perf(gateway): 优化负载感知调度

主要改进：
- 优化负载感知调度的准确性和响应速度
- 将 AccountUsageService 的包级缓存改为依赖注入
- 修复 SSE/JSON 转义和 nil 安全问题
- 恢复 Google One 功能兼容性

backend/internal/repository/concurrency_cache.go

@@ -27,14 +27,8 @@ const (
 	accountSlotKeyPrefix = "concurrency:account:"
 	// 格式: concurrency:user:{userID}
 	userSlotKeyPrefix = "concurrency:user:"
-
-	// Wait queue keys (global structures)
-	// - total: integer total queue depth across all users
-	// - updated: sorted set of userID -> lastUpdateUnixSec (for TTL cleanup)
-	// - counts: hash of userID -> current wait count
-	waitQueueTotalKey   = "concurrency:wait:total"
-	waitQueueUpdatedKey = "concurrency:wait:updated"
-	waitQueueCountsKey  = "concurrency:wait:counts"
+	// 等待队列计数器格式: concurrency:wait:{userID}
+	waitQueueKeyPrefix = "concurrency:wait:"
 	// 账号级等待队列计数器格式: wait:account:{accountID}
 	accountWaitKeyPrefix = "wait:account:"
 
@@ -100,55 +94,27 @@ var (
 	`)
 
 	// incrementWaitScript - only sets TTL on first creation to avoid refreshing
-	// KEYS[1] = total key
-	// KEYS[2] = updated zset key
-	// KEYS[3] = counts hash key
-	// ARGV[1] = userID
-	// ARGV[2] = maxWait
-	// ARGV[3] = TTL in seconds
-	// ARGV[4] = cleanup limit
+	// KEYS[1] = wait queue key
+	// ARGV[1] = maxWait
+	// ARGV[2] = TTL in seconds
 	incrementWaitScript = redis.NewScript(`
-		local totalKey = KEYS[1]
-		local updatedKey = KEYS[2]
-		local countsKey = KEYS[3]
-
-		local userID = ARGV[1]
-		local maxWait = tonumber(ARGV[2])
-		local ttl = tonumber(ARGV[3])
-		local cleanupLimit = tonumber(ARGV[4])
-
-		redis.call('SETNX', totalKey, 0)
-
-		local timeResult = redis.call('TIME')
-		local now = tonumber(timeResult[1])
-		local expireBefore = now - ttl
-
-		-- Cleanup expired users (bounded)
-		local expired = redis.call('ZRANGEBYSCORE', updatedKey, '-inf', expireBefore, 'LIMIT', 0, cleanupLimit)
-		for _, uid in ipairs(expired) do
-			local c = tonumber(redis.call('HGET', countsKey, uid) or '0')
-			if c > 0 then
-				redis.call('DECRBY', totalKey, c)
-			end
-			redis.call('HDEL', countsKey, uid)
-			redis.call('ZREM', updatedKey, uid)
+		local current = redis.call('GET', KEYS[1])
+		if current == false then
+			current = 0
+		else
+			current = tonumber(current)
 		end
 
-		local current = tonumber(redis.call('HGET', countsKey, userID) or '0')
-		if current >= maxWait then
+		if current >= tonumber(ARGV[1]) then
 			return 0
 		end
 
-		local newVal = current + 1
-		redis.call('HSET', countsKey, userID, newVal)
-		redis.call('ZADD', updatedKey, now, userID)
-		redis.call('INCR', totalKey)
+		local newVal = redis.call('INCR', KEYS[1])
 
-		-- Keep global structures from living forever in totally idle deployments.
-		local ttlKeep = ttl * 2
-		redis.call('EXPIRE', totalKey, ttlKeep)
-		redis.call('EXPIRE', updatedKey, ttlKeep)
-		redis.call('EXPIRE', countsKey, ttlKeep)
+		-- Only set TTL on first creation to avoid refreshing zombie data
+		if newVal == 1 then
+			redis.call('EXPIRE', KEYS[1], ARGV[2])
+		end
 
 			return 1
 		`)
@@ -178,111 +144,6 @@ var (
 
 	// decrementWaitScript - same as before
 	decrementWaitScript = redis.NewScript(`
-		local totalKey = KEYS[1]
-		local updatedKey = KEYS[2]
-		local countsKey = KEYS[3]
-
-		local userID = ARGV[1]
-		local ttl = tonumber(ARGV[2])
-		local cleanupLimit = tonumber(ARGV[3])
-
-		redis.call('SETNX', totalKey, 0)
-
-		local timeResult = redis.call('TIME')
-		local now = tonumber(timeResult[1])
-		local expireBefore = now - ttl
-
-		-- Cleanup expired users (bounded)
-		local expired = redis.call('ZRANGEBYSCORE', updatedKey, '-inf', expireBefore, 'LIMIT', 0, cleanupLimit)
-		for _, uid in ipairs(expired) do
-			local c = tonumber(redis.call('HGET', countsKey, uid) or '0')
-			if c > 0 then
-				redis.call('DECRBY', totalKey, c)
-			end
-			redis.call('HDEL', countsKey, uid)
-			redis.call('ZREM', updatedKey, uid)
-		end
-
-		local current = tonumber(redis.call('HGET', countsKey, userID) or '0')
-		if current <= 0 then
-			return 1
-		end
-
-		local newVal = current - 1
-		if newVal <= 0 then
-			redis.call('HDEL', countsKey, userID)
-			redis.call('ZREM', updatedKey, userID)
-		else
-			redis.call('HSET', countsKey, userID, newVal)
-			redis.call('ZADD', updatedKey, now, userID)
-		end
-		redis.call('DECR', totalKey)
-
-		local ttlKeep = ttl * 2
-		redis.call('EXPIRE', totalKey, ttlKeep)
-		redis.call('EXPIRE', updatedKey, ttlKeep)
-		redis.call('EXPIRE', countsKey, ttlKeep)
-
-		return 1
-	`)
-
-	// getTotalWaitScript returns the global wait depth with TTL cleanup.
-	// KEYS[1] = total key
-	// KEYS[2] = updated zset key
-	// KEYS[3] = counts hash key
-	// ARGV[1] = TTL in seconds
-	// ARGV[2] = cleanup limit
-	getTotalWaitScript = redis.NewScript(`
-		local totalKey = KEYS[1]
-		local updatedKey = KEYS[2]
-		local countsKey = KEYS[3]
-
-		local ttl = tonumber(ARGV[1])
-		local cleanupLimit = tonumber(ARGV[2])
-
-		redis.call('SETNX', totalKey, 0)
-
-		local timeResult = redis.call('TIME')
-		local now = tonumber(timeResult[1])
-		local expireBefore = now - ttl
-
-		-- Cleanup expired users (bounded)
-		local expired = redis.call('ZRANGEBYSCORE', updatedKey, '-inf', expireBefore, 'LIMIT', 0, cleanupLimit)
-		for _, uid in ipairs(expired) do
-			local c = tonumber(redis.call('HGET', countsKey, uid) or '0')
-			if c > 0 then
-				redis.call('DECRBY', totalKey, c)
-			end
-			redis.call('HDEL', countsKey, uid)
-			redis.call('ZREM', updatedKey, uid)
-		end
-
-		-- If totalKey got lost but counts exist (e.g. Redis restart), recompute once.
-		local total = redis.call('GET', totalKey)
-		if total == false then
-			total = 0
-			local vals = redis.call('HVALS', countsKey)
-			for _, v in ipairs(vals) do
-				total = total + tonumber(v)
-			end
-			redis.call('SET', totalKey, total)
-		end
-
-		local ttlKeep = ttl * 2
-		redis.call('EXPIRE', totalKey, ttlKeep)
-		redis.call('EXPIRE', updatedKey, ttlKeep)
-		redis.call('EXPIRE', countsKey, ttlKeep)
-
-		local result = tonumber(redis.call('GET', totalKey) or '0')
-		if result < 0 then
-			result = 0
-			redis.call('SET', totalKey, 0)
-		end
-		return result
-	`)
-
-	// decrementAccountWaitScript - account-level wait queue decrement
-	decrementAccountWaitScript = redis.NewScript(`
 			local current = redis.call('GET', KEYS[1])
 			if current ~= false and tonumber(current) > 0 then
 				redis.call('DECR', KEYS[1])
@@ -383,9 +244,7 @@ func userSlotKey(userID int64) string {
 }
 
 func waitQueueKey(userID int64) string {
-	// Historical: per-user string keys were used.
-	// Now we use global structures keyed by userID string.
-	return strconv.FormatInt(userID, 10)
+	return fmt.Sprintf("%s%d", waitQueueKeyPrefix, userID)
 }
 
 func accountWaitKey(accountID int64) string {
@@ -449,16 +308,8 @@ func (c *concurrencyCache) GetUserConcurrency(ctx context.Context, userID int64)
 // Wait queue operations
 
 func (c *concurrencyCache) IncrementWaitCount(ctx context.Context, userID int64, maxWait int) (bool, error) {
-	userKey := waitQueueKey(userID)
-	result, err := incrementWaitScript.Run(
-		ctx,
-		c.rdb,
-		[]string{waitQueueTotalKey, waitQueueUpdatedKey, waitQueueCountsKey},
-		userKey,
-		maxWait,
-		c.waitQueueTTLSeconds,
-		200, // cleanup limit per call
-	).Int()
+	key := waitQueueKey(userID)
+	result, err := incrementWaitScript.Run(ctx, c.rdb, []string{key}, maxWait, c.waitQueueTTLSeconds).Int()
 	if err != nil {
 		return false, err
 	}
@@ -466,35 +317,11 @@ func (c *concurrencyCache) IncrementWaitCount(ctx context.Context, userID int64,
 }
 
 func (c *concurrencyCache) DecrementWaitCount(ctx context.Context, userID int64) error {
-	userKey := waitQueueKey(userID)
-	_, err := decrementWaitScript.Run(
-		ctx,
-		c.rdb,
-		[]string{waitQueueTotalKey, waitQueueUpdatedKey, waitQueueCountsKey},
-		userKey,
-		c.waitQueueTTLSeconds,
-		200, // cleanup limit per call
-	).Result()
+	key := waitQueueKey(userID)
+	_, err := decrementWaitScript.Run(ctx, c.rdb, []string{key}).Result()
 	return err
 }
 
-func (c *concurrencyCache) GetTotalWaitCount(ctx context.Context) (int, error) {
-	if c.rdb == nil {
-		return 0, nil
-	}
-	total, err := getTotalWaitScript.Run(
-		ctx,
-		c.rdb,
-		[]string{waitQueueTotalKey, waitQueueUpdatedKey, waitQueueCountsKey},
-		c.waitQueueTTLSeconds,
-		500, // cleanup limit per query (rare)
-	).Int64()
-	if err != nil {
-		return 0, err
-	}
-	return int(total), nil
-}
-
 // Account wait queue operations
 
 func (c *concurrencyCache) IncrementAccountWaitCount(ctx context.Context, accountID int64, maxWait int) (bool, error) {
@@ -508,7 +335,7 @@ func (c *concurrencyCache) IncrementAccountWaitCount(ctx context.Context, accoun
 
 func (c *concurrencyCache) DecrementAccountWaitCount(ctx context.Context, accountID int64) error {
 	key := accountWaitKey(accountID)
-	_, err := decrementAccountWaitScript.Run(ctx, c.rdb, []string{key}).Result()
+	_, err := decrementWaitScript.Run(ctx, c.rdb, []string{key}).Result()
 	return err
 }
 

backend/internal/repository/concurrency_cache_integration_test.go

@@ -158,7 +158,7 @@ func (s *ConcurrencyCacheSuite) TestUserSlot_TTL() {
 
 func (s *ConcurrencyCacheSuite) TestWaitQueue_IncrementAndDecrement() {
 	userID := int64(20)
-	userKey := waitQueueKey(userID)
+	waitKey := fmt.Sprintf("%s%d", waitQueueKeyPrefix, userID)
 
 	ok, err := s.cache.IncrementWaitCount(s.ctx, userID, 2)
 	require.NoError(s.T(), err, "IncrementWaitCount 1")
@@ -172,31 +172,31 @@ func (s *ConcurrencyCacheSuite) TestWaitQueue_IncrementAndDecrement() {
 	require.NoError(s.T(), err, "IncrementWaitCount 3")
 	require.False(s.T(), ok, "expected wait increment over max to fail")
 
-	ttl, err := s.rdb.TTL(s.ctx, waitQueueTotalKey).Result()
-	require.NoError(s.T(), err, "TTL wait total key")
-	s.AssertTTLWithin(ttl, 1*time.Second, testSlotTTL*2)
+	ttl, err := s.rdb.TTL(s.ctx, waitKey).Result()
+	require.NoError(s.T(), err, "TTL waitKey")
+	s.AssertTTLWithin(ttl, 1*time.Second, testSlotTTL)
 
 	require.NoError(s.T(), s.cache.DecrementWaitCount(s.ctx, userID), "DecrementWaitCount")
 
-	val, err := s.rdb.HGet(s.ctx, waitQueueCountsKey, userKey).Int()
-	require.NoError(s.T(), err, "HGET wait queue count")
+	val, err := s.rdb.Get(s.ctx, waitKey).Int()
+	if !errors.Is(err, redis.Nil) {
+		require.NoError(s.T(), err, "Get waitKey")
+	}
 	require.Equal(s.T(), 1, val, "expected wait count 1")
-
-	total, err := s.rdb.Get(s.ctx, waitQueueTotalKey).Int()
-	require.NoError(s.T(), err, "GET wait queue total")
-	require.Equal(s.T(), 1, total, "expected total wait count 1")
 }
 
 func (s *ConcurrencyCacheSuite) TestWaitQueue_DecrementNoNegative() {
 	userID := int64(300)
-	userKey := waitQueueKey(userID)
+	waitKey := fmt.Sprintf("%s%d", waitQueueKeyPrefix, userID)
 
 	// Test decrement on non-existent key - should not error and should not create negative value
 	require.NoError(s.T(), s.cache.DecrementWaitCount(s.ctx, userID), "DecrementWaitCount on non-existent key")
 
-	// Verify count remains zero / absent.
-	val, err := s.rdb.HGet(s.ctx, waitQueueCountsKey, userKey).Int()
-	require.True(s.T(), errors.Is(err, redis.Nil))
+	// Verify no key was created or it's not negative
+	val, err := s.rdb.Get(s.ctx, waitKey).Int()
+	if !errors.Is(err, redis.Nil) {
+		require.NoError(s.T(), err, "Get waitKey")
+	}
 	require.GreaterOrEqual(s.T(), val, 0, "expected non-negative wait count after decrement on empty")
 
 	// Set count to 1, then decrement twice
@@ -210,15 +210,12 @@ func (s *ConcurrencyCacheSuite) TestWaitQueue_DecrementNoNegative() {
 	// Decrement again on 0 - should not go negative
 	require.NoError(s.T(), s.cache.DecrementWaitCount(s.ctx, userID), "DecrementWaitCount on zero")
 
-	// Verify per-user count is absent and total is non-negative.
-	_, err = s.rdb.HGet(s.ctx, waitQueueCountsKey, userKey).Result()
-	require.True(s.T(), errors.Is(err, redis.Nil), "expected count field removed on zero")
-
-	total, err := s.rdb.Get(s.ctx, waitQueueTotalKey).Int()
+	// Verify count is 0, not negative
+	val, err = s.rdb.Get(s.ctx, waitKey).Int()
 	if !errors.Is(err, redis.Nil) {
-		require.NoError(s.T(), err)
+		require.NoError(s.T(), err, "Get waitKey after double decrement")
 	}
-	require.GreaterOrEqual(s.T(), total, 0, "expected non-negative total wait count")
+	require.GreaterOrEqual(s.T(), val, 0, "expected non-negative wait count")
 }
 
 func (s *ConcurrencyCacheSuite) TestAccountWaitQueue_IncrementAndDecrement() {

backend/internal/service/account_usage_service.go

@@ -4,9 +4,11 @@ import (
 	"context"
 	"fmt"
 	"log"
+	"strings"
 	"sync"
 	"time"
 
+	"github.com/Wei-Shaw/sub2api/internal/pkg/antigravity"
 	"github.com/Wei-Shaw/sub2api/internal/pkg/pagination"
 	"github.com/Wei-Shaw/sub2api/internal/pkg/usagestats"
 )
@@ -17,11 +19,11 @@ type UsageLogRepository interface {
 	Delete(ctx context.Context, id int64) error
 
 	ListByUser(ctx context.Context, userID int64, params pagination.PaginationParams) ([]UsageLog, *pagination.PaginationResult, error)
-	ListByAPIKey(ctx context.Context, apiKeyID int64, params pagination.PaginationParams) ([]UsageLog, *pagination.PaginationResult, error)
+	ListByApiKey(ctx context.Context, apiKeyID int64, params pagination.PaginationParams) ([]UsageLog, *pagination.PaginationResult, error)
 	ListByAccount(ctx context.Context, accountID int64, params pagination.PaginationParams) ([]UsageLog, *pagination.PaginationResult, error)
 
 	ListByUserAndTimeRange(ctx context.Context, userID int64, startTime, endTime time.Time) ([]UsageLog, *pagination.PaginationResult, error)
-	ListByAPIKeyAndTimeRange(ctx context.Context, apiKeyID int64, startTime, endTime time.Time) ([]UsageLog, *pagination.PaginationResult, error)
+	ListByApiKeyAndTimeRange(ctx context.Context, apiKeyID int64, startTime, endTime time.Time) ([]UsageLog, *pagination.PaginationResult, error)
 	ListByAccountAndTimeRange(ctx context.Context, accountID int64, startTime, endTime time.Time) ([]UsageLog, *pagination.PaginationResult, error)
 	ListByModelAndTimeRange(ctx context.Context, modelName string, startTime, endTime time.Time) ([]UsageLog, *pagination.PaginationResult, error)
 
@@ -32,10 +34,10 @@ type UsageLogRepository interface {
 	GetDashboardStats(ctx context.Context) (*usagestats.DashboardStats, error)
 	GetUsageTrendWithFilters(ctx context.Context, startTime, endTime time.Time, granularity string, userID, apiKeyID int64) ([]usagestats.TrendDataPoint, error)
 	GetModelStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID int64) ([]usagestats.ModelStat, error)
-	GetAPIKeyUsageTrend(ctx context.Context, startTime, endTime time.Time, granularity string, limit int) ([]usagestats.APIKeyUsageTrendPoint, error)
+	GetApiKeyUsageTrend(ctx context.Context, startTime, endTime time.Time, granularity string, limit int) ([]usagestats.ApiKeyUsageTrendPoint, error)
 	GetUserUsageTrend(ctx context.Context, startTime, endTime time.Time, granularity string, limit int) ([]usagestats.UserUsageTrendPoint, error)
 	GetBatchUserUsageStats(ctx context.Context, userIDs []int64) (map[int64]*usagestats.BatchUserUsageStats, error)
-	GetBatchAPIKeyUsageStats(ctx context.Context, apiKeyIDs []int64) (map[int64]*usagestats.BatchAPIKeyUsageStats, error)
+	GetBatchApiKeyUsageStats(ctx context.Context, apiKeyIDs []int64) (map[int64]*usagestats.BatchApiKeyUsageStats, error)
 
 	// User dashboard stats
 	GetUserDashboardStats(ctx context.Context, userID int64) (*usagestats.UserDashboardStats, error)
@@ -51,7 +53,7 @@ type UsageLogRepository interface {
 
 	// Aggregated stats (optimized)
 	GetUserStatsAggregated(ctx context.Context, userID int64, startTime, endTime time.Time) (*usagestats.UsageStats, error)
-	GetAPIKeyStatsAggregated(ctx context.Context, apiKeyID int64, startTime, endTime time.Time) (*usagestats.UsageStats, error)
+	GetApiKeyStatsAggregated(ctx context.Context, apiKeyID int64, startTime, endTime time.Time) (*usagestats.UsageStats, error)
 	GetAccountStatsAggregated(ctx context.Context, accountID int64, startTime, endTime time.Time) (*usagestats.UsageStats, error)
 	GetModelStatsAggregated(ctx context.Context, modelName string, startTime, endTime time.Time) (*usagestats.UsageStats, error)
 	GetDailyStatsAggregated(ctx context.Context, userID int64, startTime, endTime time.Time) ([]map[string]any, error)
@@ -69,13 +71,33 @@ type windowStatsCache struct {
 	timestamp time.Time
 }
 
-var (
-	apiCacheMap         = sync.Map{} // 缓存 API 响应
-	windowStatsCacheMap = sync.Map{} // 缓存窗口统计
+// antigravityUsageCache 缓存 Antigravity 额度数据
+type antigravityUsageCache struct {
+	usageInfo *UsageInfo
+	timestamp time.Time
+}
+
+const (
 	apiCacheTTL         = 10 * time.Minute
 	windowStatsCacheTTL = 1 * time.Minute
 )
 
+// UsageCache 封装账户使用量相关的缓存
+type UsageCache struct {
+	apiCache         *sync.Map // accountID -> *apiUsageCache
+	windowStatsCache *sync.Map // accountID -> *windowStatsCache
+	antigravityCache *sync.Map // accountID -> *antigravityUsageCache
+}
+
+// NewUsageCache 创建 UsageCache 实例
+func NewUsageCache() *UsageCache {
+	return &UsageCache{
+		apiCache:         &sync.Map{},
+		antigravityCache: &sync.Map{},
+		windowStatsCache: &sync.Map{},
+	}
+}
+
 // WindowStats 窗口期统计
 type WindowStats struct {
 	Requests int64   `json:"requests"`
@@ -91,6 +113,12 @@ type UsageProgress struct {
 	WindowStats      *WindowStats `json:"window_stats,omitempty"` // 窗口期统计（从窗口开始到当前的使用量）
 }
 
+// AntigravityModelQuota Antigravity 单个模型的配额信息
+type AntigravityModelQuota struct {
+	Utilization int    `json:"utilization"` // 使用率 0-100
+	ResetTime   string `json:"reset_time"`  // 重置时间 ISO8601
+}
+
 // UsageInfo 账号使用量信息
 type UsageInfo struct {
 	UpdatedAt        *time.Time     `json:"updated_at,omitempty"`         // 更新时间
@@ -99,6 +127,9 @@ type UsageInfo struct {
 	SevenDaySonnet   *UsageProgress `json:"seven_day_sonnet,omitempty"`   // 7天Sonnet窗口
 	GeminiProDaily   *UsageProgress `json:"gemini_pro_daily,omitempty"`   // Gemini Pro 日配额
 	GeminiFlashDaily *UsageProgress `json:"gemini_flash_daily,omitempty"` // Gemini Flash 日配额
+
+	// Antigravity 多模型配额
+	AntigravityQuota map[string]*AntigravityModelQuota `json:"antigravity_quota,omitempty"`
 }
 
 // ClaudeUsageResponse Anthropic API返回的usage结构
@@ -124,19 +155,51 @@ type ClaudeUsageFetcher interface {
 
 // AccountUsageService 账号使用量查询服务
 type AccountUsageService struct {
-	accountRepo        AccountRepository
-	usageLogRepo       UsageLogRepository
-	usageFetcher       ClaudeUsageFetcher
-	geminiQuotaService *GeminiQuotaService
+	accountRepo             AccountRepository
+	usageLogRepo            UsageLogRepository
+	usageFetcher            ClaudeUsageFetcher
+	geminiQuotaService      *GeminiQuotaService
+	antigravityQuotaFetcher QuotaFetcher
+	cache                   *UsageCache
 }
 
 // NewAccountUsageService 创建AccountUsageService实例
-func NewAccountUsageService(accountRepo AccountRepository, usageLogRepo UsageLogRepository, usageFetcher ClaudeUsageFetcher, geminiQuotaService *GeminiQuotaService) *AccountUsageService {
+func NewAccountUsageService(
+	accountRepo AccountRepository,
+	usageLogRepo UsageLogRepository,
+	usageFetcher ClaudeUsageFetcher,
+	geminiQuotaService *GeminiQuotaService,
+	antigravityQuotaFetcher *AntigravityQuotaFetcher,
+	cache *UsageCache,
+) *AccountUsageService {
+	if cache == nil {
+		cache = &UsageCache{
+			apiCache:         &sync.Map{},
+			antigravityCache: &sync.Map{},
+			windowStatsCache: &sync.Map{},
+		}
+	}
+	if cache.apiCache == nil {
+		cache.apiCache = &sync.Map{}
+	}
+	if cache.antigravityCache == nil {
+		cache.antigravityCache = &sync.Map{}
+	}
+	if cache.windowStatsCache == nil {
+		cache.windowStatsCache = &sync.Map{}
+	}
+
+	var quotaFetcher QuotaFetcher
+	if antigravityQuotaFetcher != nil {
+		quotaFetcher = antigravityQuotaFetcher
+	}
 	return &AccountUsageService{
-		accountRepo:        accountRepo,
-		usageLogRepo:       usageLogRepo,
-		usageFetcher:       usageFetcher,
-		geminiQuotaService: geminiQuotaService,
+		accountRepo:             accountRepo,
+		usageLogRepo:            usageLogRepo,
+		usageFetcher:            usageFetcher,
+		geminiQuotaService:      geminiQuotaService,
+		antigravityQuotaFetcher: quotaFetcher,
+		cache:                   cache,
 	}
 }
 
@@ -154,12 +217,17 @@ func (s *AccountUsageService) GetUsage(ctx context.Context, accountID int64) (*U
 		return s.getGeminiUsage(ctx, account)
 	}
 
+	// Antigravity 平台：使用 AntigravityQuotaFetcher 获取额度
+	if account.Platform == PlatformAntigravity {
+		return s.getAntigravityUsage(ctx, account)
+	}
+
 	// 只有oauth类型账号可以通过API获取usage（有profile scope）
 	if account.CanGetUsage() {
 		var apiResp *ClaudeUsageResponse
 
 		// 1. 检查 API 缓存（10 分钟）
-		if cached, ok := apiCacheMap.Load(accountID); ok {
+		if cached, ok := s.cache.apiCache.Load(accountID); ok {
 			if cache, ok := cached.(*apiUsageCache); ok && time.Since(cache.timestamp) < apiCacheTTL {
 				apiResp = cache.response
 			}
@@ -172,7 +240,7 @@ func (s *AccountUsageService) GetUsage(ctx context.Context, accountID int64) (*U
 				return nil, err
 			}
 			// 缓存 API 响应
-			apiCacheMap.Store(accountID, &apiUsageCache{
+			s.cache.apiCache.Store(accountID, &apiUsageCache{
 				response:  apiResp,
 				timestamp: time.Now(),
 			})
@@ -224,12 +292,70 @@ func (s *AccountUsageService) getGeminiUsage(ctx context.Context, account *Accou
 	totals := geminiAggregateUsage(stats)
 	resetAt := geminiDailyResetTime(now)
 
-	usage.GeminiProDaily = buildGeminiUsageProgress(totals.ProRequests, quota.ProRPD, resetAt, totals.ProTokens, totals.ProCost, now)
-	usage.GeminiFlashDaily = buildGeminiUsageProgress(totals.FlashRequests, quota.FlashRPD, resetAt, totals.FlashTokens, totals.FlashCost, now)
+	usage.GeminiProDaily = buildGeminiUsageProgress(totals.ProRequests, quota.ProRPD, resetAt, totals.ProTokens, totals.ProCost)
+	usage.GeminiFlashDaily = buildGeminiUsageProgress(totals.FlashRequests, quota.FlashRPD, resetAt, totals.FlashTokens, totals.FlashCost)
 
 	return usage, nil
 }
 
+// getAntigravityUsage 获取 Antigravity 账户额度
+func (s *AccountUsageService) getAntigravityUsage(ctx context.Context, account *Account) (*UsageInfo, error) {
+	if s.antigravityQuotaFetcher == nil || !s.antigravityQuotaFetcher.CanFetch(account) {
+		now := time.Now()
+		return &UsageInfo{UpdatedAt: &now}, nil
+	}
+
+	// Ensure project_id is stable for quota queries.
+	if strings.TrimSpace(account.GetCredential("project_id")) == "" {
+		projectID := antigravity.GenerateMockProjectID()
+		if account.Credentials == nil {
+			account.Credentials = map[string]any{}
+		}
+		account.Credentials["project_id"] = projectID
+		if s.accountRepo != nil {
+			_, err := s.accountRepo.BulkUpdate(ctx, []int64{account.ID}, AccountBulkUpdate{
+				Credentials: map[string]any{"project_id": projectID},
+			})
+			if err != nil {
+				log.Printf("Failed to persist antigravity project_id for account %d: %v", account.ID, err)
+			}
+		}
+	}
+
+	// 1. 检查缓存（10 分钟）
+	if cached, ok := s.cache.antigravityCache.Load(account.ID); ok {
+		if cache, ok := cached.(*antigravityUsageCache); ok && time.Since(cache.timestamp) < apiCacheTTL {
+			// 重新计算 RemainingSeconds
+			usage := cloneUsageInfo(cache.usageInfo)
+			if usage.FiveHour != nil && usage.FiveHour.ResetsAt != nil {
+				usage.FiveHour.RemainingSeconds = remainingSecondsUntil(*usage.FiveHour.ResetsAt)
+			}
+			return usage, nil
+		}
+	}
+
+	// 2. 获取代理 URL
+	proxyURL, err := s.antigravityQuotaFetcher.GetProxyURL(ctx, account)
+	if err != nil {
+		log.Printf("Failed to get proxy URL for account %d: %v", account.ID, err)
+		proxyURL = ""
+	}
+
+	// 3. 调用 API 获取额度
+	result, err := s.antigravityQuotaFetcher.FetchQuota(ctx, account, proxyURL)
+	if err != nil {
+		return nil, fmt.Errorf("fetch antigravity quota failed: %w", err)
+	}
+
+	// 4. 缓存结果
+	s.cache.antigravityCache.Store(account.ID, &antigravityUsageCache{
+		usageInfo: result.UsageInfo,
+		timestamp: time.Now(),
+	})
+
+	return result.UsageInfo, nil
+}
+
 // addWindowStats 为 usage 数据添加窗口期统计
 // 使用独立缓存（1 分钟），与 API 缓存分离
 func (s *AccountUsageService) addWindowStats(ctx context.Context, account *Account, usage *UsageInfo) {
@@ -241,7 +367,7 @@ func (s *AccountUsageService) addWindowStats(ctx context.Context, account *Accou
 
 	// 检查窗口统计缓存（1 分钟）
 	var windowStats *WindowStats
-	if cached, ok := windowStatsCacheMap.Load(account.ID); ok {
+	if cached, ok := s.cache.windowStatsCache.Load(account.ID); ok {
 		if cache, ok := cached.(*windowStatsCache); ok && time.Since(cache.timestamp) < windowStatsCacheTTL {
 			windowStats = cache.stats
 		}
@@ -269,7 +395,7 @@ func (s *AccountUsageService) addWindowStats(ctx context.Context, account *Accou
 		}
 
 		// 缓存窗口统计（1 分钟）
-		windowStatsCacheMap.Store(account.ID, &windowStatsCache{
+		s.cache.windowStatsCache.Store(account.ID, &windowStatsCache{
 			stats:     windowStats,
 			timestamp: time.Now(),
 		})
@@ -342,12 +468,12 @@ func (s *AccountUsageService) buildUsageInfo(resp *ClaudeUsageResponse, updatedA
 
 	// 5小时窗口 - 始终创建对象（即使 ResetsAt 为空）
 	info.FiveHour = &UsageProgress{
-		Utilization: resp.FiveHour.Utilization,
+		Utilization: clampFloat64(resp.FiveHour.Utilization, 0, 100),
 	}
 	if resp.FiveHour.ResetsAt != "" {
 		if fiveHourReset, err := parseTime(resp.FiveHour.ResetsAt); err == nil {
 			info.FiveHour.ResetsAt = &fiveHourReset
-			info.FiveHour.RemainingSeconds = int(time.Until(fiveHourReset).Seconds())
+			info.FiveHour.RemainingSeconds = remainingSecondsUntil(fiveHourReset)
 		} else {
 			log.Printf("Failed to parse FiveHour.ResetsAt: %s, error: %v", resp.FiveHour.ResetsAt, err)
 		}
@@ -357,14 +483,14 @@ func (s *AccountUsageService) buildUsageInfo(resp *ClaudeUsageResponse, updatedA
 	if resp.SevenDay.ResetsAt != "" {
 		if sevenDayReset, err := parseTime(resp.SevenDay.ResetsAt); err == nil {
 			info.SevenDay = &UsageProgress{
-				Utilization:      resp.SevenDay.Utilization,
+				Utilization:      clampFloat64(resp.SevenDay.Utilization, 0, 100),
 				ResetsAt:         &sevenDayReset,
-				RemainingSeconds: int(time.Until(sevenDayReset).Seconds()),
+				RemainingSeconds: remainingSecondsUntil(sevenDayReset),
 			}
 		} else {
 			log.Printf("Failed to parse SevenDay.ResetsAt: %s, error: %v", resp.SevenDay.ResetsAt, err)
 			info.SevenDay = &UsageProgress{
-				Utilization: resp.SevenDay.Utilization,
+				Utilization: clampFloat64(resp.SevenDay.Utilization, 0, 100),
 			}
 		}
 	}
@@ -373,14 +499,14 @@ func (s *AccountUsageService) buildUsageInfo(resp *ClaudeUsageResponse, updatedA
 	if resp.SevenDaySonnet.ResetsAt != "" {
 		if sonnetReset, err := parseTime(resp.SevenDaySonnet.ResetsAt); err == nil {
 			info.SevenDaySonnet = &UsageProgress{
-				Utilization:      resp.SevenDaySonnet.Utilization,
+				Utilization:      clampFloat64(resp.SevenDaySonnet.Utilization, 0, 100),
 				ResetsAt:         &sonnetReset,
-				RemainingSeconds: int(time.Until(sonnetReset).Seconds()),
+				RemainingSeconds: remainingSecondsUntil(sonnetReset),
 			}
 		} else {
 			log.Printf("Failed to parse SevenDaySonnet.ResetsAt: %s, error: %v", resp.SevenDaySonnet.ResetsAt, err)
 			info.SevenDaySonnet = &UsageProgress{
-				Utilization: resp.SevenDaySonnet.Utilization,
+				Utilization: clampFloat64(resp.SevenDaySonnet.Utilization, 0, 100),
 			}
 		}
 	}
@@ -394,10 +520,7 @@ func (s *AccountUsageService) estimateSetupTokenUsage(account *Account) *UsageIn
 
 	// 如果有session_window信息
 	if account.SessionWindowEnd != nil {
-		remaining := int(time.Until(*account.SessionWindowEnd).Seconds())
-		if remaining < 0 {
-			remaining = 0
-		}
+		remaining := remainingSecondsUntil(*account.SessionWindowEnd)
 
 		// 根据状态估算使用率 (百分比形式，100 = 100%)
 		var utilization float64
@@ -409,6 +532,7 @@ func (s *AccountUsageService) estimateSetupTokenUsage(account *Account) *UsageIn
 		default:
 			utilization = 0.0
 		}
+		utilization = clampFloat64(utilization, 0, 100)
 
 		info.FiveHour = &UsageProgress{
 			Utilization:      utilization,
@@ -427,15 +551,12 @@ func (s *AccountUsageService) estimateSetupTokenUsage(account *Account) *UsageIn
 	return info
 }
 
-func buildGeminiUsageProgress(used, limit int64, resetAt time.Time, tokens int64, cost float64, now time.Time) *UsageProgress {
+func buildGeminiUsageProgress(used, limit int64, resetAt time.Time, tokens int64, cost float64) *UsageProgress {
 	if limit <= 0 {
 		return nil
 	}
-	utilization := (float64(used) / float64(limit)) * 100
-	remainingSeconds := int(resetAt.Sub(now).Seconds())
-	if remainingSeconds < 0 {
-		remainingSeconds = 0
-	}
+	utilization := clampFloat64((float64(used)/float64(limit))*100, 0, 100)
+	remainingSeconds := remainingSecondsUntil(resetAt)
 	resetCopy := resetAt
 	return &UsageProgress{
 		Utilization:      utilization,
@@ -448,3 +569,47 @@ func buildGeminiUsageProgress(used, limit int64, resetAt time.Time, tokens int64
 		},
 	}
 }
+
+func cloneUsageInfo(src *UsageInfo) *UsageInfo {
+	if src == nil {
+		return nil
+	}
+	dst := *src
+	if src.UpdatedAt != nil {
+		t := *src.UpdatedAt
+		dst.UpdatedAt = &t
+	}
+	dst.FiveHour = cloneUsageProgress(src.FiveHour)
+	dst.SevenDay = cloneUsageProgress(src.SevenDay)
+	dst.SevenDaySonnet = cloneUsageProgress(src.SevenDaySonnet)
+	dst.GeminiProDaily = cloneUsageProgress(src.GeminiProDaily)
+	dst.GeminiFlashDaily = cloneUsageProgress(src.GeminiFlashDaily)
+	if src.AntigravityQuota != nil {
+		dst.AntigravityQuota = make(map[string]*AntigravityModelQuota, len(src.AntigravityQuota))
+		for k, v := range src.AntigravityQuota {
+			if v == nil {
+				dst.AntigravityQuota[k] = nil
+				continue
+			}
+			copyVal := *v
+			dst.AntigravityQuota[k] = &copyVal
+		}
+	}
+	return &dst
+}
+
+func cloneUsageProgress(src *UsageProgress) *UsageProgress {
+	if src == nil {
+		return nil
+	}
+	dst := *src
+	if src.ResetsAt != nil {
+		t := *src.ResetsAt
+		dst.ResetsAt = &t
+	}
+	if src.WindowStats != nil {
+		statsCopy := *src.WindowStats
+		dst.WindowStats = &statsCopy
+	}
+	return &dst
+}

backend/internal/service/concurrency_service.go

@@ -32,7 +32,6 @@ type ConcurrencyCache interface {
 	// 等待队列计数（只在首次创建时设置 TTL）
 	IncrementWaitCount(ctx context.Context, userID int64, maxWait int) (bool, error)
 	DecrementWaitCount(ctx context.Context, userID int64) error
-	GetTotalWaitCount(ctx context.Context) (int, error)
 
 	// 批量负载查询（只读）
 	GetAccountsLoadBatch(ctx context.Context, accounts []AccountWithConcurrency) (map[int64]*AccountLoadInfo, error)
@@ -201,14 +200,6 @@ func (s *ConcurrencyService) DecrementWaitCount(ctx context.Context, userID int6
 	}
 }
 
-// GetTotalWaitCount returns the total wait queue depth across users.
-func (s *ConcurrencyService) GetTotalWaitCount(ctx context.Context) (int, error) {
-	if s.cache == nil {
-		return 0, nil
-	}
-	return s.cache.GetTotalWaitCount(ctx)
-}
-
 // IncrementAccountWaitCount increments the wait queue counter for an account.
 func (s *ConcurrencyService) IncrementAccountWaitCount(ctx context.Context, accountID int64, maxWait int) (bool, error) {
 	if s.cache == nil {

backend/internal/service/gateway_multiplatform_test.go

@@ -136,6 +136,12 @@ func (m *mockAccountRepoForPlatform) SetRateLimited(ctx context.Context, id int6
 func (m *mockAccountRepoForPlatform) SetOverloaded(ctx context.Context, id int64, until time.Time) error {
 	return nil
 }
+func (m *mockAccountRepoForPlatform) SetTempUnschedulable(ctx context.Context, id int64, until time.Time, reason string) error {
+	return nil
+}
+func (m *mockAccountRepoForPlatform) ClearTempUnschedulable(ctx context.Context, id int64) error {
+	return nil
+}
 func (m *mockAccountRepoForPlatform) ClearRateLimit(ctx context.Context, id int64) error {
 	return nil
 }
@@ -276,7 +282,7 @@ func TestGatewayService_SelectAccountForModelWithPlatform_GeminiOAuthPreference(
 
 	repo := &mockAccountRepoForPlatform{
 		accounts: []Account{
-			{ID: 1, Platform: PlatformGemini, Priority: 1, Status: StatusActive, Schedulable: true, Type: AccountTypeAPIKey},
+			{ID: 1, Platform: PlatformGemini, Priority: 1, Status: StatusActive, Schedulable: true, Type: AccountTypeApiKey},
 			{ID: 2, Platform: PlatformGemini, Priority: 1, Status: StatusActive, Schedulable: true, Type: AccountTypeOAuth},
 		},
 		accountsByID: map[int64]*Account{},
@@ -617,7 +623,7 @@ func TestGatewayService_selectAccountWithMixedScheduling(t *testing.T) {
 	t.Run("混合调度-Gemini优先选择OAuth账户", func(t *testing.T) {
 		repo := &mockAccountRepoForPlatform{
 			accounts: []Account{
-				{ID: 1, Platform: PlatformGemini, Priority: 1, Status: StatusActive, Schedulable: true, Type: AccountTypeAPIKey},
+				{ID: 1, Platform: PlatformGemini, Priority: 1, Status: StatusActive, Schedulable: true, Type: AccountTypeApiKey},
 				{ID: 2, Platform: PlatformGemini, Priority: 1, Status: StatusActive, Schedulable: true, Type: AccountTypeOAuth},
 			},
 			accountsByID: map[int64]*Account{},

backend/internal/service/gateway_request.go

@@ -75,9 +75,19 @@ func ParseGatewayRequest(body []byte) (*ParsedRequest, error) {
 // FilterThinkingBlocks removes thinking blocks from request body
 // Returns filtered body or original body if filtering fails (fail-safe)
 // This prevents 400 errors from invalid thinking block signatures
+//
+// Strategy:
+//   - When thinking.type != "enabled": Remove all thinking blocks
+//   - When thinking.type == "enabled": Only remove thinking blocks without valid signatures
+//     (blocks with missing/empty/dummy signatures that would cause 400 errors)
 func FilterThinkingBlocks(body []byte) []byte {
 	// Fast path: if body doesn't contain "thinking", skip parsing
-	if !bytes.Contains(body, []byte("thinking")) {
+	if !bytes.Contains(body, []byte(`"type":"thinking"`)) &&
+		!bytes.Contains(body, []byte(`"type": "thinking"`)) &&
+		!bytes.Contains(body, []byte(`"type":"redacted_thinking"`)) &&
+		!bytes.Contains(body, []byte(`"type": "redacted_thinking"`)) &&
+		!bytes.Contains(body, []byte(`"thinking":`)) &&
+		!bytes.Contains(body, []byte(`"thinking" :`)) {
 		return body
 	}
 
@@ -86,6 +96,14 @@ func FilterThinkingBlocks(body []byte) []byte {
 		return body // Return original on parse error
 	}
 
+	// Check if thinking is enabled
+	thinkingEnabled := false
+	if thinking, ok := req["thinking"].(map[string]any); ok {
+		if thinkType, ok := thinking["type"].(string); ok && thinkType == "enabled" {
+			thinkingEnabled = true
+		}
+	}
+
 	messages, ok := req["messages"].([]any)
 	if !ok {
 		return body // No messages array
@@ -98,6 +116,7 @@ func FilterThinkingBlocks(body []byte) []byte {
 			continue
 		}
 
+		role, _ := msgMap["role"].(string)
 		content, ok := msgMap["content"].([]any)
 		if !ok {
 			continue
@@ -106,6 +125,7 @@ func FilterThinkingBlocks(body []byte) []byte {
 		// Filter thinking blocks from content array
 		newContent := make([]any, 0, len(content))
 		filteredThisMessage := false
+
 		for _, block := range content {
 			blockMap, ok := block.(map[string]any)
 			if !ok {
@@ -114,22 +134,34 @@ func FilterThinkingBlocks(body []byte) []byte {
 			}
 
 			blockType, _ := blockMap["type"].(string)
-			// Explicit Anthropic-style thinking block: {"type":"thinking", ...}
-			if blockType == "thinking" {
+
+			// Handle thinking/redacted_thinking blocks
+			if blockType == "thinking" || blockType == "redacted_thinking" {
+				// When thinking is enabled and this is an assistant message,
+				// only keep thinking blocks with valid (non-empty, non-dummy) signatures
+				if thinkingEnabled && role == "assistant" {
+					signature, _ := blockMap["signature"].(string)
+					// Keep blocks with valid signatures, remove those without
+					if signature != "" && signature != "skip_thought_signature_validator" {
+						newContent = append(newContent, block)
+						continue
+					}
+				}
+
 				filtered = true
 				filteredThisMessage = true
-				continue // Skip thinking blocks
+				continue
 			}
 
 			// Some clients send the "thinking" object without a "type" discriminator.
-			// Vertex/Claude still expects a signature for any thinking block, so we drop it.
 			// We intentionally do not drop other typed blocks (e.g. tool_use) that might
 			// legitimately contain a "thinking" key inside their payload.
 			if blockType == "" {
-				if _, hasThinking := blockMap["thinking"]; hasThinking {
+				if thinkingContent, hasThinking := blockMap["thinking"]; hasThinking {
+					_ = thinkingContent
 					filtered = true
 					filteredThisMessage = true
-					continue // Skip thinking blocks
+					continue
 				}
 			}
 

backend/internal/service/gateway_service.go

@@ -905,7 +905,7 @@ func (s *GatewayService) GetAccessToken(ctx context.Context, account *Account) (
 	case AccountTypeOAuth, AccountTypeSetupToken:
 		// Both oauth and setup-token use OAuth token flow
 		return s.getOAuthToken(ctx, account)
-	case AccountTypeAPIKey:
+	case AccountTypeApiKey:
 		apiKey := account.GetCredential("api_key")
 		if apiKey == "" {
 			return "", "", errors.New("api_key not found in credentials")
@@ -976,7 +976,7 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A
 
 	// 应用模型映射（仅对apikey类型账号）
 	originalModel := reqModel
-	if account.Type == AccountTypeAPIKey {
+	if account.Type == AccountTypeApiKey {
 		mappedModel := account.GetMappedModel(reqModel)
 		if mappedModel != reqModel {
 			// 替换请求体中的模型名
@@ -1013,8 +1013,37 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A
 			return nil, fmt.Errorf("upstream request failed: %w", err)
 		}
 
-		// 检查是否需要重试
-		if resp.StatusCode >= 400 && s.shouldRetryUpstreamError(account, resp.StatusCode) {
+		// 优先检测thinking block签名错误（400）并重试一次
+		if resp.StatusCode == 400 {
+			respBody, readErr := io.ReadAll(io.LimitReader(resp.Body, 2<<20))
+			if readErr == nil {
+				_ = resp.Body.Close()
+
+				if s.isThinkingBlockSignatureError(respBody) {
+					log.Printf("Account %d: detected thinking block signature error, retrying with filtered thinking blocks", account.ID)
+
+					// 过滤thinking blocks并重试
+					filteredBody := FilterThinkingBlocks(body)
+					retryReq, buildErr := s.buildUpstreamRequest(ctx, c, account, filteredBody, token, tokenType, reqModel)
+					if buildErr == nil {
+						retryResp, retryErr := s.httpUpstream.Do(retryReq, proxyURL, account.ID, account.Concurrency)
+						if retryErr == nil {
+							// 使用重试后的响应，继续后续处理
+							resp = retryResp
+							break
+						}
+					}
+					// 重试失败，恢复原始响应体继续处理
+					resp.Body = io.NopCloser(bytes.NewReader(respBody))
+					break
+				}
+				// 不是thinking签名错误，恢复响应体
+				resp.Body = io.NopCloser(bytes.NewReader(respBody))
+			}
+		}
+
+		// 检查是否需要通用重试（排除400，因为400已经在上面特殊处理过了）
+		if resp.StatusCode >= 400 && resp.StatusCode != 400 && s.shouldRetryUpstreamError(account, resp.StatusCode) {
 			if attempt < maxRetries {
 				log.Printf("Account %d: upstream error %d, retry %d/%d after %v",
 					account.ID, resp.StatusCode, attempt, maxRetries, retryDelay)
@@ -1047,13 +1076,13 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A
 	}
 
 	// 处理错误响应（不可重试的错误）
-	if resp.StatusCode >= 400 {
-		// 可选：对部分 400 触发 failover（默认关闭以保持语义）
-		if resp.StatusCode == 400 && s.cfg != nil && s.cfg.Gateway.FailoverOn400 {
-			respBody, readErr := io.ReadAll(resp.Body)
-			if readErr != nil {
-				// ReadAll failed, fall back to normal error handling without consuming the stream
-				return s.handleErrorResponse(ctx, resp, c, account)
+		if resp.StatusCode >= 400 {
+			// 可选：对部分 400 触发 failover（默认关闭以保持语义）
+			if resp.StatusCode == 400 && s.cfg != nil && s.cfg.Gateway.FailoverOn400 {
+				respBody, readErr := io.ReadAll(io.LimitReader(resp.Body, 2<<20))
+				if readErr != nil {
+					// ReadAll failed, fall back to normal error handling without consuming the stream
+					return s.handleErrorResponse(ctx, resp, c, account)
 			}
 			_ = resp.Body.Close()
 			resp.Body = io.NopCloser(bytes.NewReader(respBody))
@@ -1110,7 +1139,7 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A
 func (s *GatewayService) buildUpstreamRequest(ctx context.Context, c *gin.Context, account *Account, body []byte, token, tokenType, modelID string) (*http.Request, error) {
 	// 确定目标URL
 	targetURL := claudeAPIURL
-	if account.Type == AccountTypeAPIKey {
+	if account.Type == AccountTypeApiKey {
 		baseURL := account.GetBaseURL()
 		targetURL = baseURL + "/v1/messages"
 	}
@@ -1136,10 +1165,6 @@ func (s *GatewayService) buildUpstreamRequest(ctx context.Context, c *gin.Contex
 		}
 	}
 
-	// Filter thinking blocks from request body (prevents 400 errors from missing/invalid signatures).
-	// We apply this for the main /v1/messages path as well as count_tokens.
-	body = FilterThinkingBlocks(body)
-
 	req, err := http.NewRequestWithContext(ctx, "POST", targetURL, bytes.NewReader(body))
 	if err != nil {
 		return nil, err
@@ -1178,10 +1203,10 @@ func (s *GatewayService) buildUpstreamRequest(ctx context.Context, c *gin.Contex
 	// 处理anthropic-beta header（OAuth账号需要特殊处理）
 	if tokenType == "oauth" {
 		req.Header.Set("anthropic-beta", s.getBetaHeader(modelID, c.GetHeader("anthropic-beta")))
-	} else if s.cfg != nil && s.cfg.Gateway.InjectBetaForAPIKey && req.Header.Get("anthropic-beta") == "" {
+	} else if s.cfg != nil && s.cfg.Gateway.InjectBetaForApiKey && req.Header.Get("anthropic-beta") == "" {
 		// API-key：仅在请求显式使用 beta 特性且客户端未提供时，按需补齐（默认关闭）
 		if requestNeedsBetaFeatures(body) {
-			if beta := defaultAPIKeyBetaHeader(body); beta != "" {
+			if beta := defaultApiKeyBetaHeader(body); beta != "" {
 				req.Header.Set("anthropic-beta", beta)
 			}
 		}
@@ -1248,12 +1273,12 @@ func requestNeedsBetaFeatures(body []byte) bool {
 	return false
 }
 
-func defaultAPIKeyBetaHeader(body []byte) string {
+func defaultApiKeyBetaHeader(body []byte) string {
 	modelID := gjson.GetBytes(body, "model").String()
 	if strings.Contains(strings.ToLower(modelID), "haiku") {
-		return claude.APIKeyHaikuBetaHeader
+		return claude.ApiKeyHaikuBetaHeader
 	}
-	return claude.APIKeyBetaHeader
+	return claude.ApiKeyBetaHeader
 }
 
 func truncateForLog(b []byte, maxBytes int) string {
@@ -1270,6 +1295,20 @@ func truncateForLog(b []byte, maxBytes int) string {
 	return s
 }
 
+// isThinkingBlockSignatureError 检测是否是thinking block签名错误
+// 这类错误可以通过过滤thinking blocks并重试来解决
+func (s *GatewayService) isThinkingBlockSignatureError(respBody []byte) bool {
+	msg := strings.ToLower(strings.TrimSpace(extractUpstreamErrorMessage(respBody)))
+	if msg == "" {
+		return false
+	}
+
+	// 检测thinking block签名相关的错误
+	// 例如: "Invalid `signature` in `thinking` block"
+	return (strings.Contains(msg, "thinking") || strings.Contains(msg, "thought")) &&
+		strings.Contains(msg, "signature")
+}
+
 func (s *GatewayService) shouldFailoverOn400(respBody []byte) bool {
 	// 只对“可能是兼容性差异导致”的 400 允许切换，避免无意义重试。
 	// 默认保守：无法识别则不切换。
@@ -1318,7 +1357,13 @@ func (s *GatewayService) handleErrorResponse(ctx context.Context, resp *http.Res
 	body, _ := io.ReadAll(resp.Body)
 
 	// 处理上游错误，标记账号状态
-	s.rateLimitService.HandleUpstreamError(ctx, account, resp.StatusCode, resp.Header, body)
+	shouldDisable := false
+	if s.rateLimitService != nil {
+		shouldDisable = s.rateLimitService.HandleUpstreamError(ctx, account, resp.StatusCode, resp.Header, body)
+	}
+	if shouldDisable {
+		return nil, &UpstreamFailoverError{StatusCode: resp.StatusCode}
+	}
 
 	// 根据状态码返回适当的自定义错误响应（不透传上游详细信息）
 	var errType, errMsg string
@@ -1630,7 +1675,7 @@ func (s *GatewayService) replaceModelInResponseBody(body []byte, fromModel, toMo
 // RecordUsageInput 记录使用量的输入参数
 type RecordUsageInput struct {
 	Result       *ForwardResult
-	APIKey       *APIKey
+	ApiKey       *ApiKey
 	User         *User
 	Account      *Account
 	Subscription *UserSubscription // 可选：订阅信息
@@ -1639,7 +1684,7 @@ type RecordUsageInput struct {
 // RecordUsage 记录使用量并扣费（或更新订阅用量）
 func (s *GatewayService) RecordUsage(ctx context.Context, input *RecordUsageInput) error {
 	result := input.Result
-	apiKey := input.APIKey
+	apiKey := input.ApiKey
 	user := input.User
 	account := input.Account
 	subscription := input.Subscription
@@ -1676,7 +1721,7 @@ func (s *GatewayService) RecordUsage(ctx context.Context, input *RecordUsageInpu
 	durationMs := int(result.Duration.Milliseconds())
 	usageLog := &UsageLog{
 		UserID:              user.ID,
-		APIKeyID:            apiKey.ID,
+		ApiKeyID:            apiKey.ID,
 		AccountID:           account.ID,
 		RequestID:           result.RequestID,
 		Model:               result.Model,
@@ -1754,15 +1799,14 @@ func (s *GatewayService) ForwardCountTokens(ctx context.Context, c *gin.Context,
 	body := parsed.Body
 	reqModel := parsed.Model
 
-	// Antigravity 账户不支持 count_tokens 转发，返回估算值
-	// 参考 Antigravity-Manager 和 proxycast 实现
+	// Antigravity 账户不支持 count_tokens 转发，直接返回空值
 	if account.Platform == PlatformAntigravity {
-		c.JSON(http.StatusOK, gin.H{"input_tokens": 100})
+		c.JSON(http.StatusOK, gin.H{"input_tokens": 0})
 		return nil
 	}
 
 	// 应用模型映射（仅对 apikey 类型账号）
-	if account.Type == AccountTypeAPIKey {
+	if account.Type == AccountTypeApiKey {
 		if reqModel != "" {
 			mappedModel := account.GetMappedModel(reqModel)
 			if mappedModel != reqModel {
@@ -1848,7 +1892,7 @@ func (s *GatewayService) ForwardCountTokens(ctx context.Context, c *gin.Context,
 func (s *GatewayService) buildCountTokensRequest(ctx context.Context, c *gin.Context, account *Account, body []byte, token, tokenType, modelID string) (*http.Request, error) {
 	// 确定目标 URL
 	targetURL := claudeAPICountTokensURL
-	if account.Type == AccountTypeAPIKey {
+	if account.Type == AccountTypeApiKey {
 		baseURL := account.GetBaseURL()
 		targetURL = baseURL + "/v1/messages/count_tokens"
 	}
@@ -1866,9 +1910,6 @@ func (s *GatewayService) buildCountTokensRequest(ctx context.Context, c *gin.Con
 		}
 	}
 
-	// Filter thinking blocks from request body (prevents 400 errors from invalid signatures)
-	body = FilterThinkingBlocks(body)
-
 	req, err := http.NewRequestWithContext(ctx, "POST", targetURL, bytes.NewReader(body))
 	if err != nil {
 		return nil, err
@@ -1910,10 +1951,10 @@ func (s *GatewayService) buildCountTokensRequest(ctx context.Context, c *gin.Con
 	// OAuth 账号：处理 anthropic-beta header
 	if tokenType == "oauth" {
 		req.Header.Set("anthropic-beta", s.getBetaHeader(modelID, c.GetHeader("anthropic-beta")))
-	} else if s.cfg != nil && s.cfg.Gateway.InjectBetaForAPIKey && req.Header.Get("anthropic-beta") == "" {
+	} else if s.cfg != nil && s.cfg.Gateway.InjectBetaForApiKey && req.Header.Get("anthropic-beta") == "" {
 		// API-key：与 messages 同步的按需 beta 注入（默认关闭）
 		if requestNeedsBetaFeatures(body) {
-			if beta := defaultAPIKeyBetaHeader(body); beta != "" {
+			if beta := defaultApiKeyBetaHeader(body); beta != "" {
 				req.Header.Set("anthropic-beta", beta)
 			}
 		}