalGB,
+  postgresStorageTotalGB,
   sensitiveHeadersStr,
   loadSystemConfig,
   loadSystemVersion,
   saveSiteInfo,
   saveProxyConfig,
   saveBasicConfig,
+  saveMonitoringConfig,
   saveLogConfig,
   saveCleanupConfig,
   handleAutoCleanupToggle,

frontend/src/views/admin/UserStats.vue

@@ -1,394 +0,0 @@
-<template>
-  <div class="space-y-6 px-4 sm:px-6 lg:px-0">
-    <div class="flex flex-col sm:flex-row sm:items-center sm:justify-between gap-3">
-      <div>
-        <h1 class="text-lg font-semibold">
-          用户统计
-        </h1>
-        <p class="text-xs text-muted-foreground">
-          查看用户排行榜与使用趋势
-        </p>
-      </div>
-      <div class="flex flex-wrap items-center gap-3">
-        <TimeRangePicker
-          v-model="timeRange"
-          :allow-hourly="true"
-        />
-        <Select
-          v-model="selectedUserId"
-        >
-          <SelectTrigger class="h-8 text-xs w-52">
-            <SelectValue placeholder="选择用户" />
-          </SelectTrigger>
-          <SelectContent>
-            <SelectItem
-              v-for="user in users"
-              :key="user.id"
-              :value="user.id"
-            >
-              {{ user.username || user.email }}
-            </SelectItem>
-          </SelectContent>
-        </Select>
-        <Select
-          v-model="compareUserId"
-        >
-          <SelectTrigger class="h-8 text-xs w-52">
-            <SelectValue placeholder="对比用户（可选）" />
-          </SelectTrigger>
-          <SelectContent>
-            <SelectItem value="__none__">
-              不对比
-            </SelectItem>
-            <SelectItem
-              v-for="user in users"
-              :key="`compare-${user.id}`"
-              :value="user.id"
-            >
-              {{ user.username || user.email }}
-            </SelectItem>
-          </SelectContent>
-        </Select>
-      </div>
-    </div>
-
-    <div class="grid grid-cols-1 lg:grid-cols-2 gap-4">
-      <LeaderboardTable
-        title="用户排行榜"
-        :items="leaderboard"
-        :metric="metric"
-        :loading="leaderboardLoading"
-        @update:metric="metric = $event"
-      />
-
-      <Card class="p-4 space-y-3">
-        <h3 class="text-sm font-semibold">
-          用户摘要
-        </h3>
-        <div
-          v-if="summaryLoading"
-          class="p-6"
-        >
-          <LoadingState />
-        </div>
-        <div
-          v-else
-          class="grid grid-cols-2 gap-3 text-sm"
-        >
-          <div>
-            <div class="text-xs text-muted-foreground">
-              请求数
-            </div>
-            <div class="font-semibold">
-              {{ userSummary?.total_requests ?? 0 }}
-            </div>
-          </div>
-          <div>
-            <div class="text-xs text-muted-foreground">
-              Tokens
-            </div>
-            <div class="font-semibold">
-              {{ formatTokens(userSummary?.total_tokens ?? 0) }}
-            </div>
-          </div>
-          <div>
-            <div class="text-xs text-muted-foreground">
-              成本
-            </div>
-            <div class="font-semibold">
-              {{ formatCurrency(userSummary?.total_cost ?? 0) }}
-            </div>
-          </div>
-          <div>
-            <div class="text-xs text-muted-foreground">
-              错误率
-            </div>
-            <div class="font-semibold">
-              {{ userSummary?.error_rate ?? 0 }}%
-            </div>
-          </div>
-        </div>
-      </Card>
-    </div>
-
-    <Card class="p-4 space-y-4">
-      <h3 class="text-sm font-semibold">
-        用户使用趋势
-      </h3>
-      <div
-        v-if="seriesLoading"
-        class="p-6"
-      >
-        <LoadingState />
-      </div>
-      <div
-        v-else
-        class="h-[280px]"
-      >
-        <LineChart :data="seriesChartData" />
-      </div>
-    </Card>
-
-    <Card
-      v-if="comparisonSeries.length > 0"
-      class="p-4 space-y-4"
-    >
-      <h3 class="text-sm font-semibold">
-        用户对比趋势
-      </h3>
-      <div class="h-[280px]">
-        <LineChart :data="comparisonChartData" />
-      </div>
-    </Card>
-  </div>
-</template>
-
-<script setup lang="ts">
-import { ref, computed, onMounted, onUnmounted, watch } from 'vue'
-import { Card, Select, SelectContent, SelectItem, SelectTrigger, SelectValue } from '@/components/ui'
-import LineChart from '@/components/charts/LineChart.vue'
-import { LoadingState, TimeRangePicker } from '@/components/common'
-import { LeaderboardTable } from '@/components/stats'
-import { adminApi, type LeaderboardItem } from '@/api/admin'
-import { usersApi, type User } from '@/api/users'
-import { usageApi } from '@/api/usage'
-import { formatCurrency, formatTokens } from '@/utils/format'
-import { getDateRangeFromPeriod } from '@/features/usage/composables'
-import type { DateRangeParams } from '@/features/usage/types'
-
-const timeRange = ref<DateRangeParams>(getDateRangeFromPeriod('last7days'))
-const metric = ref<'requests' | 'tokens' | 'cost'>('requests')
-
-const users = ref<User[]>([])
-const selectedUserId = ref<string | null>(null)
-const compareUserId = ref<string>('__none__')
-
-const leaderboard = ref<LeaderboardItem[]>([])
-const leaderboardLoading = ref(false)
-
-interface UsageSummary {
-  total_requests: number
-  total_tokens: number
-  total_cost: number
-  error_rate: number
-}
-
-interface TimeSeriesItem {
-  date: string
-  total_cost: number
-}
-
-const userSummary = ref<UsageSummary | null>(null)
-const summaryLoading = ref(false)
-
-const series = ref<TimeSeriesItem[]>([])
-const comparisonSeries = ref<TimeSeriesItem[]>([])
-const seriesLoading = ref(false)
-let leaderboardRequestId = 0
-let summaryRequestId = 0
-let seriesRequestId = 0
-let leaderboardLoadPromise: Promise<void> | null = null
-let hasPendingLeaderboardLoad = false
-let leaderboardDebounceTimer: ReturnType<typeof setTimeout> | null = null
-let userPanelsLoadPromise: Promise<void> | null = null
-let hasPendingUserPanelsLoad = false
-let userPanelsDebounceTimer: ReturnType<typeof setTimeout> | null = null
-
-function buildTimeRangeParams() {
-  return {
-    start_date: timeRange.value.start_date,
-    end_date: timeRange.value.end_date,
-    preset: timeRange.value.preset,
-    timezone: timeRange.value.timezone,
-    tz_offset_minutes: timeRange.value.tz_offset_minutes,
-    granularity: timeRange.value.granularity || 'day'
-  }
-}
-
-async function loadUsers() {
-  users.value = await usersApi.getAllUsers()
-  if (!selectedUserId.value && users.value.length > 0) {
-    selectedUserId.value = users.value[0].id
-  }
-}
-
-async function loadLeaderboard() {
-  if (leaderboardLoadPromise) {
-    hasPendingLeaderboardLoad = true
-    return leaderboardLoadPromise
-  }
-  leaderboardLoadPromise = (async () => {
-  const requestId = ++leaderboardRequestId
-  leaderboardLoading.value = true
-  try {
-    const response = await adminApi.getLeaderboardUsers({
-      ...buildTimeRangeParams(),
-      metric: metric.value,
-      limit: 10
-    })
-    if (requestId !== leaderboardRequestId) return
-    leaderboard.value = response.items
-  } finally {
-    if (requestId === leaderboardRequestId) {
-      leaderboardLoading.value = false
-    }
-  }
-  })().finally(() => {
-    leaderboardLoadPromise = null
-    if (hasPendingLeaderboardLoad) {
-      hasPendingLeaderboardLoad = false
-      void loadLeaderboard()
-    }
-  })
-  return leaderboardLoadPromise
-}
-
-async function loadSummary() {
-  if (!selectedUserId.value) return
-  const requestId = ++summaryRequestId
-  summaryLoading.value = true
-  try {
-    const summary = await usageApi.getUsageStats({
-      ...buildTimeRangeParams(),
-      user_id: selectedUserId.value
-    })
-    if (requestId !== summaryRequestId) return
-    userSummary.value = summary
-  } finally {
-    if (requestId === summaryRequestId) {
-      summaryLoading.value = false
-    }
-  }
-}
-
-async function loadSeries() {
-  if (!selectedUserId.value) return
-  const requestId = ++seriesRequestId
-  seriesLoading.value = true
-  try {
-    const baseParams = {
-      ...buildTimeRangeParams(),
-      user_id: selectedUserId.value
-    }
-    const shouldCompare = Boolean(compareUserId.value && compareUserId.value !== '__none__')
-    const comparePromise: Promise<TimeSeriesItem[]> = shouldCompare
-      ? adminApi.getTimeSeries({
-        ...buildTimeRangeParams(),
-        user_id: compareUserId.value
-      })
-      : Promise.resolve([])
-
-    const [primarySeries, compareSeries] = await Promise.all([
-      adminApi.getTimeSeries(baseParams),
-      comparePromise
-    ])
-
-    if (requestId !== seriesRequestId) return
-    series.value = primarySeries
-    comparisonSeries.value = compareSeries
-  } finally {
-    if (requestId === seriesRequestId) {
-      seriesLoading.value = false
-    }
-  }
-}
-
-async function loadUserPanels() {
-  if (userPanelsLoadPromise) {
-    hasPendingUserPanelsLoad = true
-    return userPanelsLoadPromise
-  }
-  userPanelsLoadPromise = Promise.all([loadSummary(), loadSeries()])
-    .then(() => undefined)
-    .finally(() => {
-      userPanelsLoadPromise = null
-      if (hasPendingUserPanelsLoad) {
-        hasPendingUserPanelsLoad = false
-        void loadUserPanels()
-      }
-    })
-  return userPanelsLoadPromise
-}
-
-const seriesChartData = computed(() => ({
-  labels: series.value.map(item => item.date),
-  datasets: [
-    {
-      label: '成本',
-      data: series.value.map(item => item.total_cost),
-      borderColor: 'rgb(59, 130, 246)',
-      tension: 0.25,
-      pointRadius: 2
-    }
-  ]
-}))
-
-const comparisonChartData = computed(() => ({
-  labels: series.value.map(item => item.date),
-  datasets: [
-    {
-      label: '当前用户',
-      data: series.value.map(item => item.total_cost),
-      borderColor: 'rgb(59, 130, 246)',
-      tension: 0.25,
-      pointRadius: 2
-    },
-    {
-      label: '对比用户',
-      data: comparisonSeries.value.map(item => item.total_cost),
-      borderColor: 'rgb(234, 179, 8)',
-      tension: 0.25,
-      pointRadius: 2
-    }
-  ]
-}))
-
-function scheduleLeaderboardLoad() {
-  if (leaderboardDebounceTimer) {
-    clearTimeout(leaderboardDebounceTimer)
-  }
-  leaderboardDebounceTimer = setTimeout(() => {
-    leaderboardDebounceTimer = null
-    void loadLeaderboard()
-  }, 120)
-}
-
-function scheduleUserPanelsLoad() {
-  if (userPanelsDebounceTimer) {
-    clearTimeout(userPanelsDebounceTimer)
-  }
-  userPanelsDebounceTimer = setTimeout(() => {
-    userPanelsDebounceTimer = null
-    void loadUserPanels()
-  }, 120)
-}
-
-watch([timeRange, metric], scheduleLeaderboardLoad, { deep: true })
-watch([timeRange, selectedUserId, compareUserId], scheduleUserPanelsLoad, { deep: true })
-
-onMounted(async () => {
-  await loadUsers()
-  await Promise.all([
-    loadLeaderboard(),
-    loadUserPanels()
-  ])
-})
-
-onUnmounted(() => {
-  if (leaderboardDebounceTimer) {
-    clearTimeout(leaderboardDebounceTimer)
-    leaderboardDebounceTimer = null
-  }
-  if (userPanelsDebounceTimer) {
-    clearTimeout(userPanelsDebounceTimer)
-    userPanelsDebounceTimer = null
-  }
-  hasPendingLeaderboardLoad = false
-  hasPendingUserPanelsLoad = false
-  leaderboardLoadPromise = null
-  userPanelsLoadPromise = null
-  leaderboardRequestId += 1
-  summaryRequestId += 1
-  seriesRequestId += 1
-})
-</script>

frontend/src/views/admin/Users.vue

@@ -1019,7 +1019,7 @@ import { adminWalletApi, type AdminWallet } from '@/api/admin-wallets'
 import { useToast } from '@/composables/useToast'
 import { useConfirm } from '@/composables/useConfirm'
 import { useClipboard } from '@/composables/useClipboard'
-import { usageApi, type UsageByUser } from '@/api/usage'
+import { analyticsApi } from '@/api/analytics'
 import { adminApi } from '@/api/admin'
 import { walletStatusBadge, walletStatusLabel } from '@/utils/walletDisplay'
 
@@ -1102,7 +1102,14 @@ const userApiKeyForm = ref({
 })
 
 // 用户统计
-const userStats = ref<Record<string, UsageByUser>>({})
+interface UserUsageSummary {
+  user_id: string
+  request_count: number
+  total_tokens: number
+  total_cost: number
+}
+
+const userStats = ref<Record<string, UserUsageSummary>>({})
 const loadingStats = ref(false)
 let userStatsRequestId = 0
 const userWalletMap = ref<Record<string, AdminWallet>>({})
@@ -1181,10 +1188,26 @@ async function loadUserStats() {
   const requestId = ++userStatsRequestId
   loadingStats.value = true
   try {
-    const data = await usageApi.getUsageByUser()
+    const response = await analyticsApi.getBreakdown({
+      scope: { kind: 'global' },
+      time_range: {
+        preset: 'last30days',
+        granularity: 'day',
+        timezone: 'UTC',
+        tz_offset_minutes: 0,
+      },
+      dimension: 'user',
+      metric: 'requests_total',
+      limit: 200,
+    })
     if (requestId !== userStatsRequestId) return
-    userStats.value = data.reduce((acc: Record<string, UsageByUser>, stat: UsageByUser) => {
-      acc[stat.user_id] = stat
+    userStats.value = response.rows.reduce((acc: Record<string, UserUsageSummary>, row) => {
+      acc[row.key] = {
+        user_id: row.key,
+        request_count: row.requests_total,
+        total_tokens: row.total_tokens,
+        total_cost: row.total_cost_usd,
+      }
       return acc
     }, {})
   } catch (err) {

frontend/src/views/admin/components/analytics/LeaderboardTab.vue

@@ -0,0 +1,496 @@
+<template>
+  <div class="space-y-4">
+    <div class="flex flex-col gap-2 sm:flex-row sm:items-center sm:justify-between">
+      <p class="text-xs text-muted-foreground">
+        固定按费用排序，点击{{ dimensionLabel }}名称即可在右侧查看同时间范围下的趋势。
+      </p>
+      <button
+        class="inline-flex h-8 items-center rounded-lg border border-border/70 bg-background px-3 text-[10px] font-medium text-muted-foreground transition-colors hover:bg-muted/40 hover:text-foreground"
+        @click="exportLeaderboard"
+      >
+        导出 CSV
+      </button>
+    </div>
+
+    <div class="grid gap-4 xl:grid-cols-[minmax(0,390px)_minmax(0,1fr)] 2xl:grid-cols-[minmax(0,420px)_minmax(0,1.18fr)]">
+      <LeaderboardTable
+        :title="leaderboardTitle"
+        :items="paginatedLeaderboard"
+        :selected-id="selectedId"
+        :loading="leaderboardLoading"
+        :unavailable="leaderboardUnavailable"
+        :has-loaded="leaderboardHasLoaded"
+        :current-page="leaderboardCurrentPage"
+        :total-items="leaderboard.length"
+        :page-size="LEADERBOARD_PAGE_SIZE"
+        :show-pagination="true"
+        @select="selectedId = $event"
+        @update:current-page="leaderboardCurrentPage = $event"
+      />
+
+      <Card class="overflow-hidden">
+        <div class="border-b border-border/60 px-4 py-4">
+          <div class="flex flex-col gap-3 lg:flex-row lg:items-start lg:justify-between">
+            <div class="min-w-0">
+              <div class="flex flex-wrap items-center gap-2">
+                <Badge
+                  variant="outline"
+                  class="border-amber-500/35 bg-amber-500/8 text-[10px] text-amber-700 dark:text-amber-300"
+                >
+                  {{ selectedItem ? `#${selectedItem.rank}` : '未选择' }}
+                </Badge>
+                <h3 class="truncate text-sm font-semibold">
+                  {{ selectedLabel || `${dimensionLabel}趋势` }}
+                </h3>
+              </div>
+              <p class="mt-1 text-[11px] text-muted-foreground">
+                {{ trendLeadText }}
+              </p>
+            </div>
+
+            <Tabs
+              v-model="currentTrendGranularitySelection"
+              class="self-start"
+            >
+              <div class="flex flex-col items-start gap-2">
+                <div
+                  v-if="isPanelRefreshing"
+                  class="inline-flex h-7 items-center rounded-lg border border-border/70 bg-background px-2.5 text-[10px] text-muted-foreground"
+                >
+                  更新中
+                </div>
+                <TabsList class="tabs-button-list flex-wrap justify-start gap-1">
+                  <TabsTrigger
+                    v-for="option in trendGranularityTabStates"
+                    :key="option.value"
+                    :value="option.value"
+                    :disabled="option.disabled"
+                    class="min-w-[38px] px-2 py-0.5 text-[10px]"
+                    :class="option.disabled ? 'cursor-not-allowed opacity-40' : ''"
+                  >
+                    {{ option.label }}
+                  </TabsTrigger>
+                </TabsList>
+              </div>
+            </Tabs>
+          </div>
+
+          <div
+            v-if="selectedSummary"
+            class="mt-4 grid grid-cols-2 gap-2 transition-opacity sm:grid-cols-4"
+            :class="{ 'opacity-60': isPanelRefreshing }"
+          >
+            <div class="rounded-xl border border-border/60 bg-muted/15 px-3 py-2.5">
+              <div class="text-[10px] uppercase tracking-[0.18em] text-muted-foreground">
+                请求数
+              </div>
+              <div class="mt-1 text-sm font-semibold tabular-nums">
+                {{ selectedSummary.total_requests.toLocaleString() }}
+              </div>
+            </div>
+            <div class="rounded-xl border border-border/60 bg-muted/15 px-3 py-2.5">
+              <div class="text-[10px] uppercase tracking-[0.18em] text-muted-foreground">
+                总计 Tokens
+              </div>
+              <div class="mt-1 text-sm font-semibold tabular-nums">
+                {{ formatTokens(selectedSummary.total_tokens) }}
+              </div>
+            </div>
+            <div class="rounded-xl border border-border/60 bg-muted/15 px-3 py-2.5">
+              <div class="text-[10px] uppercase tracking-[0.18em] text-muted-foreground">
+                费用
+              </div>
+              <div class="mt-1 text-sm font-semibold tabular-nums">
+                {{ formatCurrency(selectedSummary.total_cost) }}
+              </div>
+            </div>
+            <div class="rounded-xl border border-border/60 bg-muted/15 px-3 py-2.5">
+              <div class="text-[10px] uppercase tracking-[0.18em] text-muted-foreground">
+                成本
+              </div>
+              <div class="mt-1 text-sm font-semibold tabular-nums">
+                {{ formatCurrency(selectedSummary.total_actual_cost) }}
+              </div>
+            </div>
+          </div>
+        </div>
+
+        <LineChart
+          v-if="selectedItem && trendChartData"
+          :class="{ 'opacity-60 transition-opacity': isPanelRefreshing }"
+          :data="trendChartData"
+          :options="trendChartOptions"
+          :height="320"
+        />
+        <div
+          v-else
+          class="flex h-[320px] items-center justify-center px-6 text-center text-xs text-muted-foreground"
+        >
+          {{ trendEmptyText }}
+        </div>
+      </Card>
+    </div>
+  </div>
+</template>
+
+<script setup lang="ts">
+import { computed, ref, toRef, watch } from 'vue'
+import type { ChartData, ChartOptions } from 'chart.js'
+import { Badge, Card, Tabs, TabsList, TabsTrigger } from '@/components/ui'
+import { LeaderboardTable } from '@/components/stats'
+import LineChart from '@/components/charts/LineChart.vue'
+import { useInjectedAnalyticsFilters } from '@/composables/useAnalyticsFilters'
+import { useLeaderboardData, type LeaderboardDimension } from '@/composables/analytics'
+import {
+  analyticsGranularityTabs,
+  formatAnalyticsTrendBucketLabel,
+  formatAnalyticsTrendDescription,
+  formatAnalyticsTrendTooltipTitle,
+  isAnalyticsSingleDayRange,
+} from '@/utils/analyticsGranularity'
+import { downloadCsvFromObjects } from '@/utils/csvExport'
+import { formatCurrency, formatTokens } from '@/utils/format'
+
+const props = defineProps<{
+  dimension: LeaderboardDimension
+}>()
+
+const dimension = toRef(props, 'dimension')
+const filters = useInjectedAnalyticsFilters()
+const {
+  leaderboard,
+  leaderboardLoading,
+  leaderboardError,
+  leaderboardHasLoaded,
+  selectedId,
+  selectedItem,
+  selectedLabel,
+  selectedSummary,
+  selectedTrendBuckets,
+  panelLoading,
+  panelError,
+  panelHasLoaded,
+  resolvedTrendGranularity,
+  currentTrendGranularitySelection,
+} = useLeaderboardData(dimension)
+
+const trendAxisTickCount = 6
+const LEADERBOARD_PAGE_SIZE = 5
+const leaderboardCurrentPage = ref(1)
+
+const dimensionLabel = computed(() => (
+  props.dimension === 'users' ? '用户' : 'API Key'
+))
+
+const leaderboardTitle = computed(() => `${dimensionLabel.value}费用排行榜`)
+const leaderboardUnavailable = computed(() => (
+  leaderboardError.value && !leaderboardHasLoaded.value
+))
+const leaderboardTotalPages = computed(() => (
+  Math.max(1, Math.ceil(leaderboard.value.length / LEADERBOARD_PAGE_SIZE))
+))
+const paginatedLeaderboard = computed(() => {
+  const start = (leaderboardCurrentPage.value - 1) * LEADERBOARD_PAGE_SIZE
+  return leaderboard.value.slice(start, start + LEADERBOARD_PAGE_SIZE)
+})
+
+const trendGranularityTabStates = computed(() => (
+  analyticsGranularityTabs.map(option => ({
+    ...option,
+    disabled: option.value === 'hour' ? !isAnalyticsSingleDayRange(filters.timeRange.value) : false,
+  }))
+))
+
+const trendChartDescription = computed(() => (
+  formatAnalyticsTrendDescription(
+    resolvedTrendGranularity.value,
+    '请求量、费用、成本与总计 Tokens',
+  )
+))
+
+const trendLeadText = computed(() => {
+  if (leaderboardUnavailable.value) {
+    return `${dimensionLabel.value}排行榜接口未返回结果，请稍后重试。`
+  }
+  if (!selectedItem.value) {
+    return `从左侧选择${dimensionLabel.value}后，这里会显示与首页仪表盘一致的趋势视图。`
+  }
+  return `${selectedLabel.value} · ${trendChartDescription.value}`
+})
+
+const trendEmptyText = computed(() => {
+  if (leaderboardUnavailable.value) {
+    return `${dimensionLabel.value}排行榜暂不可用`
+  }
+  if (!selectedItem.value) {
+    return `请选择一个${dimensionLabel.value}查看趋势`
+  }
+  if (panelError.value && !panelHasLoaded.value) {
+    return `${dimensionLabel.value}趋势暂不可用`
+  }
+  return isInitialPanelLoading.value ? '加载中...' : '暂无趋势数据'
+})
+
+const isInitialPanelLoading = computed(() => (
+  panelLoading.value && !panelHasLoaded.value && selectedTrendBuckets.value.length === 0
+))
+
+const isPanelRefreshing = computed(() => (
+  panelLoading.value && panelHasLoaded.value && selectedTrendBuckets.value.length > 0
+))
+
+watch(
+  () => [
+    props.dimension,
+    filters.timeRange.value.preset,
+    filters.timeRange.value.start_date,
+    filters.timeRange.value.end_date,
+  ],
+  () => {
+    leaderboardCurrentPage.value = 1
+  },
+)
+
+watch(
+  leaderboardTotalPages,
+  (totalPages) => {
+    if (leaderboardCurrentPage.value > totalPages) {
+      leaderboardCurrentPage.value = totalPages
+    }
+  },
+)
+
+watch(
+  [paginatedLeaderboard, selectedId],
+  ([items, currentSelectedId]) => {
+    if (!items.length) return
+    if (!items.some(item => item.id === currentSelectedId)) {
+      selectedId.value = items[0].id
+    }
+  },
+  { immediate: true },
+)
+
+function computeNiceAxisMax(maxValue: number, intervals: number): number {
+  if (!Number.isFinite(maxValue) || maxValue <= 0) return 1
+  const rawStep = maxValue / intervals
+  const magnitude = 10 ** Math.floor(Math.log10(rawStep))
+  const normalized = rawStep / magnitude
+  const niceNormalized = normalized <= 1
+    ? 1
+    : normalized <= 2
+      ? 2
+      : normalized <= 2.5
+        ? 2.5
+        : normalized <= 5
+          ? 5
+          : 10
+  return niceNormalized * magnitude * intervals
+}
+
+const trendScaleMaxima = computed(() => {
+  const intervals = Math.max(trendAxisTickCount - 1, 1)
+  const requestMax = Math.max(0, ...selectedTrendBuckets.value.map(bucket => bucket.requests_total))
+  const tokenMax = Math.max(0, ...selectedTrendBuckets.value.map(bucket => bucket.total_tokens))
+  const costMax = Math.max(
+    0,
+    ...selectedTrendBuckets.value.map(bucket => Math.max(bucket.total_cost_usd, bucket.actual_total_cost_usd)),
+  )
+
+  return {
+    requests: computeNiceAxisMax(requestMax, intervals),
+    tokens: computeNiceAxisMax(tokenMax, intervals),
+    cost: computeNiceAxisMax(costMax, intervals),
+  }
+})
+
+const trendChartData = computed<ChartData<'line'> | null>(() => {
+  if (!selectedTrendBuckets.value.length) return null
+
+  return {
+    labels: selectedTrendBuckets.value.map(bucket => (
+      formatAnalyticsTrendBucketLabel(bucket.bucket_start, resolvedTrendGranularity.value)
+    )),
+    datasets: [
+      {
+        label: '费用',
+        data: selectedTrendBuckets.value.map(bucket => bucket.total_cost_usd),
+        borderColor: 'rgb(232, 145, 89)',
+        backgroundColor: 'rgba(232, 145, 89, 0.12)',
+        borderWidth: 2.2,
+        fill: false,
+        tension: 0.28,
+        pointRadius: 0,
+        pointHoverRadius: 4,
+        pointHitRadius: 14,
+        yAxisID: 'y2',
+      },
+      {
+        label: '成本',
+        data: selectedTrendBuckets.value.map(bucket => bucket.actual_total_cost_usd),
+        borderColor: 'rgb(196, 92, 124)',
+        backgroundColor: 'rgba(196, 92, 124, 0.12)',
+        borderWidth: 2,
+        borderDash: [7, 5],
+        fill: false,
+        tension: 0.28,
+        pointRadius: 0,
+        pointHoverRadius: 4,
+        pointHitRadius: 14,
+        yAxisID: 'y2',
+      },
+      {
+        label: '总计 Tokens',
+        data: selectedTrendBuckets.value.map(bucket => bucket.total_tokens),
+        borderColor: 'rgb(103, 149, 230)',
+        backgroundColor: 'rgba(103, 149, 230, 0.16)',
+        borderWidth: 2.6,
+        fill: true,
+        tension: 0.3,
+        pointRadius: 0,
+        pointHoverRadius: 4,
+        pointHitRadius: 14,
+        yAxisID: 'y1',
+      },
+      {
+        label: '请求数',
+        data: selectedTrendBuckets.value.map(bucket => bucket.requests_total),
+        borderColor: 'rgb(146, 126, 214)',
+        backgroundColor: 'rgba(146, 126, 214, 0.08)',
+        borderWidth: 2.2,
+        borderDash: [7, 5],
+        fill: false,
+        tension: 0.28,
+        pointRadius: 0,
+        pointHoverRadius: 4,
+        pointHitRadius: 14,
+        yAxisID: 'y',
+      },
+    ],
+  }
+})
+
+const trendChartOptions = computed<ChartOptions<'line'>>(() => ({
+  responsive: true,
+  maintainAspectRatio: false,
+  interaction: {
+    mode: 'index',
+    intersect: false,
+  },
+  scales: {
+    x: {
+      grid: {
+        color: 'rgba(156, 163, 175, 0.05)',
+      },
+      ticks: {
+        color: 'rgb(92, 98, 109)',
+        maxRotation: 0,
+        autoSkip: true,
+        maxTicksLimit: resolvedTrendGranularity.value === 'hour' ? 12 : undefined,
+      },
+    },
+    y: {
+      type: 'linear',
+      display: true,
+      position: 'left',
+      min: 0,
+      max: trendScaleMaxima.value.requests,
+      beginAtZero: true,
+      alignToPixels: true,
+      bounds: 'ticks',
+      grid: {
+        color: 'rgba(156, 163, 175, 0.05)',
+      },
+      ticks: {
+        color: 'rgb(92, 98, 109)',
+        count: trendAxisTickCount,
+      },
+    },
+    y1: {
+      type: 'linear',
+      display: true,
+      position: 'right',
+      min: 0,
+      max: trendScaleMaxima.value.tokens,
+      beginAtZero: true,
+      alignToPixels: true,
+      bounds: 'ticks',
+      grid: {
+        drawOnChartArea: false,
+      },
+      ticks: {
+        color: 'rgb(92, 98, 109)',
+        count: trendAxisTickCount,
+        callback: value => {
+          const tokenValue = Number(value)
+          const tokenMax = trendScaleMaxima.value.tokens
+          const costMax = trendScaleMaxima.value.cost
+          const ratio = tokenMax > 0 ? tokenValue / tokenMax : 0
+          const costValue = ratio * costMax
+          return [formatTokens(tokenValue), formatCurrency(costValue)]
+        },
+      },
+    },
+    y2: {
+      type: 'linear',
+      display: false,
+      position: 'right',
+      min: 0,
+      max: trendScaleMaxima.value.cost,
+      beginAtZero: true,
+      alignToPixels: true,
+      bounds: 'ticks',
+      grid: {
+        drawOnChartArea: false,
+      },
+      ticks: {
+        color: 'rgb(92, 98, 109)',
+        count: trendAxisTickCount,
+        callback: value => formatCurrency(Number(value)),
+      },
+    },
+  },
+  plugins: {
+    legend: {
+      labels: {
+        usePointStyle: true,
+      },
+    },
+    tooltip: {
+      callbacks: {
+        title: items => {
+          const bucket = selectedTrendBuckets.value[items[0]?.dataIndex ?? -1]
+          if (!bucket) return ''
+          return formatAnalyticsTrendTooltipTitle(bucket.bucket_start, resolvedTrendGranularity.value)
+        },
+        label: context => {
+          const value = Number(context.raw ?? 0)
+          if (context.dataset.label === '总计 Tokens') {
+            return `${context.dataset.label}: ${formatTokens(value)}`
+          }
+          if (context.dataset.label === '费用' || context.dataset.label === '成本') {
+            return `${context.dataset.label}: ${formatCurrency(value)}`
+          }
+          return `${context.dataset.label}: ${value.toLocaleString()}`
+        },
+      },
+    },
+  },
+}))
+
+function exportLeaderboard() {
+  if (!leaderboard.value.length) return
+  downloadCsvFromObjects(
+    `leaderboard-${props.dimension}-${new Date().toISOString().slice(0, 10)}`,
+    leaderboard.value as unknown as Record<string, unknown>[],
+    [
+      { key: 'rank', label: '排名' },
+      { key: 'name', label: '名称' },
+      { key: 'requests', label: '请求数' },
+      { key: 'tokens', label: 'Tokens' },
+      { key: 'cost', label: '费用' },
+      { key: 'actualCost', label: '成本' },
+    ],
+  )
+}
+</script>

frontend/src/views/admin/components/analytics/PerformanceTab.vue

@@ -0,0 +1,921 @@
+<template>
+  <div class="space-y-5">
+    <div class="space-y-3">
+      <div class="flex flex-col gap-2 sm:flex-row sm:items-start sm:justify-between">
+        <div>
+          <h3 class="text-sm font-medium">
+            系统资源监控
+          </h3>
+          <p class="mt-1 text-[11px] text-muted-foreground">
+            6 项核心资源的实时压力指数，越满表示越紧张
+          </p>
+        </div>
+        <div
+          v-if="systemStatusTimestamp"
+          class="inline-flex h-7 items-center rounded-lg border border-border/70 bg-background px-2.5 text-[10px] text-muted-foreground"
+        >
+          {{ systemStatusTimestamp }}
+        </div>
+      </div>
+
+      <div class="grid gap-3 sm:grid-cols-2 xl:grid-cols-3">
+        <Card
+          v-for="item in systemMetricCards"
+          :key="item.key"
+          class="min-h-[148px] border-border/60 bg-card p-3.5"
+          :class="{ 'opacity-60 transition-opacity': isRefreshing }"
+        >
+          <div class="flex items-start justify-between gap-2.5">
+            <div class="min-w-0">
+              <div class="text-[10px] font-medium uppercase tracking-[0.14em] text-muted-foreground/85">
+                {{ item.label }}
+              </div>
+              <div class="mt-1.5 text-xl font-semibold tracking-tight text-foreground">
+                {{ item.value }}
+              </div>
+              <div class="mt-1 line-clamp-2 text-[10px] text-muted-foreground">
+                {{ item.detail }}
+              </div>
+            </div>
+
+            <div
+              class="flex h-8 w-8 shrink-0 items-center justify-center rounded-xl border"
+              :class="item.iconClass"
+            >
+              <component :is="item.icon" class="h-3.5 w-3.5" />
+            </div>
+          </div>
+
+          <div class="mt-2.5 space-y-2">
+            <div class="flex items-center justify-between gap-2">
+              <span class="min-w-0 truncate text-[10px] text-muted-foreground">
+                {{ item.meta }}
+              </span>
+              <span
+                class="inline-flex shrink-0 rounded-full border px-2 py-0.5 text-[10px] font-medium"
+                :class="item.badgeClass"
+              >
+                {{ item.badge }}
+              </span>
+            </div>
+
+            <div class="h-1.5 overflow-hidden rounded-full bg-muted/50">
+              <div
+                class="h-full rounded-full transition-all"
+                :class="item.progressClass"
+                :style="{ width: item.progressWidth }"
+              />
+            </div>
+
+            <div class="flex items-center justify-between gap-2 text-[10px] text-muted-foreground">
+              <span>{{ item.caption }}</span>
+              <span class="tabular-nums">{{ item.captionValue }}</span>
+            </div>
+          </div>
+        </Card>
+      </div>
+    </div>
+
+    <div class="space-y-3">
+      <div class="flex flex-col gap-2 sm:flex-row sm:items-start sm:justify-between">
+        <div>
+          <h3 class="text-sm font-medium">
+            运行概览
+          </h3>
+          <p class="mt-1 text-[11px] text-muted-foreground">
+            汇总健康状态、今日负载与当前风险信号
+          </p>
+        </div>
+        <div
+          v-if="isRefreshing"
+          class="inline-flex h-7 items-center rounded-lg border border-border/70 bg-background px-2.5 text-[10px] text-muted-foreground"
+        >
+          更新中
+        </div>
+      </div>
+
+      <div class="grid gap-3 sm:grid-cols-2 xl:grid-cols-3">
+        <Card
+          v-for="item in statusCards"
+          :key="item.key"
+          class="min-h-[138px] border-border/60 bg-card p-4"
+          :class="{ 'opacity-60 transition-opacity': isRefreshing }"
+        >
+          <div class="flex items-start justify-between gap-3">
+            <div class="min-w-0">
+              <div class="text-[10px] font-medium uppercase tracking-[0.14em] text-muted-foreground/85">
+                {{ item.label }}
+              </div>
+              <div class="mt-2 text-2xl font-semibold tracking-tight text-foreground">
+                {{ item.value }}
+              </div>
+              <div class="mt-1.5 text-[11px] leading-relaxed text-muted-foreground">
+                {{ item.detail }}
+              </div>
+            </div>
+
+            <div
+              class="flex h-10 w-10 shrink-0 items-center justify-center rounded-2xl border"
+              :class="item.iconClass"
+            >
+              <component :is="item.icon" class="h-4 w-4" />
+            </div>
+          </div>
+
+          <div class="mt-3 flex items-center justify-between gap-3 border-t border-border/50 pt-3">
+            <span class="min-w-0 truncate text-[11px] text-muted-foreground">
+              {{ item.caption }}
+            </span>
+            <span
+              class="inline-flex rounded-full border px-2 py-0.5 text-[10px] font-medium"
+              :class="item.badgeClass"
+            >
+              {{ item.badge }}
+            </span>
+          </div>
+        </Card>
+      </div>
+    </div>
+
+    <div class="grid gap-4 lg:grid-cols-2">
+      <Card class="overflow-hidden border-border/60 bg-card p-4">
+        <div :class="{ 'opacity-60 transition-opacity': isRefreshing }">
+          <PercentileChart
+            title="响应延迟百分位"
+            subtitle="观察 P50 / P90 / P99 响应耗时的分布变化"
+            :series="percentiles"
+            mode="response"
+            :loading="isInitialLoading"
+            :unavailable="performanceUnavailable"
+            unavailable-text="性能数据暂不可用"
+          />
+        </div>
+      </Card>
+
+      <Card class="overflow-hidden border-border/60 bg-card p-4">
+        <div :class="{ 'opacity-60 transition-opacity': isRefreshing }">
+          <PercentileChart
+            title="首字节延迟百分位"
+            subtitle="聚焦首包返回速度，判断链路是否出现抖动"
+            :series="percentiles"
+            mode="ttfb"
+            :loading="isInitialLoading"
+            :unavailable="performanceUnavailable"
+            unavailable-text="性能数据暂不可用"
+          />
+        </div>
+      </Card>
+
+      <Card class="overflow-hidden border-border/60 bg-card p-4">
+        <div :class="{ 'opacity-60 transition-opacity': isRefreshing }">
+          <ErrorDistributionChart
+            title="错误分布"
+            subtitle="按错误类别观察主要失败来源"
+            :distribution="errorDistribution"
+            :loading="isInitialLoading"
+            :unavailable="performanceUnavailable"
+            unavailable-text="性能数据暂不可用"
+          />
+        </div>
+      </Card>
+
+      <Card class="overflow-hidden border-border/60 bg-card p-4">
+        <div class="space-y-4" :class="{ 'opacity-60 transition-opacity': isRefreshing }">
+          <div class="flex flex-col gap-3 sm:flex-row sm:items-start sm:justify-between">
+            <div>
+              <h3 class="text-sm font-semibold">
+                错误趋势
+              </h3>
+              <p class="mt-1 text-[11px] text-muted-foreground">
+                观察错误量在当前时间范围内的波动与峰值
+              </p>
+            </div>
+
+            <div
+              v-if="!performanceUnavailable"
+              class="flex flex-wrap gap-2"
+            >
+              <div class="rounded-full border border-rose-500/20 bg-rose-500/8 px-2.5 py-1 text-[10px] font-medium text-rose-700 dark:text-rose-300">
+                总错误 {{ errorTotalCount }}
+              </div>
+              <div
+                v-if="errorPeakLabel"
+                class="rounded-full border border-border/70 bg-muted/20 px-2.5 py-1 text-[10px] font-medium text-muted-foreground"
+              >
+                峰值 {{ errorPeakLabel }}
+              </div>
+            </div>
+          </div>
+
+          <div
+            v-if="isInitialLoading"
+            class="p-6"
+          >
+            <LoadingState />
+          </div>
+          <div
+            v-else-if="performanceUnavailable"
+            class="flex h-[272px] items-center justify-center rounded-2xl border border-dashed border-border/70 bg-muted/10 text-xs text-muted-foreground"
+          >
+            性能数据暂不可用
+          </div>
+          <div
+            v-else-if="errorTrendChartData"
+            class="rounded-2xl border border-border/60 bg-background p-3"
+          >
+            <LineChart
+              :data="errorTrendChartData"
+              :options="errorTrendChartOptions"
+              :height="248"
+            />
+          </div>
+          <div
+            v-else
+            class="flex h-[272px] items-center justify-center rounded-2xl border border-dashed border-border/70 bg-muted/10 text-xs text-muted-foreground"
+          >
+            暂无错误趋势数据
+          </div>
+        </div>
+      </Card>
+    </div>
+
+    <div>
+      <div class="mb-3 flex items-center justify-between">
+        <h3 class="text-sm font-medium">
+          提供商健康度
+        </h3>
+        <div
+          v-if="isRefreshing"
+          class="inline-flex h-7 items-center rounded-lg border border-border/70 bg-background px-2.5 text-[10px] text-muted-foreground"
+        >
+          更新中
+        </div>
+      </div>
+      <div
+        v-if="performanceUnavailable"
+        class="text-xs text-muted-foreground"
+      >
+        性能数据暂不可用
+      </div>
+      <div
+        v-else-if="providerStatus.length === 0 && !loading"
+        class="text-xs text-muted-foreground"
+      >
+        暂无提供商状态数据
+      </div>
+      <div class="grid grid-cols-1 gap-2.5 sm:grid-cols-2 lg:grid-cols-5">
+        <Card
+          v-for="provider in providerStatus"
+          :key="provider.provider_name"
+          class="space-y-2 border-border/60 bg-card p-3"
+          :class="{ 'opacity-60 transition-opacity': isRefreshing }"
+        >
+          <div class="flex items-start justify-between gap-2">
+            <div class="min-w-0">
+              <div class="truncate text-xs font-semibold" :title="provider.provider_name">
+                {{ provider.provider_name }}
+              </div>
+              <div class="mt-1 text-[10px] text-muted-foreground tabular-nums">
+                {{ provider.requests_total?.toLocaleString() ?? 0 }} 请求
+              </div>
+            </div>
+            <Badge
+              :variant="provider.error_rate > 5 ? 'destructive' : provider.success_rate >= 98 ? 'success' : 'warning'"
+              class="h-5 px-1.5 text-[9px]"
+            >
+              {{ provider.error_rate > 5 ? '异常' : provider.success_rate >= 98 ? '稳定' : '波动' }}
+            </Badge>
+          </div>
+          <div class="grid grid-cols-2 gap-x-3 gap-y-2 text-[10px] text-muted-foreground">
+            <div>
+              成功率
+              <div class="mt-0.5 font-semibold text-foreground tabular-nums">
+                {{ provider.success_rate.toFixed(1) }}%
+              </div>
+            </div>
+            <div>
+              错误率
+              <div class="mt-0.5 font-semibold text-foreground tabular-nums">
+                {{ provider.error_rate.toFixed(1) }}%
+              </div>
+            </div>
+            <div>
+              响应
+              <div class="mt-0.5 font-semibold text-foreground tabular-nums">
+                {{ formatLatency(provider.avg_response_time_ms) }}
+              </div>
+            </div>
+            <div>
+              TTFB
+              <div class="mt-0.5 font-semibold text-foreground tabular-nums">
+                {{ formatLatency(provider.avg_first_byte_time_ms) }}
+              </div>
+            </div>
+          </div>
+        </Card>
+      </div>
+    </div>
+  </div>
+</template>
+
+<script setup lang="ts">
+import { computed, type Component } from 'vue'
+import type { ChartData, ChartOptions } from 'chart.js'
+import { Activity, AlertTriangle, Cpu, Database, Gauge, HardDrive, Key } from 'lucide-vue-next'
+import type { MonitoringMetricStatus } from '@/api/audit'
+import { Badge, Card } from '@/components/ui'
+import { LoadingState } from '@/components/common'
+import { PercentileChart, ErrorDistributionChart } from '@/components/stats'
+import LineChart from '@/components/charts/LineChart.vue'
+import { usePerformanceData } from '@/composables/analytics'
+import { formatCurrency, formatTokens } from '@/utils/format'
+
+const {
+  latency,
+  percentiles,
+  errorDistribution,
+  errorTrend,
+  errorTotal: errorTotalCount,
+  errorRate,
+  providerStatus,
+  healthSummary,
+  systemStatus,
+  loading,
+  loadError,
+  hasLoaded,
+} = usePerformanceData()
+
+const isInitialLoading = computed(() => loading.value && !hasLoaded.value)
+const isRefreshing = computed(() => loading.value && hasLoaded.value)
+const performanceUnavailable = computed(() => loadError.value && !hasLoaded.value)
+const errorPeakPoint = computed(() => (
+  errorTrend.value.slice().sort((left, right) => right.total - left.total)[0] ?? null
+))
+const errorPeakLabel = computed(() => (
+  errorPeakPoint.value && errorPeakPoint.value.total > 0
+    ? `${errorPeakPoint.value.date.slice(5)} · ${errorPeakPoint.value.total}`
+    : null
+))
+
+function formatLatency(value: number) {
+  if (!value || value <= 0) return '-'
+  if (value < 1000) return `${Math.round(value)}ms`
+  return `${(value / 1000).toFixed(2)}s`
+}
+
+function formatCount(value: number | null | undefined, ready: boolean) {
+  return ready && value != null ? value.toLocaleString() : '--'
+}
+
+function formatRatio(active: number | null | undefined, total: number | null | undefined, ready: boolean) {
+  return ready && active != null && total != null ? `${active}/${total}` : '--'
+}
+
+type StatusTone = 'success' | 'warning' | 'danger' | 'info' | 'neutral'
+
+interface StatusCardItem {
+  key: string
+  label: string
+  value: string
+  detail: string
+  caption: string
+  badge: string
+  badgeClass: string
+  iconClass: string
+  icon: Component
+}
+
+interface SystemMetricCardItem {
+  key: string
+  label: string
+  value: string
+  detail: string
+  meta: string
+  caption: string
+  captionValue: string
+  badge: string
+  badgeClass: string
+  iconClass: string
+  progressClass: string
+  progressWidth: string
+  icon: Component
+}
+
+function toneStyle(tone: StatusTone) {
+  switch (tone) {
+    case 'success':
+      return {
+        badgeClass: 'border-emerald-500/20 bg-emerald-500/8 text-emerald-700 dark:text-emerald-300',
+        iconClass: 'border-emerald-500/20 bg-emerald-500/8 text-emerald-700 dark:text-emerald-300',
+      }
+    case 'warning':
+      return {
+        badgeClass: 'border-amber-500/20 bg-amber-500/8 text-amber-700 dark:text-amber-300',
+        iconClass: 'border-amber-500/20 bg-amber-500/8 text-amber-700 dark:text-amber-300',
+      }
+    case 'danger':
+      return {
+        badgeClass: 'border-rose-500/20 bg-rose-500/8 text-rose-700 dark:text-rose-300',
+        iconClass: 'border-rose-500/20 bg-rose-500/8 text-rose-700 dark:text-rose-300',
+      }
+    case 'info':
+      return {
+        badgeClass: 'border-sky-500/20 bg-sky-500/8 text-sky-700 dark:text-sky-300',
+        iconClass: 'border-sky-500/20 bg-sky-500/8 text-sky-700 dark:text-sky-300',
+      }
+    default:
+      return {
+        badgeClass: 'border-border/70 bg-muted/20 text-muted-foreground',
+        iconClass: 'border-border/70 bg-muted/20 text-muted-foreground',
+      }
+  }
+}
+
+function toneProgressClass(tone: StatusTone) {
+  switch (tone) {
+    case 'success':
+      return 'bg-emerald-500/80'
+    case 'warning':
+      return 'bg-amber-500/80'
+    case 'danger':
+      return 'bg-rose-500/80'
+    case 'info':
+      return 'bg-sky-500/80'
+    default:
+      return 'bg-muted-foreground/40'
+  }
+}
+
+function healthTone(unhealthy: number, total: number): StatusTone {
+  if (total <= 0) return 'neutral'
+  if (unhealthy <= 0) return 'success'
+  if (unhealthy / total < 0.25) return 'warning'
+  return 'danger'
+}
+
+function availabilityTone(active: number, total: number): StatusTone {
+  if (total <= 0) return 'neutral'
+  if (active === total) return 'success'
+  if (active > 0) return 'warning'
+  return 'danger'
+}
+
+function latencyTone(avgMs: number): StatusTone {
+  if (!avgMs || avgMs <= 0) return 'neutral'
+  if (avgMs < 1500) return 'success'
+  if (avgMs < 3000) return 'warning'
+  return 'danger'
+}
+
+function monitoringTone(status: MonitoringMetricStatus | undefined): StatusTone {
+  switch (status) {
+    case 'ok':
+      return 'success'
+    case 'warning':
+    case 'degraded':
+      return 'warning'
+    case 'danger':
+    case 'error':
+      return 'danger'
+    default:
+      return 'neutral'
+  }
+}
+
+function formatPercent(value: number | null | undefined) {
+  return value == null ? '--' : `${value.toFixed(1)}%`
+}
+
+function formatCompactCount(value: number | null | undefined) {
+  return value == null ? '--' : value.toLocaleString()
+}
+
+function formatBytes(value: number | null | undefined) {
+  if (value == null) return '--'
+  if (value === 0) return '0 B'
+
+  const units = ['B', 'KB', 'MB', 'GB', 'TB']
+  let current = value
+  let unitIndex = 0
+
+  while (current >= 1024 && unitIndex < units.length - 1) {
+    current /= 1024
+    unitIndex += 1
+  }
+
+  const digits = current >= 100 ? 0 : current >= 10 ? 1 : 2
+  return `${current.toFixed(digits)} ${units[unitIndex]}`
+}
+
+function formatTimeLabel(timestamp: string | null | undefined) {
+  if (!timestamp) return null
+  return `更新 ${new Date(timestamp).toLocaleTimeString('zh-CN', {
+    hour: '2-digit',
+    minute: '2-digit',
+  })}`
+}
+
+function clampPercent(value: number | null | undefined) {
+  if (value == null) return null
+  return Math.max(0, Math.min(100, value))
+}
+
+function progressWidth(value: number | null | undefined) {
+  const safeValue = clampPercent(value)
+  const width = safeValue == null ? 12 : Math.max(6, safeValue)
+  return `${width}%`
+}
+
+function remainingToPressurePercent(value: number | null | undefined) {
+  const safeValue = clampPercent(value)
+  return safeValue == null ? null : Math.max(0, 100 - safeValue)
+}
+
+function formatPressureValue(value: number | null | undefined) {
+  return formatPercent(clampPercent(value))
+}
+
+function latencyPressureValue(latencyMs: number | null | undefined, status: MonitoringMetricStatus | undefined) {
+  if (latencyMs == null) {
+    return statusPressureValue(status)
+  }
+
+  if (latencyMs >= 200) return 100
+  if (latencyMs >= 120) return 88
+  if (latencyMs >= 80) return 72
+  if (latencyMs >= 40) return 34
+  if (latencyMs >= 20) return 24
+  return 16
+}
+
+function statusPressureValue(status: MonitoringMetricStatus | undefined) {
+  switch (status) {
+    case 'ok':
+      return 16
+    case 'warning':
+      return 72
+    case 'degraded':
+      return 82
+    case 'danger':
+      return 92
+    case 'error':
+      return 100
+    default:
+      return null
+  }
+}
+
+const PRESSURE_CAPTION = '压力指数'
+
+function pressureMetric(value: number | null | undefined) {
+  return clampPercent(value)
+}
+
+function storagePressureMetric(value: number | null | undefined) {
+  return remainingToPressurePercent(value)
+}
+
+function fallbackPressureWidth() {
+  return progressWidth(null)
+}
+
+function fallbackPressureCaptionValue() {
+  return '--'
+}
+
+const statusCards = computed<StatusCardItem[]>(() => {
+  if (performanceUnavailable.value) {
+    return [
+      { key: 'endpoint-health', label: 'Endpoint 健康', icon: Database },
+      { key: 'key-health', label: 'Key 健康', icon: Key },
+      { key: 'provider-status', label: 'Provider 在线', icon: Cpu },
+      { key: 'today-traffic', label: '今日流量', icon: Activity },
+      { key: 'recent-alerts', label: '近 1h 告警', icon: AlertTriangle },
+      { key: 'latency', label: '平均延迟', icon: Gauge },
+    ].map(item => ({
+      ...item,
+      value: '不可用',
+      detail: '性能接口未返回结果',
+      caption: '请稍后重试',
+      badge: '不可用',
+      ...toneStyle('neutral'),
+    }))
+  }
+
+  const healthReady = healthSummary.value !== null
+  const systemReady = systemStatus.value !== null
+
+  const endpointTotal = healthSummary.value?.endpoints.total ?? 0
+  const endpointActive = healthSummary.value?.endpoints.active ?? 0
+  const endpointUnhealthy = healthSummary.value?.endpoints.unhealthy ?? 0
+  const endpointTone = toneStyle(healthTone(endpointUnhealthy, endpointTotal))
+
+  const keyTotal = healthSummary.value?.keys.total ?? systemStatus.value?.api_keys.total ?? 0
+  const keyActive = healthSummary.value?.keys.active ?? systemStatus.value?.api_keys.active ?? 0
+  const keyUnhealthy = healthSummary.value?.keys.unhealthy ?? Math.max(keyTotal - keyActive, 0)
+  const keyCircuitOpen = healthSummary.value?.keys.circuit_open ?? 0
+  const keyTone = toneStyle(healthTone(keyUnhealthy, keyTotal))
+
+  const providerTotal = systemStatus.value?.providers.total ?? 0
+  const providerActive = systemStatus.value?.providers.active ?? 0
+  const userActive = systemStatus.value?.users.active ?? 0
+  const userTotal = systemStatus.value?.users.total ?? 0
+  const providerTone = toneStyle(availabilityTone(providerActive, providerTotal))
+
+  const todayRequests = systemStatus.value?.today_stats.requests ?? 0
+  const todayTokens = systemStatus.value?.today_stats.tokens ?? 0
+  const todayCost = systemStatus.value?.today_stats.cost_usd ?? 0
+  const throughputTone = toneStyle(todayRequests > 0 ? 'info' : 'neutral')
+
+  const recentAlerts = systemStatus.value?.recent_errors ?? 0
+  const alertsTone = toneStyle(recentAlerts <= 0 && errorRate.value < 2 ? 'success' : recentAlerts < 5 && errorRate.value < 5 ? 'warning' : 'danger')
+
+  const responseAvg = latency.value.response_time_ms.avg ?? 0
+  const ttfbAvg = latency.value.first_byte_time_ms.avg ?? 0
+  const latencyStyle = toneStyle(latencyTone(responseAvg))
+
+  return [
+    {
+      key: 'endpoint-health',
+      label: 'Endpoint 健康',
+      value: formatRatio(endpointActive, endpointTotal, healthReady),
+      detail: healthReady ? `异常端点 ${endpointUnhealthy}` : '健康摘要载入中',
+      caption: healthReady ? '当前可用端点覆盖' : '等待健康摘要返回',
+      badge: healthReady ? (endpointUnhealthy > 0 ? '关注中' : '稳定') : '载入中',
+      ...endpointTone,
+      icon: Database,
+    },
+    {
+      key: 'key-health',
+      label: 'Key 健康',
+      value: formatRatio(keyActive, keyTotal, healthReady || systemReady),
+      detail: healthReady ? `异常 Key ${keyUnhealthy} · 熔断 ${keyCircuitOpen}` : '健康摘要载入中',
+      caption: healthReady ? '密钥池可用状态' : '等待健康摘要返回',
+      badge: healthReady ? (keyUnhealthy > 0 || keyCircuitOpen > 0 ? '波动' : '健康') : '载入中',
+      ...keyTone,
+      icon: Key,
+    },
+    {
+      key: 'provider-status',
+      label: 'Provider 在线',
+      value: formatRatio(providerActive, providerTotal, systemReady),
+      detail: systemReady ? `活跃用户 ${userActive}/${userTotal}` : '系统状态载入中',
+      caption: systemReady ? '提供商接入状态' : '等待系统状态返回',
+      badge: systemReady ? (providerActive === providerTotal ? '在线' : '需关注') : '载入中',
+      ...providerTone,
+      icon: Cpu,
+    },
+    {
+      key: 'today-traffic',
+      label: '今日流量',
+      value: formatCount(todayRequests, systemReady),
+      detail: systemReady ? `Tokens ${formatTokens(todayTokens)}` : '系统状态载入中',
+      caption: systemReady ? `成本 ${formatCurrency(todayCost)}` : '等待系统状态返回',
+      badge: systemReady ? (todayRequests > 0 ? '活跃' : '空闲') : '载入中',
+      ...throughputTone,
+      icon: Activity,
+    },
+    {
+      key: 'recent-alerts',
+      label: '近 1h 告警',
+      value: formatCount(recentAlerts, systemReady),
+      detail: `当前范围失败 ${errorTotalCount.value.toLocaleString()}`,
+      caption: `错误率 ${errorRate.value.toFixed(1)}%`,
+      badge: systemReady
+        ? recentAlerts <= 0 && errorRate.value < 2 ? '低风险' : recentAlerts < 5 && errorRate.value < 5 ? '需关注' : '偏高'
+        : '载入中',
+      ...alertsTone,
+      icon: AlertTriangle,
+    },
+    {
+      key: 'latency-baseline',
+      label: '延迟基线',
+      value: formatLatency(responseAvg),
+      detail: `TTFB ${formatLatency(ttfbAvg)}`,
+      caption: `P50 ${formatLatency(latency.value.response_time_ms.p50 ?? 0)}`,
+      badge: responseAvg > 0 ? (responseAvg < 1500 ? '顺畅' : responseAvg < 3000 ? '一般' : '偏慢') : '无数据',
+      ...latencyStyle,
+      icon: Gauge,
+    },
+  ]
+})
+
+const systemStatusTimestamp = computed(() => formatTimeLabel(systemStatus.value?.timestamp))
+
+const systemMetricCards = computed<SystemMetricCardItem[]>(() => {
+  const metrics = systemStatus.value?.system_metrics
+
+  if (!metrics) {
+    const detail = systemStatus.value ? 'system_metrics 未返回' : '系统状态载入中'
+    const tone = toneStyle('neutral')
+    const progressClass = toneProgressClass('neutral')
+    return [
+      { key: 'cpu', label: 'CPU', icon: Cpu },
+      { key: 'memory', label: '内存', icon: HardDrive },
+      { key: 'redis-latency', label: 'Redis', icon: Activity },
+      { key: 'redis-memory', label: 'Redis 内存', icon: HardDrive },
+      { key: 'postgres-pool', label: 'PostgreSQL', icon: Database },
+      { key: 'postgres-storage', label: 'PostgreSQL 空间', icon: HardDrive },
+    ].map(item => ({
+      ...item,
+      value: '--',
+      detail,
+      meta: '等待监控数据',
+      caption: PRESSURE_CAPTION,
+      captionValue: fallbackPressureCaptionValue(),
+      badge: '未知',
+      progressWidth: fallbackPressureWidth(),
+      progressClass,
+      ...tone,
+    }))
+  }
+
+  const cpuTone = monitoringTone(metrics.cpu.status)
+  const memoryTone = monitoringTone(metrics.memory.status)
+  const redisTone = monitoringTone(metrics.redis.status)
+  const redisMemoryTone = monitoringTone(metrics.redis.memory_status ?? metrics.redis.status)
+  const postgresTone = monitoringTone(metrics.postgres.status)
+  const postgresStorageTone = monitoringTone(metrics.postgres.storage_status ?? metrics.postgres.status)
+  const redisMemoryCeiling = metrics.redis.memory_ceiling_bytes ?? metrics.redis.maxmemory_bytes ?? null
+  const redisMemorySource = metrics.redis.memory_source ?? 'unknown'
+  const redisMemorySourceLabel = redisMemorySource === 'configured'
+    ? '手动容量'
+    : redisMemorySource === 'maxmemory'
+        ? '上限'
+      : redisMemorySource === 'system'
+        ? '估算上限'
+        : '未配置容量'
+  const cpuPressure = pressureMetric(metrics.cpu.usage_percent ?? metrics.cpu.load_percent)
+  const memoryPressure = pressureMetric(metrics.memory.used_percent)
+  const redisHealthPressure = latencyPressureValue(metrics.redis.latency_ms, metrics.redis.status)
+  const redisMemoryPressure = pressureMetric(metrics.redis.memory_percent ?? statusPressureValue(metrics.redis.memory_status))
+  const postgresPoolPressure = pressureMetric(
+    metrics.postgres.usage_percent ?? metrics.postgres.pool_usage_percent ?? metrics.postgres.server_usage_percent,
+  )
+  const postgresStoragePressure = storagePressureMetric(metrics.postgres.storage_free_percent)
+
+  return [
+    {
+      key: 'cpu',
+      label: 'CPU',
+      value: formatPercent(metrics.cpu.usage_percent ?? metrics.cpu.load_percent),
+      detail: metrics.cpu.usage_percent != null
+        ? `实时占用 · Load ${formatPercent(metrics.cpu.load_percent)}`
+        : (metrics.cpu.message ?? 'CPU 监控数据不可用'),
+      meta: metrics.cpu.core_count > 0 ? `${metrics.cpu.core_count} 核` : '核心数未知',
+      caption: PRESSURE_CAPTION,
+      captionValue: formatPressureValue(cpuPressure),
+      badge: metrics.cpu.label,
+      progressWidth: progressWidth(cpuPressure),
+      progressClass: toneProgressClass(cpuTone),
+      ...toneStyle(cpuTone),
+      icon: Cpu,
+    },
+    {
+      key: 'memory',
+      label: '内存',
+      value: formatPercent(metrics.memory.used_percent),
+      detail: metrics.memory.total_bytes != null
+        ? `${formatBytes(metrics.memory.used_bytes)} / ${formatBytes(metrics.memory.total_bytes)}`
+        : (metrics.memory.message ?? '内存监控数据不可用'),
+      meta: metrics.memory.available_bytes != null ? `可用 ${formatBytes(metrics.memory.available_bytes)}` : '等待系统指标',
+      caption: PRESSURE_CAPTION,
+      captionValue: formatPressureValue(memoryPressure),
+      badge: metrics.memory.label,
+      progressWidth: progressWidth(memoryPressure),
+      progressClass: toneProgressClass(memoryTone),
+      ...toneStyle(memoryTone),
+      icon: HardDrive,
+    },
+    {
+      key: 'redis-latency',
+      label: 'Redis',
+      value: metrics.redis.latency_ms != null ? formatLatency(metrics.redis.latency_ms) : metrics.redis.label,
+      detail: metrics.redis.message ?? '缓存链路实时探测',
+      meta: metrics.redis.latency_ms != null ? 'PING 延迟' : '连接状态',
+      caption: PRESSURE_CAPTION,
+      captionValue: formatPressureValue(redisHealthPressure),
+      badge: metrics.redis.label,
+      progressWidth: progressWidth(redisHealthPressure),
+      progressClass: toneProgressClass(redisTone),
+      ...toneStyle(redisTone),
+      icon: Activity,
+    },
+    {
+      key: 'redis-memory',
+      label: 'Redis 内存',
+      value: formatBytes(metrics.redis.used_memory_bytes),
+      detail: metrics.redis.peak_memory_bytes != null
+        ? `峰值 ${formatBytes(metrics.redis.peak_memory_bytes)}`
+        : (redisMemorySource === 'unknown' ? '可在系统设置中手动填写 Redis 总内存' : '等待 Redis 内存指标'),
+      meta: redisMemoryCeiling != null
+        ? `${redisMemorySourceLabel} ${formatBytes(redisMemoryCeiling)}`
+        : '未提供内存上限',
+      caption: PRESSURE_CAPTION,
+      captionValue: formatPressureValue(redisMemoryPressure),
+      badge: metrics.redis.memory_label ?? metrics.redis.label,
+      progressWidth: progressWidth(redisMemoryPressure),
+      progressClass: toneProgressClass(redisMemoryTone),
+      ...toneStyle(redisMemoryTone),
+      icon: HardDrive,
+    },
+    {
+      key: 'postgres-pool',
+      label: 'PostgreSQL',
+      value: formatPercent(metrics.postgres.usage_percent),
+      detail: metrics.postgres.message
+        ?? (metrics.postgres.server_connections != null && metrics.postgres.server_max_connections != null
+          ? `服务端 ${formatCompactCount(metrics.postgres.server_connections)} / ${formatCompactCount(metrics.postgres.server_max_connections)} 连接`
+          : `${formatCompactCount(metrics.postgres.checked_out)} / ${formatCompactCount(metrics.postgres.max_capacity)} 连接占用`),
+      meta: metrics.postgres.server_connections != null && metrics.postgres.server_max_connections != null
+        ? `应用池 ${formatCompactCount(metrics.postgres.checked_out)} / ${formatCompactCount(metrics.postgres.max_capacity)}`
+        : `池 ${formatCompactCount(metrics.postgres.pool_size)} · 溢出 ${formatCompactCount(metrics.postgres.overflow)}`,
+      caption: PRESSURE_CAPTION,
+      captionValue: formatPressureValue(postgresPoolPressure),
+      badge: metrics.postgres.label,
+      progressWidth: progressWidth(postgresPoolPressure),
+      progressClass: toneProgressClass(postgresTone),
+      ...toneStyle(postgresTone),
+      icon: Database,
+    },
+    {
+      key: 'postgres-storage',
+      label: 'PostgreSQL 空间',
+      value: formatBytes(metrics.postgres.storage_free_bytes),
+      detail: metrics.postgres.database_size_bytes != null
+        ? `库体积 ${formatBytes(metrics.postgres.database_size_bytes)}`
+        : (metrics.postgres.storage_message ?? '等待 PostgreSQL 空间指标'),
+      meta: metrics.postgres.storage_total_bytes != null
+        ? `总容量 ${formatBytes(metrics.postgres.storage_total_bytes)}`
+        : (metrics.postgres.storage_message ?? '数据目录总量未知'),
+      caption: PRESSURE_CAPTION,
+      captionValue: formatPressureValue(postgresStoragePressure),
+      badge: metrics.postgres.storage_label ?? metrics.postgres.label,
+      progressWidth: progressWidth(postgresStoragePressure),
+      progressClass: toneProgressClass(postgresStorageTone),
+      ...toneStyle(postgresStorageTone),
+      icon: HardDrive,
+    },
+  ]
+})
+
+const errorTrendChartData = computed<ChartData<'line'> | null>(() => {
+  if (!errorTrend.value.length) return null
+  return {
+    labels: errorTrend.value.map(d => d.date?.slice(5) ?? ''),
+    datasets: [{
+      label: '错误数',
+      data: errorTrend.value.map(d => d.total),
+      borderColor: 'rgb(226, 85, 118)',
+      backgroundColor: 'rgba(226, 85, 118, 0.16)',
+      fill: true,
+      tension: 0.34,
+      pointRadius: 0,
+      pointHoverRadius: 4,
+      pointHitRadius: 14,
+      borderWidth: 2.4,
+    }],
+  }
+})
+
+const errorTrendChartOptions = computed<ChartOptions<'line'>>(() => ({
+  responsive: true,
+  maintainAspectRatio: false,
+  interaction: {
+    mode: 'index',
+    intersect: false,
+  },
+  plugins: {
+    legend: {
+      display: false,
+    },
+    tooltip: {
+      callbacks: {
+        title: items => errorTrend.value[items[0]?.dataIndex ?? -1]?.date ?? '',
+        label: context => `错误数: ${Number(context.raw ?? 0).toLocaleString()}`,
+      },
+    },
+  },
+  scales: {
+    x: {
+      grid: {
+        color: 'rgba(148, 163, 184, 0.06)',
+      },
+      ticks: {
+        color: 'rgb(100, 116, 139)',
+        maxRotation: 0,
+      },
+    },
+    y: {
+      beginAtZero: true,
+      grid: {
+        color: 'rgba(148, 163, 184, 0.07)',
+      },
+      ticks: {
+        color: 'rgb(100, 116, 139)',
+        precision: 0,
+      },
+    },
+  },
+}))
+</script>

frontend/src/views/admin/system-settings/MonitoringCapacitySection.vue

@@ -0,0 +1,81 @@
+<template>
+  <CardSection
+    title="监控容量"
+    description="为远程或托管 Redis / PostgreSQL 手动填写总容量，便于监控面板计算剩余空间"
+  >
+    <template #actions>
+      <Button
+        size="sm"
+        :disabled="loading || !hasChanges"
+        @click="$emit('save')"
+      >
+        {{ loading ? '保存中...' : '保存' }}
+      </Button>
+    </template>
+    <div class="grid grid-cols-1 md:grid-cols-2 gap-6">
+      <div>
+        <Label
+          for="redis-memory-total-gb"
+          class="block text-sm font-medium"
+        >
+          Redis 总内存 (GB)
+        </Label>
+        <Input
+          id="redis-memory-total-gb"
+          :model-value="redisMemoryTotalGB"
+          type="number"
+          step="0.1"
+          min="0"
+          placeholder="0"
+          class="mt-1"
+          @update:model-value="$emit('update:redisMemoryTotalGB', Number($event))"
+        />
+        <p class="mt-1 text-xs text-muted-foreground">
+          用于远程 Redis 未上报完整容量时计算剩余内存，0 表示继续自动探测
+        </p>
+      </div>
+
+      <div>
+        <Label
+          for="postgres-storage-total-gb"
+          class="block text-sm font-medium"
+        >
+          PostgreSQL 总空间 (GB)
+        </Label>
+        <Input
+          id="postgres-storage-total-gb"
+          :model-value="postgresStorageTotalGB"
+          type="number"
+          step="0.1"
+          min="0"
+          placeholder="0"
+          class="mt-1"
+          @update:model-value="$emit('update:postgresStorageTotalGB', Number($event))"
+        />
+        <p class="mt-1 text-xs text-muted-foreground">
+          监控面板会用“总空间 - 当前数据库体积”估算剩余空间，0 表示未设置
+        </p>
+      </div>
+    </div>
+  </CardSection>
+</template>
+
+<script setup lang="ts">
+import Button from '@/components/ui/button.vue'
+import Input from '@/components/ui/input.vue'
+import Label from '@/components/ui/label.vue'
+import { CardSection } from '@/components/layout'
+
+defineProps<{
+  redisMemoryTotalGB: number
+  postgresStorageTotalGB: number
+  loading: boolean
+  hasChanges: boolean
+}>()
+
+defineEmits<{
+  save: []
+  'update:redisMemoryTotalGB': [value: number]
+  'update:postgresStorageTotalGB': [value: number]
+}>()
+</script>

frontend/src/views/admin/system-settings/composables/useSystemConfig.ts

@@ -19,6 +19,9 @@ export interface SystemConfig {
   auto_delete_expired_keys: boolean
   // 格式转换
   enable_format_conversion: boolean
+  // 监控容量
+  redis_memory_total_bytes: number
+  postgres_storage_total_bytes: number
   // 请求记录
   request_record_level: string
   max_request_body_size: number
@@ -55,6 +58,9 @@ const CONFIG_KEYS = [
   'auto_delete_expired_keys',
   // 格式转换
   'enable_format_conversion',
+  // 监控容量
+  'redis_memory_total_bytes',
+  'postgres_storage_total_bytes',
   // 请求记录
   'request_record_level',
   'max_request_body_size',
@@ -92,6 +98,9 @@ function createDefaultConfig(): SystemConfig {
     auto_delete_expired_keys: false,
     // 格式转换
     enable_format_conversion: false,
+    // 监控容量
+    redis_memory_total_bytes: 0,
+    postgres_storage_total_bytes: 0,
     // 请求记录
     request_record_level: 'basic',
     max_request_body_size: 1048576,
@@ -114,6 +123,18 @@ function createDefaultConfig(): SystemConfig {
   }
 }
 
+const BYTES_PER_GB = 1024 * 1024 * 1024
+
+function toBytesFromGB(value: number): number {
+  if (!Number.isFinite(value) || value <= 0) return 0
+  return Math.round(value * BYTES_PER_GB)
+}
+
+function toGBFromBytes(value: number): number {
+  if (!Number.isFinite(value) || value <= 0) return 0
+  return Math.round((value / BYTES_PER_GB) * 100) / 100
+}
+
 export function useSystemConfig() {
   const { success, error } = useToast()
   const { refreshSiteInfo } = useSiteInfo()
@@ -126,6 +147,7 @@ export function useSystemConfig() {
   const siteInfoLoading = ref(false)
   const proxyConfigLoading = ref(false)
   const basicConfigLoading = ref(false)
+  const monitoringConfigLoading = ref(false)
   const logConfigLoading = ref(false)
   const cleanupConfigLoading = ref(false)
 
@@ -166,6 +188,14 @@ export function useSystemConfig() {
     )
   })
 
+  const hasMonitoringConfigChanges = computed(() => {
+    if (!originalConfig.value) return false
+    return (
+      systemConfig.value.redis_memory_total_bytes !== originalConfig.value.redis_memory_total_bytes ||
+      systemConfig.value.postgres_storage_total_bytes !== originalConfig.value.postgres_storage_total_bytes
+    )
+  })
+
   const hasCleanupConfigChanges = computed(() => {
     if (!originalConfig.value) return false
     return (
@@ -200,6 +230,20 @@ export function useSystemConfig() {
     },
   })
 
+  const redisMemoryTotalGB = computed({
+    get: () => toGBFromBytes(systemConfig.value.redis_memory_total_bytes),
+    set: (val: number) => {
+      systemConfig.value.redis_memory_total_bytes = toBytesFromGB(val)
+    },
+  })
+
+  const postgresStorageTotalGB = computed({
+    get: () => toGBFromBytes(systemConfig.value.postgres_storage_total_bytes),
+    set: (val: number) => {
+      systemConfig.value.postgres_storage_total_bytes = toBytesFromGB(val)
+    },
+  })
+
   // 敏感请求头数组和字符串之间的转换
   const sensitiveHeadersStr = computed({
     get: () => systemConfig.value.sensitive_headers.join(', '),
@@ -397,6 +441,40 @@ export function useSystemConfig() {
     }
   }
 
+  async function saveMonitoringConfig() {
+    monitoringConfigLoading.value = true
+    try {
+      const configItems = [
+        {
+          key: 'redis_memory_total_bytes',
+          value: systemConfig.value.redis_memory_total_bytes,
+          description: 'Redis 总内存容量（字节）',
+        },
+        {
+          key: 'postgres_storage_total_bytes',
+          value: systemConfig.value.postgres_storage_total_bytes,
+          description: 'PostgreSQL 总存储空间（字节）',
+        },
+      ]
+
+      await Promise.all(
+        configItems.map((item) =>
+          adminApi.updateSystemConfig(item.key, item.value, item.description)
+        )
+      )
+      if (originalConfig.value) {
+        originalConfig.value.redis_memory_total_bytes = systemConfig.value.redis_memory_total_bytes
+        originalConfig.value.postgres_storage_total_bytes = systemConfig.value.postgres_storage_total_bytes
+      }
+      success('监控容量配置已保存')
+    } catch (err) {
+      error('保存监控容量配置失败')
+      log.error('保存监控容量配置失败:', err)
+    } finally {
+      monitoringConfigLoading.value = false
+    }
+  }
+
   async function saveCleanupConfig() {
     cleanupConfigLoading.value = true
     try {
@@ -497,17 +575,21 @@ export function useSystemConfig() {
     siteInfoLoading,
     proxyConfigLoading,
     basicConfigLoading,
+    monitoringConfigLoading,
     logConfigLoading,
     cleanupConfigLoading,
     // 变动检测
     hasSiteInfoChanges,
     hasProxyConfigChanges,
     hasBasicConfigChanges,
+    hasMonitoringConfigChanges,
     hasLogConfigChanges,
     hasCleanupConfigChanges,
     // 计算属性
     maxRequestBodySizeKB,
     maxResponseBodySizeKB,
+    redisMemoryTotalGB,
+    postgresStorageTotalGB,
     sensitiveHeadersStr,
     // 加载函数
     loadSystemConfig,
@@ -516,6 +598,7 @@ export function useSystemConfig() {
     saveSiteInfo,
     saveProxyConfig,
     saveBasicConfig,
+    saveMonitoringConfig,
     saveLogConfig,
     saveCleanupConfig,
     handleAutoCleanupToggle,

frontend/src/views/public/home-config.ts

@@ -63,6 +63,7 @@ disable_response_storage = true
 name = "OpenAI"
 base_url = "${baseUrl.value}/v1"
 wire_api = "responses"
+supports_websockets = true
 requires_openai_auth = true`)
 
   const codexAuthConfig = computed(() => `{

frontend/src/views/shared/HealthMonitor.vue

@@ -1,5 +1,5 @@
 <template>
-  <div class="space-y-6 pb-8">
+  <div class="space-y-6 px-4 sm:px-6 lg:px-0 pb-8">
     <HealthMonitorCard
       title="健康监控"
       :is-admin="isAdminPage"
@@ -11,8 +11,10 @@
 <script setup lang="ts">
 import { computed } from 'vue'
 import { useRoute } from 'vue-router'
+import { useAuthStore } from '@/stores/auth'
 import HealthMonitorCard from '@/features/providers/components/HealthMonitorCard.vue'
 
 const route = useRoute()
-const isAdminPage = computed(() => route.path.startsWith('/admin'))
+const authStore = useAuthStore()
+const isAdminPage = computed(() => authStore.isAdmin || route.path.startsWith('/admin'))
 </script>

frontend/src/views/shared/Usage.vue

@@ -1,79 +1,13 @@
 <template>
   <div class="space-y-6 pb-8">
-    <!-- 面包屑旁的折叠按钮 -->
-    <Teleport
-      to="#header-actions-right"
-      defer
+    <!-- 跳转数据分析链接 -->
+    <div class="text-right">
+      <RouterLink
+        :to="isAdminPage ? '/admin/analytics' : '/dashboard/reports'"
+        class="text-xs text-muted-foreground hover:text-foreground transition-colors"
       >
-      <button
-        class="flex h-9 w-9 items-center justify-center rounded-lg text-muted-foreground hover:text-foreground hover:bg-muted/50 transition"
-        :title="statsExpanded ? '收起用量分析' : '展开用量分析'"
-        @click="statsExpanded = !statsExpanded"
-      >
-        <PanelTopClose
-          v-if="statsExpanded"
-          class="h-4 w-4"
-        />
-        <PanelTopOpen
-          v-else
-          class="h-4 w-4"
-        />
-      </button>
-    </Teleport>
-
-    <!-- 用量分析面板（可折叠） -->
-    <div
-      v-if="statsExpanded"
-      class="space-y-4"
-    >
-      <!-- 活跃度热图 + 请求间隔时间线 -->
-      <div class="grid grid-cols-1 xl:grid-cols-2 gap-4">
-        <ActivityHeatmapCard
-          :data="activityHeatmapData"
-          :title="isAdminPage ? '总体活跃天数' : '我的活跃天数'"
-          :is-loading="isLoadingHeatmap"
-          :has-error="heatmapError"
-        />
-        <IntervalTimelineCard
-          :title="isAdminPage ? '请求间隔时间线' : '我的请求间隔'"
-          :is-admin="isAdminPage"
-          :hours="24"
-        />
-      </div>
-
-      <!-- 分析统计 -->
-      <!-- 管理员：模型 + 提供商 + API格式（3列） -->
-      <div
-        v-if="isAdminPage"
-        class="grid grid-cols-1 lg:grid-cols-3 gap-4"
-      >
-        <UsageModelTable
-          :data="enhancedModelStats"
-          :is-admin="authStore.isAdmin"
-        />
-        <UsageProviderTable
-          :data="providerStats"
-          :is-admin="authStore.isAdmin"
-        />
-        <UsageApiFormatTable
-          :data="apiFormatStats"
-          :is-admin="authStore.isAdmin"
-        />
-      </div>
-      <!-- 用户：模型 + API格式（2列） -->
-      <div
-        v-else
-        class="grid grid-cols-1 lg:grid-cols-2 gap-4"
-      >
-        <UsageModelTable
-          :data="enhancedModelStats"
-          :is-admin="authStore.isAdmin"
-        />
-        <UsageApiFormatTable
-          :data="apiFormatStats"
-          :is-admin="false"
-        />
-      </div>
+        {{ isAdminPage ? '查看数据分析 →' : '查看综合报表 →' }}
+      </RouterLink>
     </div>
 
     <!-- 使用记录 -->
@@ -85,13 +19,17 @@
       :time-range="timeRange"
       :filter-search="filterSearch"
       :filter-user="filterUser"
+      :filter-api-key="filterApiKey"
       :filter-model="filterModel"
       :filter-provider="filterProvider"
       :filter-api-format="filterApiFormat"
       :filter-status="filterStatus"
       :available-users="availableUsers"
+      :available-api-keys="availableApiKeys"
       :available-models="availableModels"
+      :available-api-formats="availableApiFormats"
       :available-providers="availableProviders"
+      :available-statuses="availableStatuses"
       :current-page="currentPage"
       :page-size="pageSize"
       :total-records="effectiveTotalRecords"
@@ -100,6 +38,7 @@
       @update:time-range="handleTimeRangeChange"
       @update:filter-search="handleFilterSearchChange"
       @update:filter-user="handleFilterUserChange"
+      @update:filter-api-key="handleFilterApiKeyChange"
       @update:filter-model="handleFilterModelChange"
       @update:filter-provider="handleFilterProviderChange"
       @update:filter-api-format="handleFilterApiFormatChange"
@@ -124,29 +63,20 @@
 <script setup lang="ts">
 import { ref, computed, onMounted, onUnmounted, watch } from 'vue'
 import { useRoute } from 'vue-router'
-import { useLocalStorage } from '@vueuse/core'
+// useLocalStorage no longer needed (analytics panel removed)
 import { useAuthStore } from '@/stores/auth'
-import { usageApi } from '@/api/usage'
-import { usersApi } from '@/api/users'
-import { meApi } from '@/api/me'
-import { PanelTopClose, PanelTopOpen } from 'lucide-vue-next'
+import { analyticsApi, type AnalyticsFilterOption } from '@/api/analytics'
 import {
-  UsageModelTable,
-  UsageProviderTable,
-  UsageApiFormatTable,
   UsageRecordsTable,
-  ActivityHeatmapCard,
   RequestDetailDrawer,
-  IntervalTimelineCard
 } from '@/features/usage/components'
 import {
   useUsageData,
   getDateRangeFromPeriod
 } from '@/features/usage/composables'
 import type { DateRangeParams, FilterStatusValue } from '@/features/usage/types'
-import type { UserOption } from '@/features/usage/components/UsageRecordsTable.vue'
 import { log } from '@/utils/logger'
-import type { ActivityHeatmap } from '@/types/activity'
+// ActivityHeatmap moved to analytics page
 import { useToast } from '@/composables/useToast'
 
 const route = useRoute()
@@ -156,9 +86,6 @@ const authStore = useAuthStore()
 // 判断是否是管理员页面
 const isAdminPage = computed(() => route.path.startsWith('/admin'))
 
-// 用量分析面板折叠状态（默认展开，持久化到 localStorage）
-const statsExpanded = useLocalStorage('usage-stats-expanded', true)
-
 // 时间范围选择
 const timeRange = ref<DateRangeParams>(getDateRangeFromPeriod('today'))
 
@@ -170,155 +97,28 @@ const pageSizeOptions = [10, 20, 50, 100]
 // 筛选状态
 const filterSearch = ref('')
 const filterUser = ref('__all__')
+const filterApiKey = ref('__all__')
 const filterModel = ref('__all__')
 const filterProvider = ref('__all__')
 const filterApiFormat = ref('__all__')
 const filterStatus = ref<FilterStatusValue>('__all__')
 
-// 用户列表（仅管理员页面使用）
-const availableUsers = ref<UserOption[]>([])
-
 // 使用 composables
 const {
   isLoadingRecords,
-  providerStats,
-  apiFormatStats,
   currentRecords,
   totalRecords,
-  enhancedModelStats,
+  availableUsers,
+  availableApiKeys,
   availableModels,
+  availableApiFormats,
   availableProviders,
+  availableStatuses,
   loadStats,
   loadRecords
 } = useUsageData({ isAdminPage })
 
-// 热力图状态
-const activityHeatmapData = ref<ActivityHeatmap | null>(null)
-const isLoadingHeatmap = ref(false)
-const heatmapError = ref(false)
-const ADMIN_ANALYTICS_REFRESH_INTERVAL = 60000
-let adminAnalyticsRefreshInFlight: Promise<void> | null = null
-let lastAdminAnalyticsRefreshAt = 0
-
-// 加载热力图数据
-async function loadHeatmapData() {
-  isLoadingHeatmap.value = true
-  heatmapError.value = false
-  try {
-    if (isAdminPage.value) {
-      activityHeatmapData.value = await usageApi.getActivityHeatmap()
-    } else {
-      activityHeatmapData.value = await meApi.getActivityHeatmap()
-    }
-  } catch (error) {
-    log.error('加载热力图数据失败:', error)
-    heatmapError.value = true
-  } finally {
-    isLoadingHeatmap.value = false
-  }
-}
-
-async function loadAdminUsers() {
-  try {
-    const users = await usersApi.getAllUsers()
-    availableUsers.value = users.map(u => ({ id: u.id, username: u.username, email: u.email }))
-  } catch (error) {
-    log.error('加载用户列表失败:', error)
-  }
-}
-
-async function refreshAdminAnalytics(options: { force?: boolean } = {}) {
-  if (!isAdminPage.value) return
-  if (!options.force && !isPageVisible.value) return
-
-  const now = Date.now()
-  if (!options.force && now - lastAdminAnalyticsRefreshAt < ADMIN_ANALYTICS_REFRESH_INTERVAL) {
-    return
-  }
-  if (adminAnalyticsRefreshInFlight) {
-    return adminAnalyticsRefreshInFlight
-  }
-
-  adminAnalyticsRefreshInFlight = (async () => {
-    let hasSuccessfulRefresh = false
-
-    try {
-      await loadStats(timeRange.value)
-      hasSuccessfulRefresh = true
-    } catch (error) {
-      log.error('加载统计数据失败:', error)
-      warning('统计数据加载失败，请刷新重试')
-    }
-
-    try {
-      await loadHeatmapData()
-      hasSuccessfulRefresh = true
-    } catch (error) {
-      log.error('加载热力图数据失败:', error)
-    }
-
-    if (hasSuccessfulRefresh) {
-      lastAdminAnalyticsRefreshAt = Date.now()
-    }
-  })()
-
-  try {
-    await adminAnalyticsRefreshInFlight
-  } finally {
-    adminAnalyticsRefreshInFlight = null
-  }
-}
-
-// 用户页面需要前端筛选
-const filteredRecords = computed(() => {
-  if (!isAdminPage.value) {
-    let records = [...currentRecords.value]
-
-    if (filterModel.value !== '__all__') {
-      records = records.filter(record => record.model === filterModel.value)
-    }
-
-    if (filterProvider.value !== '__all__') {
-      records = records.filter(record => record.provider === filterProvider.value)
-    }
-
-    if (filterApiFormat.value !== '__all__') {
-      records = records.filter(record =>
-        record.api_format?.toUpperCase() === filterApiFormat.value.toUpperCase()
-      )
-    }
-
-    if (filterStatus.value !== '__all__') {
-      if (filterStatus.value === 'stream') {
-        records = records.filter(record =>
-          record.is_stream && !record.error_message && (!record.status_code || record.status_code === 200)
-        )
-      } else if (filterStatus.value === 'standard') {
-        records = records.filter(record =>
-          !record.is_stream && !record.error_message && (!record.status_code || record.status_code === 200)
-        )
-      } else if (filterStatus.value === 'active') {
-        records = records.filter(record =>
-          record.status === 'pending' || record.status === 'streaming'
-        )
-      } else if (filterStatus.value === 'failed') {
-        // 失败请求需要同时考虑新旧两种判断方式：
-        // 1. 新方式：status = "failed"
-        // 2. 旧方式：status_code >= 400 或 error_message 不为空
-        records = records.filter(record =>
-          record.status === 'failed' ||
-          (record.status_code && record.status_code >= 400) ||
-          record.error_message
-        )
-      } else if (filterStatus.value === 'cancelled') {
-        records = records.filter(record => record.status === 'cancelled')
-      }
-    }
-
-    return records
-  }
-  return currentRecords.value
-})
+// 热力图已移至数据分析页
 
 // 获取活跃请求的 ID 列表
 const activeRequestIds = computed(() => {
@@ -349,11 +149,10 @@ async function pollActiveRequests() {
   pollInFlight = true
 
   try {
-    // 根据页面类型选择不同的 API
-    const idsParam = activeRequestIds.value.join(',')
-    const { requests } = isAdminPage.value
-      ? await usageApi.getActiveRequests(activeRequestIds.value)
-      : await meApi.getActiveRequests(idsParam)
+    const { requests } = await analyticsApi.getActiveRequests({
+      scope: isAdminPage.value ? { kind: 'global' } : { kind: 'me' },
+      ids: activeRequestIds.value,
+    })
 
     let shouldRefresh = false
 
@@ -512,23 +311,8 @@ onUnmounted(() => {
   stopGlobalAutoRefresh()
 })
 
-// 用户页面的前端分页（后端一次性返回所有记录，前端分页+筛选）
-const paginatedRecords = computed(() => {
-  if (!isAdminPage.value) {
-    const start = (currentPage.value - 1) * pageSize.value
-    const end = start + pageSize.value
-    return filteredRecords.value.slice(start, end)
-  }
-  return currentRecords.value
-})
-
-// 用户页面使用前端筛选后的总数，管理员页面使用后端返回的总数
-const effectiveTotalRecords = computed(() => {
-  if (!isAdminPage.value) {
-    return filteredRecords.value.length
-  }
-  return totalRecords.value
-})
+const paginatedRecords = computed(() => currentRecords.value)
+const effectiveTotalRecords = computed(() => totalRecords.value)
 
 // 显示的记录
 const displayRecords = computed(() => paginatedRecords.value)
@@ -538,33 +322,64 @@ const displayRecords = computed(() => paginatedRecords.value)
 const detailModalOpen = ref(false)
 const selectedRequestId = ref<string | null>(null)
 
+function hasOption(options: AnalyticsFilterOption[], value: string) {
+  return options.some(option => option.value === value)
+}
+
+function normalizeFilterSelections() {
+  let changed = false
+
+  if (filterUser.value !== '__all__' && !hasOption(availableUsers.value, filterUser.value)) {
+    filterUser.value = '__all__'
+    changed = true
+  }
+  if (filterApiKey.value !== '__all__' && !hasOption(availableApiKeys.value, filterApiKey.value)) {
+    filterApiKey.value = '__all__'
+    changed = true
+  }
+  if (filterModel.value !== '__all__' && !hasOption(availableModels.value, filterModel.value)) {
+    filterModel.value = '__all__'
+    changed = true
+  }
+  if (filterProvider.value !== '__all__' && !hasOption(availableProviders.value, filterProvider.value)) {
+    filterProvider.value = '__all__'
+    changed = true
+  }
+  if (filterApiFormat.value !== '__all__' && !hasOption(availableApiFormats.value, filterApiFormat.value)) {
+    filterApiFormat.value = '__all__'
+    changed = true
+  }
+  if (filterStatus.value !== '__all__' && !hasOption(availableStatuses.value, filterStatus.value)) {
+    filterStatus.value = '__all__'
+    changed = true
+  }
+
+  return changed
+}
+
+async function syncFilterOptions() {
+  await loadStats(timeRange.value, getCurrentFilters())
+
+  if (!normalizeFilterSelections()) {
+    return
+  }
+
+  await loadStats(timeRange.value, getCurrentFilters())
+}
+
 // 初始化加载
 onMounted(async () => {
   document.addEventListener('visibilitychange', handleVisibilityChange)
 
-  if (isAdminPage.value) {
-    // 管理员页面优先加载记录，统计面板在后台顺序刷新，避免瞬时并发打满后端。
-    await loadRecords(
-      { page: currentPage.value, pageSize: pageSize.value },
-      getCurrentFilters(),
-      timeRange.value
-    )
-    void (async () => {
-      await refreshAdminAnalytics({ force: true })
-      await loadAdminUsers()
-    })()
-  } else {
-    // 用户页面：loadStats 已包含记录加载，不需要单独调用 loadRecords
-    await Promise.allSettled([
-      loadStats(timeRange.value).catch(err => {
+  const statsTask = syncFilterOptions().catch(err => {
     log.error('加载统计数据失败:', err)
     warning('统计数据加载失败，请刷新重试')
-      }),
-      loadHeatmapData().catch(err => {
-        log.error('加载热力图数据失败:', err)
   })
-    ])
-  }
+  await statsTask
+  await loadRecords(
+    { page: currentPage.value, pageSize: pageSize.value },
+    getCurrentFilters()
+  )
 
   if (globalAutoRefresh.value && isPageVisible.value) {
     startGlobalAutoRefresh()
@@ -575,32 +390,21 @@ onMounted(async () => {
 async function handleTimeRangeChange(value: DateRangeParams) {
   timeRange.value = value
   currentPage.value = 1 // 重置到第一页
-  if (isAdminPage.value) {
-    await loadRecords({ page: 1, pageSize: pageSize.value }, getCurrentFilters(), timeRange.value)
-    await refreshAdminAnalytics({ force: true })
-    return
-  }
-  await loadStats(timeRange.value)
-  // 用户页面：loadStats 已包含记录加载
+  await syncFilterOptions()
+  await loadRecords({ page: 1, pageSize: pageSize.value }, getCurrentFilters())
 }
 
 // 处理分页变化
 async function handlePageChange(page: number) {
   currentPage.value = page
-  if (isAdminPage.value) {
-    await loadRecords({ page, pageSize: pageSize.value }, getCurrentFilters(), timeRange.value)
-  }
-  // 用户页面使用前端分页，无需重新请求
+  await loadRecords({ page, pageSize: pageSize.value }, getCurrentFilters())
 }
 
 // 处理每页大小变化
 async function handlePageSizeChange(size: number) {
   pageSize.value = size
   currentPage.value = 1  // 重置到第一页
-  if (isAdminPage.value) {
-    await loadRecords({ page: 1, pageSize: size }, getCurrentFilters(), timeRange.value)
-  }
-  // 用户页面使用前端分页，无需重新请求
+  await loadRecords({ page: 1, pageSize: size }, getCurrentFilters())
 }
 
 // 获取当前筛选参数
@@ -608,6 +412,7 @@ function getCurrentFilters() {
   return {
     search: filterSearch.value.trim() || undefined,
     user_id: filterUser.value !== '__all__' ? filterUser.value : undefined,
+    api_key_id: filterApiKey.value !== '__all__' ? filterApiKey.value : undefined,
     model: filterModel.value !== '__all__' ? filterModel.value : undefined,
     provider: filterProvider.value !== '__all__' ? filterProvider.value : undefined,
     api_format: filterApiFormat.value !== '__all__' ? filterApiFormat.value : undefined,
@@ -619,57 +424,49 @@ function getCurrentFilters() {
 async function handleFilterSearchChange(value: string) {
   filterSearch.value = value
   currentPage.value = 1
-
-  if (isAdminPage.value) {
-    await loadRecords({ page: 1, pageSize: pageSize.value }, getCurrentFilters(), timeRange.value)
-  }
-  // 用户页面：search 需要重新从后端拉取数据（后端支持 search 参数）
-  // 但通过 filteredRecords 做前端过滤已覆盖，无需额外请求
+  await loadRecords({ page: 1, pageSize: pageSize.value }, getCurrentFilters())
 }
 
 async function handleFilterUserChange(value: string) {
   filterUser.value = value
   currentPage.value = 1  // 重置到第一页
-
-  if (isAdminPage.value) {
-    await loadRecords({ page: 1, pageSize: pageSize.value }, getCurrentFilters(), timeRange.value)
+  await syncFilterOptions()
+  await loadRecords({ page: 1, pageSize: pageSize.value }, getCurrentFilters())
 }
+
+async function handleFilterApiKeyChange(value: string) {
+  filterApiKey.value = value
+  currentPage.value = 1
+  await syncFilterOptions()
+  await loadRecords({ page: 1, pageSize: pageSize.value }, getCurrentFilters())
 }
 
 async function handleFilterModelChange(value: string) {
   filterModel.value = value
   currentPage.value = 1  // 重置到第一页
-
-  if (isAdminPage.value) {
-    await loadRecords({ page: 1, pageSize: pageSize.value }, getCurrentFilters(), timeRange.value)
-  }
+  await syncFilterOptions()
+  await loadRecords({ page: 1, pageSize: pageSize.value }, getCurrentFilters())
 }
 
 async function handleFilterProviderChange(value: string) {
   filterProvider.value = value
   currentPage.value = 1
-
-  if (isAdminPage.value) {
-    await loadRecords({ page: 1, pageSize: pageSize.value }, getCurrentFilters(), timeRange.value)
-  }
+  await syncFilterOptions()
+  await loadRecords({ page: 1, pageSize: pageSize.value }, getCurrentFilters())
 }
 
 async function handleFilterApiFormatChange(value: string) {
   filterApiFormat.value = value
   currentPage.value = 1
-
-  if (isAdminPage.value) {
-    await loadRecords({ page: 1, pageSize: pageSize.value }, getCurrentFilters(), timeRange.value)
-  }
+  await syncFilterOptions()
+  await loadRecords({ page: 1, pageSize: pageSize.value }, getCurrentFilters())
 }
 
 async function handleFilterStatusChange(value: string) {
   filterStatus.value = value as FilterStatusValue
   currentPage.value = 1
-
-  if (isAdminPage.value) {
-    await loadRecords({ page: 1, pageSize: pageSize.value }, getCurrentFilters(), timeRange.value)
-  }
+  await syncFilterOptions()
+  await loadRecords({ page: 1, pageSize: pageSize.value }, getCurrentFilters())
 }
 
 // 刷新数据
@@ -678,18 +475,8 @@ async function refreshData() {
   if (refreshInFlight) return refreshInFlight
 
   refreshInFlight = (async () => {
-    if (isAdminPage.value) {
-      await loadRecords(
-        { page: currentPage.value, pageSize: pageSize.value },
-        getCurrentFilters(),
-        timeRange.value
-      )
-      void refreshAdminAnalytics()
-      return
-    }
-
-    await loadStats(timeRange.value)
-    // 用户页面：loadStats 已包含记录加载
+    await syncFilterOptions()
+    await loadRecords({ page: currentPage.value, pageSize: pageSize.value }, getCurrentFilters())
   })()
 
   try {

pyproject.toml

@@ -46,6 +46,7 @@ dependencies = [
     "aiosmtplib>=5.1.0",
     "redis>=7.1.0",
     "prometheus-client>=0.24.1",
+    "psutil>=5.9.8",
     "apscheduler>=3.11.2",
     "ldap3>=2.9.1",
     "msgpack>=1.1.2",

src/api/admin/__init__.py

@@ -17,7 +17,6 @@ from .provider_query import router as provider_query_router
 from .provider_strategy import router as provider_strategy_router
 from .providers import router as providers_router
 from .security import router as security_router
-from .stats import router as stats_router
 from .system import router as system_router
 from .usage import router as usage_router
 from .users import router as users_router
@@ -39,7 +38,6 @@ router.include_router(provider_oauth_router)
 router.include_router(adaptive_router)
 router.include_router(models_router)
 router.include_router(security_router)
-router.include_router(stats_router)
 router.include_router(provider_query_router)
 router.include_router(modules_router)
 router.include_router(pool_router)

src/api/admin/endpoints/health.py

@@ -138,7 +138,7 @@ async def get_health_summary(
     - `total_endpoints`: Endpoint 总数
     - `active_endpoints`: 活跃 Endpoint 数量
     - `total_keys`: Key 总数
-    - `active_keys`: 活跃 Key 数量
+    - `active_keys`: 当前可调度 Key 数量（Provider 活跃 + Key 启用 + 存在活跃 Endpoint + 非熔断）
     - `circuit_breaker_open_keys`: 熔断的 Key 数量
     """
     adapter = AdminHealthSummaryAdapter()

src/api/admin/monitoring/audit.py

@@ -2,14 +2,21 @@
 
 from __future__ import annotations
 
+import os
+import time
 from dataclasses import dataclass
 from datetime import datetime, timedelta, timezone
 from typing import Any
 
 from fastapi import APIRouter, Depends, HTTPException, Query, Request
-from sqlalchemy import case, func
+from sqlalchemy import case, func, text
 from sqlalchemy.orm import Session
 
+try:
+    import psutil
+except ModuleNotFoundError:
+    psutil = None
+
 from src.api.base.admin_adapter import AdminApiAdapter
 from src.api.base.context import ApiRequestContext
 from src.api.base.pagination import PaginationMeta, build_pagination_payload
@@ -17,6 +24,7 @@ from src.api.base.pipeline import get_pipeline
 from src.config.constants import CacheTTL
 from src.core.logger import logger
 from src.database import get_db
+from src.database.database import get_pool_status
 from src.models.database import (
     ApiKey,
     AuditEventType,
@@ -25,7 +33,9 @@ from src.models.database import (
     Usage,
 )
 from src.models.database import User as DBUser
+from src.clients.redis_client import get_redis_client
 from src.services.health.monitor import HealthMonitor
+from src.services.system.config import SystemConfigService
 from src.services.system.audit import audit_service
 from src.utils.cache_decorator import cache_result
 from src.utils.database_helpers import escape_like_pattern
@@ -34,6 +44,334 @@ router = APIRouter(prefix="/api/admin/monitoring", tags=["Admin - Monitoring"])
 pipeline = get_pipeline()
 
 
+def _unavailable_metric(message: str, **fields: Any) -> dict[str, Any]:
+    return {
+        "status": "unknown",
+        "label": _label_for_status("unknown"),
+        **fields,
+        "message": message,
+    }
+
+
+def _round_metric(value: float | int | None, digits: int = 1) -> float | None:
+    if value is None:
+        return None
+    return round(float(value), digits)
+
+
+def _metric_status(percent: float | None, *, warning: float, danger: float) -> str:
+    if percent is None:
+        return "unknown"
+    if percent >= danger:
+        return "danger"
+    if percent >= warning:
+        return "warning"
+    return "ok"
+
+
+def _remaining_status(percent: float | None, *, warning: float, danger: float) -> str:
+    if percent is None:
+        return "unknown"
+    if percent <= danger:
+        return "danger"
+    if percent <= warning:
+        return "warning"
+    return "ok"
+
+
+def _label_for_status(status: str) -> str:
+    return {
+        "ok": "正常",
+        "warning": "偏高",
+        "danger": "紧张",
+        "degraded": "降级",
+        "error": "异常",
+        "unknown": "未知",
+    }.get(status, "未知")
+
+
+def _optional_int(value: Any) -> int | None:
+    if value is None:
+        return None
+    try:
+        return int(value)
+    except (TypeError, ValueError):
+        return None
+
+
+def _get_manual_capacity_bytes(db: Session, key: str) -> int | None:
+    value = _optional_int(SystemConfigService.get_config(db, key, default=0))
+    if value is None or value <= 0:
+        return None
+    return value
+
+
+def _build_cpu_metric() -> dict[str, Any]:
+    core_count = int((psutil.cpu_count() if psutil is not None else os.cpu_count()) or 0)
+    load_percent: float | None = None
+
+    try:
+        if hasattr(os, "getloadavg") and core_count > 0:
+            load_percent = _round_metric((os.getloadavg()[0] / core_count) * 100)
+    except OSError:
+        load_percent = None
+
+    if psutil is None:
+        status = _metric_status(load_percent, warning=70, danger=85)
+        return {
+            "status": status,
+            "label": _label_for_status(status),
+            "usage_percent": None,
+            "load_percent": load_percent,
+            "core_count": core_count,
+            "message": "psutil is not installed",
+        }
+
+    cpu_percent = _round_metric(psutil.cpu_percent(interval=0.05))
+    status = _metric_status(cpu_percent, warning=70, danger=85)
+    return {
+        "status": status,
+        "label": _label_for_status(status),
+        "usage_percent": cpu_percent,
+        "load_percent": load_percent,
+        "core_count": core_count,
+    }
+
+
+def _build_memory_metric() -> dict[str, Any]:
+    if psutil is None:
+        return _unavailable_metric(
+            "psutil is not installed",
+            used_percent=None,
+            used_bytes=None,
+            available_bytes=None,
+            total_bytes=None,
+        )
+
+    memory = psutil.virtual_memory()
+    used_percent = _round_metric(memory.percent)
+    status = _metric_status(used_percent, warning=75, danger=90)
+    return {
+        "status": status,
+        "label": _label_for_status(status),
+        "used_percent": used_percent,
+        "used_bytes": int(memory.used),
+        "available_bytes": int(memory.available),
+        "total_bytes": int(memory.total),
+    }
+
+
+async def _build_redis_metric(db: Session) -> dict[str, Any]:
+    started = time.perf_counter()
+    try:
+        redis = await get_redis_client(require_redis=False)
+        if redis is None:
+            status = "degraded"
+            return {
+                "status": status,
+                "label": _label_for_status(status),
+                "latency_ms": None,
+                "memory_status": "unknown",
+                "memory_label": _label_for_status("unknown"),
+                "used_memory_bytes": None,
+                "peak_memory_bytes": None,
+                "maxmemory_bytes": None,
+                "memory_ceiling_bytes": None,
+                "memory_source": "unknown",
+                "available_memory_bytes": None,
+                "memory_percent": None,
+                "message": "Redis client not initialized",
+            }
+
+        await redis.ping()
+        latency_ms = _round_metric((time.perf_counter() - started) * 1000)
+        # 优先使用 Redis 自身的 INFO 输出，避免依赖应用所在机器的本地内存视角。
+        memory_info = await redis.info(section="memory")
+        used_memory_bytes = _optional_int(memory_info.get("used_memory"))
+        peak_memory_bytes = _optional_int(memory_info.get("used_memory_peak"))
+        maxmemory_bytes = _optional_int(memory_info.get("maxmemory"))
+        total_system_memory_bytes = _optional_int(memory_info.get("total_system_memory"))
+        configured_memory_bytes = _get_manual_capacity_bytes(db, "redis_memory_total_bytes")
+        if maxmemory_bytes is not None and maxmemory_bytes <= 0:
+            maxmemory_bytes = None
+
+        memory_ceiling_bytes = configured_memory_bytes or maxmemory_bytes or total_system_memory_bytes
+        if configured_memory_bytes is not None:
+            memory_source = "configured"
+        elif maxmemory_bytes is not None:
+            memory_source = "maxmemory"
+        elif total_system_memory_bytes is not None:
+            memory_source = "system"
+        else:
+            memory_source = "unknown"
+        available_memory_bytes = (
+            max(int(memory_ceiling_bytes) - int(used_memory_bytes), 0)
+            if memory_ceiling_bytes is not None and used_memory_bytes is not None
+            else None
+        )
+        memory_percent = _round_metric(
+            (used_memory_bytes / memory_ceiling_bytes * 100)
+            if used_memory_bytes is not None and memory_ceiling_bytes not in (None, 0)
+            else None
+        )
+        status = "warning" if latency_ms is not None and latency_ms >= 80 else "ok"
+        memory_status = _metric_status(memory_percent, warning=75, danger=90)
+        return {
+            "status": status,
+            "label": _label_for_status(status),
+            "latency_ms": latency_ms,
+            "memory_status": memory_status,
+            "memory_label": _label_for_status(memory_status),
+            "used_memory_bytes": used_memory_bytes,
+            "peak_memory_bytes": peak_memory_bytes,
+            "maxmemory_bytes": maxmemory_bytes,
+            "memory_ceiling_bytes": memory_ceiling_bytes,
+            "memory_source": memory_source,
+            "available_memory_bytes": available_memory_bytes,
+            "memory_percent": memory_percent,
+            "message": None,
+        }
+    except Exception as exc:
+        return {
+            "status": "error",
+            "label": _label_for_status("error"),
+            "latency_ms": None,
+            "memory_status": "unknown",
+            "memory_label": _label_for_status("unknown"),
+            "used_memory_bytes": None,
+            "peak_memory_bytes": None,
+            "maxmemory_bytes": None,
+            "memory_ceiling_bytes": None,
+            "memory_source": "unknown",
+            "available_memory_bytes": None,
+            "memory_percent": None,
+            "message": str(exc),
+        }
+
+
+def _build_postgres_metric(db: Session) -> dict[str, Any]:
+    try:
+        pool_status = get_pool_status()
+        checked_out = int(pool_status.get("checked_out") or 0)
+        max_capacity = int(pool_status.get("max_capacity") or 0)
+        pool_usage_percent = _round_metric(
+            (checked_out / max_capacity * 100) if max_capacity > 0 else 0
+        )
+        usage_percent = pool_usage_percent
+        storage_total_bytes: int | None = None
+        storage_free_bytes: int | None = None
+        storage_free_percent: float | None = None
+        database_size_bytes: int | None = None
+        server_connections: int | None = None
+        server_max_connections: int | None = None
+        server_usage_percent: float | None = None
+        message: str | None = None
+        storage_message: str | None = None
+        storage_total_bytes = _get_manual_capacity_bytes(db, "postgres_storage_total_bytes")
+
+        bind = db.get_bind()
+        dialect_name = bind.dialect.name if bind is not None else ""
+        if dialect_name == "postgresql":
+            try:
+                # 只读取 PostgreSQL 原生 SQL 可见的指标，避免依赖本机数据目录或宿主机磁盘。
+                stats = (
+                    db.execute(
+                        text(
+                            """
+                            SELECT
+                                pg_database_size(current_database()) AS database_size_bytes,
+                                current_setting('max_connections')::int AS server_max_connections,
+                                COALESCE(
+                                    (
+                                        SELECT numbackends
+                                        FROM pg_stat_database
+                                        WHERE datname = current_database()
+                                    ),
+                                    0
+                                ) AS server_connections
+                            """
+                        )
+                    )
+                    .mappings()
+                    .one()
+                )
+                database_size_bytes = _optional_int(stats.get("database_size_bytes"))
+                server_max_connections = _optional_int(stats.get("server_max_connections"))
+                server_connections = _optional_int(stats.get("server_connections"))
+                server_usage_percent = _round_metric(
+                    (server_connections / server_max_connections * 100)
+                    if server_connections is not None and server_max_connections not in (None, 0)
+                    else None
+                )
+                if server_usage_percent is not None:
+                    usage_percent = server_usage_percent
+                if storage_total_bytes is not None and database_size_bytes is not None:
+                    storage_free_bytes = max(storage_total_bytes - database_size_bytes, 0)
+                    storage_free_percent = _round_metric(
+                        (storage_free_bytes / storage_total_bytes * 100)
+                        if storage_total_bytes > 0
+                        else None
+                    )
+                else:
+                    storage_message = "请在系统设置中填写 PostgreSQL 总空间"
+            except Exception as exc:
+                message = f"PostgreSQL 原生指标不可用: {exc}"
+                if storage_total_bytes is not None:
+                    storage_message = "已配置 PostgreSQL 总空间，但当前库体积不可用"
+                else:
+                    storage_message = "请在系统设置中填写 PostgreSQL 总空间"
+        else:
+            storage_message = "当前数据库不是 PostgreSQL"
+
+        status = _metric_status(usage_percent, warning=70, danger=90)
+        storage_status = _remaining_status(storage_free_percent, warning=20, danger=10)
+        return {
+            "status": status,
+            "label": _label_for_status(status),
+            "usage_percent": usage_percent,
+            "pool_usage_percent": pool_usage_percent,
+            "checked_out": checked_out,
+            "pool_size": int(pool_status.get("pool_size") or 0),
+            "overflow": int(pool_status.get("overflow") or 0),
+            "max_capacity": max_capacity,
+            "pool_timeout": int(pool_status.get("pool_timeout") or 0),
+            "server_connections": server_connections,
+            "server_max_connections": server_max_connections,
+            "server_usage_percent": server_usage_percent,
+            "storage_status": storage_status,
+            "storage_label": _label_for_status(storage_status),
+            "storage_total_bytes": storage_total_bytes,
+            "storage_free_bytes": storage_free_bytes,
+            "storage_free_percent": storage_free_percent,
+            "database_size_bytes": database_size_bytes,
+            "storage_message": storage_message,
+            "message": message,
+        }
+    except Exception as exc:
+        return {
+            "status": "error",
+            "label": _label_for_status("error"),
+            "usage_percent": None,
+            "pool_usage_percent": None,
+            "checked_out": 0,
+            "pool_size": 0,
+            "overflow": 0,
+            "max_capacity": 0,
+            "pool_timeout": 0,
+            "server_connections": None,
+            "server_max_connections": None,
+            "server_usage_percent": None,
+            "storage_status": "unknown",
+            "storage_label": _label_for_status("unknown"),
+            "storage_total_bytes": None,
+            "storage_free_bytes": None,
+            "storage_free_percent": None,
+            "database_size_bytes": None,
+            "storage_message": None,
+            "message": str(exc),
+        }
+
+
 @router.get("/audit-logs")
 async def get_audit_logs(
     request: Request,
@@ -96,6 +434,7 @@ async def get_system_status(request: Request, db: Session = Depends(get_db)) ->
     - `api_keys`: API Key 统计（total: 总数, active: 活跃数）
     - `today_stats`: 今日统计（requests: 请求数, tokens: token 数, cost_usd: 成本）
     - `recent_errors`: 最近 1 小时内的错误数
+    - `system_metrics`: 系统资源摘要（cpu / memory / redis / postgres）
     """
     adapter = AdminSystemStatusAdapter()
     return await pipeline.run(adapter=adapter, http_request=request, db=db, mode=adapter.mode)
@@ -381,6 +720,11 @@ class AdminSystemStatusAdapter(AdminApiAdapter):
             recent_errors=int(recent_errors or 0),
         )
 
+        cpu_metric = _build_cpu_metric()
+        memory_metric = _build_memory_metric()
+        redis_metric = await _build_redis_metric(db)
+        postgres_metric = _build_postgres_metric(db)
+
         return {
             "timestamp": datetime.now(timezone.utc).isoformat(),
             "users": {"total": total_users, "active": active_users},
@@ -392,6 +736,12 @@ class AdminSystemStatusAdapter(AdminApiAdapter):
                 "cost_usd": f"${today_cost:.4f}",
             },
             "recent_errors": recent_errors,
+            "system_metrics": {
+                "cpu": cpu_metric,
+                "memory": memory_metric,
+                "redis": redis_metric,
+                "postgres": postgres_metric,
+            },
         }
 
 

src/api/admin/provider_oauth.py

@@ -36,7 +36,7 @@ from src.clients.redis_client import get_redis_client
 from src.core.crypto import crypto_service
 from src.core.exceptions import InvalidRequestException, NotFoundException
 from src.core.logger import logger
-from src.core.provider_oauth_utils import enrich_auth_config, post_oauth_token
+from src.core.provider_oauth_utils import enrich_auth_config, parse_codex_id_token, post_oauth_token
 from src.core.provider_templates.fixed_providers import FIXED_PROVIDERS
 from src.core.provider_templates.types import ProviderType
 from src.database import get_db_context
@@ -150,6 +150,10 @@ def _mark_refresh_failed_sync(key_id: str, reason: str) -> None:
         if not key:
             raise NotFoundException("Key 不存在", "key")
         current_reason = str(getattr(key, "oauth_invalid_reason", None) or "").strip()
+        from src.services.provider.oauth_token import is_account_level_block
+
+        if is_account_level_block(current_reason):
+            return
         merged_reason = _merge_refresh_failure_reason(current_reason, reason)
         if merged_reason is None:
             return
@@ -1577,11 +1581,62 @@ def _coerce_import_str(value: Any) -> str | None:
     return normalized or None
 
 
+def _decode_unverified_jwt_json_part(part: str) -> dict[str, Any] | None:
+    try:
+        padding = "=" * (-len(part) % 4)
+        decoded = base64.urlsafe_b64decode(f"{part}{padding}")
+        payload = json.loads(decoded.decode("utf-8"))
+    except Exception:
+        return None
+    return payload if isinstance(payload, dict) else None
+
+
+def _looks_like_access_token(token: str) -> bool:
+    """识别手工粘贴的 JWT Access Token，避免误当 Refresh Token 去刷新验证。"""
+    parts = token.strip().split(".")
+    if len(parts) != 3 or not all(parts):
+        return False
+
+    header = _decode_unverified_jwt_json_part(parts[0])
+    payload = _decode_unverified_jwt_json_part(parts[1])
+    if not header or not payload:
+        return False
+
+    token_type = str(header.get("typ") or "").lower()
+    if token_type and token_type not in {"jwt", "at+jwt"}:
+        return False
+
+    return any(key in payload for key in ("exp", "aud", "iss", "scope", "scp"))
+
+
+def _build_standard_oauth_import_entry_from_token(token: str) -> dict[str, Any]:
+    if _looks_like_access_token(token):
+        return {"access_token": token}
+    return {"refresh_token": token}
+
+
+def _normalize_single_import_tokens(
+    *,
+    refresh_token: str | None,
+    access_token: str | None,
+) -> tuple[str, str]:
+    refresh_token_value = (refresh_token or "").strip()
+    access_token_value = (access_token or "").strip()
+    if (
+        refresh_token_value
+        and not access_token_value
+        and _looks_like_access_token(refresh_token_value)
+    ):
+        access_token_value = refresh_token_value
+        refresh_token_value = ""
+    return refresh_token_value, access_token_value
+
+
 def _extract_standard_oauth_import_entry(item: Any) -> dict[str, Any] | None:
     if isinstance(item, str):
         token = _coerce_import_str(item)
         if token:
-            return {"refresh_token": token}
+            return _build_standard_oauth_import_entry_from_token(token)
         return None
     if not isinstance(item, dict):
         return None
@@ -1589,10 +1644,17 @@ def _extract_standard_oauth_import_entry(item: Any) -> dict[str, Any] | None:
     refresh_token = _coerce_import_str(item.get("refresh_token")) or _coerce_import_str(
         item.get("refreshToken")
     )
-    if not refresh_token:
+    access_token = _coerce_import_str(item.get("access_token")) or _coerce_import_str(
+        item.get("accessToken")
+    )
+    if not refresh_token and not access_token:
         return None
 
-    entry: dict[str, Any] = {"refresh_token": refresh_token}
+    entry: dict[str, Any] = {}
+    if refresh_token:
+        entry["refresh_token"] = refresh_token
+    if access_token:
+        entry["access_token"] = access_token
 
     account_id = (
         _coerce_import_str(item.get("account_id"))
@@ -1678,14 +1740,18 @@ def _parse_standard_oauth_import_entries(raw_input: str) -> list[dict[str, Any]]
     for line in raw.splitlines():
         token = line.strip()
         if token and not token.startswith("#"):
-            result.append({"refresh_token": token})
+            result.append(_build_standard_oauth_import_entry_from_token(token))
 
     return result
 
 
 def _parse_tokens_input(raw_input: str) -> list[str]:
     """兼容旧逻辑：仅返回 refresh_token 列表。"""
-    return [entry["refresh_token"] for entry in _parse_standard_oauth_import_entries(raw_input)]
+    return [
+        str(entry["refresh_token"])
+        for entry in _parse_standard_oauth_import_entries(raw_input)
+        if entry.get("refresh_token")
+    ]
 
 
 def _parse_kiro_import_input(raw_input: str) -> list[dict[str, Any]]:
@@ -1788,7 +1854,8 @@ def _parse_kiro_import_input(raw_input: str) -> list[dict[str, Any]]:
 
 
 class ImportRefreshTokenRequest(BaseModel):
-    refresh_token: str = Field(..., min_length=1, description="Refresh Token")
+    refresh_token: str | None = Field(None, min_length=1, description="Refresh Token")
+    access_token: str | None = Field(None, min_length=1, description="Access Token（Codex 可选）")
     name: str | None = Field(None, max_length=100, description="账号名称（可选）")
     proxy_node_id: str | None = Field(
         None,
@@ -1944,6 +2011,243 @@ def _apply_codex_import_hints(auth_config: dict[str, Any], import_entry: dict[st
             auth_config[field] = value
 
 
+def _decode_access_token_expires_at(access_token: str) -> int | None:
+    """从 access token JWT payload 中解析 exp。"""
+    try:
+        import jwt
+
+        claims = jwt.decode(
+            access_token,
+            options={
+                "verify_signature": False,
+                "verify_aud": False,
+            },
+        )
+    except Exception:
+        return None
+    if not isinstance(claims, dict):
+        return None
+    try:
+        exp = int(claims.get("exp"))
+    except Exception:
+        return None
+    return exp if exp > 0 else None
+
+
+def _build_access_token_import_payload(
+    *,
+    provider_type: str,
+    access_token: str,
+    refresh_token: str | None = None,
+    refresh_error: str | None = None,
+) -> tuple[dict[str, Any], dict[str, Any]]:
+    expires_at = _decode_access_token_expires_at(access_token)
+    token_data: dict[str, Any] = {
+        "access_token": access_token,
+        "token_type": "Bearer",
+    }
+    if expires_at is not None:
+        token_data["expires_at"] = expires_at
+
+    auth_config: dict[str, Any] = {
+        "provider_type": provider_type,
+        "token_type": "Bearer",
+        "refresh_token": refresh_token or None,
+        "expires_at": expires_at,
+        "scope": None,
+        "updated_at": int(time.time()),
+        "imported_from_access_token": True,
+        "access_token_import_temporary": not bool(refresh_token),
+    }
+    if refresh_error:
+        auth_config["refresh_token_import_error"] = refresh_error[:300]
+    return token_data, auth_config
+
+
+async def _build_codex_access_token_import_auth_config(
+    *,
+    provider_type: str,
+    access_token: str,
+    refresh_token: str | None,
+    proxy_config: dict[str, Any] | None,
+    import_entry: dict[str, Any] | None = None,
+    refresh_error: str | None = None,
+) -> tuple[str, dict[str, Any], int | None]:
+    if provider_type != ProviderType.CODEX.value:
+        raise InvalidRequestException("仅 Codex 支持 Access Token 导入")
+    access_token = access_token.strip()
+    if len(access_token) < 10:
+        raise InvalidRequestException("Access Token 无效或过短")
+
+    token_data, auth_config = _build_access_token_import_payload(
+        provider_type=provider_type,
+        access_token=access_token,
+        refresh_token=refresh_token,
+        refresh_error=refresh_error,
+    )
+    try:
+        auth_config = await enrich_auth_config(
+            provider_type=provider_type,
+            auth_config=auth_config,
+            token_response=token_data,
+            access_token=access_token,
+            proxy_config=proxy_config,
+        )
+    except Exception as exc:
+        logger.warning("Codex Access Token 导入: enrich_auth_config 失败: {}", exc)
+        parsed = parse_codex_id_token(access_token)
+        for key, value in parsed.items():
+            if value and not auth_config.get(key):
+                auth_config[key] = value
+
+    if import_entry:
+        _apply_codex_import_hints(auth_config, import_entry)
+    return access_token, auth_config, auth_config.get("expires_at")
+
+
+async def _exchange_refresh_token_for_import(
+    *,
+    provider_type: str,
+    template: Any,
+    refresh_token: str,
+    proxy_config: dict[str, Any] | None,
+    timeout_seconds: float,
+) -> tuple[str, dict[str, Any], int | None]:
+    token_url = template.oauth.token_url
+    is_json = "anthropic.com" in token_url
+    scope_str = " ".join(template.oauth.scopes) if template.oauth.scopes else ""
+
+    if is_json:
+        body: dict[str, Any] = {
+            "grant_type": "refresh_token",
+            "client_id": template.oauth.client_id,
+            "refresh_token": refresh_token,
+        }
+        if scope_str:
+            body["scope"] = scope_str
+        headers = {"Content-Type": "application/json", "Accept": "application/json"}
+        data = None
+        json_body = body
+    else:
+        form: dict[str, str] = {
+            "grant_type": "refresh_token",
+            "client_id": template.oauth.client_id,
+            "refresh_token": refresh_token,
+        }
+        if scope_str:
+            form["scope"] = scope_str
+        if template.oauth.client_secret:
+            form["client_secret"] = template.oauth.client_secret
+        headers = {
+            "Content-Type": "application/x-www-form-urlencoded",
+            "Accept": "application/json",
+        }
+        data = form
+        json_body = None
+
+    resp = await post_oauth_token(
+        provider_type=provider_type,
+        token_url=token_url,
+        headers=headers,
+        data=data,
+        json_body=json_body,
+        proxy_config=proxy_config,
+        timeout_seconds=timeout_seconds,
+    )
+
+    if resp.status_code < 200 or resp.status_code >= 300:
+        error_reason = f"HTTP {resp.status_code}"
+        try:
+            error_body = resp.json()
+            if "error" in error_body:
+                error_reason = str(error_body.get("error_description") or error_body.get("error"))
+        except Exception:
+            error_reason = resp.text[:100] if resp.text else f"HTTP {resp.status_code}"
+        raise InvalidRequestException(f"Refresh Token 验证失败: {error_reason}")
+
+    token_data = resp.json()
+    access_token = str(token_data.get("access_token") or "")
+    new_refresh_token = str(token_data.get("refresh_token") or "") or refresh_token
+
+    if not access_token:
+        raise InvalidRequestException("token refresh 返回缺少 access_token")
+
+    expires_in = token_data.get("expires_in")
+    expires_at: int | None = None
+    try:
+        if expires_in is not None:
+            expires_at = int(time.time()) + int(expires_in)
+    except Exception:
+        expires_at = None
+
+    auth_config: dict[str, Any] = {
+        "provider_type": provider_type,
+        "token_type": token_data.get("token_type"),
+        "refresh_token": new_refresh_token or None,
+        "expires_at": expires_at,
+        "scope": token_data.get("scope"),
+        "updated_at": int(time.time()),
+    }
+
+    auth_config = await enrich_auth_config(
+        provider_type=provider_type,
+        auth_config=auth_config,
+        token_response=token_data,
+        access_token=access_token,
+        proxy_config=proxy_config,
+    )
+    return access_token, auth_config, expires_at
+
+
+async def _build_standard_oauth_import_credentials(
+    *,
+    provider_type: str,
+    template: Any,
+    import_entry: dict[str, Any],
+    proxy_config: dict[str, Any] | None,
+    timeout_seconds: float,
+) -> tuple[str, dict[str, Any], int | None]:
+    refresh_token = str(import_entry.get("refresh_token") or "").strip()
+    access_token_input = str(import_entry.get("access_token") or "").strip()
+
+    if refresh_token:
+        try:
+            access_token, auth_config, expires_at = await _exchange_refresh_token_for_import(
+                provider_type=provider_type,
+                template=template,
+                refresh_token=refresh_token,
+                proxy_config=proxy_config,
+                timeout_seconds=timeout_seconds,
+            )
+            if provider_type == ProviderType.CODEX.value:
+                _apply_codex_import_hints(auth_config, import_entry)
+            return access_token, auth_config, expires_at
+        except Exception as exc:
+            if access_token_input and provider_type == ProviderType.CODEX.value:
+                return await _build_codex_access_token_import_auth_config(
+                    provider_type=provider_type,
+                    access_token=access_token_input,
+                    refresh_token=refresh_token,
+                    proxy_config=proxy_config,
+                    import_entry=import_entry,
+                    refresh_error=str(exc),
+                )
+            raise
+
+    if access_token_input:
+        if provider_type != ProviderType.CODEX.value:
+            raise InvalidRequestException("仅 Codex 支持 Access Token 导入")
+        return await _build_codex_access_token_import_auth_config(
+            provider_type=provider_type,
+            access_token=access_token_input,
+            refresh_token=None,
+            proxy_config=proxy_config,
+            import_entry=import_entry,
+        )
+
+    raise InvalidRequestException("Token 无效或过短")
+
+
 @router.post(
     "/providers/{provider_id}/import-refresh-token",
     response_model=ProviderCompleteOAuthResponse,
@@ -1969,9 +2273,9 @@ async def import_refresh_token(
     )
 
     if provider_type == ProviderType.KIRO.value:
-        raw_import = payload.refresh_token.strip()
+        raw_import = (payload.refresh_token or "").strip()
         if not raw_import:
-            raise InvalidRequestException("Refresh Token 不能为空")
+            raise InvalidRequestException("Kiro 导入需要 Refresh Token 或完整凭据 JSON")
 
         # 使用统一的解析函数
         credentials = _parse_kiro_import_input(raw_import)
@@ -2045,91 +2349,41 @@ async def import_refresh_token(
 
     template = _require_oauth_template(provider_type)
 
-    # 用 refresh_token 换取 access_token
-    refresh_token = payload.refresh_token.strip()
-    token_url = template.oauth.token_url
-    is_json = "anthropic.com" in token_url
-    scope_str = " ".join(template.oauth.scopes) if template.oauth.scopes else ""
+    refresh_token, access_token_input = _normalize_single_import_tokens(
+        refresh_token=payload.refresh_token,
+        access_token=payload.access_token,
+    )
+    if not refresh_token and not access_token_input:
+        raise InvalidRequestException("缺少 refresh_token 或 access_token")
+    if access_token_input and provider_type != ProviderType.CODEX.value:
+        raise InvalidRequestException("仅 Codex 支持 Access Token 导入")
 
-    if is_json:
-        body: dict[str, Any] = {
-            "grant_type": "refresh_token",
-            "client_id": template.oauth.client_id,
-            "refresh_token": refresh_token,
-        }
-        if scope_str:
-            body["scope"] = scope_str
-        headers = {"Content-Type": "application/json", "Accept": "application/json"}
-        data = None
-        json_body = body
-    else:
-        form: dict[str, str] = {
-            "grant_type": "refresh_token",
-            "client_id": template.oauth.client_id,
-            "refresh_token": refresh_token,
-        }
-        if scope_str:
-            form["scope"] = scope_str
-        if template.oauth.client_secret:
-            form["client_secret"] = template.oauth.client_secret
-        headers = {
-            "Content-Type": "application/x-www-form-urlencoded",
-            "Accept": "application/json",
-        }
-        data = form
-        json_body = None
-
-    # proxy_config 和 key_proxy 已在上方 Kiro 分支之前统一解析
-
-    resp = await post_oauth_token(
+    refresh_error: str | None = None
+    if refresh_token:
+        try:
+            access_token, auth_config, expires_at = await _exchange_refresh_token_for_import(
                 provider_type=provider_type,
-        token_url=token_url,
-        headers=headers,
-        data=data,
-        json_body=json_body,
+                template=template,
+                refresh_token=refresh_token,
                 proxy_config=proxy_config,
                 timeout_seconds=30.0,
             )
-
-    if resp.status_code < 200 or resp.status_code >= 300:
-        error_reason = f"HTTP {resp.status_code}"
-        try:
-            error_body = resp.json()
-            if "error" in error_body:
-                error_reason = str(error_body.get("error_description") or error_body.get("error"))
-        except Exception:
-            error_reason = resp.text[:100] if resp.text else f"HTTP {resp.status_code}"
-        raise InvalidRequestException(f"Refresh Token 验证失败: {error_reason}")
-
-    token = resp.json()
-    access_token = str(token.get("access_token") or "")
-    new_refresh_token = str(token.get("refresh_token") or "") or refresh_token
-    expires_in = token.get("expires_in")
-    expires_at: int | None = None
-    try:
-        if expires_in is not None:
-            expires_at = int(time.time()) + int(expires_in)
-    except Exception:
-        expires_at = None
-
-    if not access_token:
-        raise InvalidRequestException("token refresh 返回缺少 access_token")
-
-    # 构建 auth_config
-    auth_config: dict[str, Any] = {
-        "provider_type": provider_type,
-        "token_type": token.get("token_type"),
-        "refresh_token": new_refresh_token or None,
-        "expires_at": expires_at,
-        "scope": token.get("scope"),
-        "updated_at": int(time.time()),
-    }
-
-    auth_config = await enrich_auth_config(
+        except Exception as exc:
+            if not access_token_input or provider_type != ProviderType.CODEX.value:
+                raise
+            refresh_error = str(exc)
+            access_token, auth_config, expires_at = await _build_codex_access_token_import_auth_config(
                 provider_type=provider_type,
-        auth_config=auth_config,
-        token_response=token,
-        access_token=access_token,
+                access_token=access_token_input,
+                refresh_token=refresh_token,
+                proxy_config=proxy_config,
+                refresh_error=refresh_error,
+            )
+    else:
+        access_token, auth_config, expires_at = await _build_codex_access_token_import_auth_config(
+            provider_type=provider_type,
+            access_token=access_token_input,
+            refresh_token=None,
             proxy_config=proxy_config,
         )
 
@@ -2162,7 +2416,7 @@ async def import_refresh_token(
         key_id=str(new_key.id),
         provider_type=provider_type,
         expires_at=expires_at,
-        has_refresh_token=bool(new_refresh_token),
+        has_refresh_token=bool(auth_config.get("refresh_token")),
         email=auth_config.get("email"),
         replaced=replaced,
     )
@@ -2284,10 +2538,6 @@ async def _batch_import_standard_oauth_internal(
         raise InvalidRequestException("未找到有效的 Token 数据")
 
     api_formats = _get_provider_api_formats(provider)
-    token_url = template.oauth.token_url
-    is_json = "anthropic.com" in token_url
-    scope_str = " ".join(template.oauth.scopes) if template.oauth.scopes else ""
-
     total = len(import_entries)
     results: list[BatchImportResultItem] = [None] * total  # type: ignore[list-item]
     success_count = 0
@@ -2305,152 +2555,32 @@ async def _batch_import_standard_oauth_internal(
 
         async with sem:
             try:
-                refresh_token = import_entry.get("refresh_token", "")
-                if not refresh_token or len(refresh_token) < 10:
-                    result_item = BatchImportResultItem(
-                        index=idx,
-                        status="error",
-                        error="Token 无效或过短",
-                    )
-                    failed_count += 1
-                else:
-                    if is_json:
-                        body: dict[str, Any] = {
-                            "grant_type": "refresh_token",
-                            "client_id": template.oauth.client_id,
-                            "refresh_token": refresh_token,
-                        }
-                        if scope_str:
-                            body["scope"] = scope_str
-                        headers = {
-                            "Content-Type": "application/json",
-                            "Accept": "application/json",
-                        }
-                        data = None
-                        json_body = body
-                    else:
-                        form: dict[str, str] = {
-                            "grant_type": "refresh_token",
-                            "client_id": template.oauth.client_id,
-                            "refresh_token": refresh_token,
-                        }
-                        if scope_str:
-                            form["scope"] = scope_str
-                        if template.oauth.client_secret:
-                            form["client_secret"] = template.oauth.client_secret
-                        headers = {
-                            "Content-Type": "application/x-www-form-urlencoded",
-                            "Accept": "application/json",
-                        }
-                        data = form
-                        json_body = None
-
                 try:
                     _release_batch_import_db_connection_before_await(db)
-                        resp = await post_oauth_token(
+                    access_token, auth_config, _expires_at = (
+                        await _build_standard_oauth_import_credentials(
                             provider_type=provider_type,
-                            token_url=token_url,
-                            headers=headers,
-                            data=data,
-                            json_body=json_body,
+                            template=template,
+                            import_entry=import_entry,
                             proxy_config=proxy_config,
                             timeout_seconds=timeout_seconds,
                         )
-                    except Exception as exc:
-                        result_item = BatchImportResultItem(
-                            index=idx,
-                            status="error",
-                            error=f"Token 刷新请求失败: {exc}",
-                        )
-                        failed_count += 1
-                        processed_count += 1
-                        results[idx] = result_item
-                        if progress_hook is not None:
-                            _release_batch_import_db_connection_before_await(db)
-                            await progress_hook(
-                                total, processed_count, success_count, failed_count, result_item
-                            )
-                        return
-
-                    if resp.status_code < 200 or resp.status_code >= 300:
-                        error_reason = f"HTTP {resp.status_code}"
-                        try:
-                            error_body = resp.json()
-                            if "error" in error_body:
-                                error_reason = str(
-                                    error_body.get("error_description") or error_body.get("error")
-                                )
-                        except Exception:
-                            error_reason = (
-                                resp.text[:100] if resp.text else f"HTTP {resp.status_code}"
-                            )
-
-                        result_item = BatchImportResultItem(
-                            index=idx,
-                            status="error",
-                            error=f"Token 验证失败: {error_reason}",
-                        )
-                        failed_count += 1
-                        processed_count += 1
-                        results[idx] = result_item
-                        if progress_hook is not None:
-                            _release_batch_import_db_connection_before_await(db)
-                            await progress_hook(
-                                total, processed_count, success_count, failed_count, result_item
-                            )
-                        return
-
-                    token_data = resp.json()
-                    access_token = str(token_data.get("access_token") or "")
-                    new_refresh_token = str(token_data.get("refresh_token") or "") or refresh_token
-
-                    if not access_token:
-                        result_item = BatchImportResultItem(
-                            index=idx,
-                            status="error",
-                            error="Token 刷新返回缺少 access_token",
-                        )
-                        failed_count += 1
-                        processed_count += 1
-                        results[idx] = result_item
-                        if progress_hook is not None:
-                            _release_batch_import_db_connection_before_await(db)
-                            await progress_hook(
-                                total, processed_count, success_count, failed_count, result_item
-                            )
-                        return
-
-                    expires_in = token_data.get("expires_in")
-                    expires_at: int | None = None
-                    try:
-                        if expires_in is not None:
-                            expires_at = int(time.time()) + int(expires_in)
-                    except Exception:
-                        expires_at = None
-
-                    auth_config: dict[str, Any] = {
-                        "provider_type": provider_type,
-                        "token_type": token_data.get("token_type"),
-                        "refresh_token": new_refresh_token or None,
-                        "expires_at": expires_at,
-                        "scope": token_data.get("scope"),
-                        "updated_at": int(time.time()),
-                    }
-
-                    try:
-                        _release_batch_import_db_connection_before_await(db)
-                        auth_config = await enrich_auth_config(
-                            provider_type=provider_type,
-                            auth_config=auth_config,
-                            token_response=token_data,
-                            access_token=access_token,
-                            proxy_config=proxy_config,
                     )
                 except Exception as exc:
-                        logger.warning("批量导入: enrich_auth_config 失败 (index={}): {}", idx, exc)
-
-                    if provider_type == ProviderType.CODEX.value:
-                        _apply_codex_import_hints(auth_config, import_entry)
+                    result_item = BatchImportResultItem(
+                        index=idx,
+                        status="error",
+                        error=f"Token 验证失败: {exc}",
+                    )
+                    failed_count += 1
+                    processed_count += 1
+                    results[idx] = result_item
+                    if progress_hook is not None:
+                        _release_batch_import_db_connection_before_await(db)
+                        await progress_hook(
+                            total, processed_count, success_count, failed_count, result_item
+                        )
+                    return
 
                 async with db_lock:
                     try:

src/api/admin/provider_query.py

@@ -1840,6 +1840,7 @@ async def _run_concurrent_test(
                 endpoints_by_id = {
                     str(endpoint.id): endpoint
                     for endpoint in preload_db.query(ProviderEndpoint)
+                    .options(joinedload(ProviderEndpoint.provider))
                     .filter(ProviderEndpoint.id.in_(endpoint_ids))
                     .all()
                 }
@@ -1847,6 +1848,7 @@ async def _run_concurrent_test(
                 keys_by_id = {
                     str(key.id): key
                     for key in preload_db.query(ProviderAPIKey)
+                    .options(joinedload(ProviderAPIKey.provider))
                     .filter(ProviderAPIKey.id.in_(key_ids))
                     .all()
                 }

src/api/admin/stats/__init__.py

@@ -1,22 +0,0 @@
-"""Stats admin routes export."""
-
-from fastapi import APIRouter
-
-from .comparison import router as comparison_router
-from .cost import router as cost_router
-from .errors import router as errors_router
-from .leaderboard import router as leaderboard_router
-from .performance import router as performance_router
-from .quota import router as quota_router
-from .time_series import router as time_series_router
-
-router = APIRouter(prefix="/api/admin/stats", tags=["Admin - Stats"])
-router.include_router(leaderboard_router)
-router.include_router(time_series_router)
-router.include_router(cost_router)
-router.include_router(quota_router)
-router.include_router(performance_router)
-router.include_router(errors_router)
-router.include_router(comparison_router)
-
-__all__ = ["router"]

src/api/admin/stats/common.py

@@ -1,196 +0,0 @@
-"""Shared helpers for admin stats routes."""
-
-from __future__ import annotations
-
-import hashlib
-import json
-from datetime import date, datetime, time, timedelta, timezone
-from typing import Any, Literal
-
-from fastapi import HTTPException
-from sqlalchemy import and_, or_
-
-from src.api.base.pipeline import get_pipeline
-from src.config.settings import config
-from src.models.database import Usage
-from src.services.system.time_range import TimeRangeParams
-
-pipeline = get_pipeline()
-
-
-def _apply_admin_default_range(
-    params: TimeRangeParams | None,
-) -> TimeRangeParams | None:
-    """Apply a default range to avoid unbounded scans."""
-    if params is not None:
-        return params
-
-    days = int(getattr(config, "admin_usage_default_days", 0) or 0)
-    if days <= 0:
-        return None
-
-    today = datetime.now(timezone.utc).date()
-    start_date = today - timedelta(days=days - 1)
-    return TimeRangeParams(
-        start_date=start_date,
-        end_date=today,
-        timezone="UTC",
-        tz_offset_minutes=0,
-    ).validate_and_resolve()
-
-
-def _build_time_range_params(
-    start_date: date | None,
-    end_date: date | None,
-    preset: str | None,
-    timezone_name: str | None,
-    tz_offset_minutes: int | None,
-) -> TimeRangeParams | None:
-    if not preset and start_date is None and end_date is None:
-        return None
-    try:
-        return TimeRangeParams(
-            start_date=start_date,
-            end_date=end_date,
-            preset=preset,
-            timezone=timezone_name,
-            tz_offset_minutes=tz_offset_minutes or 0,
-        ).validate_and_resolve()
-    except Exception as exc:
-        raise HTTPException(status_code=400, detail=str(exc)) from exc
-
-
-def _hash_filters(filters: dict[str, Any]) -> str:
-    raw = json.dumps(filters, sort_keys=True, ensure_ascii=False, default=str)
-    return hashlib.sha1(raw.encode("utf-8")).hexdigest()[:16]
-
-
-def _build_time_range_from_days(
-    days: int, timezone_name: str | None, tz_offset_minutes: int | None
-) -> TimeRangeParams:
-    base = TimeRangeParams(
-        preset="today",
-        timezone=timezone_name,
-        tz_offset_minutes=tz_offset_minutes or 0,
-    ).validate_and_resolve()
-    user_today = base.start_date
-    start_date = user_today - timedelta(days=days - 1)
-    return TimeRangeParams(
-        start_date=start_date,
-        end_date=user_today,
-        timezone=timezone_name,
-        tz_offset_minutes=tz_offset_minutes or 0,
-    ).validate_and_resolve()
-
-
-def _linear_regression(values: list[float]) -> tuple[float, float]:
-    n = len(values)
-    if n <= 1:
-        return 0.0, values[0] if values else 0.0
-    xs = list(range(n))
-    sum_x = sum(xs)
-    sum_y = sum(values)
-    sum_x2 = sum(x * x for x in xs)
-    sum_xy = sum(x * y for x, y in zip(xs, values))
-    denom = n * sum_x2 - sum_x * sum_x
-    if denom == 0:
-        return 0.0, values[-1]
-    slope = (n * sum_xy - sum_x * sum_y) / denom
-    intercept = (sum_y - slope * sum_x) / n
-    return slope, intercept
-
-
-def _build_cache_key(
-    leaderboard_type: str,
-    metric: str,
-    time_range: TimeRangeParams | None,
-    filters: dict[str, Any],
-) -> str:
-    start_value = time_range.start_date.isoformat() if time_range else "all"
-    end_value = time_range.end_date.isoformat() if time_range else "all"
-    tz_value = time_range.timezone if time_range else "utc"
-    offset_value = time_range.tz_offset_minutes if time_range else 0
-    return (
-        f"leaderboard:{leaderboard_type}:{metric}:{start_value}:{end_value}:"
-        f"{tz_value}:{offset_value}:{_hash_filters(filters)}"
-    )
-
-
-def _is_today_range(time_range: TimeRangeParams | None) -> bool:
-    if not time_range:
-        return False
-    try:
-        user_today = time_range._get_user_today()
-    except Exception:
-        return False
-    return time_range.end_date == user_today
-
-
-def _split_daily_and_usage_segments(
-    time_range: TimeRangeParams | None,
-    use_daily: bool,
-) -> tuple[tuple[datetime, datetime] | None, list[tuple[datetime, datetime]] | None]:
-    if not time_range:
-        return None, None
-
-    start_utc, end_utc = time_range.to_utc_datetime_range()
-    if not use_daily:
-        return None, [(start_utc, end_utc)]
-
-    complete_dates, head_boundary, tail_boundary = time_range.get_complete_utc_dates()
-    daily_range = None
-    if complete_dates:
-        daily_start = datetime.combine(complete_dates[0], time.min, tzinfo=timezone.utc)
-        daily_end = datetime.combine(
-            complete_dates[-1] + timedelta(days=1), time.min, tzinfo=timezone.utc
-        )
-        daily_range = (daily_start, daily_end)
-
-    usage_segments: list[tuple[datetime, datetime]] = []
-    if head_boundary:
-        usage_segments.append(head_boundary)
-    if tail_boundary:
-        usage_segments.append(tail_boundary)
-
-    if not daily_range and not usage_segments:
-        usage_segments = [(start_utc, end_utc)]
-
-    return daily_range, usage_segments
-
-
-def _apply_usage_time_segments(
-    query: Any, segments: list[tuple[datetime, datetime]] | None
-) -> Any | None:
-    if segments is None:
-        return query
-    if not segments:
-        return None
-
-    conditions = []
-    for start_utc, end_utc in segments:
-        if start_utc >= end_utc:
-            continue
-        conditions.append(and_(Usage.created_at >= start_utc, Usage.created_at < end_utc))
-
-    if not conditions:
-        return None
-
-    return query.filter(or_(*conditions))
-
-
-def _union_queries(queries: list[Any]) -> Any | None:
-    base = None
-    for query in queries:
-        if query is None:
-            continue
-        if base is None:
-            base = query
-        else:
-            base = base.union_all(query)
-    return base
-
-
-def _metric_order(
-    metric: Literal["requests", "tokens", "cost"], order: Literal["asc", "desc"], expr: Any
-) -> Any:
-    return expr.asc() if order == "asc" else expr.desc()

src/api/admin/stats/comparison.py

@@ -1,142 +0,0 @@
-"""Admin comparison stats routes."""
-
-from __future__ import annotations
-
-from datetime import date, timedelta
-from typing import Any, Literal
-
-from fastapi import APIRouter, Depends, HTTPException, Query, Request
-from sqlalchemy.orm import Session
-
-from src.api.base.admin_adapter import AdminApiAdapter
-from src.api.base.context import ApiRequestContext
-from src.config.constants import CacheTTL
-from src.database import get_db
-from src.services.system.stats_aggregator import AggregatedStats, StatsFilter, query_stats_hybrid
-from src.services.system.time_range import TimeRangeParams
-from src.utils.cache_decorator import cache_result
-
-from .common import pipeline
-
-router = APIRouter()
-
-
-class AdminComparisonAdapter(AdminApiAdapter):
-    def __init__(
-        self,
-        current_start: date,
-        current_end: date,
-        comparison_type: Literal["period", "year"],
-        timezone_name: str | None,
-        tz_offset_minutes: int | None,
-    ) -> None:
-        self.current_start = current_start
-        self.current_end = current_end
-        self.comparison_type = comparison_type
-        self.timezone_name = timezone_name
-        self.tz_offset_minutes = tz_offset_minutes
-
-    @cache_result(
-        key_prefix="admin:stats:comparison",
-        ttl=CacheTTL.ADMIN_USAGE_AGGREGATION,
-        user_specific=False,
-        vary_by=[
-            "current_start",
-            "current_end",
-            "comparison_type",
-            "timezone_name",
-            "tz_offset_minutes",
-        ],
-    )
-    async def handle(self, context: ApiRequestContext) -> Any:  # type: ignore[override]
-        if self.current_start > self.current_end:
-            raise HTTPException(status_code=400, detail="current_start must be <= current_end")
-
-        days = (self.current_end - self.current_start).days + 1
-
-        def _safe_year_shift(value: date) -> date:
-            try:
-                return value.replace(year=value.year - 1)
-            except ValueError:
-                return value.replace(year=value.year - 1, day=28)
-
-        if self.comparison_type == "period":
-            comparison_end = self.current_start - timedelta(days=1)
-            comparison_start = comparison_end - timedelta(days=days - 1)
-        else:
-            comparison_start = _safe_year_shift(self.current_start)
-            comparison_end = _safe_year_shift(self.current_end)
-
-        current_range = TimeRangeParams(
-            start_date=self.current_start,
-            end_date=self.current_end,
-            timezone=self.timezone_name,
-            tz_offset_minutes=self.tz_offset_minutes or 0,
-        ).validate_and_resolve()
-        comparison_range = TimeRangeParams(
-            start_date=comparison_start,
-            end_date=comparison_end,
-            timezone=self.timezone_name,
-            tz_offset_minutes=self.tz_offset_minutes or 0,
-        ).validate_and_resolve()
-
-        current_stats = query_stats_hybrid(context.db, current_range, filters=StatsFilter())
-        comparison_stats = query_stats_hybrid(context.db, comparison_range, filters=StatsFilter())
-
-        def _stats_payload(stats: AggregatedStats) -> dict[str, Any]:
-            total_tokens = (
-                stats.input_tokens
-                + stats.output_tokens
-                + stats.cache_creation_tokens
-                + stats.cache_read_tokens
-            )
-            return {
-                "total_requests": stats.total_requests,
-                "total_tokens": total_tokens,
-                "total_cost": float(stats.total_cost),
-                "actual_total_cost": float(stats.actual_total_cost),
-                "avg_response_time_ms": float(stats.avg_response_time_ms),
-                "error_requests": stats.error_requests,
-            }
-
-        def _pct_change(current: float, previous: float) -> float | None:
-            if previous == 0:
-                return None if current != 0 else 0.0
-            return round((current - previous) / previous * 100, 2)
-
-        current_payload = _stats_payload(current_stats)
-        comparison_payload = _stats_payload(comparison_stats)
-        changes = {
-            key: _pct_change(float(current_payload[key]), float(comparison_payload[key]))
-            for key in current_payload.keys()
-        }
-
-        return {
-            "current": current_payload,
-            "comparison": comparison_payload,
-            "change_percent": changes,
-            "current_start": self.current_start.isoformat(),
-            "current_end": self.current_end.isoformat(),
-            "comparison_start": comparison_start.isoformat(),
-            "comparison_end": comparison_end.isoformat(),
-        }
-
-
-@router.get("/comparison")
-async def get_comparison(
-    request: Request,
-    db: Session = Depends(get_db),
-    current_start: date = Query(...),
-    current_end: date = Query(...),
-    comparison_type: Literal["period", "year"] = Query("period"),
-    timezone_name: str | None = Query(None, alias="timezone"),
-    tz_offset_minutes: int | None = Query(0),
-) -> Any:
-    adapter = AdminComparisonAdapter(
-        current_start=current_start,
-        current_end=current_end,
-        comparison_type=comparison_type,
-        timezone_name=timezone_name,
-        tz_offset_minutes=tz_offset_minutes,
-    )
-    return await pipeline.run(adapter=adapter, http_request=request, db=db, mode=adapter.mode)

src/api/admin/stats/cost.py

@@ -1,215 +0,0 @@
-"""Admin cost stats routes."""
-
-from __future__ import annotations
-
-from datetime import date, timedelta
-from typing import Any
-
-from fastapi import APIRouter, Depends, HTTPException, Query, Request
-from sqlalchemy import func
-from sqlalchemy.orm import Session
-
-from src.api.base.admin_adapter import AdminApiAdapter
-from src.api.base.context import ApiRequestContext
-from src.config.constants import CacheTTL
-from src.database import get_db
-from src.models.database import Usage
-from src.services.system.stats_aggregator import query_time_series
-from src.services.system.time_range import TimeRangeParams
-from src.utils.cache_decorator import cache_result
-
-from .common import (
-    _apply_admin_default_range,
-    _build_time_range_from_days,
-    _build_time_range_params,
-    _linear_regression,
-    pipeline,
-)
-
-router = APIRouter()
-
-
-class AdminCostForecastAdapter(AdminApiAdapter):
-    def __init__(
-        self,
-        time_range: TimeRangeParams | None,
-        days: int,
-        forecast_days: int,
-        timezone_name: str | None,
-        tz_offset_minutes: int | None,
-    ) -> None:
-        self.time_range = time_range
-        self.days = days
-        self.forecast_days = forecast_days
-        self.timezone_name = timezone_name
-        self.fallback_tz_offset_minutes = tz_offset_minutes
-
-    @cache_result(
-        key_prefix="admin:stats:cost:forecast",
-        ttl=CacheTTL.ADMIN_USAGE_AGGREGATION,
-        user_specific=False,
-        vary_by=[
-            "time_range.start_date",
-            "time_range.end_date",
-            "time_range.preset",
-            "time_range.timezone",
-            "time_range.tz_offset_minutes",
-            "days",
-            "forecast_days",
-            "timezone_name",
-            "fallback_tz_offset_minutes",
-        ],
-    )
-    async def handle(self, context: ApiRequestContext) -> Any:  # type: ignore[override]
-        time_range = self.time_range or _build_time_range_from_days(
-            self.days, self.timezone_name, self.fallback_tz_offset_minutes
-        )
-        time_range.granularity = "day"
-        try:
-            series = query_time_series(context.db, time_range)
-        except ValueError as exc:
-            raise HTTPException(status_code=400, detail=str(exc)) from exc
-        history = [
-            {"date": item["date"], "total_cost": float(item.get("total_cost", 0.0))}
-            for item in series
-        ]
-
-        values = [item["total_cost"] for item in history]
-        slope, intercept = _linear_regression(values)
-
-        forecast = []
-        if history:
-            last_date = date.fromisoformat(history[-1]["date"])
-        else:
-            last_date = time_range.end_date
-        for i in range(self.forecast_days):
-            idx = len(values) + i
-            predicted = max(0.0, slope * idx + intercept)
-            forecast.append(
-                {
-                    "date": (last_date + timedelta(days=i + 1)).isoformat(),
-                    "total_cost": round(predicted, 4),
-                }
-            )
-
-        return {
-            "history": history,
-            "forecast": forecast,
-            "slope": round(slope, 6),
-            "intercept": round(intercept, 6),
-            "start_date": time_range.start_date.isoformat(),
-            "end_date": time_range.end_date.isoformat(),
-        }
-
-
-@router.get("/cost/forecast")
-async def get_cost_forecast(
-    request: Request,
-    db: Session = Depends(get_db),
-    start_date: date | None = Query(None),
-    end_date: date | None = Query(None),
-    preset: str | None = Query(None),
-    timezone_name: str | None = Query(None, alias="timezone"),
-    tz_offset_minutes: int | None = Query(0),
-    days: int = Query(30, ge=7, le=365),
-    forecast_days: int = Query(7, ge=1, le=90),
-) -> Any:
-    time_range = _build_time_range_params(
-        start_date, end_date, preset, timezone_name, tz_offset_minutes
-    )
-    adapter = AdminCostForecastAdapter(
-        time_range=time_range,
-        days=days,
-        forecast_days=forecast_days,
-        timezone_name=timezone_name,
-        tz_offset_minutes=tz_offset_minutes,
-    )
-    return await pipeline.run(adapter=adapter, http_request=request, db=db, mode=adapter.mode)
-
-
-class AdminCostSavingsAdapter(AdminApiAdapter):
-    def __init__(
-        self,
-        time_range: TimeRangeParams | None,
-        provider_name: str | None,
-        model: str | None,
-    ) -> None:
-        self.time_range = _apply_admin_default_range(time_range)
-        self.provider_name = provider_name
-        self.model = model
-
-    @cache_result(
-        key_prefix="admin:stats:cost:savings",
-        ttl=CacheTTL.ADMIN_USAGE_AGGREGATION,
-        user_specific=False,
-        vary_by=[
-            "time_range.start_date",
-            "time_range.end_date",
-            "time_range.preset",
-            "time_range.timezone",
-            "time_range.tz_offset_minutes",
-            "provider_name",
-            "model",
-        ],
-    )
-    async def handle(self, context: ApiRequestContext) -> Any:  # type: ignore[override]
-        if not self.time_range:
-            return {
-                "cache_read_tokens": 0,
-                "cache_read_cost": 0.0,
-                "cache_creation_cost": 0.0,
-                "estimated_full_cost": 0.0,
-                "cache_savings": 0.0,
-            }
-
-        start_utc, end_utc = self.time_range.to_utc_datetime_range()
-        query = context.db.query(
-            func.sum(Usage.cache_read_input_tokens).label("cache_read_tokens"),
-            func.sum(Usage.cache_read_cost_usd).label("cache_read_cost"),
-            func.sum(Usage.cache_creation_cost_usd).label("cache_creation_cost"),
-            func.sum(
-                func.coalesce(Usage.output_price_per_1m, 0) * Usage.cache_read_input_tokens
-            ).label("estimated_full_cost_raw"),
-        ).filter(Usage.created_at >= start_utc, Usage.created_at < end_utc)
-        if self.provider_name:
-            query = query.filter(Usage.provider_name == self.provider_name)
-        if self.model:
-            query = query.filter(Usage.model == self.model)
-
-        row = query.first()
-        cache_read_tokens = int(getattr(row, "cache_read_tokens", 0) or 0)
-        cache_read_cost = float(getattr(row, "cache_read_cost", 0) or 0.0)
-        cache_creation_cost = float(getattr(row, "cache_creation_cost", 0) or 0.0)
-        estimated_full_cost = float(getattr(row, "estimated_full_cost_raw", 0) or 0.0) / 1_000_000
-        if estimated_full_cost <= 0 and cache_read_cost > 0:
-            estimated_full_cost = cache_read_cost * 10
-        cache_savings = estimated_full_cost - cache_read_cost
-
-        return {
-            "cache_read_tokens": cache_read_tokens,
-            "cache_read_cost": round(cache_read_cost, 6),
-            "cache_creation_cost": round(cache_creation_cost, 6),
-            "estimated_full_cost": round(estimated_full_cost, 6),
-            "cache_savings": round(cache_savings, 6),
-        }
-
-
-@router.get("/cost/savings")
-async def get_cost_savings(
-    request: Request,
-    db: Session = Depends(get_db),
-    start_date: date | None = Query(None),
-    end_date: date | None = Query(None),
-    preset: str | None = Query(None),
-    timezone_name: str | None = Query(None, alias="timezone"),
-    tz_offset_minutes: int | None = Query(0),
-    provider_name: str | None = Query(None),
-    model: str | None = Query(None),
-) -> Any:
-    time_range = _build_time_range_params(
-        start_date, end_date, preset, timezone_name, tz_offset_minutes
-    )
-    adapter = AdminCostSavingsAdapter(
-        time_range=time_range, provider_name=provider_name, model=model
-    )
-    return await pipeline.run(adapter=adapter, http_request=request, db=db, mode=adapter.mode)

src/api/admin/stats/errors.py

@@ -1,125 +0,0 @@
-"""Admin error stats routes."""
-
-from __future__ import annotations
-
-from datetime import date, timezone
-from typing import Any
-
-from fastapi import APIRouter, Depends, Query, Request
-from sqlalchemy import func
-from sqlalchemy.orm import Session
-
-from src.api.base.admin_adapter import AdminApiAdapter
-from src.api.base.context import ApiRequestContext
-from src.config.constants import CacheTTL
-from src.database import get_db
-from src.models.database import StatsDailyError, Usage
-from src.services.system.time_range import TimeRangeParams
-from src.utils.cache_decorator import cache_result
-
-from .common import _apply_admin_default_range, _build_time_range_params, pipeline
-
-router = APIRouter()
-
-
-class AdminErrorDistributionAdapter(AdminApiAdapter):
-    def __init__(self, time_range: TimeRangeParams | None) -> None:
-        self.time_range = _apply_admin_default_range(time_range)
-
-    @cache_result(
-        key_prefix="admin:stats:errors:distribution",
-        ttl=CacheTTL.ADMIN_USAGE_AGGREGATION,
-        user_specific=False,
-        vary_by=[
-            "time_range.start_date",
-            "time_range.end_date",
-            "time_range.preset",
-            "time_range.timezone",
-            "time_range.tz_offset_minutes",
-        ],
-    )
-    async def handle(self, context: ApiRequestContext) -> Any:  # type: ignore[override]
-        if not self.time_range:
-            return {"distribution": [], "trend": []}
-
-        time_range = self.time_range
-        is_utc = (time_range.timezone in {None, "UTC"}) and time_range.tz_offset_minutes == 0
-        distribution: dict[str, int] = {}
-        trend: dict[str, dict[str, int]] = {}
-
-        if is_utc:
-            start_utc, end_utc = time_range.to_utc_datetime_range()
-            rows = (
-                context.db.query(StatsDailyError)
-                .filter(StatsDailyError.date >= start_utc, StatsDailyError.date < end_utc)
-                .all()
-            )
-            for row in rows:
-                date_str = (
-                    row.date.astimezone(timezone.utc).date().isoformat()
-                    if row.date.tzinfo
-                    else row.date.date().isoformat()
-                )
-                distribution[row.error_category] = distribution.get(row.error_category, 0) + int(
-                    row.count or 0
-                )
-                trend.setdefault(date_str, {})
-                trend[date_str][row.error_category] = trend[date_str].get(
-                    row.error_category, 0
-                ) + int(row.count or 0)
-        else:
-            for local_date, day_start_utc, day_end_utc in time_range.get_local_day_hours():
-                rows = (
-                    context.db.query(
-                        Usage.error_category,
-                        func.count(Usage.id).label("count"),
-                    )
-                    .filter(
-                        Usage.created_at >= day_start_utc,
-                        Usage.created_at < day_end_utc,
-                        Usage.error_category.isnot(None),
-                    )
-                    .group_by(Usage.error_category)
-                    .all()
-                )
-                date_str = local_date.isoformat()
-                for row in rows:
-                    distribution[row.error_category] = distribution.get(
-                        row.error_category, 0
-                    ) + int(row.count or 0)
-                    trend.setdefault(date_str, {})
-                    trend[date_str][row.error_category] = trend[date_str].get(
-                        row.error_category, 0
-                    ) + int(row.count or 0)
-
-        trend_items = []
-        for day in sorted(trend.keys()):
-            counts = trend[day]
-            total = sum(counts.values())
-            trend_items.append({"date": day, "total": total, "categories": counts})
-
-        distribution_items = [
-            {"category": category, "count": count}
-            for category, count in sorted(
-                distribution.items(), key=lambda item: item[1], reverse=True
-            )
-        ]
-
-        return {"distribution": distribution_items, "trend": trend_items}
-
-
-@router.get("/errors/distribution")
-async def get_error_distribution(
-    request: Request,
-    db: Session = Depends(get_db),
-    start_date: date | None = Query(None),
-    end_date: date | None = Query(None),
-    preset: str | None = Query(None),
-    timezone_name: str | None = Query(None, alias="timezone"),
-    tz_offset_minutes: int | None = Query(0),
-) -> Any:
-    time_range = _build_time_range_params(
-        start_date, end_date, preset, timezone_name, tz_offset_minutes
-    )
-    adapter = AdminErrorDistributionAdapter(time_range=time_range)
-    return await pipeline.run(adapter=adapter, http_request=request, db=db, mode=adapter.mode)

src/api/admin/stats/leaderboard.py

@@ -1,759 +0,0 @@
-"""Admin leaderboard stats routes."""
-
-from __future__ import annotations
-
-import json
-from datetime import date
-from typing import Any, Literal
-
-from fastapi import APIRouter, Depends, Query, Request
-from sqlalchemy import func
-from sqlalchemy.orm import Session
-
-from src.api.base.admin_adapter import AdminApiAdapter
-from src.api.base.context import ApiRequestContext
-from src.clients.redis_client import get_redis_client_sync
-from src.config.constants import CacheTTL
-from src.core.enums import UserRole
-from src.database import get_db
-from src.models.database import (
-    ApiKey,
-    StatsDailyApiKey,
-    StatsDailyModel,
-    StatsUserDaily,
-    Usage,
-    User,
-)
-from src.services.system.time_range import TimeRangeParams
-
-from .common import (
-    _apply_admin_default_range,
-    _apply_usage_time_segments,
-    _build_cache_key,
-    _build_time_range_params,
-    _is_today_range,
-    _metric_order,
-    _split_daily_and_usage_segments,
-    _union_queries,
-    pipeline,
-)
-
-router = APIRouter()
-
-
-class AdminUserLeaderboardAdapter(AdminApiAdapter):
-    def __init__(
-        self,
-        time_range: TimeRangeParams | None,
-        metric: Literal["requests", "tokens", "cost"],
-        order: Literal["asc", "desc"],
-        limit: int,
-        offset: int,
-        provider_name: str | None,
-        model: str | None,
-        include_inactive: bool,
-        exclude_admin: bool,
-    ) -> None:
-        self.time_range = _apply_admin_default_range(time_range)
-        self.metric = metric
-        self.order = order
-        self.limit = limit
-        self.offset = offset
-        self.provider_name = provider_name
-        self.model = model
-        self.include_inactive = include_inactive
-        self.exclude_admin = exclude_admin
-
-    async def handle(self, context: ApiRequestContext) -> Any:  # type: ignore[override]
-        db = context.db
-        cacheable = not _is_today_range(self.time_range)
-        redis_client = get_redis_client_sync()
-        cache_key = None
-        if cacheable and redis_client:
-            cache_key = _build_cache_key(
-                "users",
-                self.metric,
-                self.time_range,
-                {
-                    "order": self.order,
-                    "limit": self.limit,
-                    "offset": self.offset,
-                    "provider_name": self.provider_name,
-                    "model": self.model,
-                    "include_inactive": self.include_inactive,
-                    "exclude_admin": self.exclude_admin,
-                },
-            )
-            cached = await redis_client.get(cache_key)
-            if cached:
-                try:
-                    return json.loads(cached)
-                except Exception:
-                    pass
-
-        use_daily = self.time_range is not None and not self.provider_name and not self.model
-        daily_range, usage_segments = _split_daily_and_usage_segments(self.time_range, use_daily)
-
-        daily_query = None
-        if daily_range:
-            daily_query = (
-                db.query(
-                    StatsUserDaily.user_id.label("entity_id"),
-                    func.sum(StatsUserDaily.total_requests).label("requests"),
-                    func.sum(
-                        StatsUserDaily.input_tokens
-                        + StatsUserDaily.output_tokens
-                        + StatsUserDaily.cache_creation_tokens
-                        + StatsUserDaily.cache_read_tokens
-                    ).label("tokens"),
-                    func.sum(StatsUserDaily.total_cost).label("cost"),
-                )
-                .filter(StatsUserDaily.date >= daily_range[0], StatsUserDaily.date < daily_range[1])
-                .group_by(StatsUserDaily.user_id)
-            )
-
-        usage_query = db.query(
-            Usage.user_id.label("entity_id"),
-            func.count(Usage.id).label("requests"),
-            func.sum(
-                Usage.input_tokens
-                + Usage.output_tokens
-                + Usage.cache_creation_input_tokens
-                + Usage.cache_read_input_tokens
-            ).label("tokens"),
-            func.sum(Usage.total_cost_usd).label("cost"),
-        ).filter(
-            Usage.user_id.isnot(None),
-            Usage.status.notin_(["pending", "streaming"]),
-            Usage.provider_name.notin_(["unknown", "pending"]),
-        )
-        if self.provider_name:
-            usage_query = usage_query.filter(Usage.provider_name == self.provider_name)
-        if self.model:
-            usage_query = usage_query.filter(Usage.model == self.model)
-        usage_query = _apply_usage_time_segments(usage_query, usage_segments)
-        if usage_query is not None:
-            usage_query = usage_query.group_by(Usage.user_id)
-
-        union_query = _union_queries([daily_query, usage_query])
-        if union_query is None:
-            return {
-                "items": [],
-                "total": 0,
-                "metric": self.metric,
-                "start_date": self.time_range.start_date.isoformat() if self.time_range else None,
-                "end_date": self.time_range.end_date.isoformat() if self.time_range else None,
-            }
-
-        union_subq = union_query.subquery()
-        agg_subq = (
-            db.query(
-                union_subq.c.entity_id.label("entity_id"),
-                func.sum(union_subq.c.requests).label("requests"),
-                func.sum(union_subq.c.tokens).label("tokens"),
-                func.sum(union_subq.c.cost).label("cost"),
-            )
-            .group_by(union_subq.c.entity_id)
-            .subquery()
-        )
-
-        base_query = (
-            db.query(
-                User.id.label("id"),
-                User.username,
-                User.email,
-                agg_subq.c.requests,
-                agg_subq.c.tokens,
-                agg_subq.c.cost,
-            )
-            .join(agg_subq, agg_subq.c.entity_id == User.id)
-            .filter(User.is_deleted.is_(False))
-        )
-        if not self.include_inactive:
-            base_query = base_query.filter(User.is_active.is_(True))
-        if self.exclude_admin:
-            base_query = base_query.filter(User.role != UserRole.ADMIN)
-
-        metric_expr = {
-            "requests": agg_subq.c.requests,
-            "tokens": agg_subq.c.tokens,
-            "cost": agg_subq.c.cost,
-        }[self.metric]
-        order_expr = _metric_order(self.metric, self.order, metric_expr)
-        rank_expr = func.dense_rank().over(order_by=order_expr).label("rank")
-
-        total = db.query(func.count()).select_from(base_query.subquery()).scalar() or 0
-        rows = (
-            base_query.add_columns(rank_expr, metric_expr.label("metric_value"))
-            .order_by(order_expr)
-            .offset(self.offset)
-            .limit(self.limit)
-            .all()
-        )
-
-        items = []
-        for row in rows:
-            name = row.username or row.email or str(row.id)
-            value = row.metric_value or 0
-            if self.metric in {"requests", "tokens"}:
-                value = int(value)
-            else:
-                value = float(value)
-            items.append(
-                {
-                    "rank": int(row.rank),
-                    "id": row.id,
-                    "name": name,
-                    "value": value,
-                    "requests": int(row.requests or 0),
-                    "tokens": int(row.tokens or 0),
-                    "cost": float(row.cost or 0.0),
-                }
-            )
-
-        context.add_audit_metadata(
-            action="leaderboard_users",
-            start_date=self.time_range.start_date.isoformat() if self.time_range else None,
-            end_date=self.time_range.end_date.isoformat() if self.time_range else None,
-            preset=self.time_range.preset if self.time_range else None,
-            timezone=self.time_range.timezone if self.time_range else None,
-            metric=self.metric,
-            order=self.order,
-            limit=self.limit,
-            offset=self.offset,
-            provider_name=self.provider_name,
-            model=self.model,
-            include_inactive=self.include_inactive,
-            exclude_admin=self.exclude_admin,
-            result_count=len(items),
-            total=total,
-        )
-
-        result = {
-            "items": items,
-            "total": total,
-            "metric": self.metric,
-            "start_date": self.time_range.start_date.isoformat() if self.time_range else None,
-            "end_date": self.time_range.end_date.isoformat() if self.time_range else None,
-        }
-
-        if cacheable and redis_client and cache_key:
-            try:
-                await redis_client.setex(
-                    cache_key, CacheTTL.ADMIN_LEADERBOARD, json.dumps(result, ensure_ascii=False)
-                )
-            except Exception:
-                pass
-
-        return result
-
-
-class AdminApiKeyLeaderboardAdapter(AdminApiAdapter):
-    def __init__(
-        self,
-        time_range: TimeRangeParams | None,
-        metric: Literal["requests", "tokens", "cost"],
-        order: Literal["asc", "desc"],
-        limit: int,
-        offset: int,
-        provider_name: str | None,
-        model: str | None,
-        include_inactive: bool,
-        exclude_admin: bool,
-    ) -> None:
-        self.time_range = _apply_admin_default_range(time_range)
-        self.metric = metric
-        self.order = order
-        self.limit = limit
-        self.offset = offset
-        self.provider_name = provider_name
-        self.model = model
-        self.include_inactive = include_inactive
-        self.exclude_admin = exclude_admin
-
-    async def handle(self, context: ApiRequestContext) -> Any:  # type: ignore[override]
-        db = context.db
-        cacheable = not _is_today_range(self.time_range)
-        redis_client = get_redis_client_sync()
-        cache_key = None
-        if cacheable and redis_client:
-            cache_key = _build_cache_key(
-                "api_keys",
-                self.metric,
-                self.time_range,
-                {
-                    "order": self.order,
-                    "limit": self.limit,
-                    "offset": self.offset,
-                    "provider_name": self.provider_name,
-                    "model": self.model,
-                    "include_inactive": self.include_inactive,
-                    "exclude_admin": self.exclude_admin,
-                },
-            )
-            cached = await redis_client.get(cache_key)
-            if cached:
-                try:
-                    return json.loads(cached)
-                except Exception:
-                    pass
-
-        use_daily = self.time_range is not None and not self.provider_name and not self.model
-        daily_range, usage_segments = _split_daily_and_usage_segments(self.time_range, use_daily)
-
-        daily_query = None
-        if daily_range:
-            daily_query = (
-                db.query(
-                    StatsDailyApiKey.api_key_id.label("entity_id"),
-                    func.sum(StatsDailyApiKey.total_requests).label("requests"),
-                    func.sum(
-                        StatsDailyApiKey.input_tokens
-                        + StatsDailyApiKey.output_tokens
-                        + StatsDailyApiKey.cache_creation_tokens
-                        + StatsDailyApiKey.cache_read_tokens
-                    ).label("tokens"),
-                    func.sum(StatsDailyApiKey.total_cost).label("cost"),
-                )
-                .filter(
-                    StatsDailyApiKey.date >= daily_range[0], StatsDailyApiKey.date < daily_range[1]
-                )
-                .group_by(StatsDailyApiKey.api_key_id)
-            )
-
-        usage_query = db.query(
-            Usage.api_key_id.label("entity_id"),
-            func.count(Usage.id).label("requests"),
-            func.sum(
-                Usage.input_tokens
-                + Usage.output_tokens
-                + Usage.cache_creation_input_tokens
-                + Usage.cache_read_input_tokens
-            ).label("tokens"),
-            func.sum(Usage.total_cost_usd).label("cost"),
-        ).filter(
-            Usage.api_key_id.isnot(None),
-            Usage.status.notin_(["pending", "streaming"]),
-            Usage.provider_name.notin_(["unknown", "pending"]),
-        )
-        if self.provider_name:
-            usage_query = usage_query.filter(Usage.provider_name == self.provider_name)
-        if self.model:
-            usage_query = usage_query.filter(Usage.model == self.model)
-
-        usage_query = _apply_usage_time_segments(usage_query, usage_segments)
-        if usage_query is None:
-            return {
-                "items": [],
-                "total": 0,
-                "metric": self.metric,
-                "start_date": self.time_range.start_date.isoformat() if self.time_range else None,
-                "end_date": self.time_range.end_date.isoformat() if self.time_range else None,
-            }
-        usage_query = usage_query.group_by(Usage.api_key_id)
-
-        union_query = _union_queries([daily_query, usage_query])
-        if union_query is None:
-            return {
-                "items": [],
-                "total": 0,
-                "metric": self.metric,
-                "start_date": self.time_range.start_date.isoformat() if self.time_range else None,
-                "end_date": self.time_range.end_date.isoformat() if self.time_range else None,
-            }
-
-        union_subq = union_query.subquery()
-        agg_subq = (
-            db.query(
-                union_subq.c.entity_id.label("entity_id"),
-                func.sum(union_subq.c.requests).label("requests"),
-                func.sum(union_subq.c.tokens).label("tokens"),
-                func.sum(union_subq.c.cost).label("cost"),
-            )
-            .group_by(union_subq.c.entity_id)
-            .subquery()
-        )
-
-        base_query = (
-            db.query(
-                ApiKey,
-                User,
-                agg_subq.c.requests,
-                agg_subq.c.tokens,
-                agg_subq.c.cost,
-            )
-            .join(agg_subq, agg_subq.c.entity_id == ApiKey.id)
-            .join(User, User.id == ApiKey.user_id)
-            .filter(User.is_deleted.is_(False))
-        )
-        if not self.include_inactive:
-            base_query = base_query.filter(ApiKey.is_active.is_(True))
-        if self.exclude_admin:
-            base_query = base_query.filter(User.role != UserRole.ADMIN)
-
-        metric_expr = {
-            "requests": agg_subq.c.requests,
-            "tokens": agg_subq.c.tokens,
-            "cost": agg_subq.c.cost,
-        }[self.metric]
-        order_expr = _metric_order(self.metric, self.order, metric_expr)
-        rank_expr = func.dense_rank().over(order_by=order_expr).label("rank")
-
-        total = db.query(func.count()).select_from(base_query.subquery()).scalar() or 0
-        rows = (
-            base_query.add_columns(rank_expr, metric_expr.label("metric_value"))
-            .order_by(order_expr)
-            .offset(self.offset)
-            .limit(self.limit)
-            .all()
-        )
-
-        items = []
-        for row in rows:
-            api_key = row.ApiKey
-            name = api_key.name or api_key.get_display_key()
-            value = row.metric_value or 0
-            if self.metric in {"requests", "tokens"}:
-                value = int(value)
-            else:
-                value = float(value)
-            items.append(
-                {
-                    "rank": int(row.rank),
-                    "id": api_key.id,
-                    "name": name,
-                    "value": value,
-                    "requests": int(row.requests or 0),
-                    "tokens": int(row.tokens or 0),
-                    "cost": float(row.cost or 0.0),
-                }
-            )
-
-        context.add_audit_metadata(
-            action="leaderboard_api_keys",
-            start_date=self.time_range.start_date.isoformat() if self.time_range else None,
-            end_date=self.time_range.end_date.isoformat() if self.time_range else None,
-            preset=self.time_range.preset if self.time_range else None,
-            timezone=self.time_range.timezone if self.time_range else None,
-            metric=self.metric,
-            order=self.order,
-            limit=self.limit,
-            offset=self.offset,
-            provider_name=self.provider_name,
-            model=self.model,
-            include_inactive=self.include_inactive,
-            exclude_admin=self.exclude_admin,
-            result_count=len(items),
-            total=total,
-        )
-
-        result = {
-            "items": items,
-            "total": total,
-            "metric": self.metric,
-            "start_date": self.time_range.start_date.isoformat() if self.time_range else None,
-            "end_date": self.time_range.end_date.isoformat() if self.time_range else None,
-        }
-
-        if cacheable and redis_client and cache_key:
-            try:
-                await redis_client.setex(
-                    cache_key, CacheTTL.ADMIN_LEADERBOARD, json.dumps(result, ensure_ascii=False)
-                )
-            except Exception:
-                pass
-
-        return result
-
-
-class AdminModelLeaderboardAdapter(AdminApiAdapter):
-    def __init__(
-        self,
-        time_range: TimeRangeParams | None,
-        metric: Literal["requests", "tokens", "cost"],
-        order: Literal["asc", "desc"],
-        limit: int,
-        offset: int,
-        provider_name: str | None,
-        model: str | None,
-    ) -> None:
-        self.time_range = _apply_admin_default_range(time_range)
-        self.metric = metric
-        self.order = order
-        self.limit = limit
-        self.offset = offset
-        self.provider_name = provider_name
-        self.model = model
-
-    async def handle(self, context: ApiRequestContext) -> Any:  # type: ignore[override]
-        db = context.db
-        cacheable = not _is_today_range(self.time_range)
-        redis_client = get_redis_client_sync()
-        cache_key = None
-        if cacheable and redis_client:
-            cache_key = _build_cache_key(
-                "models",
-                self.metric,
-                self.time_range,
-                {
-                    "order": self.order,
-                    "limit": self.limit,
-                    "offset": self.offset,
-                    "provider_name": self.provider_name,
-                    "model": self.model,
-                },
-            )
-            cached = await redis_client.get(cache_key)
-            if cached:
-                try:
-                    return json.loads(cached)
-                except Exception:
-                    pass
-
-        use_daily = self.time_range is not None and not self.provider_name
-        daily_range, usage_segments = _split_daily_and_usage_segments(self.time_range, use_daily)
-
-        daily_query = None
-        if daily_range:
-            daily_query = (
-                db.query(
-                    StatsDailyModel.model.label("entity_id"),
-                    func.sum(StatsDailyModel.total_requests).label("requests"),
-                    func.sum(
-                        StatsDailyModel.input_tokens
-                        + StatsDailyModel.output_tokens
-                        + StatsDailyModel.cache_creation_tokens
-                        + StatsDailyModel.cache_read_tokens
-                    ).label("tokens"),
-                    func.sum(StatsDailyModel.total_cost).label("cost"),
-                )
-                .filter(
-                    StatsDailyModel.date >= daily_range[0], StatsDailyModel.date < daily_range[1]
-                )
-                .group_by(StatsDailyModel.model)
-            )
-            if self.model:
-                daily_query = daily_query.filter(StatsDailyModel.model == self.model)
-
-        usage_query = db.query(
-            Usage.model.label("entity_id"),
-            func.count(Usage.id).label("requests"),
-            func.sum(
-                Usage.input_tokens
-                + Usage.output_tokens
-                + Usage.cache_creation_input_tokens
-                + Usage.cache_read_input_tokens
-            ).label("tokens"),
-            func.sum(Usage.total_cost_usd).label("cost"),
-        ).filter(
-            Usage.status.notin_(["pending", "streaming"]),
-            Usage.provider_name.notin_(["unknown", "pending"]),
-        )
-        if self.provider_name:
-            usage_query = usage_query.filter(Usage.provider_name == self.provider_name)
-        if self.model:
-            usage_query = usage_query.filter(Usage.model == self.model)
-        usage_query = _apply_usage_time_segments(usage_query, usage_segments)
-        if usage_query is not None:
-            usage_query = usage_query.group_by(Usage.model)
-
-        union_query = _union_queries([daily_query, usage_query])
-        if union_query is None:
-            return {
-                "items": [],
-                "total": 0,
-                "metric": self.metric,
-                "start_date": self.time_range.start_date.isoformat() if self.time_range else None,
-                "end_date": self.time_range.end_date.isoformat() if self.time_range else None,
-            }
-
-        union_subq = union_query.subquery()
-        agg_subq = (
-            db.query(
-                union_subq.c.entity_id.label("entity_id"),
-                func.sum(union_subq.c.requests).label("requests"),
-                func.sum(union_subq.c.tokens).label("tokens"),
-                func.sum(union_subq.c.cost).label("cost"),
-            )
-            .group_by(union_subq.c.entity_id)
-            .subquery()
-        )
-
-        base_query = db.query(
-            agg_subq.c.entity_id.label("id"),
-            agg_subq.c.entity_id.label("name"),
-            agg_subq.c.requests,
-            agg_subq.c.tokens,
-            agg_subq.c.cost,
-        )
-
-        metric_expr = {
-            "requests": agg_subq.c.requests,
-            "tokens": agg_subq.c.tokens,
-            "cost": agg_subq.c.cost,
-        }[self.metric]
-        order_expr = _metric_order(self.metric, self.order, metric_expr)
-        rank_expr = func.dense_rank().over(order_by=order_expr).label("rank")
-
-        total = db.query(func.count()).select_from(base_query.subquery()).scalar() or 0
-        rows = (
-            base_query.add_columns(rank_expr, metric_expr.label("metric_value"))
-            .order_by(order_expr)
-            .offset(self.offset)
-            .limit(self.limit)
-            .all()
-        )
-
-        items = []
-        for row in rows:
-            value = row.metric_value or 0
-            if self.metric in {"requests", "tokens"}:
-                value = int(value)
-            else:
-                value = float(value)
-            items.append(
-                {
-                    "rank": int(row.rank),
-                    "id": row.id,
-                    "name": row.name,
-                    "value": value,
-                    "requests": int(row.requests or 0),
-                    "tokens": int(row.tokens or 0),
-                    "cost": float(row.cost or 0.0),
-                }
-            )
-
-        context.add_audit_metadata(
-            action="leaderboard_models",
-            start_date=self.time_range.start_date.isoformat() if self.time_range else None,
-            end_date=self.time_range.end_date.isoformat() if self.time_range else None,
-            preset=self.time_range.preset if self.time_range else None,
-            timezone=self.time_range.timezone if self.time_range else None,
-            metric=self.metric,
-            order=self.order,
-            limit=self.limit,
-            offset=self.offset,
-            provider_name=self.provider_name,
-            model=self.model,
-            result_count=len(items),
-            total=total,
-        )
-
-        result = {
-            "items": items,
-            "total": total,
-            "metric": self.metric,
-            "start_date": self.time_range.start_date.isoformat() if self.time_range else None,
-            "end_date": self.time_range.end_date.isoformat() if self.time_range else None,
-        }
-
-        if cacheable and redis_client and cache_key:
-            try:
-                await redis_client.setex(
-                    cache_key, CacheTTL.ADMIN_LEADERBOARD, json.dumps(result, ensure_ascii=False)
-                )
-            except Exception:
-                pass
-
-        return result
-
-
-@router.get("/leaderboard/users")
-async def get_user_leaderboard(
-    request: Request,
-    db: Session = Depends(get_db),
-    start_date: date | None = Query(None),
-    end_date: date | None = Query(None),
-    preset: str | None = Query(None),
-    timezone_name: str | None = Query(None, alias="timezone"),
-    tz_offset_minutes: int | None = Query(0),
-    metric: Literal["requests", "tokens", "cost"] = Query("requests"),
-    order: Literal["desc", "asc"] = Query("desc"),
-    limit: int = Query(10, ge=1, le=100),
-    offset: int = Query(0, ge=0),
-    provider_name: str | None = Query(None),
-    model: str | None = Query(None),
-    include_inactive: bool = Query(False),
-    exclude_admin: bool = Query(False),
-) -> Any:
-    time_range = _build_time_range_params(
-        start_date, end_date, preset, timezone_name, tz_offset_minutes
-    )
-    adapter = AdminUserLeaderboardAdapter(
-        time_range=time_range,
-        metric=metric,
-        order=order,
-        limit=limit,
-        offset=offset,
-        provider_name=provider_name,
-        model=model,
-        include_inactive=include_inactive,
-        exclude_admin=exclude_admin,
-    )
-    return await pipeline.run(adapter=adapter, http_request=request, db=db, mode=adapter.mode)
-
-
-@router.get("/leaderboard/api-keys")
-async def get_api_key_leaderboard(
-    request: Request,
-    db: Session = Depends(get_db),
-    start_date: date | None = Query(None),
-    end_date: date | None = Query(None),
-    preset: str | None = Query(None),
-    timezone_name: str | None = Query(None, alias="timezone"),
-    tz_offset_minutes: int | None = Query(0),
-    metric: Literal["requests", "tokens", "cost"] = Query("requests"),
-    order: Literal["desc", "asc"] = Query("desc"),
-    limit: int = Query(10, ge=1, le=100),
-    offset: int = Query(0, ge=0),
-    provider_name: str | None = Query(None),
-    model: str | None = Query(None),
-    include_inactive: bool = Query(False),
-    exclude_admin: bool = Query(False),
-) -> Any:
-    time_range = _build_time_range_params(
-        start_date, end_date, preset, timezone_name, tz_offset_minutes
-    )
-    adapter = AdminApiKeyLeaderboardAdapter(
-        time_range=time_range,
-        metric=metric,
-        order=order,
-        limit=limit,
-        offset=offset,
-        provider_name=provider_name,
-        model=model,
-        include_inactive=include_inactive,
-        exclude_admin=exclude_admin,
-    )
-    return await pipeline.run(adapter=adapter, http_request=request, db=db, mode=adapter.mode)
-
-
-@router.get("/leaderboard/models")
-async def get_model_leaderboard(
-    request: Request,
-    db: Session = Depends(get_db),
-    start_date: date | None = Query(None),
-    end_date: date | None = Query(None),
-    preset: str | None = Query(None),
-    timezone_name: str | None = Query(None, alias="timezone"),
-    tz_offset_minutes: int | None = Query(0),
-    metric: Literal["requests", "tokens", "cost"] = Query("requests"),
-    order: Literal["desc", "asc"] = Query("desc"),
-    limit: int = Query(10, ge=1, le=100),
-    offset: int = Query(0, ge=0),
-    provider_name: str | None = Query(None),
-    model: str | None = Query(None),
-) -> Any:
-    time_range = _build_time_range_params(
-        start_date, end_date, preset, timezone_name, tz_offset_minutes
-    )
-    adapter = AdminModelLeaderboardAdapter(
-        time_range=time_range,
-        metric=metric,
-        order=order,
-        limit=limit,
-        offset=offset,
-        provider_name=provider_name,
-        model=model,
-    )
-    return await pipeline.run(adapter=adapter, http_request=request, db=db, mode=adapter.mode)

src/api/admin/stats/performance.py

@@ -1,93 +0,0 @@
-"""Admin performance stats routes."""
-
-from __future__ import annotations
-
-from datetime import date, timezone
-from typing import Any
-
-from fastapi import APIRouter, Depends, Query, Request
-from sqlalchemy.orm import Session
-
-from src.api.base.admin_adapter import AdminApiAdapter
-from src.api.base.context import ApiRequestContext
-from src.config.constants import CacheTTL
-from src.database import get_db
-from src.models.database import StatsDaily
-from src.services.system.stats_aggregator import StatsAggregatorService
-from src.services.system.time_range import TimeRangeParams
-from src.utils.cache_decorator import cache_result
-
-from .common import _apply_admin_default_range, _build_time_range_params, pipeline
-
-router = APIRouter()
-
-
-class AdminPercentilesAdapter(AdminApiAdapter):
-    def __init__(self, time_range: TimeRangeParams | None) -> None:
-        self.time_range = _apply_admin_default_range(time_range)
-
-    @cache_result(
-        key_prefix="admin:stats:performance:percentiles",
-        ttl=CacheTTL.ADMIN_USAGE_AGGREGATION,
-        user_specific=False,
-        vary_by=[
-            "time_range.start_date",
-            "time_range.end_date",
-            "time_range.preset",
-            "time_range.timezone",
-            "time_range.tz_offset_minutes",
-        ],
-    )
-    async def handle(self, context: ApiRequestContext) -> Any:  # type: ignore[override]
-        if not self.time_range:
-            return []
-
-        time_range = self.time_range
-        is_utc = (time_range.timezone in {None, "UTC"}) and time_range.tz_offset_minutes == 0
-
-        if is_utc:
-            start_utc, end_utc = time_range.to_utc_datetime_range()
-            rows = (
-                context.db.query(StatsDaily)
-                .filter(StatsDaily.date >= start_utc, StatsDaily.date < end_utc)
-                .order_by(StatsDaily.date.asc())
-                .all()
-            )
-            result = []
-            for row in rows:
-                date_str = (
-                    row.date.astimezone(timezone.utc).date().isoformat()
-                    if row.date.tzinfo
-                    else row.date.date().isoformat()
-                )
-                result.append(
-                    {
-                        "date": date_str,
-                        "p50_response_time_ms": row.p50_response_time_ms,
-                        "p90_response_time_ms": row.p90_response_time_ms,
-                        "p99_response_time_ms": row.p99_response_time_ms,
-                        "p50_first_byte_time_ms": row.p50_first_byte_time_ms,
-                        "p90_first_byte_time_ms": row.p90_first_byte_time_ms,
-                        "p99_first_byte_time_ms": row.p99_first_byte_time_ms,
-                    }
-                )
-            return result
-
-        return StatsAggregatorService.compute_percentiles_by_local_day(context.db, time_range)
-
-
-@router.get("/performance/percentiles")
-async def get_percentiles(
-    request: Request,
-    db: Session = Depends(get_db),
-    start_date: date | None = Query(None),
-    end_date: date | None = Query(None),
-    preset: str | None = Query(None),
-    timezone_name: str | None = Query(None, alias="timezone"),
-    tz_offset_minutes: int | None = Query(0),
-) -> Any:
-    time_range = _build_time_range_params(
-        start_date, end_date, preset, timezone_name, tz_offset_minutes
-    )
-    adapter = AdminPercentilesAdapter(time_range=time_range)
-    return await pipeline.run(adapter=adapter, http_request=request, db=db, mode=adapter.mode)

src/api/admin/stats/quota.py

@@ -1,90 +0,0 @@
-"""Admin quota usage stats routes."""
-
-from __future__ import annotations
-
-from datetime import datetime, timedelta, timezone
-from typing import Any
-
-from fastapi import APIRouter, Depends, Request
-from sqlalchemy.orm import Session
-
-from src.api.base.admin_adapter import AdminApiAdapter
-from src.api.base.context import ApiRequestContext
-from src.config.constants import CacheTTL
-from src.core.enums import ProviderBillingType
-from src.database import get_db
-from src.models.database import Provider
-from src.utils.cache_decorator import cache_result
-
-from .common import pipeline
-
-router = APIRouter()
-
-
-class AdminQuotaUsageAdapter(AdminApiAdapter):
-    @cache_result(
-        key_prefix="admin:stats:providers:quota_usage",
-        ttl=CacheTTL.ADMIN_USAGE_AGGREGATION,
-        user_specific=False,
-    )
-    async def handle(self, context: ApiRequestContext) -> Any:  # type: ignore[override]
-        db = context.db
-        providers = (
-            db.query(Provider)
-            .filter(
-                (Provider.billing_type == ProviderBillingType.MONTHLY_QUOTA)
-                | (Provider.monthly_quota_usd.isnot(None))
-            )
-            .all()
-        )
-        now = datetime.now(timezone.utc)
-
-        result = []
-        for provider in providers:
-            quota = float(provider.monthly_quota_usd or 0)
-            used = float(provider.monthly_used_usd or 0)
-            remaining = max(quota - used, 0.0)
-            usage_percent = round((used / quota) * 100, 2) if quota > 0 else 0.0
-
-            reset_at = provider.quota_last_reset_at
-            if reset_at:
-                days_elapsed = max(1, (now - reset_at).days)
-            else:
-                days_elapsed = max(1, now.day - 1)
-
-            daily_rate = used / days_elapsed if used > 0 else 0.0
-            estimated_exhaust_at = None
-            if daily_rate > 0 and remaining > 0:
-                estimated_exhaust_at = now + timedelta(days=remaining / daily_rate)
-            if provider.quota_expires_at:
-                if not estimated_exhaust_at or provider.quota_expires_at < estimated_exhaust_at:
-                    estimated_exhaust_at = provider.quota_expires_at
-
-            result.append(
-                {
-                    "id": provider.id,
-                    "name": provider.name,
-                    "quota_usd": float(quota),
-                    "used_usd": float(used),
-                    "remaining_usd": float(remaining),
-                    "usage_percent": usage_percent,
-                    "quota_expires_at": (
-                        provider.quota_expires_at.isoformat() if provider.quota_expires_at else None
-                    ),
-                    "estimated_exhaust_at": (
-                        estimated_exhaust_at.isoformat() if estimated_exhaust_at else None
-                    ),
-                }
-            )
-
-        result.sort(key=lambda x: x["usage_percent"], reverse=True)
-        return {"providers": result}
-
-
-@router.get("/providers/quota-usage")
-async def get_quota_usage(
-    request: Request,
-    db: Session = Depends(get_db),
-) -> Any:
-    adapter = AdminQuotaUsageAdapter()
-    return await pipeline.run(adapter=adapter, http_request=request, db=db, mode=adapter.mode)