From 72ccd1b11145a1d6b019716a691755b9cc93b32c Mon Sep 17 00:00:00 2001
From: jianjian <jjaw@jjaw.cn>
Date: Thu, 2 Jul 2026 20:22:17 +0000
Subject: [PATCH 01/28] fix: batch group capacity summaries

---
 backend/internal/repository/account_repo.go   |  89 +++++++++
 backend/internal/repository/group_repo.go     |  43 +++++
 .../service/group_capacity_service.go         | 180 +++++++++++++++++-
 .../service/group_capacity_service_test.go    | 179 +++++++++++++++++
 4 files changed, 487 insertions(+), 4 deletions(-)
 create mode 100644 backend/internal/service/group_capacity_service_test.go

diff --git a/backend/internal/repository/account_repo.go b/backend/internal/repository/account_repo.go
index 2a2f5e0576..be03a6225b 100644
--- a/backend/internal/repository/account_repo.go
+++ b/backend/internal/repository/account_repo.go
@@ -1026,6 +1026,95 @@ func (r *accountRepository) ListSchedulableByGroupID(ctx context.Context, groupI
 	})
 }
 
+func (r *accountRepository) ListSchedulableCapacityByGroupIDs(ctx context.Context, groupIDs []int64) ([]service.GroupAccountCapacityRow, error) {
+	groupIDs = uniquePositiveInt64s(groupIDs)
+	if len(groupIDs) == 0 {
+		return []service.GroupAccountCapacityRow{}, nil
+	}
+	if r.sql == nil {
+		rows := make([]service.GroupAccountCapacityRow, 0)
+		for _, groupID := range groupIDs {
+			accounts, err := r.ListSchedulableByGroupID(ctx, groupID)
+			if err != nil {
+				return nil, err
+			}
+			for i := range accounts {
+				acc := &accounts[i]
+				rows = append(rows, service.GroupAccountCapacityRow{
+					GroupID:             groupID,
+					AccountID:           acc.ID,
+					Concurrency:         acc.Concurrency,
+					Extra:               copyJSONMap(acc.Extra),
+					SessionWindowStart:  acc.SessionWindowStart,
+					SessionWindowEnd:    acc.SessionWindowEnd,
+					SessionWindowStatus: acc.SessionWindowStatus,
+				})
+			}
+		}
+		return rows, nil
+	}
+
+	rows, err := r.sql.QueryContext(ctx, `
+		SELECT
+			ag.group_id,
+			a.id AS account_id,
+			a.concurrency,
+			COALESCE(a.extra, '{}'::jsonb)::text AS extra,
+			a.session_window_start,
+			a.session_window_end,
+			COALESCE(a.session_window_status, '') AS session_window_status
+		FROM account_groups ag
+		JOIN accounts a ON a.id = ag.account_id
+		LEFT JOIN proxies p
+			ON p.id = a.proxy_id
+			AND p.deleted_at IS NULL
+			AND p.status = $2
+		WHERE ag.group_id = ANY($1)
+			AND a.deleted_at IS NULL
+			AND a.status = $2
+			AND a.schedulable = TRUE
+			AND (a.owner_user_id IS NULL OR p.id IS NOT NULL)
+			AND (a.temp_unschedulable_until IS NULL OR a.temp_unschedulable_until <= $3)
+			AND (a.expires_at IS NULL OR a.expires_at > $3 OR a.auto_pause_on_expired = FALSE)
+			AND (a.overload_until IS NULL OR a.overload_until <= $3)
+			AND (a.rate_limit_reset_at IS NULL OR a.rate_limit_reset_at <= $3)
+		ORDER BY ag.group_id ASC, ag.priority ASC, a.priority ASC, a.id ASC
+	`, pq.Array(groupIDs), service.StatusActive, time.Now())
+	if err != nil {
+		return nil, err
+	}
+	defer func() { _ = rows.Close() }()
+
+	out := make([]service.GroupAccountCapacityRow, 0)
+	for rows.Next() {
+		var row service.GroupAccountCapacityRow
+		var extraRaw string
+		if err := rows.Scan(
+			&row.GroupID,
+			&row.AccountID,
+			&row.Concurrency,
+			&extraRaw,
+			&row.SessionWindowStart,
+			&row.SessionWindowEnd,
+			&row.SessionWindowStatus,
+		); err != nil {
+			return nil, err
+		}
+		if extraRaw != "" && extraRaw != "null" {
+			var extra map[string]any
+			if err := json.Unmarshal([]byte(extraRaw), &extra); err != nil {
+				return nil, err
+			}
+			row.Extra = extra
+		}
+		out = append(out, row)
+	}
+	if err := rows.Err(); err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
 func (r *accountRepository) ListSchedulableByPlatform(ctx context.Context, platform string) ([]service.Account, error) {
 	now := time.Now()
 	accounts, err := r.client.Account.Query().
diff --git a/backend/internal/repository/group_repo.go b/backend/internal/repository/group_repo.go
index 4e839b6a12..a4e173006e 100644
--- a/backend/internal/repository/group_repo.go
+++ b/backend/internal/repository/group_repo.go
@@ -466,6 +466,49 @@ func (r *groupRepository) ListActive(ctx context.Context) ([]service.Group, erro
 	return outGroups, nil
 }
 
+func (r *groupRepository) ListActiveIDs(ctx context.Context) ([]int64, error) {
+	if r.sql != nil {
+		rows, err := r.sql.QueryContext(ctx, `
+			SELECT id
+			FROM groups
+			WHERE status = $1
+			  AND deleted_at IS NULL
+			ORDER BY sort_order ASC, id ASC
+		`, service.StatusActive)
+		if err != nil {
+			return nil, err
+		}
+		defer func() { _ = rows.Close() }()
+
+		ids := make([]int64, 0)
+		for rows.Next() {
+			var id int64
+			if err := rows.Scan(&id); err != nil {
+				return nil, err
+			}
+			ids = append(ids, id)
+		}
+		if err := rows.Err(); err != nil {
+			return nil, err
+		}
+		return ids, nil
+	}
+
+	groups, err := r.client.Group.Query().
+		Where(group.StatusEQ(service.StatusActive)).
+		Select(group.FieldID).
+		Order(dbent.Asc(group.FieldSortOrder), dbent.Asc(group.FieldID)).
+		All(ctx)
+	if err != nil {
+		return nil, err
+	}
+	ids := make([]int64, 0, len(groups))
+	for i := range groups {
+		ids = append(ids, groups[i].ID)
+	}
+	return ids, nil
+}
+
 func (r *groupRepository) ListActiveByPlatform(ctx context.Context, platform string) ([]service.Group, error) {
 	groups, err := r.client.Group.Query().
 		Where(group.StatusEQ(service.StatusActive), group.PlatformEQ(platform)).
diff --git a/backend/internal/service/group_capacity_service.go b/backend/internal/service/group_capacity_service.go
index 459084dc59..b77b35773b 100644
--- a/backend/internal/service/group_capacity_service.go
+++ b/backend/internal/service/group_capacity_service.go
@@ -16,6 +16,26 @@ type GroupCapacitySummary struct {
 	RPMMax          int   `json:"rpm_max"`
 }
 
+// GroupAccountCapacityRow is the lightweight account projection needed for
+// capacity summary aggregation.
+type GroupAccountCapacityRow struct {
+	GroupID             int64
+	AccountID           int64
+	Concurrency         int
+	Extra               map[string]any
+	SessionWindowStart  *time.Time
+	SessionWindowEnd    *time.Time
+	SessionWindowStatus string
+}
+
+type groupCapacityActiveGroupIDLister interface {
+	ListActiveIDs(ctx context.Context) ([]int64, error)
+}
+
+type groupCapacityAccountLister interface {
+	ListSchedulableCapacityByGroupIDs(ctx context.Context, groupIDs []int64) ([]GroupAccountCapacityRow, error)
+}
+
 // GroupCapacityService aggregates per-group capacity from runtime data.
 type GroupCapacityService struct {
 	accountRepo        AccountRepository
@@ -44,24 +64,176 @@ func NewGroupCapacityService(
 
 // GetAllGroupCapacity returns capacity summary for all active groups.
 func (s *GroupCapacityService) GetAllGroupCapacity(ctx context.Context) ([]GroupCapacitySummary, error) {
-	groups, err := s.groupRepo.ListActive(ctx)
+	groupIDs, err := s.listActiveGroupIDs(ctx)
 	if err != nil {
 		return nil, err
 	}
 
-	results := make([]GroupCapacitySummary, 0, len(groups))
+	if lister, ok := s.accountRepo.(groupCapacityAccountLister); ok {
+		return s.getGroupCapacitiesBatch(ctx, groupIDs, lister)
+	}
+
+	return s.getGroupCapacitiesSequential(ctx, groupIDs), nil
+}
+
+func (s *GroupCapacityService) listActiveGroupIDs(ctx context.Context) ([]int64, error) {
+	if lister, ok := s.groupRepo.(groupCapacityActiveGroupIDLister); ok {
+		return lister.ListActiveIDs(ctx)
+	}
+
+	groups, err := s.groupRepo.ListActive(ctx)
+	if err != nil {
+		return nil, err
+	}
+	groupIDs := make([]int64, 0, len(groups))
 	for i := range groups {
-		cap, err := s.getGroupCapacity(ctx, groups[i].ID)
+		groupIDs = append(groupIDs, groups[i].ID)
+	}
+	return groupIDs, nil
+}
+
+func (s *GroupCapacityService) getGroupCapacitiesSequential(ctx context.Context, groupIDs []int64) []GroupCapacitySummary {
+	results := make([]GroupCapacitySummary, 0, len(groupIDs))
+	for _, groupID := range groupIDs {
+		cap, err := s.getGroupCapacity(ctx, groupID)
 		if err != nil {
 			// Skip groups with errors, return partial results
 			continue
 		}
-		cap.GroupID = groups[i].ID
+		cap.GroupID = groupID
 		results = append(results, cap)
 	}
+	return results
+}
+
+type groupCapacityAccountRef struct {
+	groupID   int64
+	accountID int64
+}
+
+func (s *GroupCapacityService) getGroupCapacitiesBatch(ctx context.Context, groupIDs []int64, lister groupCapacityAccountLister) ([]GroupCapacitySummary, error) {
+	results := make([]GroupCapacitySummary, len(groupIDs))
+	groupIndex := make(map[int64]int, len(groupIDs))
+	for i, groupID := range groupIDs {
+		results[i].GroupID = groupID
+		groupIndex[groupID] = i
+	}
+	if len(groupIDs) == 0 {
+		return results, nil
+	}
+
+	rows, err := lister.ListSchedulableCapacityByGroupIDs(ctx, groupIDs)
+	if err != nil {
+		return nil, err
+	}
+	if len(rows) == 0 {
+		return results, nil
+	}
+
+	refs := make([]groupCapacityAccountRef, 0, len(rows))
+	seenGroupAccount := make(map[groupCapacityAccountRef]struct{}, len(rows))
+	accountIDSet := make(map[int64]struct{}, len(rows))
+	accountIDs := make([]int64, 0, len(rows))
+	sessionTimeouts := make(map[int64]time.Duration)
+
+	for _, row := range rows {
+		idx, ok := groupIndex[row.GroupID]
+		if !ok || row.AccountID <= 0 {
+			continue
+		}
+
+		ref := groupCapacityAccountRef{groupID: row.GroupID, accountID: row.AccountID}
+		if _, ok := seenGroupAccount[ref]; ok {
+			continue
+		}
+		seenGroupAccount[ref] = struct{}{}
+		refs = append(refs, ref)
+
+		if _, ok := accountIDSet[row.AccountID]; !ok {
+			accountIDSet[row.AccountID] = struct{}{}
+			accountIDs = append(accountIDs, row.AccountID)
+		}
+
+		acc := Account{
+			ID:                  row.AccountID,
+			Concurrency:         row.Concurrency,
+			Extra:               row.Extra,
+			SessionWindowStart:  row.SessionWindowStart,
+			SessionWindowEnd:    row.SessionWindowEnd,
+			SessionWindowStatus: row.SessionWindowStatus,
+		}
+
+		results[idx].ConcurrencyMax += acc.Concurrency
+
+		if maxSessions := acc.GetMaxSessions(); maxSessions > 0 {
+			results[idx].SessionsMax += maxSessions
+			timeout := time.Duration(acc.GetSessionIdleTimeoutMinutes()) * time.Minute
+			if timeout <= 0 {
+				timeout = 5 * time.Minute
+			}
+			sessionTimeouts[acc.ID] = timeout
+		}
+
+		if rpm := acc.GetBaseRPM(); rpm > 0 {
+			results[idx].RPMMax += rpm
+		}
+	}
+
+	if len(accountIDs) == 0 {
+		return results, nil
+	}
+
+	concurrencyMap := map[int64]int{}
+	if s.concurrencyService != nil {
+		concurrencyMap, _ = s.concurrencyService.GetAccountConcurrencyBatch(ctx, accountIDs)
+	}
+
+	sessionAccountIDs := accountIDsForGroupsWithLimit(refs, groupIndex, results, func(summary GroupCapacitySummary) bool {
+		return summary.SessionsMax > 0
+	})
+	var sessionsMap map[int64]int
+	if len(sessionAccountIDs) > 0 && s.sessionLimitCache != nil {
+		sessionsMap, _ = s.sessionLimitCache.GetActiveSessionCountBatch(ctx, sessionAccountIDs, sessionTimeouts)
+	}
+
+	rpmAccountIDs := accountIDsForGroupsWithLimit(refs, groupIndex, results, func(summary GroupCapacitySummary) bool {
+		return summary.RPMMax > 0
+	})
+	var rpmMap map[int64]int
+	if len(rpmAccountIDs) > 0 && s.rpmCache != nil {
+		rpmMap, _ = s.rpmCache.GetRPMBatch(ctx, rpmAccountIDs)
+	}
+
+	for _, ref := range refs {
+		idx := groupIndex[ref.groupID]
+		results[idx].ConcurrencyUsed += concurrencyMap[ref.accountID]
+		if sessionsMap != nil && results[idx].SessionsMax > 0 {
+			results[idx].SessionsUsed += sessionsMap[ref.accountID]
+		}
+		if rpmMap != nil && results[idx].RPMMax > 0 {
+			results[idx].RPMUsed += rpmMap[ref.accountID]
+		}
+	}
 	return results, nil
 }
 
+func accountIDsForGroupsWithLimit(refs []groupCapacityAccountRef, groupIndex map[int64]int, summaries []GroupCapacitySummary, include func(GroupCapacitySummary) bool) []int64 {
+	seen := make(map[int64]struct{})
+	accountIDs := make([]int64, 0)
+	for _, ref := range refs {
+		idx, ok := groupIndex[ref.groupID]
+		if !ok || !include(summaries[idx]) {
+			continue
+		}
+		if _, ok := seen[ref.accountID]; ok {
+			continue
+		}
+		seen[ref.accountID] = struct{}{}
+		accountIDs = append(accountIDs, ref.accountID)
+	}
+	return accountIDs
+}
+
 func (s *GroupCapacityService) getGroupCapacity(ctx context.Context, groupID int64) (GroupCapacitySummary, error) {
 	accounts, err := s.accountRepo.ListSchedulableByGroupID(ctx, groupID)
 	if err != nil {
diff --git a/backend/internal/service/group_capacity_service_test.go b/backend/internal/service/group_capacity_service_test.go
new file mode 100644
index 0000000000..73927307d2
--- /dev/null
+++ b/backend/internal/service/group_capacity_service_test.go
@@ -0,0 +1,179 @@
+package service
+
+import (
+	"context"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/require"
+)
+
+type groupCapacityAccountRepoStub struct {
+	AccountRepository
+	rows      []GroupAccountCapacityRow
+	requested []int64
+}
+
+func (s *groupCapacityAccountRepoStub) ListSchedulableCapacityByGroupIDs(_ context.Context, groupIDs []int64) ([]GroupAccountCapacityRow, error) {
+	s.requested = append([]int64(nil), groupIDs...)
+	return append([]GroupAccountCapacityRow(nil), s.rows...), nil
+}
+
+type groupCapacityGroupRepoStub struct {
+	GroupRepository
+	groupIDs  []int64
+	listCalls int
+}
+
+func (s *groupCapacityGroupRepoStub) ListActiveIDs(context.Context) ([]int64, error) {
+	s.listCalls++
+	return append([]int64(nil), s.groupIDs...), nil
+}
+
+type groupCapacityConcurrencyCacheStub struct {
+	ConcurrencyCache
+	counts    map[int64]int
+	requested []int64
+}
+
+func (s *groupCapacityConcurrencyCacheStub) GetAccountConcurrencyBatch(_ context.Context, accountIDs []int64) (map[int64]int, error) {
+	s.requested = append([]int64(nil), accountIDs...)
+	out := make(map[int64]int, len(accountIDs))
+	for _, id := range accountIDs {
+		out[id] = s.counts[id]
+	}
+	return out, nil
+}
+
+type groupCapacitySessionCacheStub struct {
+	SessionLimitCache
+	counts       map[int64]int
+	requested    []int64
+	idleTimeouts map[int64]time.Duration
+}
+
+func (s *groupCapacitySessionCacheStub) GetActiveSessionCountBatch(_ context.Context, accountIDs []int64, idleTimeouts map[int64]time.Duration) (map[int64]int, error) {
+	s.requested = append([]int64(nil), accountIDs...)
+	s.idleTimeouts = make(map[int64]time.Duration, len(idleTimeouts))
+	for id, timeout := range idleTimeouts {
+		s.idleTimeouts[id] = timeout
+	}
+	out := make(map[int64]int, len(accountIDs))
+	for _, id := range accountIDs {
+		out[id] = s.counts[id]
+	}
+	return out, nil
+}
+
+type groupCapacityRPMCacheStub struct {
+	RPMCache
+	counts    map[int64]int
+	requested []int64
+}
+
+func (s *groupCapacityRPMCacheStub) GetRPMBatch(_ context.Context, accountIDs []int64) (map[int64]int, error) {
+	s.requested = append([]int64(nil), accountIDs...)
+	out := make(map[int64]int, len(accountIDs))
+	for _, id := range accountIDs {
+		out[id] = s.counts[id]
+	}
+	return out, nil
+}
+
+func TestGetAllGroupCapacityBatchAggregatesRuntimeAndLimits(t *testing.T) {
+	accountRepo := &groupCapacityAccountRepoStub{
+		rows: []GroupAccountCapacityRow{
+			{
+				GroupID:     10,
+				AccountID:   1,
+				Concurrency: 2,
+				Extra: map[string]any{
+					"max_sessions":                 3,
+					"session_idle_timeout_minutes": 7,
+					"base_rpm":                     11,
+				},
+			},
+			{
+				GroupID:     20,
+				AccountID:   1,
+				Concurrency: 2,
+				Extra: map[string]any{
+					"max_sessions":                 3,
+					"session_idle_timeout_minutes": 7,
+					"base_rpm":                     11,
+				},
+			},
+			{
+				GroupID:     20,
+				AccountID:   2,
+				Concurrency: 4,
+				Extra: map[string]any{
+					"max_sessions":                 1,
+					"session_idle_timeout_minutes": 9,
+					"base_rpm":                     13,
+				},
+			},
+		},
+	}
+	groupRepo := &groupCapacityGroupRepoStub{groupIDs: []int64{10, 20}}
+	concurrencyCache := &groupCapacityConcurrencyCacheStub{counts: map[int64]int{1: 1, 2: 2}}
+	sessionCache := &groupCapacitySessionCacheStub{counts: map[int64]int{1: 2, 2: 1}}
+	rpmCache := &groupCapacityRPMCacheStub{counts: map[int64]int{1: 5, 2: 7}}
+	svc := NewGroupCapacityService(
+		accountRepo,
+		groupRepo,
+		NewConcurrencyService(concurrencyCache),
+		sessionCache,
+		rpmCache,
+	)
+
+	results, err := svc.GetAllGroupCapacity(context.Background())
+	require.NoError(t, err)
+
+	require.Equal(t, 1, groupRepo.listCalls)
+	require.Equal(t, []int64{10, 20}, accountRepo.requested)
+	require.Equal(t, []int64{1, 2}, concurrencyCache.requested)
+	require.ElementsMatch(t, []int64{1, 2}, sessionCache.requested)
+	require.ElementsMatch(t, []int64{1, 2}, rpmCache.requested)
+	require.Equal(t, 7*time.Minute, sessionCache.idleTimeouts[1])
+	require.Equal(t, 9*time.Minute, sessionCache.idleTimeouts[2])
+
+	require.Equal(t, []GroupCapacitySummary{
+		{
+			GroupID:         10,
+			ConcurrencyUsed: 1,
+			ConcurrencyMax:  2,
+			SessionsUsed:    2,
+			SessionsMax:     3,
+			RPMUsed:         5,
+			RPMMax:          11,
+		},
+		{
+			GroupID:         20,
+			ConcurrencyUsed: 3,
+			ConcurrencyMax:  6,
+			SessionsUsed:    3,
+			SessionsMax:     4,
+			RPMUsed:         12,
+			RPMMax:          24,
+		},
+	}, results)
+}
+
+func TestGetAllGroupCapacityBatchKeepsEmptyGroupRows(t *testing.T) {
+	accountRepo := &groupCapacityAccountRepoStub{
+		rows: []GroupAccountCapacityRow{
+			{GroupID: 20, AccountID: 2, Concurrency: 4},
+		},
+	}
+	groupRepo := &groupCapacityGroupRepoStub{groupIDs: []int64{10, 20}}
+	svc := NewGroupCapacityService(accountRepo, groupRepo, nil, nil, nil)
+
+	results, err := svc.GetAllGroupCapacity(context.Background())
+	require.NoError(t, err)
+
+	require.Equal(t, []GroupCapacitySummary{
+		{GroupID: 10},
+		{GroupID: 20, ConcurrencyMax: 4},
+	}, results)
+}

From 36d5f4e4caf87fe88c7c52445c22c3a83d42b56a Mon Sep 17 00:00:00 2001
From: weiness <840833211@qq.com>
Date: Fri, 3 Jul 2026 10:29:33 +0800
Subject: [PATCH 02/28] feat: make setup migration timeout configurable

---
 backend/internal/setup/setup.go      | 26 ++++++++++++++++++--------
 backend/internal/setup/setup_test.go | 17 +++++++++++++++++
 deploy/.env.example                  |  7 +++++++
 deploy/docker-compose.dev.yml        |  1 +
 deploy/docker-compose.local.yml      |  5 +++++
 deploy/docker-compose.standalone.yml |  5 +++++
 deploy/docker-compose.yml            |  5 +++++
 7 files changed, 58 insertions(+), 8 deletions(-)

diff --git a/backend/internal/setup/setup.go b/backend/internal/setup/setup.go
index 51baf3dfe6..a2c4e2847a 100644
--- a/backend/internal/setup/setup.go
+++ b/backend/internal/setup/setup.go
@@ -28,6 +28,7 @@ const (
 	InstallLockFile            = ".installed"
 	defaultUserConcurrency     = 5
 	simpleModeAdminConcurrency = 30
+	defaultMigrationTimeout    = 60 * time.Second
 )
 
 func setupDefaultAdminConcurrency() int {
@@ -73,12 +74,13 @@ func GetInstallLockPath() string {
 
 // SetupConfig holds the setup configuration
 type SetupConfig struct {
-	Database DatabaseConfig `json:"database" yaml:"database"`
-	Redis    RedisConfig    `json:"redis" yaml:"redis"`
-	Admin    AdminConfig    `json:"admin" yaml:"-"` // Not stored in config file
-	Server   ServerConfig   `json:"server" yaml:"server"`
-	JWT      JWTConfig      `json:"jwt" yaml:"jwt"`
-	Timezone string         `json:"timezone" yaml:"timezone"` // e.g. "Asia/Shanghai", "UTC"
+	Database                DatabaseConfig `json:"database" yaml:"database"`
+	Redis                   RedisConfig    `json:"redis" yaml:"redis"`
+	Admin                   AdminConfig    `json:"admin" yaml:"-"` // Not stored in config file
+	Server                  ServerConfig   `json:"server" yaml:"server"`
+	JWT                     JWTConfig      `json:"jwt" yaml:"jwt"`
+	Timezone                string         `json:"timezone" yaml:"timezone"` // e.g. "Asia/Shanghai", "UTC"
+	MigrationTimeoutSeconds int            `json:"migration_timeout_seconds" yaml:"migration_timeout_seconds,omitempty"`
 }
 
 type DatabaseConfig struct {
@@ -350,11 +352,18 @@ func initializeDatabase(cfg *SetupConfig) error {
 		}
 	}()
 
-	migrationCtx, cancel := context.WithTimeout(context.Background(), 60*time.Second)
+	migrationCtx, cancel := context.WithTimeout(context.Background(), cfg.migrationTimeout())
 	defer cancel()
 	return repository.ApplyMigrations(migrationCtx, db)
 }
 
+func (cfg *SetupConfig) migrationTimeout() time.Duration {
+	if cfg != nil && cfg.MigrationTimeoutSeconds > 0 {
+		return time.Duration(cfg.MigrationTimeoutSeconds) * time.Second
+	}
+	return defaultMigrationTimeout
+}
+
 func createAdminUser(cfg *SetupConfig) (bool, string, error) {
 	dsn := fmt.Sprintf(
 		"host=%s port=%d user=%s password=%s dbname=%s sslmode=%s",
@@ -578,7 +587,8 @@ func AutoSetupFromEnv() error {
 			Secret:     getEnvOrDefault("JWT_SECRET", ""),
 			ExpireHour: getEnvIntOrDefault("JWT_EXPIRE_HOUR", 24),
 		},
-		Timezone: tz,
+		Timezone:                tz,
+		MigrationTimeoutSeconds: getEnvIntOrDefault("SETUP_MIGRATION_TIMEOUT_SECONDS", 0),
 	}
 
 	// Generate JWT secret if not provided
diff --git a/backend/internal/setup/setup_test.go b/backend/internal/setup/setup_test.go
index a2aa2f4cc1..b95c162bda 100644
--- a/backend/internal/setup/setup_test.go
+++ b/backend/internal/setup/setup_test.go
@@ -4,6 +4,7 @@ import (
 	"os"
 	"strings"
 	"testing"
+	"time"
 )
 
 func TestDecideAdminBootstrap(t *testing.T) {
@@ -70,6 +71,22 @@ func TestSetupDefaultAdminConcurrency(t *testing.T) {
 	})
 }
 
+func TestSetupMigrationTimeout(t *testing.T) {
+	t.Run("uses default timeout when unset", func(t *testing.T) {
+		cfg := &SetupConfig{}
+		if got := cfg.migrationTimeout(); got != 60*time.Second {
+			t.Fatalf("migrationTimeout()=%s, want 60s", got)
+		}
+	})
+
+	t.Run("uses configured timeout", func(t *testing.T) {
+		cfg := &SetupConfig{MigrationTimeoutSeconds: 300}
+		if got := cfg.migrationTimeout(); got != 300*time.Second {
+			t.Fatalf("migrationTimeout()=%s, want 300s", got)
+		}
+	})
+}
+
 func TestWriteConfigFileKeepsDefaultUserConcurrency(t *testing.T) {
 	t.Setenv("RUN_MODE", "simple")
 	t.Setenv("DATA_DIR", t.TempDir())
diff --git a/deploy/.env.example b/deploy/.env.example
index 59e4b44b91..d8892dbecf 100644
--- a/deploy/.env.example
+++ b/deploy/.env.example
@@ -196,6 +196,13 @@ JWT_EXPIRE_HOUR=24
 # - =0: 回退使用 JWT_EXPIRE_HOUR
 JWT_ACCESS_TOKEN_EXPIRE_MINUTES=0
 
+# -----------------------------------------------------------------------------
+# Setup Configuration
+# -----------------------------------------------------------------------------
+# Database migration timeout during initial setup, in seconds.
+# Leave 0 to use the built-in default of 60 seconds.
+SETUP_MIGRATION_TIMEOUT_SECONDS=0
+
 # -----------------------------------------------------------------------------
 # TOTP (2FA) Configuration
 # TOTP（双因素认证）配置
diff --git a/deploy/docker-compose.dev.yml b/deploy/docker-compose.dev.yml
index 7755fdbeff..07e89e0b51 100644
--- a/deploy/docker-compose.dev.yml
+++ b/deploy/docker-compose.dev.yml
@@ -38,6 +38,7 @@ services:
       - ADMIN_EMAIL=${ADMIN_EMAIL:-admin@sub2api.local}
       - ADMIN_PASSWORD=${ADMIN_PASSWORD:-}
       - JWT_SECRET=${JWT_SECRET:-}
+      - SETUP_MIGRATION_TIMEOUT_SECONDS=${SETUP_MIGRATION_TIMEOUT_SECONDS:-0}
       - TOTP_ENCRYPTION_KEY=${TOTP_ENCRYPTION_KEY:-}
       - TZ=${TZ:-Asia/Shanghai}
       # OpenAI HTTP upstream protocol/timeout
diff --git a/deploy/docker-compose.local.yml b/deploy/docker-compose.local.yml
index b15be2402d..21f46e3760 100644
--- a/deploy/docker-compose.local.yml
+++ b/deploy/docker-compose.local.yml
@@ -94,6 +94,11 @@ services:
       - JWT_SECRET=${JWT_SECRET:-}
       - JWT_EXPIRE_HOUR=${JWT_EXPIRE_HOUR:-24}
 
+      # =======================================================================
+      # Setup Configuration
+      # =======================================================================
+      - SETUP_MIGRATION_TIMEOUT_SECONDS=${SETUP_MIGRATION_TIMEOUT_SECONDS:-0}
+
       # =======================================================================
       # TOTP (2FA) Configuration
       # =======================================================================
diff --git a/deploy/docker-compose.standalone.yml b/deploy/docker-compose.standalone.yml
index 32afb28d6c..2e1d335624 100644
--- a/deploy/docker-compose.standalone.yml
+++ b/deploy/docker-compose.standalone.yml
@@ -76,6 +76,11 @@ services:
       - JWT_SECRET=${JWT_SECRET:-}
       - JWT_EXPIRE_HOUR=${JWT_EXPIRE_HOUR:-24}
 
+      # =======================================================================
+      # Setup Configuration
+      # =======================================================================
+      - SETUP_MIGRATION_TIMEOUT_SECONDS=${SETUP_MIGRATION_TIMEOUT_SECONDS:-0}
+
       # =======================================================================
       # Timezone Configuration
       # =======================================================================
diff --git a/deploy/docker-compose.yml b/deploy/docker-compose.yml
index fd682a87e6..3da3bcbfb0 100644
--- a/deploy/docker-compose.yml
+++ b/deploy/docker-compose.yml
@@ -90,6 +90,11 @@ services:
       - JWT_SECRET=${JWT_SECRET:-}
       - JWT_EXPIRE_HOUR=${JWT_EXPIRE_HOUR:-24}
 
+      # =======================================================================
+      # Setup Configuration
+      # =======================================================================
+      - SETUP_MIGRATION_TIMEOUT_SECONDS=${SETUP_MIGRATION_TIMEOUT_SECONDS:-0}
+
       # =======================================================================
       # TOTP (2FA) Configuration
       # =======================================================================

From 2b462b07debde257dff50ca59ee4d30b75b7be80 Mon Sep 17 00:00:00 2001
From: Heatherm Huang <heathermhuang@gmail.com>
Date: Fri, 3 Jul 2026 14:41:18 +0800
Subject: [PATCH 03/28] fix: preserve configured Grok OAuth concurrency

---
 .../service/admin_account_concurrency_test.go   | 17 +++++------------
 backend/internal/service/admin_service.go       |  3 ---
 2 files changed, 5 insertions(+), 15 deletions(-)

diff --git a/backend/internal/service/admin_account_concurrency_test.go b/backend/internal/service/admin_account_concurrency_test.go
index 3544f80e24..da57b5a64f 100644
--- a/backend/internal/service/admin_account_concurrency_test.go
+++ b/backend/internal/service/admin_account_concurrency_test.go
@@ -5,23 +5,16 @@ package service
 import (
 	"testing"
 
-	"github.com/Wei-Shaw/sub2api/internal/pkg/xai"
 	"github.com/stretchr/testify/require"
 )
 
-func TestNormalizeAccountConcurrencyCapsGrokOAuthUnlessUnsafe(t *testing.T) {
-	t.Setenv(xai.EnvUnsafeAllowHighConcurrency, "")
-
+func TestNormalizeAccountConcurrencyDefaultsInvalidGrokOAuthToOne(t *testing.T) {
 	require.Equal(t, 1, normalizeAccountConcurrency(PlatformGrok, AccountTypeOAuth, 0))
 	require.Equal(t, 1, normalizeAccountConcurrency(PlatformGrok, AccountTypeOAuth, -5))
-	require.Equal(t, 1, normalizeAccountConcurrency(PlatformGrok, AccountTypeOAuth, 50))
+}
+
+func TestNormalizeAccountConcurrencyPreservesExplicitValues(t *testing.T) {
+	require.Equal(t, 50, normalizeAccountConcurrency(PlatformGrok, AccountTypeOAuth, 50))
 	require.Equal(t, 2, normalizeAccountConcurrency(PlatformOpenAI, AccountTypeOAuth, 2))
 	require.Equal(t, 2, normalizeAccountConcurrency(PlatformGrok, AccountTypeAPIKey, 2))
 }
-
-func TestNormalizeAccountConcurrencyAllowsGrokOAuthUnsafeOverride(t *testing.T) {
-	t.Setenv(xai.EnvUnsafeAllowHighConcurrency, "true")
-
-	require.Equal(t, 50, normalizeAccountConcurrency(PlatformGrok, AccountTypeOAuth, 50))
-	require.Equal(t, 1, normalizeAccountConcurrency(PlatformGrok, AccountTypeOAuth, 0))
-}
diff --git a/backend/internal/service/admin_service.go b/backend/internal/service/admin_service.go
index bacd134db4..7685174f3b 100644
--- a/backend/internal/service/admin_service.go
+++ b/backend/internal/service/admin_service.go
@@ -2640,9 +2640,6 @@ func normalizeAccountConcurrency(platform, accountType string, concurrency int)
 		if concurrency <= 0 {
 			return 1
 		}
-		if concurrency > 1 && !xai.AllowUnsafeHighConcurrency() {
-			return 1
-		}
 	}
 	return concurrency
 }

From c620ad6a36376d8c010950d17d7d6f373e8b67a0 Mon Sep 17 00:00:00 2001
From: jianjian <jjaw@jjaw.cn>
Date: Fri, 3 Jul 2026 08:44:39 +0000
Subject: [PATCH 04/28] fix: align group capacity SQL with target schema

---
 backend/internal/repository/account_repo.go | 5 -----
 1 file changed, 5 deletions(-)

diff --git a/backend/internal/repository/account_repo.go b/backend/internal/repository/account_repo.go
index be03a6225b..18f25a5949 100644
--- a/backend/internal/repository/account_repo.go
+++ b/backend/internal/repository/account_repo.go
@@ -1065,15 +1065,10 @@ func (r *accountRepository) ListSchedulableCapacityByGroupIDs(ctx context.Contex
 			COALESCE(a.session_window_status, '') AS session_window_status
 		FROM account_groups ag
 		JOIN accounts a ON a.id = ag.account_id
-		LEFT JOIN proxies p
-			ON p.id = a.proxy_id
-			AND p.deleted_at IS NULL
-			AND p.status = $2
 		WHERE ag.group_id = ANY($1)
 			AND a.deleted_at IS NULL
 			AND a.status = $2
 			AND a.schedulable = TRUE
-			AND (a.owner_user_id IS NULL OR p.id IS NOT NULL)
 			AND (a.temp_unschedulable_until IS NULL OR a.temp_unschedulable_until <= $3)
 			AND (a.expires_at IS NULL OR a.expires_at > $3 OR a.auto_pause_on_expired = FALSE)
 			AND (a.overload_until IS NULL OR a.overload_until <= $3)

From b2687b22589c2f7aa12f4092bb89ea3739fc7735 Mon Sep 17 00:00:00 2001
From: Heatherm Huang <heathermhuang@gmail.com>
Date: Fri, 3 Jul 2026 19:43:54 +0800
Subject: [PATCH 05/28] fix(frontend): edit grok oauth model mapping

---
 README.md                                     |  6 +-
 .../components/account/EditAccountModal.vue   | 23 +++++---
 .../__tests__/EditAccountModal.spec.ts        | 55 +++++++++++++++++++
 frontend/src/composables/useModelWhitelist.ts | 13 ++++-
 4 files changed, 86 insertions(+), 11 deletions(-)

diff --git a/README.md b/README.md
index d960d93fa1..21ea27dee3 100644
--- a/README.md
+++ b/README.md
@@ -630,8 +630,10 @@ Sub2API supports Grok subscription accounts through xAI OAuth and forwards OpenA
 - Public Claude-compatible target: `/v1/messages`, converted to xAI Responses and returned as Anthropic Messages output for Claude CLI style clients
 - Public Chat Completions targets: `/v1/chat/completions` and `/chat/completions`, forwarded to `${XAI_BASE_URL:-https://api.x.ai/v1}/chat/completions`
 - Codex CLI style Responses WebSocket ingress is accepted on the Responses targets and bridged to xAI HTTP/SSE Responses upstream
-- Initial models: `grok-4.3`, `grok-build-0.1`, `grok-4.20-0309-reasoning`, `grok-4.20-0309-non-reasoning`, and `grok-4.20-multi-agent-0309`
-- Out of scope for this provider: image, video, TTS, transcription, browser automation, cookies, and Grok web scraping
+- Initial text models: `grok-4.3`, `grok-build-0.1`, `grok-4.20-0309-reasoning`, `grok-4.20-0309-non-reasoning`, and `grok-4.20-multi-agent-0309`
+- Media targets for Grok groups: `/v1/images/generations`, `/images/generations`, `/v1/images/edits`, `/images/edits`, `/v1/videos/generations`, `/videos/generations`, `/v1/videos/{request_id}`, and `/videos/{request_id}`. Generation requests require the group image-generation permission.
+- Media models: `grok-imagine`, `grok-imagine-image-quality`, `grok-imagine-image`, `grok-imagine-edit`, `grok-imagine-video`, and `grok-imagine-video-1.5`
+- Out of scope for this provider: TTS, transcription, browser automation, cookies, and Grok web scraping
 
 ### OAuth Configuration
 
diff --git a/frontend/src/components/account/EditAccountModal.vue b/frontend/src/components/account/EditAccountModal.vue
index b5f97753d8..cf65d96709 100644
--- a/frontend/src/components/account/EditAccountModal.vue
+++ b/frontend/src/components/account/EditAccountModal.vue
@@ -419,9 +419,9 @@
 
       </div>
 
-      <!-- OpenAI OAuth Model Mapping (OAuth 类型没有 apikey 容器，需要独立的模型映射区域) -->
+      <!-- OpenAI/Grok OAuth Model Mapping (OAuth 类型没有 apikey 容器，需要独立的模型映射区域) -->
       <div
-        v-if="account.platform === 'openai' && account.type === 'oauth'"
+        v-if="(account.platform === 'openai' || account.platform === 'grok') && account.type === 'oauth'"
         class="border-t border-gray-200 pt-4 dark:border-dark-600"
       >
         <label class="input-label">{{ t('admin.accounts.modelRestriction') }}</label>
@@ -3244,8 +3244,8 @@ const syncFormFromAccount = (newAccount: Account | null) => {
           : 'https://api.anthropic.com'
     editBaseUrl.value = platformDefaultUrl
 
-    // Load model mappings for OpenAI OAuth accounts
-    if (newAccount.platform === 'openai' && newAccount.credentials) {
+    // Load model mappings for OpenAI/Grok OAuth accounts
+    if ((newAccount.platform === 'openai' || newAccount.platform === 'grok') && newAccount.credentials) {
       const oauthCredentials = newAccount.credentials as Record<string, unknown>
       loadModelRestrictionFromMapping(oauthCredentials.model_mapping as Record<string, unknown> | undefined)
     } else {
@@ -3972,14 +3972,23 @@ const handleSubmit = async () => {
       updatePayload.credentials = newCredentials
     }
 
-    // OpenAI OAuth: persist model mapping to credentials
-    if (props.account.platform === 'openai' && props.account.type === 'oauth') {
+    // OpenAI/Grok OAuth: persist model mapping to credentials
+    if ((props.account.platform === 'openai' || props.account.platform === 'grok') && props.account.type === 'oauth') {
       const currentCredentials = isSparkShadow.value
         ? {}
         : (updatePayload.credentials as Record<string, unknown>) ||
           ((props.account.credentials as Record<string, unknown>) || {})
       const newCredentials: Record<string, unknown> = { ...currentCredentials }
-      applyOpenAIModelMappingCredentials(newCredentials)
+      if (props.account.platform === 'openai') {
+        applyOpenAIModelMappingCredentials(newCredentials)
+      } else {
+        const modelMapping = buildModelRestrictionMapping()
+        if (modelMapping) {
+          newCredentials.model_mapping = modelMapping
+        } else {
+          delete newCredentials.model_mapping
+        }
+      }
 
       updatePayload.credentials = newCredentials
     }
diff --git a/frontend/src/components/account/__tests__/EditAccountModal.spec.ts b/frontend/src/components/account/__tests__/EditAccountModal.spec.ts
index 381d512ee7..5cc22a2958 100644
--- a/frontend/src/components/account/__tests__/EditAccountModal.spec.ts
+++ b/frontend/src/components/account/__tests__/EditAccountModal.spec.ts
@@ -241,6 +241,32 @@ function buildAntigravityAccount(projectId = 'configured-project') {
   } as any
 }
 
+function buildGrokOAuthAccount() {
+  return {
+    id: 5,
+    name: 'Grok OAuth',
+    notes: '',
+    platform: 'grok',
+    type: 'oauth',
+    credentials: {
+      refresh_token: 'grok-rt',
+      base_url: 'https://api.x.ai/v1',
+      model_mapping: {
+        'grok-latest': 'grok-4.3'
+      }
+    },
+    extra: {},
+    proxy_id: null,
+    concurrency: 1,
+    priority: 1,
+    rate_multiplier: 1,
+    status: 'active',
+    group_ids: [],
+    expires_at: null,
+    auto_pause_on_expired: false
+  } as any
+}
+
 function buildOpenAISetupTokenAccount() {
   return {
     ...buildAccount(),
@@ -357,6 +383,35 @@ describe('EditAccountModal', () => {
     })
   })
 
+  it('loads and submits Grok OAuth model mapping edits', async () => {
+    const account = buildGrokOAuthAccount()
+    updateAccountMock.mockReset()
+    checkMixedChannelRiskMock.mockReset()
+    checkMixedChannelRiskMock.mockResolvedValue({ has_risk: false })
+    updateAccountMock.mockResolvedValue(account)
+
+    const wrapper = mountModal(account)
+    expect(wrapper.text()).toContain('Imagine Image')
+    expect(wrapper.text()).toContain('Imagine Video')
+
+    const inputWithValue = (value: string) => {
+      const input = wrapper
+        .findAll('input')
+        .find((input) => (input.element as HTMLInputElement).value === value)
+      expect(input).toBeTruthy()
+      return input!
+    }
+
+    await inputWithValue('grok-latest').setValue('grok')
+    await inputWithValue('grok-4.3').setValue('grok-build-0.1')
+    await wrapper.get('form#edit-account-form').trigger('submit.prevent')
+
+    expect(updateAccountMock).toHaveBeenCalledTimes(1)
+    expect(updateAccountMock.mock.calls[0]?.[1]?.credentials?.model_mapping).toEqual({
+      grok: 'grok-build-0.1'
+    })
+  })
+
   it('only submits model mapping credentials when saving an OpenAI spark shadow account', async () => {
     authIsSimpleMode.value = false
     const account = buildOpenAISparkShadowAccount()
diff --git a/frontend/src/composables/useModelWhitelist.ts b/frontend/src/composables/useModelWhitelist.ts
index 43449b09ee..654e1bb973 100644
--- a/frontend/src/composables/useModelWhitelist.ts
+++ b/frontend/src/composables/useModelWhitelist.ts
@@ -141,7 +141,13 @@ const xaiModels = [
   'grok-latest',
   'grok-build',
   'grok-4.20-reasoning',
-  'grok-4.20-non-reasoning'
+  'grok-4.20-non-reasoning',
+  'grok-imagine',
+  'grok-imagine-image-quality',
+  'grok-imagine-image',
+  'grok-imagine-edit',
+  'grok-imagine-video',
+  'grok-imagine-video-1.5'
 ]
 
 // Cohere
@@ -286,7 +292,10 @@ const grokPresetMappings = [
   { label: 'Grok Latest', from: 'grok-latest', to: 'grok-4.3', color: 'bg-emerald-100 text-emerald-700 hover:bg-emerald-200 dark:bg-emerald-900/30 dark:text-emerald-400' },
   { label: 'Build 0.1', from: 'grok-build', to: 'grok-build-0.1', color: 'bg-cyan-100 text-cyan-700 hover:bg-cyan-200 dark:bg-cyan-900/30 dark:text-cyan-400' },
   { label: '4.20 Reasoning', from: 'grok-4.20-reasoning', to: 'grok-4.20-0309-reasoning', color: 'bg-indigo-100 text-indigo-700 hover:bg-indigo-200 dark:bg-indigo-900/30 dark:text-indigo-400' },
-  { label: '4.20 Non Reasoning', from: 'grok-4.20-non-reasoning', to: 'grok-4.20-0309-non-reasoning', color: 'bg-violet-100 text-violet-700 hover:bg-violet-200 dark:bg-violet-900/30 dark:text-violet-400' }
+  { label: '4.20 Non Reasoning', from: 'grok-4.20-non-reasoning', to: 'grok-4.20-0309-non-reasoning', color: 'bg-violet-100 text-violet-700 hover:bg-violet-200 dark:bg-violet-900/30 dark:text-violet-400' },
+  { label: 'Imagine Image', from: 'grok-imagine', to: 'grok-imagine-image-quality', color: 'bg-sky-100 text-sky-700 hover:bg-sky-200 dark:bg-sky-900/30 dark:text-sky-400' },
+  { label: 'Imagine Edit', from: 'grok-imagine-edit', to: 'grok-imagine-edit', color: 'bg-rose-100 text-rose-700 hover:bg-rose-200 dark:bg-rose-900/30 dark:text-rose-400' },
+  { label: 'Imagine Video', from: 'grok-imagine-video-1.5', to: 'grok-imagine-video-1.5', color: 'bg-amber-100 text-amber-700 hover:bg-amber-200 dark:bg-amber-900/30 dark:text-amber-400' }
 ]
 
 // Antigravity 预设映射（支持通配符）

From f3a3a0869d099e64dcbad350cd30c71d87e97c81 Mon Sep 17 00:00:00 2001
From: jjaw <jjaw@jjaw.cn>
Date: Fri, 3 Jul 2026 21:02:31 +0800
Subject: [PATCH 06/28] =?UTF-8?q?=E4=BC=98=E5=8C=96=E5=B9=B6=E5=8F=91?=
 =?UTF-8?q?=E6=A7=BD=E4=BD=8D=E6=B8=85=E7=90=86?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 ...eway_handler_warmup_intercept_unit_test.go |  1 +
 .../handler/gateway_helper_fastpath_test.go   |  4 ++
 .../handler/gateway_helper_hotpath_test.go    |  4 ++
 .../internal/repository/concurrency_cache.go  | 50 +++++++++++++++++++
 .../concurrency_cache_integration_test.go     | 33 ++++++++++++
 .../internal/service/concurrency_service.go   | 19 +++----
 .../service/concurrency_service_test.go       |  4 ++
 .../service/concurrency_slot_cleanup_test.go  | 39 +++++++++++++++
 .../service/gateway_multiplatform_test.go     |  4 ++
 backend/internal/testutil/stubs.go            |  3 ++
 10 files changed, 148 insertions(+), 13 deletions(-)
 create mode 100644 backend/internal/service/concurrency_slot_cleanup_test.go

diff --git a/backend/internal/handler/gateway_handler_warmup_intercept_unit_test.go b/backend/internal/handler/gateway_handler_warmup_intercept_unit_test.go
index 09b20722c9..75e3c1aa8c 100644
--- a/backend/internal/handler/gateway_handler_warmup_intercept_unit_test.go
+++ b/backend/internal/handler/gateway_handler_warmup_intercept_unit_test.go
@@ -137,6 +137,7 @@ func (f *fakeConcurrencyCache) GetAccountConcurrencyBatch(_ context.Context, acc
 	return result, nil
 }
 func (f *fakeConcurrencyCache) CleanupExpiredAccountSlots(context.Context, int64) error { return nil }
+func (f *fakeConcurrencyCache) CleanupExpiredAccountSlotKeys(context.Context) error     { return nil }
 func (f *fakeConcurrencyCache) CleanupStaleProcessSlots(context.Context, string) error  { return nil }
 
 func newTestGatewayHandler(t *testing.T, group *service.Group, accounts []*service.Account) (*GatewayHandler, func()) {
diff --git a/backend/internal/handler/gateway_helper_fastpath_test.go b/backend/internal/handler/gateway_helper_fastpath_test.go
index c7c0fb6c9e..fecb9b071d 100644
--- a/backend/internal/handler/gateway_helper_fastpath_test.go
+++ b/backend/internal/handler/gateway_helper_fastpath_test.go
@@ -89,6 +89,10 @@ func (m *concurrencyCacheMock) CleanupExpiredAccountSlots(ctx context.Context, a
 	return nil
 }
 
+func (m *concurrencyCacheMock) CleanupExpiredAccountSlotKeys(ctx context.Context) error {
+	return nil
+}
+
 func (m *concurrencyCacheMock) CleanupStaleProcessSlots(ctx context.Context, activeRequestPrefix string) error {
 	return nil
 }
diff --git a/backend/internal/handler/gateway_helper_hotpath_test.go b/backend/internal/handler/gateway_helper_hotpath_test.go
index 65dc849683..fb17481f1c 100644
--- a/backend/internal/handler/gateway_helper_hotpath_test.go
+++ b/backend/internal/handler/gateway_helper_hotpath_test.go
@@ -140,6 +140,10 @@ func (s *helperConcurrencyCacheStub) CleanupExpiredAccountSlots(ctx context.Cont
 	return nil
 }
 
+func (s *helperConcurrencyCacheStub) CleanupExpiredAccountSlotKeys(ctx context.Context) error {
+	return nil
+}
+
 func (s *helperConcurrencyCacheStub) CleanupStaleProcessSlots(ctx context.Context, activeRequestPrefix string) error {
 	return nil
 }
diff --git a/backend/internal/repository/concurrency_cache.go b/backend/internal/repository/concurrency_cache.go
index 5e6f10062f..b530c30b90 100644
--- a/backend/internal/repository/concurrency_cache.go
+++ b/backend/internal/repository/concurrency_cache.go
@@ -174,6 +174,29 @@ var (
 		return 1
 	`)
 
+	// cleanupExpiredSlotKeysScript 批量清理实际存在的账号槽位键，避免后台任务从数据库加载全量账号。
+	// KEYS = 有序集合键列表，ARGV[1] = TTL（秒）。
+	cleanupExpiredSlotKeysScript = redis.NewScript(`
+		-- Redis 3.2-4.x compat: opt into effects replication so redis.call('TIME')
+		-- replicates correctly. No-op on Redis 5.0+ (effects replication is default).
+		redis.replicate_commands()
+		local ttl = tonumber(ARGV[1])
+		local timeResult = redis.call('TIME')
+		local now = tonumber(timeResult[1])
+		local expireBefore = now - ttl
+		local removed = 0
+		for i = 1, #KEYS do
+			local key = KEYS[i]
+			removed = removed + redis.call('ZREMRANGEBYSCORE', key, '-inf', expireBefore)
+			if redis.call('ZCARD', key) == 0 then
+				redis.call('DEL', key)
+			else
+				redis.call('EXPIRE', key, ttl)
+			end
+		end
+		return removed
+	`)
+
 	// startupCleanupScript 清理非当前进程前缀的槽位成员。
 	// KEYS 是有序集合键列表，ARGV[1] 是当前进程前缀，ARGV[2] 是槽位 TTL。
 	// 遍历每个 KEYS[i]，移除前缀不匹配的成员，清空后删 key，否则刷新 EXPIRE。
@@ -503,6 +526,10 @@ func (c *concurrencyCache) CleanupExpiredAccountSlots(ctx context.Context, accou
 	return err
 }
 
+func (c *concurrencyCache) CleanupExpiredAccountSlotKeys(ctx context.Context) error {
+	return c.cleanupExpiredSlotKeysByPattern(ctx, accountSlotKeyPrefix+"*")
+}
+
 func (c *concurrencyCache) CleanupStaleProcessSlots(ctx context.Context, activeRequestPrefix string) error {
 	if activeRequestPrefix == "" {
 		return nil
@@ -527,6 +554,29 @@ func (c *concurrencyCache) CleanupStaleProcessSlots(ctx context.Context, activeR
 	return nil
 }
 
+// cleanupExpiredSlotKeysByPattern 扫描实际存在的账号槽位键并批量清理过期成员。
+func (c *concurrencyCache) cleanupExpiredSlotKeysByPattern(ctx context.Context, pattern string) error {
+	const scanCount = 200
+	var cursor uint64
+	for {
+		keys, nextCursor, err := c.rdb.Scan(ctx, cursor, pattern, scanCount).Result()
+		if err != nil {
+			return fmt.Errorf("scan %s: %w", pattern, err)
+		}
+		if len(keys) > 0 {
+			_, err := cleanupExpiredSlotKeysScript.Run(ctx, c.rdb, keys, c.slotTTLSeconds).Result()
+			if err != nil {
+				return fmt.Errorf("cleanup expired slots %s: %w", pattern, err)
+			}
+		}
+		cursor = nextCursor
+		if cursor == 0 {
+			break
+		}
+	}
+	return nil
+}
+
 // cleanupSlotsByPattern 扫描匹配 pattern 的有序集合键，批量调用 Lua 脚本清理非当前进程成员。
 func (c *concurrencyCache) cleanupSlotsByPattern(ctx context.Context, pattern, activePrefix string) error {
 	const scanCount = 200
diff --git a/backend/internal/repository/concurrency_cache_integration_test.go b/backend/internal/repository/concurrency_cache_integration_test.go
index 5da94fc258..4aa172b252 100644
--- a/backend/internal/repository/concurrency_cache_integration_test.go
+++ b/backend/internal/repository/concurrency_cache_integration_test.go
@@ -435,6 +435,39 @@ func (s *ConcurrencyCacheSuite) TestCleanupExpiredAccountSlots_NoExpired() {
 	require.Equal(s.T(), 2, cur)
 }
 
+func (s *ConcurrencyCacheSuite) TestCleanupExpiredAccountSlotKeys() {
+	now := time.Now().Unix()
+	expiredTime := now - int64(testSlotTTL.Seconds()) - 10
+	accountKeyWithFresh := fmt.Sprintf("%s%d", accountSlotKeyPrefix, 301)
+	accountKeyExpiredOnly := fmt.Sprintf("%s%d", accountSlotKeyPrefix, 302)
+	userKey := fmt.Sprintf("%s%d", userSlotKeyPrefix, 303)
+
+	require.NoError(s.T(), s.rdb.ZAdd(s.ctx, accountKeyWithFresh,
+		redis.Z{Score: float64(expiredTime), Member: "expired"},
+		redis.Z{Score: float64(now), Member: "fresh"},
+	).Err())
+	require.NoError(s.T(), s.rdb.ZAdd(s.ctx, accountKeyExpiredOnly,
+		redis.Z{Score: float64(expiredTime), Member: "expired-only"},
+	).Err())
+	require.NoError(s.T(), s.rdb.ZAdd(s.ctx, userKey,
+		redis.Z{Score: float64(expiredTime), Member: "user-expired"},
+	).Err())
+
+	require.NoError(s.T(), s.cache.CleanupExpiredAccountSlotKeys(s.ctx))
+
+	accountMembers, err := s.rdb.ZRange(s.ctx, accountKeyWithFresh, 0, -1).Result()
+	require.NoError(s.T(), err)
+	require.Equal(s.T(), []string{"fresh"}, accountMembers)
+
+	exists, err := s.rdb.Exists(s.ctx, accountKeyExpiredOnly).Result()
+	require.NoError(s.T(), err)
+	require.EqualValues(s.T(), 0, exists)
+
+	userMembers, err := s.rdb.ZRange(s.ctx, userKey, 0, -1).Result()
+	require.NoError(s.T(), err)
+	require.Equal(s.T(), []string{"user-expired"}, userMembers)
+}
+
 func (s *ConcurrencyCacheSuite) TestCleanupStaleProcessSlots_RemovesOldPrefixesAndWaitCounters() {
 	accountID := int64(901)
 	userID := int64(902)
diff --git a/backend/internal/service/concurrency_service.go b/backend/internal/service/concurrency_service.go
index 712fc1a749..50e0077b94 100644
--- a/backend/internal/service/concurrency_service.go
+++ b/backend/internal/service/concurrency_service.go
@@ -47,6 +47,7 @@ type ConcurrencyCache interface {
 
 	// 清理过期槽位（后台任务）
 	CleanupExpiredAccountSlots(ctx context.Context, accountID int64) error
+	CleanupExpiredAccountSlotKeys(ctx context.Context) error
 
 	// 启动时清理旧进程遗留槽位与等待计数
 	CleanupStaleProcessSlots(ctx context.Context, activeRequestPrefix string) error
@@ -473,27 +474,19 @@ func (s *ConcurrencyService) CleanupExpiredAccountSlots(ctx context.Context, acc
 }
 
 // StartSlotCleanupWorker starts a background cleanup worker for expired account slots.
-func (s *ConcurrencyService) StartSlotCleanupWorker(accountRepo AccountRepository, interval time.Duration) {
-	if s == nil || s.cache == nil || accountRepo == nil || interval <= 0 {
+func (s *ConcurrencyService) StartSlotCleanupWorker(_ AccountRepository, interval time.Duration) {
+	if s == nil || s.cache == nil || interval <= 0 {
 		return
 	}
 
 	runCleanup := func() {
-		listCtx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
-		accounts, err := accountRepo.ListSchedulable(listCtx)
+		cleanupCtx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+		err := s.cache.CleanupExpiredAccountSlotKeys(cleanupCtx)
 		cancel()
 		if err != nil {
-			logger.LegacyPrintf("service.concurrency", "Warning: list schedulable accounts failed: %v", err)
+			logger.LegacyPrintf("service.concurrency", "Warning: cleanup expired account slots failed: %v", err)
 			return
 		}
-		for _, account := range accounts {
-			accountCtx, accountCancel := context.WithTimeout(context.Background(), 2*time.Second)
-			err := s.cache.CleanupExpiredAccountSlots(accountCtx, account.ID)
-			accountCancel()
-			if err != nil {
-				logger.LegacyPrintf("service.concurrency", "Warning: cleanup expired slots failed for account %d: %v", account.ID, err)
-			}
-		}
 	}
 
 	go func() {
diff --git a/backend/internal/service/concurrency_service_test.go b/backend/internal/service/concurrency_service_test.go
index 7d5f501dc5..bacad0245e 100644
--- a/backend/internal/service/concurrency_service_test.go
+++ b/backend/internal/service/concurrency_service_test.go
@@ -95,6 +95,10 @@ func (c *stubConcurrencyCacheForTest) CleanupExpiredAccountSlots(_ context.Conte
 	return c.cleanupErr
 }
 
+func (c *stubConcurrencyCacheForTest) CleanupExpiredAccountSlotKeys(_ context.Context) error {
+	return c.cleanupErr
+}
+
 func (c *stubConcurrencyCacheForTest) CleanupStaleProcessSlots(_ context.Context, _ string) error {
 	return c.cleanupErr
 }
diff --git a/backend/internal/service/concurrency_slot_cleanup_test.go b/backend/internal/service/concurrency_slot_cleanup_test.go
new file mode 100644
index 0000000000..be26bb391b
--- /dev/null
+++ b/backend/internal/service/concurrency_slot_cleanup_test.go
@@ -0,0 +1,39 @@
+package service
+
+import (
+	"context"
+	"sync/atomic"
+	"testing"
+	"time"
+)
+
+type slotCleanupCache struct {
+	ConcurrencyCache
+	calls atomic.Int64
+}
+
+func (c *slotCleanupCache) CleanupExpiredAccountSlotKeys(context.Context) error {
+	c.calls.Add(1)
+	return nil
+}
+
+func TestStartSlotCleanupWorker_UsesCacheWideCleanupWithoutAccountRepo(t *testing.T) {
+	cache := &slotCleanupCache{}
+	svc := NewConcurrencyService(cache)
+
+	svc.StartSlotCleanupWorker(nil, time.Hour)
+
+	deadline := time.After(time.Second)
+	ticker := time.NewTicker(10 * time.Millisecond)
+	defer ticker.Stop()
+	for {
+		if cache.calls.Load() > 0 {
+			return
+		}
+		select {
+		case <-deadline:
+			t.Fatal("cleanup worker did not call cache-wide account slot cleanup")
+		case <-ticker.C:
+		}
+	}
+}
diff --git a/backend/internal/service/gateway_multiplatform_test.go b/backend/internal/service/gateway_multiplatform_test.go
index 291cea53b9..f843ba3e45 100644
--- a/backend/internal/service/gateway_multiplatform_test.go
+++ b/backend/internal/service/gateway_multiplatform_test.go
@@ -2097,6 +2097,10 @@ func (m *mockConcurrencyCache) CleanupExpiredAccountSlots(ctx context.Context, a
 	return nil
 }
 
+func (m *mockConcurrencyCache) CleanupExpiredAccountSlotKeys(ctx context.Context) error {
+	return nil
+}
+
 func (m *mockConcurrencyCache) CleanupStaleProcessSlots(ctx context.Context, activeRequestPrefix string) error {
 	return nil
 }
diff --git a/backend/internal/testutil/stubs.go b/backend/internal/testutil/stubs.go
index bc572e1137..a0bd4bc669 100644
--- a/backend/internal/testutil/stubs.go
+++ b/backend/internal/testutil/stubs.go
@@ -76,6 +76,9 @@ func (c StubConcurrencyCache) GetAccountConcurrencyBatch(_ context.Context, acco
 func (c StubConcurrencyCache) CleanupExpiredAccountSlots(_ context.Context, _ int64) error {
 	return nil
 }
+func (c StubConcurrencyCache) CleanupExpiredAccountSlotKeys(_ context.Context) error {
+	return nil
+}
 func (c StubConcurrencyCache) CleanupStaleProcessSlots(_ context.Context, _ string) error {
 	return nil
 }

From a1b2b32e08df7c3c878fce53061e380e59ae57a6 Mon Sep 17 00:00:00 2001
From: shaw <dev@sub2api.org>
Date: Fri, 3 Jul 2026 21:02:32 +0800
Subject: [PATCH 07/28] fix: prevent silent usage_logs drops under queue
 overflow (#3656)
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

高并发下扣费成功但 usage_logs 被双层异步队列静默丢弃（单用户实测丢失 49%），
导致消费排行与对账缺口。根因是 6a685727 为避免溢出时单行 INSERT 踩踏数据库，
把三处"队列满"改成了立即终态丢弃。

改为有界阻塞背压，兼顾数据不丢与批量写入模式不变：

- CreateBestEffort/createBatched 入队移除 default 立即丢弃分支，
  队列满时阻塞等待（受 detached 15s ctx 期限约束）
- writeUsageLogBestEffort 收到 dropped 不再直接放弃，统一走
  repo.Create 同步兜底（仍经批处理器；ctx 耗尽时换新 detached 窗口），
  重复写入由 ON CONFLICT DO NOTHING 幂等防护
- worker 池默认溢出策略 sample→sync（viper 默认 + 池常量），
  溢出时提交方内联执行（提交点在响应写出后，不阻塞客户端），
  显式配置 sample/drop 的部署不受影响
- 附带修复 ensure*Batcher 的存量数据竞争：channel nil 检查移入 sync.Once

正常负载零行为变化；突发时短暂等待替代永久丢失；数据库只见批量 INSERT。
---
 backend/internal/config/config.go             |  5 +-
 backend/internal/config/config_test.go        |  4 +-
 backend/internal/repository/usage_log_repo.go | 27 +++++----
 .../usage_log_repo_integration_test.go        | 57 ++++++++++++++++---
 .../service/gateway_record_usage_test.go      |  8 ++-
 backend/internal/service/gateway_service.go   | 13 ++++-
 .../service/usage_record_worker_pool.go       |  9 +--
 7 files changed, 92 insertions(+), 31 deletions(-)

diff --git a/backend/internal/config/config.go b/backend/internal/config/config.go
index 18baa34881..b8b880eab0 100644
--- a/backend/internal/config/config.go
+++ b/backend/internal/config/config.go
@@ -1945,7 +1945,10 @@ func setDefaults() {
 	viper.SetDefault("gateway.usage_record.worker_count", 128)
 	viper.SetDefault("gateway.usage_record.queue_size", 16384)
 	viper.SetDefault("gateway.usage_record.task_timeout_seconds", 5)
-	viper.SetDefault("gateway.usage_record.overflow_policy", UsageRecordOverflowPolicySample)
+	// 默认 sync：队列满时由提交方内联执行（提交点在响应写出之后，不阻塞客户端）。
+	// sample/drop 会在溢出时静默丢弃计费任务，造成扣费与 usage_logs 对账缺口（issue #3656），
+	// 仅供显式配置的运维场景使用。
+	viper.SetDefault("gateway.usage_record.overflow_policy", UsageRecordOverflowPolicySync)
 	viper.SetDefault("gateway.usage_record.overflow_sample_percent", 10)
 	viper.SetDefault("gateway.usage_record.auto_scale_enabled", true)
 	viper.SetDefault("gateway.usage_record.auto_scale_min_workers", 128)
diff --git a/backend/internal/config/config_test.go b/backend/internal/config/config_test.go
index bf7a327563..804155d1a9 100644
--- a/backend/internal/config/config_test.go
+++ b/backend/internal/config/config_test.go
@@ -1903,8 +1903,8 @@ func TestLoad_DefaultGatewayUsageRecordConfig(t *testing.T) {
 	if cfg.Gateway.UsageRecord.TaskTimeoutSeconds != 5 {
 		t.Fatalf("task_timeout_seconds = %d, want 5", cfg.Gateway.UsageRecord.TaskTimeoutSeconds)
 	}
-	if cfg.Gateway.UsageRecord.OverflowPolicy != UsageRecordOverflowPolicySample {
-		t.Fatalf("overflow_policy = %s, want %s", cfg.Gateway.UsageRecord.OverflowPolicy, UsageRecordOverflowPolicySample)
+	if cfg.Gateway.UsageRecord.OverflowPolicy != UsageRecordOverflowPolicySync {
+		t.Fatalf("overflow_policy = %s, want %s", cfg.Gateway.UsageRecord.OverflowPolicy, UsageRecordOverflowPolicySync)
 	}
 	if cfg.Gateway.UsageRecord.OverflowSamplePercent != 10 {
 		t.Fatalf("overflow_sample_percent = %d, want 10", cfg.Gateway.UsageRecord.OverflowSamplePercent)
diff --git a/backend/internal/repository/usage_log_repo.go b/backend/internal/repository/usage_log_repo.go
index 885e63f9fd..24c648b0a5 100644
--- a/backend/internal/repository/usage_log_repo.go
+++ b/backend/internal/repository/usage_log_repo.go
@@ -372,12 +372,13 @@ func (r *usageLogRepository) CreateBestEffort(ctx context.Context, log *service.
 		}
 	}
 
+	// 队列满时阻塞等待而非立即丢弃：批处理器持续排空队列，短暂等待即可入队。
+	// 立即丢弃会造成“已扣费但无 usage_log”的永久数据缺口（issue #3656）；
+	// 阻塞上限由调用方 ctx 期限约束，超时后由上层同步兜底。
 	select {
 	case r.bestEffortBatchCh <- req:
 	case <-ctx.Done():
 		return service.MarkUsageLogCreateDropped(ctx.Err())
-	default:
-		return service.MarkUsageLogCreateDropped(errors.New("usage log best-effort queue full"))
 	}
 
 	select {
@@ -493,12 +494,12 @@ func (r *usageLogRepository) createBatched(ctx context.Context, log *service.Usa
 		resultCh: make(chan usageLogCreateResult, 1),
 	}
 
+	// 队列满时阻塞等待而非立即报错：本路径是 best-effort 丢弃后的最后兜底，
+	// 立即失败会让日志永久丢失；阻塞上限由调用方 ctx 期限约束。
 	select {
 	case r.createBatchCh <- req:
 	case <-ctx.Done():
 		return false, service.MarkUsageLogCreateNotPersisted(ctx.Err())
-	default:
-		return false, service.MarkUsageLogCreateNotPersisted(errors.New("usage log create batch queue full"))
 	}
 
 	select {
@@ -520,22 +521,28 @@ func (r *usageLogRepository) createBatched(ctx context.Context, log *service.Usa
 }
 
 func (r *usageLogRepository) ensureCreateBatcher() {
-	if r == nil || r.db == nil || r.createBatchCh != nil {
+	if r == nil || r.db == nil {
 		return
 	}
+	// nil 检查必须在 Once 内部：在外层做无同步快路径读会与 Once 内的写构成数据竞争。
 	r.createBatchOnce.Do(func() {
-		r.createBatchCh = make(chan usageLogCreateRequest, usageLogCreateBatchQueueCap)
-		go r.runCreateBatcher(r.db)
+		if r.createBatchCh == nil {
+			r.createBatchCh = make(chan usageLogCreateRequest, usageLogCreateBatchQueueCap)
+			go r.runCreateBatcher(r.db)
+		}
 	})
 }
 
 func (r *usageLogRepository) ensureBestEffortBatcher() {
-	if r == nil || r.db == nil || r.bestEffortBatchCh != nil {
+	if r == nil || r.db == nil {
 		return
 	}
+	// 同 ensureCreateBatcher：nil 检查放在 Once 内部以避免数据竞争。
 	r.bestEffortBatchOnce.Do(func() {
-		r.bestEffortBatchCh = make(chan usageLogBestEffortRequest, usageLogBestEffortBatchQueueCap)
-		go r.runBestEffortBatcher(r.db)
+		if r.bestEffortBatchCh == nil {
+			r.bestEffortBatchCh = make(chan usageLogBestEffortRequest, usageLogBestEffortBatchQueueCap)
+			go r.runBestEffortBatcher(r.db)
+		}
 	})
 }
 
diff --git a/backend/internal/repository/usage_log_repo_integration_test.go b/backend/internal/repository/usage_log_repo_integration_test.go
index ed3050d89c..b43c6d56ef 100644
--- a/backend/internal/repository/usage_log_repo_integration_test.go
+++ b/backend/internal/repository/usage_log_repo_integration_test.go
@@ -288,21 +288,21 @@ func TestUsageLogRepositoryCreateBestEffort_BatchPathDuplicateRequestID(t *testi
 	}, 3*time.Second, 20*time.Millisecond)
 }
 
-func TestUsageLogRepositoryCreateBestEffort_QueueFullReturnsDropped(t *testing.T) {
-	ctx := context.Background()
+func TestUsageLogRepositoryCreateBestEffort_QueueFullBlocksUntilCtxDeadline(t *testing.T) {
+	// 队列满时不再立即丢弃：阻塞等待入队，直到调用方 ctx 到期才标记 dropped（issue #3656）。
 	client := testEntClient(t)
 	repo := newUsageLogRepositoryWithSQL(client, integrationDB)
 	repo.bestEffortBatchCh = make(chan usageLogBestEffortRequest, 1)
 	repo.bestEffortBatchCh <- usageLogBestEffortRequest{}
 
-	user := mustCreateUser(t, client, &service.User{Email: fmt.Sprintf("usage-best-effort-full-%d@example.com", time.Now().UnixNano())})
-	apiKey := mustCreateApiKey(t, client, &service.APIKey{UserID: user.ID, Key: "sk-usage-best-effort-full-" + uuid.NewString(), Name: "k"})
-	account := mustCreateAccount(t, client, &service.Account{Name: "acc-usage-best-effort-full-" + uuid.NewString()})
+	ctx, cancel := context.WithTimeout(context.Background(), 200*time.Millisecond)
+	defer cancel()
 
+	start := time.Now()
 	err := repo.CreateBestEffort(ctx, &service.UsageLog{
-		UserID:       user.ID,
-		APIKeyID:     apiKey.ID,
-		AccountID:    account.ID,
+		UserID:       1,
+		APIKeyID:     2,
+		AccountID:    3,
 		RequestID:    uuid.NewString(),
 		Model:        "claude-3",
 		InputTokens:  10,
@@ -314,6 +314,40 @@ func TestUsageLogRepositoryCreateBestEffort_QueueFullReturnsDropped(t *testing.T
 
 	require.Error(t, err)
 	require.True(t, service.IsUsageLogCreateDropped(err))
+	require.GreaterOrEqual(t, time.Since(start), 150*time.Millisecond)
+}
+
+func TestUsageLogRepositoryCreateBestEffort_QueueFullWaitsForDrain(t *testing.T) {
+	// 队列满但批处理器随后排空时，阻塞的入队应成功完成而非丢弃。
+	client := testEntClient(t)
+	repo := newUsageLogRepositoryWithSQL(client, integrationDB)
+	repo.bestEffortBatchCh = make(chan usageLogBestEffortRequest, 1)
+	repo.bestEffortBatchCh <- usageLogBestEffortRequest{}
+
+	go func() {
+		time.Sleep(100 * time.Millisecond)
+		<-repo.bestEffortBatchCh // 排空占位请求，为阻塞中的入队腾出空间
+		req := <-repo.bestEffortBatchCh
+		sendUsageLogBestEffortResult(req.resultCh, nil)
+	}()
+
+	ctx, cancel := context.WithTimeout(context.Background(), 3*time.Second)
+	defer cancel()
+
+	err := repo.CreateBestEffort(ctx, &service.UsageLog{
+		UserID:       1,
+		APIKeyID:     2,
+		AccountID:    3,
+		RequestID:    uuid.NewString(),
+		Model:        "claude-3",
+		InputTokens:  10,
+		OutputTokens: 20,
+		TotalCost:    0.5,
+		ActualCost:   0.5,
+		CreatedAt:    time.Now().UTC(),
+	})
+
+	require.NoError(t, err)
 }
 
 func TestUsageLogRepositoryCreate_BatchPathCanceledContextMarksNotPersisted(t *testing.T) {
@@ -346,7 +380,7 @@ func TestUsageLogRepositoryCreate_BatchPathCanceledContextMarksNotPersisted(t *t
 }
 
 func TestUsageLogRepositoryCreate_BatchPathQueueFullMarksNotPersisted(t *testing.T) {
-	ctx := context.Background()
+	// 队列满时阻塞等待入队，直到调用方 ctx 到期才标记 not persisted（issue #3656）。
 	client := testEntClient(t)
 	repo := newUsageLogRepositoryWithSQL(client, integrationDB)
 	repo.createBatchCh = make(chan usageLogCreateRequest, 1)
@@ -356,6 +390,10 @@ func TestUsageLogRepositoryCreate_BatchPathQueueFullMarksNotPersisted(t *testing
 	apiKey := mustCreateApiKey(t, client, &service.APIKey{UserID: user.ID, Key: "sk-usage-create-full-" + uuid.NewString(), Name: "k"})
 	account := mustCreateAccount(t, client, &service.Account{Name: "acc-usage-create-full-" + uuid.NewString()})
 
+	ctx, cancel := context.WithTimeout(context.Background(), 200*time.Millisecond)
+	defer cancel()
+
+	start := time.Now()
 	inserted, err := repo.Create(ctx, &service.UsageLog{
 		UserID:       user.ID,
 		APIKeyID:     apiKey.ID,
@@ -372,6 +410,7 @@ func TestUsageLogRepositoryCreate_BatchPathQueueFullMarksNotPersisted(t *testing
 	require.False(t, inserted)
 	require.Error(t, err)
 	require.True(t, service.IsUsageLogCreateNotPersisted(err))
+	require.GreaterOrEqual(t, time.Since(start), 150*time.Millisecond)
 }
 
 func TestUsageLogRepositoryCreate_BatchPathCanceledAfterQueueMarksNotPersisted(t *testing.T) {
diff --git a/backend/internal/service/gateway_record_usage_test.go b/backend/internal/service/gateway_record_usage_test.go
index c819eeca6e..2769251820 100644
--- a/backend/internal/service/gateway_record_usage_test.go
+++ b/backend/internal/service/gateway_record_usage_test.go
@@ -440,7 +440,9 @@ func TestGatewayServiceRecordUsage_GeneratesRequestIDWhenAllSourcesMissing(t *te
 	require.Equal(t, billingRepo.lastCmd.RequestID, usageRepo.lastLog.RequestID)
 }
 
-func TestGatewayServiceRecordUsage_DroppedUsageLogDoesNotSyncFallback(t *testing.T) {
+func TestGatewayServiceRecordUsage_DroppedUsageLogFallsBackToSyncCreate(t *testing.T) {
+	// 计费成功后 best-effort 写入被丢弃（队列超时）时必须同步兜底，
+	// 否则出现“已扣费但无 usage_log”的对账缺口（issue #3656）。
 	usageRepo := &openAIRecordUsageBestEffortLogRepoStub{
 		bestEffortErr: MarkUsageLogCreateDropped(errors.New("usage log best-effort queue full")),
 	}
@@ -464,7 +466,9 @@ func TestGatewayServiceRecordUsage_DroppedUsageLogDoesNotSyncFallback(t *testing
 
 	require.NoError(t, err)
 	require.Equal(t, 1, usageRepo.bestEffortCalls)
-	require.Equal(t, 0, usageRepo.createCalls)
+	require.Equal(t, 1, usageRepo.createCalls)
+	// 兜底调用使用的 ctx 必须仍然存活，不能带着已死的 ctx 走过场。
+	require.NoError(t, usageRepo.lastCtxErr)
 }
 
 func TestGatewayServiceRecordUsage_BillingErrorSkipsUsageLogWrite(t *testing.T) {
diff --git a/backend/internal/service/gateway_service.go b/backend/internal/service/gateway_service.go
index 160a92a8e1..54035345d9 100644
--- a/backend/internal/service/gateway_service.go
+++ b/backend/internal/service/gateway_service.go
@@ -9473,10 +9473,17 @@ func writeUsageLogBestEffort(ctx context.Context, repo UsageLogRepository, usage
 	if writer, ok := repo.(usageLogBestEffortWriter); ok {
 		if err := writer.CreateBestEffort(usageCtx, usageLog); err != nil {
 			logger.LegacyPrintf(logKey, "Create usage log failed: %v", err)
-			if IsUsageLogCreateDropped(err) {
-				return
+			// 计费已在此前完成，日志必须落库：dropped（批处理队列超时）同样走同步兜底，
+			// 否则会出现“已扣费但无 usage_log”的对账缺口（issue #3656）。
+			// 重复写入由 usage_logs 的 ON CONFLICT (request_id, api_key_id) DO NOTHING 防护。
+			fallbackCtx := usageCtx
+			if usageCtx.Err() != nil {
+				// usageCtx 已耗尽（best-effort 入队阻塞到期限）：换新的 detached 窗口，避免兜底必然失败。
+				var fallbackCancel context.CancelFunc
+				fallbackCtx, fallbackCancel = detachedBillingContext(context.Background())
+				defer fallbackCancel()
 			}
-			if _, syncErr := repo.Create(usageCtx, usageLog); syncErr != nil {
+			if _, syncErr := repo.Create(fallbackCtx, usageLog); syncErr != nil {
 				logger.LegacyPrintf(logKey, "Create usage log sync fallback failed: %v", syncErr)
 			}
 		}
diff --git a/backend/internal/service/usage_record_worker_pool.go b/backend/internal/service/usage_record_worker_pool.go
index 5da0b89023..bb5ae452c8 100644
--- a/backend/internal/service/usage_record_worker_pool.go
+++ b/backend/internal/service/usage_record_worker_pool.go
@@ -15,10 +15,11 @@ import (
 )
 
 const (
-	defaultUsageRecordWorkerCount          = 128
-	defaultUsageRecordQueueSize            = 16384
-	defaultUsageRecordTaskTimeoutSeconds   = 5
-	defaultUsageRecordOverflowPolicy       = config.UsageRecordOverflowPolicySample
+	defaultUsageRecordWorkerCount        = 128
+	defaultUsageRecordQueueSize          = 16384
+	defaultUsageRecordTaskTimeoutSeconds = 5
+	// 默认 sync：溢出时提交方内联执行，保证计费任务不被静默丢弃（issue #3656）。
+	defaultUsageRecordOverflowPolicy       = config.UsageRecordOverflowPolicySync
 	defaultUsageRecordOverflowSampleRatio  = 10
 	defaultUsageRecordAutoScaleEnabled     = true
 	defaultUsageRecordAutoScaleMinWorkers  = 128

From f385cdceb78ba53108a8cf917277df10137be13b Mon Sep 17 00:00:00 2001
From: Anderson <andersonby@163.com>
Date: Fri, 3 Jul 2026 22:36:39 +0800
Subject: [PATCH 08/28] feat: add Codex image tool strip policy

---
 .../service/codex_image_generation_bridge.go  | 45 +++++++++++
 .../service/openai_codex_transform.go         | 39 ++++++---
 .../service/openai_gateway_service.go         | 21 ++++-
 .../openai_image_generation_controls_test.go  | 39 +++++++++
 .../internal/service/openai_ws_forwarder.go   | 20 ++++-
 .../openai_ws_forwarder_ingress_test.go       | 20 +++++
 .../components/account/EditAccountModal.vue   | 79 +++++++++++++++++++
 .../__tests__/EditAccountModal.spec.ts        | 35 ++++++++
 frontend/src/i18n/locales/en.ts               |  7 ++
 frontend/src/i18n/locales/zh.ts               |  6 ++
 10 files changed, 296 insertions(+), 15 deletions(-)

diff --git a/backend/internal/service/codex_image_generation_bridge.go b/backend/internal/service/codex_image_generation_bridge.go
index c7a894a792..68989d67dc 100644
--- a/backend/internal/service/codex_image_generation_bridge.go
+++ b/backend/internal/service/codex_image_generation_bridge.go
@@ -4,6 +4,13 @@ import "strings"
 
 const featureKeyCodexImageGenerationBridge = "codex_image_generation_bridge"
 
+const (
+	featureKeyCodexImageGenerationExplicitToolPolicy = "codex_image_generation_explicit_tool_policy"
+
+	codexImageGenerationExplicitToolPolicyAllow = "allow"
+	codexImageGenerationExplicitToolPolicyStrip = "strip"
+)
+
 func boolOverridePtr(v bool) *bool {
 	return &v
 }
@@ -20,6 +27,27 @@ func boolOverrideFromMap(values map[string]any, keys ...string) *bool {
 	return nil
 }
 
+func stringOverrideFromMap(values map[string]any, keys ...string) (string, bool) {
+	if values == nil {
+		return "", false
+	}
+	for _, key := range keys {
+		if v, ok := values[key].(string); ok {
+			return v, true
+		}
+	}
+	return "", false
+}
+
+func normalizeCodexImageGenerationExplicitToolPolicy(value string) string {
+	switch strings.ToLower(strings.TrimSpace(value)) {
+	case codexImageGenerationExplicitToolPolicyStrip, "remove", "drop":
+		return codexImageGenerationExplicitToolPolicyStrip
+	default:
+		return codexImageGenerationExplicitToolPolicyAllow
+	}
+}
+
 func platformBoolOverride(values map[string]any, key string, platform string) *bool {
 	if values == nil {
 		return nil
@@ -62,3 +90,20 @@ func (a *Account) CodexImageGenerationBridgeOverride() *bool {
 	openaiConfig, _ := a.Extra[PlatformOpenAI].(map[string]any)
 	return boolOverrideFromMap(openaiConfig, featureKeyCodexImageGenerationBridge, "codex_image_generation_bridge_enabled")
 }
+
+// CodexImageGenerationExplicitToolPolicy returns the account-level policy for
+// client-provided Codex /responses image_generation tools. Unknown or unset
+// values default to allow to preserve existing behavior.
+func (a *Account) CodexImageGenerationExplicitToolPolicy() string {
+	if a == nil || a.Platform != PlatformOpenAI || a.Extra == nil {
+		return codexImageGenerationExplicitToolPolicyAllow
+	}
+	if policy, ok := stringOverrideFromMap(a.Extra, featureKeyCodexImageGenerationExplicitToolPolicy); ok {
+		return normalizeCodexImageGenerationExplicitToolPolicy(policy)
+	}
+	openaiConfig, _ := a.Extra[PlatformOpenAI].(map[string]any)
+	if policy, ok := stringOverrideFromMap(openaiConfig, featureKeyCodexImageGenerationExplicitToolPolicy); ok {
+		return normalizeCodexImageGenerationExplicitToolPolicy(policy)
+	}
+	return codexImageGenerationExplicitToolPolicyAllow
+}
diff --git a/backend/internal/service/openai_codex_transform.go b/backend/internal/service/openai_codex_transform.go
index 0ece5e44ff..c33ac97f0f 100644
--- a/backend/internal/service/openai_codex_transform.go
+++ b/backend/internal/service/openai_codex_transform.go
@@ -607,18 +607,21 @@ func hasOpenAIImageGenerationTool(reqBody map[string]any) bool {
 	return false
 }
 
-// stripCodexSparkImageGenerationTools removes image_generation tool entries from
-// reqBody["tools"]. gpt-5.3-codex-spark rejects that tool upstream with HTTP 400
-// (invalid_request_error, param=tools), and Codex CLI advertises it by default, so
-// it must be dropped for spark. When the tools list becomes empty the key is removed.
-// Returns true when the body was modified.
-func stripCodexSparkImageGenerationTools(reqBody map[string]any) bool {
+func stripOpenAIImageGenerationTools(reqBody map[string]any) bool {
 	rawTools, ok := reqBody["tools"]
 	if !ok || rawTools == nil {
+		if openAIAnyToolChoiceSelectsImageGeneration(reqBody["tool_choice"]) {
+			delete(reqBody, "tool_choice")
+			return true
+		}
 		return false
 	}
 	tools, ok := rawTools.([]any)
 	if !ok {
+		if openAIAnyToolChoiceSelectsImageGeneration(reqBody["tool_choice"]) {
+			delete(reqBody, "tool_choice")
+			return true
+		}
 		return false
 	}
 	filtered := make([]any, 0, len(tools))
@@ -631,17 +634,31 @@ func stripCodexSparkImageGenerationTools(reqBody map[string]any) bool {
 		}
 		filtered = append(filtered, rawTool)
 	}
-	if !removed {
+	if !removed && !openAIAnyToolChoiceSelectsImageGeneration(reqBody["tool_choice"]) {
 		return false
 	}
-	if len(filtered) == 0 {
-		delete(reqBody, "tools")
-	} else {
-		reqBody["tools"] = filtered
+	if removed {
+		if len(filtered) == 0 {
+			delete(reqBody, "tools")
+		} else {
+			reqBody["tools"] = filtered
+		}
+	}
+	if openAIAnyToolChoiceSelectsImageGeneration(reqBody["tool_choice"]) {
+		delete(reqBody, "tool_choice")
 	}
 	return true
 }
 
+// stripCodexSparkImageGenerationTools removes image_generation tool entries from
+// reqBody["tools"]. gpt-5.3-codex-spark rejects that tool upstream with HTTP 400
+// (invalid_request_error, param=tools), and Codex CLI advertises it by default, so
+// it must be dropped for spark. When the tools list becomes empty the key is removed.
+// Returns true when the body was modified.
+func stripCodexSparkImageGenerationTools(reqBody map[string]any) bool {
+	return stripOpenAIImageGenerationTools(reqBody)
+}
+
 func hasOpenAIInputImage(reqBody map[string]any) bool {
 	if reqBody == nil {
 		return false
diff --git a/backend/internal/service/openai_gateway_service.go b/backend/internal/service/openai_gateway_service.go
index cd84790b98..6bf04ef20f 100644
--- a/backend/internal/service/openai_gateway_service.go
+++ b/backend/internal/service/openai_gateway_service.go
@@ -2738,8 +2738,25 @@ func (s *OpenAIGatewayService) Forward(ctx context.Context, c *gin.Context, acco
 	if apiKey != nil {
 		imageGenerationAllowed = GroupAllowsImageGeneration(apiKey.Group)
 	}
-	codexImageGenerationBridgeEnabled := isCodexCLI && imageGenerationAllowed && s.isCodexImageGenerationBridgeEnabled(ctx, account, apiKey)
-	imageIntent := IsImageGenerationIntent(openAIResponsesEndpoint, reqModel, body)
+	codexImageGenerationExplicitToolPolicy := codexImageGenerationExplicitToolPolicyAllow
+	if isCodexCLI {
+		codexImageGenerationExplicitToolPolicy = account.CodexImageGenerationExplicitToolPolicy()
+	}
+	codexImageGenerationBridgeEnabled := isCodexCLI && imageGenerationAllowed && codexImageGenerationExplicitToolPolicy != codexImageGenerationExplicitToolPolicyStrip && s.isCodexImageGenerationBridgeEnabled(ctx, account, apiKey)
+	var imageIntent bool
+	if isCodexCLI && codexImageGenerationExplicitToolPolicy == codexImageGenerationExplicitToolPolicyStrip {
+		decoded, decodeErr := ensureReqBody()
+		if decodeErr != nil {
+			return nil, decodeErr
+		}
+		if stripOpenAIImageGenerationTools(decoded) {
+			markDecodedModified()
+			logger.LegacyPrintf("service.openai_gateway", "[OpenAI] Stripped /responses image_generation tool for Codex client by account policy")
+		}
+		imageIntent = IsImageGenerationIntentMap(openAIResponsesEndpoint, reqModel, decoded)
+	} else {
+		imageIntent = IsImageGenerationIntent(openAIResponsesEndpoint, reqModel, body)
+	}
 	if imageIntent && !imageGenerationAllowed {
 		MarkOpsClientBusinessLimited(c, OpsClientBusinessLimitedReasonLocalFeatureGate)
 		c.JSON(http.StatusForbidden, gin.H{"error": gin.H{"type": "permission_error", "message": ImageGenerationPermissionMessage()}})
diff --git a/backend/internal/service/openai_image_generation_controls_test.go b/backend/internal/service/openai_image_generation_controls_test.go
index 6061a0bf9c..31edd36097 100644
--- a/backend/internal/service/openai_image_generation_controls_test.go
+++ b/backend/internal/service/openai_image_generation_controls_test.go
@@ -152,6 +152,45 @@ func TestOpenAIGatewayServiceForward_ExplicitImageToolWorksWithBridgeDisabled(t
 	require.NotContains(t, instructions, "image_generation")
 }
 
+func TestOpenAIGatewayServiceForward_AccountPolicyStripsExplicitImageTool(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+
+	upstream := &httpUpstreamRecorder{
+		resp: &http.Response{
+			StatusCode: http.StatusOK,
+			Header:     http.Header{"Content-Type": []string{"application/json"}},
+			Body:       io.NopCloser(strings.NewReader(`{"id":"resp_stripped_image","model":"gpt-5.4","usage":{"input_tokens":2,"output_tokens":1}}`)),
+		},
+	}
+	svc := newOpenAIImageGenerationControlTestService(upstream)
+	c, _ := newOpenAIImageGenerationControlTestContext(true, "codex_cli_rs/0.98.0")
+	account := newOpenAIImageGenerationControlTestAccount()
+	account.Extra = map[string]any{
+		featureKeyCodexImageGenerationExplicitToolPolicy: codexImageGenerationExplicitToolPolicyStrip,
+	}
+	body := []byte(`{
+		"model":"gpt-5.4",
+		"input":"draw",
+		"stream":false,
+		"tools":[
+			{"type":"function","name":"shell","parameters":{"type":"object"}},
+			{"type":"image_generation","format":"jpeg"}
+		],
+		"tool_choice":{"type":"image_generation"}
+	}`)
+
+	result, err := svc.Forward(context.Background(), c, account, body)
+
+	require.NoError(t, err)
+	require.NotNil(t, result)
+	require.NotNil(t, upstream.lastReq)
+	require.False(t, gjson.GetBytes(upstream.lastBody, `tools.#(type=="image_generation")`).Exists())
+	require.True(t, gjson.GetBytes(upstream.lastBody, `tools.#(type=="function")`).Exists())
+	require.False(t, gjson.GetBytes(upstream.lastBody, "tool_choice").Exists())
+	instructions := gjson.GetBytes(upstream.lastBody, "instructions").String()
+	require.NotContains(t, instructions, "image_generation")
+}
+
 func TestOpenAIGatewayServiceForward_ChannelBridgeOverrideEnablesCodexInjection(t *testing.T) {
 	gin.SetMode(gin.TestMode)
 
diff --git a/backend/internal/service/openai_ws_forwarder.go b/backend/internal/service/openai_ws_forwarder.go
index 067eeb6029..9261db5544 100644
--- a/backend/internal/service/openai_ws_forwarder.go
+++ b/backend/internal/service/openai_ws_forwarder.go
@@ -2448,11 +2448,15 @@ func stripCodexSparkImageGenerationToolFromRawPayload(payload []byte, model stri
 	if !isCodexSparkModel(model) || !openAIRequestBodyHasImageGenerationTool(payload) {
 		return payload, false, nil
 	}
+	return stripOpenAIImageGenerationToolFromRawPayload(payload)
+}
+
+func stripOpenAIImageGenerationToolFromRawPayload(payload []byte) ([]byte, bool, error) {
 	payloadMap := make(map[string]any)
 	if err := json.Unmarshal(payload, &payloadMap); err != nil {
 		return payload, false, err
 	}
-	if !stripCodexSparkImageGenerationTools(payloadMap) {
+	if !stripOpenAIImageGenerationTools(payloadMap) {
 		return payload, false, nil
 	}
 	rebuilt, err := json.Marshal(payloadMap)
@@ -2671,7 +2675,11 @@ func (s *OpenAIGatewayService) ProxyResponsesWebSocketFromClient(
 		}
 		apiKey := getAPIKeyFromContext(c)
 		imageGenerationAllowed := GroupAllowsImageGeneration(apiKeyGroup(apiKey))
-		codexBridgeEnabled := isCodexCLI && imageGenerationAllowed && s.isCodexImageGenerationBridgeEnabled(ctx, account, apiKey)
+		codexImageGenerationExplicitToolPolicy := codexImageGenerationExplicitToolPolicyAllow
+		if isCodexCLI {
+			codexImageGenerationExplicitToolPolicy = account.CodexImageGenerationExplicitToolPolicy()
+		}
+		codexBridgeEnabled := isCodexCLI && imageGenerationAllowed && codexImageGenerationExplicitToolPolicy != codexImageGenerationExplicitToolPolicyStrip && s.isCodexImageGenerationBridgeEnabled(ctx, account, apiKey)
 		if codexBridgeEnabled {
 			payloadMap := make(map[string]any)
 			if err := json.Unmarshal(normalized, &payloadMap); err != nil {
@@ -2709,6 +2717,14 @@ func (s *OpenAIGatewayService) ProxyResponsesWebSocketFromClient(
 			}
 			normalized = next
 		}
+		if isCodexCLI && codexImageGenerationExplicitToolPolicy == codexImageGenerationExplicitToolPolicyStrip {
+			if stripped, changed, stripErr := stripOpenAIImageGenerationToolFromRawPayload(normalized); stripErr != nil {
+				return openAIWSClientPayload{}, NewOpenAIWSClientCloseError(coderws.StatusPolicyViolation, "invalid websocket request payload", stripErr)
+			} else if changed {
+				normalized = stripped
+				logOpenAIWSModeInfo("ingress_ws_codex_image_tool_stripped_by_policy account_id=%d", account.ID)
+			}
+		}
 		if stripped, changed, stripErr := stripCodexSparkImageGenerationToolFromRawPayload(normalized, upstreamModel); stripErr != nil {
 			return openAIWSClientPayload{}, NewOpenAIWSClientCloseError(coderws.StatusPolicyViolation, "invalid websocket request payload", stripErr)
 		} else if changed {
diff --git a/backend/internal/service/openai_ws_forwarder_ingress_test.go b/backend/internal/service/openai_ws_forwarder_ingress_test.go
index 0d19a189b0..ca7c36aaa7 100644
--- a/backend/internal/service/openai_ws_forwarder_ingress_test.go
+++ b/backend/internal/service/openai_ws_forwarder_ingress_test.go
@@ -169,6 +169,26 @@ func TestStripCodexSparkImageGenerationToolFromRawPayload(t *testing.T) {
 	})
 }
 
+func TestStripOpenAIImageGenerationToolFromRawPayload(t *testing.T) {
+	payload := []byte(`{
+		"type":"response.create",
+		"model":"gpt-5.4",
+		"tools":[
+			{"type":"function","name":"shell"},
+			{"type":"image_generation","output_format":"png"}
+		],
+		"tool_choice":{"type":"image_generation"}
+	}`)
+
+	updated, changed, err := stripOpenAIImageGenerationToolFromRawPayload(payload)
+
+	require.NoError(t, err)
+	require.True(t, changed)
+	require.False(t, gjson.GetBytes(updated, `tools.#(type=="image_generation")`).Exists())
+	require.True(t, gjson.GetBytes(updated, `tools.#(type=="function")`).Exists())
+	require.False(t, gjson.GetBytes(updated, "tool_choice").Exists())
+}
+
 func TestAlignStoreDisabledPreviousResponseID(t *testing.T) {
 	t.Parallel()
 
diff --git a/frontend/src/components/account/EditAccountModal.vue b/frontend/src/components/account/EditAccountModal.vue
index b5f97753d8..57723d4903 100644
--- a/frontend/src/components/account/EditAccountModal.vue
+++ b/frontend/src/components/account/EditAccountModal.vue
@@ -1430,6 +1430,58 @@
         </div>
       </div>
 
+      <!-- OpenAI Codex 显式 image_generation tool 策略 -->
+      <div
+        v-if="account?.platform === 'openai' && (account?.type === 'oauth' || account?.type === 'setup-token' || account?.type === 'apikey')"
+        class="border-t border-gray-200 pt-4 dark:border-dark-600"
+      >
+        <div class="overflow-hidden rounded-lg border border-gray-200 bg-white shadow-sm dark:border-dark-600 dark:bg-dark-800">
+          <div class="flex items-start gap-3 px-4 py-3">
+            <div class="mt-0.5 flex h-9 w-9 shrink-0 items-center justify-center rounded-md bg-gray-50 text-slate-600 ring-1 ring-gray-200 dark:bg-dark-700 dark:text-slate-300 dark:ring-dark-600">
+              <Icon name="filter" size="sm" />
+            </div>
+            <div class="min-w-0 flex-1">
+              <label class="input-label mb-0">{{ t('admin.accounts.openai.codexImageToolPolicy') }}</label>
+              <p class="mt-1 text-xs leading-5 text-slate-600 dark:text-slate-300">
+                {{ t('admin.accounts.openai.codexImageToolPolicyDesc') }}
+              </p>
+            </div>
+          </div>
+          <div class="border-t border-gray-200 bg-gray-50/70 p-2 dark:border-dark-600 dark:bg-dark-700/50">
+            <div class="grid grid-cols-1 gap-2 sm:grid-cols-2">
+              <button
+                v-for="option in codexImageToolPolicyOptions"
+                :key="option.value"
+                type="button"
+                :data-testid="`codex-image-tool-policy-${option.value}`"
+                @click="codexImageToolPolicyMode = option.value"
+                :class="[
+                  'group flex min-h-[62px] items-start gap-2 rounded-md border px-3 py-2 text-left transition-all',
+                  codexImageToolPolicyMode === option.value
+                    ? 'border-slate-300 bg-white text-slate-900 shadow-sm ring-1 ring-slate-200 dark:border-slate-600 dark:bg-dark-800 dark:text-slate-100 dark:ring-dark-500'
+                    : 'border-transparent bg-transparent text-slate-600 hover:border-gray-200 hover:bg-white dark:text-slate-300 dark:hover:border-dark-500 dark:hover:bg-dark-800'
+                ]"
+              >
+                <span
+                  :class="[
+                    'mt-0.5 flex h-5 w-5 shrink-0 items-center justify-center rounded-full border transition-colors',
+                    codexImageToolPolicyMode === option.value
+                      ? 'border-slate-600 bg-slate-700 text-white dark:border-slate-400 dark:bg-slate-500'
+                      : 'border-gray-300 text-transparent group-hover:border-gray-400 dark:border-dark-500'
+                  ]"
+                >
+                  <Icon name="check" size="xs" :stroke-width="2" />
+                </span>
+                <span class="min-w-0">
+                  <span class="block text-sm font-medium">{{ option.label }}</span>
+                  <span class="mt-0.5 block text-xs leading-4 text-slate-500 dark:text-slate-400">{{ option.description }}</span>
+                </span>
+              </button>
+            </div>
+          </div>
+        </div>
+      </div>
+
       <!-- OpenAI WS Mode 三态（off/ctx_pool/passthrough） -->
       <div
         v-if="account?.platform === 'openai' && (account?.type === 'oauth' || account?.type === 'setup-token' || account?.type === 'apikey')"
@@ -2629,6 +2681,8 @@ const codexCLIOnlyEnabled = ref(false)
 const codexCLIOnlyAppServerEnabled = ref(false)
 type CodexImageGenerationBridgeMode = 'inherit' | 'enabled' | 'disabled'
 const codexImageGenerationBridgeMode = ref<CodexImageGenerationBridgeMode>('inherit')
+type CodexImageToolPolicyMode = 'allow' | 'strip'
+const codexImageToolPolicyMode = ref<CodexImageToolPolicyMode>('allow')
 type AnthropicAPIKeyAuthScheme = 'x_api_key' | 'authorization_bearer'
 const anthropicPassthroughEnabled = ref(false)
 const anthropicAPIKeyAuthScheme = ref<AnthropicAPIKeyAuthScheme>('x_api_key')
@@ -2703,6 +2757,22 @@ const codexImageGenerationBridgeOptions = computed<Array<{
     description: t('admin.accounts.openai.codexImageGenerationBridgeDisabledDesc')
   }
 ])
+const codexImageToolPolicyOptions = computed<Array<{
+  value: CodexImageToolPolicyMode
+  label: string
+  description: string
+}>>(() => [
+  {
+    value: 'allow',
+    label: t('admin.accounts.openai.codexImageToolPolicyAllow'),
+    description: t('admin.accounts.openai.codexImageToolPolicyAllowDesc')
+  },
+  {
+    value: 'strip',
+    label: t('admin.accounts.openai.codexImageToolPolicyStrip'),
+    description: t('admin.accounts.openai.codexImageToolPolicyStripDesc')
+  }
+])
 const codexImageGenerationBridgeBadgeLabel = computed(() => {
   switch (codexImageGenerationBridgeMode.value) {
     case 'enabled':
@@ -3036,6 +3106,7 @@ const syncFormFromAccount = (newAccount: Account | null) => {
   codexCLIOnlyEnabled.value = false
   codexCLIOnlyAppServerEnabled.value = false
   codexImageGenerationBridgeMode.value = 'inherit'
+  codexImageToolPolicyMode.value = 'allow'
   anthropicPassthroughEnabled.value = false
   anthropicAPIKeyAuthScheme.value = 'x_api_key'
   webSearchEmulationMode.value = 'default'
@@ -3059,6 +3130,9 @@ const syncFormFromAccount = (newAccount: Account | null) => {
     } else if (codexImageGenerationBridgeValue === false) {
       codexImageGenerationBridgeMode.value = 'disabled'
     }
+    codexImageToolPolicyMode.value = extra?.codex_image_generation_explicit_tool_policy === 'strip'
+      ? 'strip'
+      : 'allow'
     openaiOAuthResponsesWebSocketV2Mode.value = resolveOpenAIWSModeFromExtra(extra, {
       modeKey: 'openai_oauth_responses_websockets_v2_mode',
       enabledKey: 'openai_oauth_responses_websockets_v2_enabled',
@@ -4199,6 +4273,11 @@ const handleSubmit = async () => {
       } else {
         newExtra.codex_image_generation_bridge = codexImageGenerationBridgeMode.value === 'enabled'
       }
+      if (codexImageToolPolicyMode.value === 'strip') {
+        newExtra.codex_image_generation_explicit_tool_policy = 'strip'
+      } else {
+        delete newExtra.codex_image_generation_explicit_tool_policy
+      }
 
       if (props.account.type === 'oauth' || props.account.type === 'setup-token') {
         if (codexCLIOnlyEnabled.value) {
diff --git a/frontend/src/components/account/__tests__/EditAccountModal.spec.ts b/frontend/src/components/account/__tests__/EditAccountModal.spec.ts
index 381d512ee7..c3646a8e51 100644
--- a/frontend/src/components/account/__tests__/EditAccountModal.spec.ts
+++ b/frontend/src/components/account/__tests__/EditAccountModal.spec.ts
@@ -577,6 +577,41 @@ describe('EditAccountModal', () => {
     expect(updateAccountMock.mock.calls[0]?.[1]?.extra).not.toHaveProperty('codex_image_generation_bridge_enabled')
   })
 
+  it('submits account-level Codex image_generation tool strip policy', async () => {
+    const account = buildAccount()
+    updateAccountMock.mockReset()
+    checkMixedChannelRiskMock.mockReset()
+    checkMixedChannelRiskMock.mockResolvedValue({ has_risk: false })
+    updateAccountMock.mockResolvedValue(account)
+
+    const wrapper = mountModal(account)
+
+    await wrapper.get('button[data-testid="codex-image-tool-policy-strip"]').trigger('click')
+    await wrapper.get('form#edit-account-form').trigger('submit.prevent')
+
+    expect(updateAccountMock).toHaveBeenCalledTimes(1)
+    expect(updateAccountMock.mock.calls[0]?.[1]?.extra?.codex_image_generation_explicit_tool_policy).toBe('strip')
+  })
+
+  it('keeps account-level Codex image_generation tool policy unset when allowing', async () => {
+    const account = buildAccount()
+    account.extra = {
+      codex_image_generation_explicit_tool_policy: 'strip'
+    }
+    updateAccountMock.mockReset()
+    checkMixedChannelRiskMock.mockReset()
+    checkMixedChannelRiskMock.mockResolvedValue({ has_risk: false })
+    updateAccountMock.mockResolvedValue(account)
+
+    const wrapper = mountModal(account)
+
+    await wrapper.get('button[data-testid="codex-image-tool-policy-allow"]').trigger('click')
+    await wrapper.get('form#edit-account-form').trigger('submit.prevent')
+
+    expect(updateAccountMock).toHaveBeenCalledTimes(1)
+    expect(updateAccountMock.mock.calls[0]?.[1]?.extra).not.toHaveProperty('codex_image_generation_explicit_tool_policy')
+  })
+
   it('setup-token account can select and submit OAuth WS mode', async () => {
     const account = buildOpenAISetupTokenAccount()
     updateAccountMock.mockReset()
diff --git a/frontend/src/i18n/locales/en.ts b/frontend/src/i18n/locales/en.ts
index 831b7d7eae..4b6ec9f35f 100644
--- a/frontend/src/i18n/locales/en.ts
+++ b/frontend/src/i18n/locales/en.ts
@@ -3569,6 +3569,13 @@ export default {
         codexImageGenerationBridgeBadgeInherit: 'Channel policy',
         codexImageGenerationBridgeBadgeEnabled: 'Account on',
         codexImageGenerationBridgeBadgeDisabled: 'Account off',
+        codexImageToolPolicy: 'Codex image tool forwarding',
+        codexImageToolPolicyDesc:
+          'Only applies to client-provided image_generation tools in Codex /responses text requests; standalone image-generation endpoints are unaffected.',
+        codexImageToolPolicyAllow: 'Allow',
+        codexImageToolPolicyAllowDesc: 'Default behavior. Keep client-provided image_generation tools.',
+        codexImageToolPolicyStrip: 'Remove',
+        codexImageToolPolicyStripDesc: 'Delete image_generation tools and matching tool_choice before forwarding.',
         compactMode: 'Compact mode',
         compactModeDesc:
           'Controls how this account participates in /responses/compact routing. Auto follows probe results, Force On always allows, Force Off always excludes.',
diff --git a/frontend/src/i18n/locales/zh.ts b/frontend/src/i18n/locales/zh.ts
index 64c62f39f8..9c76037d9d 100644
--- a/frontend/src/i18n/locales/zh.ts
+++ b/frontend/src/i18n/locales/zh.ts
@@ -3742,6 +3742,12 @@ export default {
         codexImageGenerationBridgeBadgeInherit: '渠道策略',
         codexImageGenerationBridgeBadgeEnabled: '账号开启',
         codexImageGenerationBridgeBadgeDisabled: '账号关闭',
+        codexImageToolPolicy: 'Codex 图片工具传递策略',
+        codexImageToolPolicyDesc: '仅作用于 Codex /responses 文本端点中客户端显式携带的 image_generation tool；不影响独立图片生成接口。',
+        codexImageToolPolicyAllow: '放行',
+        codexImageToolPolicyAllowDesc: '默认行为，保留客户端显式发送的 image_generation tool。',
+        codexImageToolPolicyStrip: '移除',
+        codexImageToolPolicyStripDesc: '转发前删除 image_generation tool 和指向它的 tool_choice。',
         compactMode: 'Compact 模式',
         compactModeDesc:
           '控制本账号在 /responses/compact 调度中的参与方式。Auto 跟随探测结果，Force On 强制允许，Force Off 强制排除。',

From ebbdc70311781325d9db2ece7ef1936b487f6053 Mon Sep 17 00:00:00 2001
From: DaydreamCoding <22166516+DaydreamCoding@users.noreply.github.com>
Date: Fri, 3 Jul 2026 23:02:27 +0800
Subject: [PATCH 09/28] =?UTF-8?q?feat(usage):=20=E9=94=99=E8=AF=AF?=
 =?UTF-8?q?=E8=AF=B7=E6=B1=82=E5=AF=B9=E9=BD=90=E7=94=A8=E9=87=8F=E6=98=8E?=
 =?UTF-8?q?=E7=BB=86(UI/=E6=8E=92=E5=BA=8F/=E7=AD=9B=E9=80=89/=E5=88=97?=
 =?UTF-8?q?=E8=AE=BE=E7=BD=AE)?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

错误请求列表(/admin/usage 错误 tab、Ops 弹窗、用户端 /usage 错误 tab)
对齐用量明细的交互与信息密度。Squash of:

- feat(usage): 错误请求全面对齐用量明细(UI/列序/排序/筛选/列设置/新列)
- refactor(usage): 错误表提取共享徽章工具与 IP 批量工具条,后端排序解析归并 SetSort
- feat(usage): /admin/usage 错误请求新增分类过滤
- fix(ops): 错误列表 phase=upstream 过滤生效,守卫豁免改为显式 opt-in
- fix(ops): 错误列表用户列回退显示已删除 KEY 所有者
- fix(usage): 错误请求状态码排序对齐展示/过滤,筛选项改固定常用码
---
 backend/internal/handler/admin/ops_handler.go |  49 +-
 backend/internal/handler/usage_handler.go     |   3 +
 .../repository/ops_error_where_test.go        |  17 +-
 backend/internal/repository/ops_repo.go       |  60 ++-
 backend/internal/service/ops_models.go        |  42 +-
 backend/internal/service/ops_service.go       |   8 +-
 .../service/ops_service_user_error_test.go    |  18 +-
 backend/internal/service/ops_user_error.go    |  21 +-
 .../internal/service/ops_user_error_test.go   |  20 +-
 frontend/src/__tests__/setup.ts               |  14 +
 frontend/src/api/admin/ops.ts                 |  18 +-
 frontend/src/api/admin/usage.ts               |   4 +
 .../components/admin/usage/UsageFilters.vue   |  73 ++-
 frontend/src/components/common/DataTable.vue  |   7 +
 .../components/common/IpGeoBatchToolbar.vue   |  59 +++
 .../user/UserErrorRequestsTable.vue           | 281 +++++-----
 frontend/src/i18n/locales/en.ts               |   3 +-
 frontend/src/i18n/locales/zh.ts               |   3 +-
 frontend/src/types/index.ts                   |   8 +
 frontend/src/utils/errorBadges.ts             |  61 +++
 frontend/src/utils/errorCategory.ts           |  31 ++
 frontend/src/views/admin/UsageView.vue        |  97 +++-
 .../ops/components/OpsErrorDetailsModal.vue   |  15 +-
 .../admin/ops/components/OpsErrorLogTable.vue | 497 +++++++++---------
 frontend/src/views/user/UsageView.vue         | 158 +++++-
 25 files changed, 1108 insertions(+), 459 deletions(-)
 create mode 100644 frontend/src/components/common/IpGeoBatchToolbar.vue
 create mode 100644 frontend/src/utils/errorBadges.ts
 create mode 100644 frontend/src/utils/errorCategory.ts

diff --git a/backend/internal/handler/admin/ops_handler.go b/backend/internal/handler/admin/ops_handler.go
index b9558b97b1..e820aef0c6 100644
--- a/backend/internal/handler/admin/ops_handler.go
+++ b/backend/internal/handler/admin/ops_handler.go
@@ -73,6 +73,13 @@ func NewOpsHandler(opsService *service.OpsService) *OpsHandler {
 }
 
 // GetErrorLogs lists ops error logs.
+// applyOpsErrorSortParams reads sort_by/sort_order query params into the filter.
+// Column whitelist and order normalization live in the repository; unknown
+// values degrade to the default (created_at DESC), mirroring the usage list.
+func applyOpsErrorSortParams(c *gin.Context, filter *service.OpsErrorLogFilter) {
+	filter.SetSort(c.Query("sort_by"), c.Query("sort_order"))
+}
+
 // GET /api/v1/admin/ops/errors
 func (h *OpsHandler) GetErrorLogs(c *gin.Context) {
 	if h.opsService == nil {
@@ -114,10 +121,17 @@ func (h *OpsHandler) GetErrorLogs(c *gin.Context) {
 	// buildOpsErrorLogsWhere 以 COALESCE(requested_model, model) 比对。
 	filter.Model = strings.TrimSpace(c.Query("model"))
 
-	// Force request errors: client-visible status >= 400.
-	// buildOpsErrorLogsWhere already applies this for non-upstream phase.
-	if strings.EqualFold(strings.TrimSpace(filter.Phase), "upstream") {
-		filter.Phase = ""
+	// 请求错误语义:client-visible status>=400 守卫恒生效（未设
+	// IncludeRecoveredUpstream 时 phase=upstream 不再绕过守卫），故
+	// phase=upstream 作为普通过滤条件保留——此前这里清空该值，导致
+	// 错误类型下拉选「上游」等于不过滤。
+
+	// 分类(用户侧粗分类码)→ phase/type ANY 条件,与用户端 /usage/errors 同一映射;
+	// 未知分类返回空切片 = 不过滤。与 phase 参数可同时设置(AND 语义)。
+	if cat := strings.TrimSpace(c.Query("category")); cat != "" {
+		phases, types := service.CategoryToFilter(cat)
+		filter.ErrorPhasesAny = phases
+		filter.ErrorTypesAny = types
 	}
 
 	if platform := strings.TrimSpace(c.Query("platform")); platform != "" {
@@ -187,6 +201,8 @@ func (h *OpsHandler) GetErrorLogs(c *gin.Context) {
 		filter.StatusCodes = out
 	}
 
+	applyOpsErrorSortParams(c, filter)
+
 	result, err := h.opsService.GetErrorLogs(c.Request.Context(), filter)
 	if err != nil {
 		response.ErrorFrom(c, err)
@@ -234,10 +250,17 @@ func (h *OpsHandler) ListRequestErrors(c *gin.Context) {
 	// buildOpsErrorLogsWhere 以 COALESCE(requested_model, model) 比对。
 	filter.Model = strings.TrimSpace(c.Query("model"))
 
-	// Force request errors: client-visible status >= 400.
-	// buildOpsErrorLogsWhere already applies this for non-upstream phase.
-	if strings.EqualFold(strings.TrimSpace(filter.Phase), "upstream") {
-		filter.Phase = ""
+	// 请求错误语义:client-visible status>=400 守卫恒生效（未设
+	// IncludeRecoveredUpstream 时 phase=upstream 不再绕过守卫），故
+	// phase=upstream 作为普通过滤条件保留——此前这里清空该值，导致
+	// 错误类型下拉选「上游」等于不过滤。
+
+	// 分类(用户侧粗分类码)→ phase/type ANY 条件,与用户端 /usage/errors 同一映射;
+	// 未知分类返回空切片 = 不过滤。与 phase 参数可同时设置(AND 语义)。
+	if cat := strings.TrimSpace(c.Query("category")); cat != "" {
+		phases, types := service.CategoryToFilter(cat)
+		filter.ErrorPhasesAny = phases
+		filter.ErrorTypesAny = types
 	}
 
 	if platform := strings.TrimSpace(c.Query("platform")); platform != "" {
@@ -291,6 +314,8 @@ func (h *OpsHandler) ListRequestErrors(c *gin.Context) {
 		filter.StatusCodes = out
 	}
 
+	applyOpsErrorSortParams(c, filter)
+
 	result, err := h.opsService.GetErrorLogs(c.Request.Context(), filter)
 	if err != nil {
 		response.ErrorFrom(c, err)
@@ -362,6 +387,8 @@ func (h *OpsHandler) ListRequestErrorUpstreamErrors(c *gin.Context) {
 	}
 	filter.View = "all"
 	filter.Phase = "upstream"
+	// 上游错误列表需含 status<400 的 recovered 行,显式豁免客户端可见守卫。
+	filter.IncludeRecoveredUpstream = true
 	filter.Owner = "provider"
 	filter.Source = strings.TrimSpace(c.Query("error_source"))
 	filter.Query = strings.TrimSpace(c.Query("q"))
@@ -377,6 +404,8 @@ func (h *OpsHandler) ListRequestErrorUpstreamErrors(c *gin.Context) {
 		filter.ClientRequestID = clientRequestID
 	}
 
+	applyOpsErrorSortParams(c, filter)
+
 	result, err := h.opsService.GetErrorLogs(c.Request.Context(), filter)
 	if err != nil {
 		response.ErrorFrom(c, err)
@@ -442,6 +471,8 @@ func (h *OpsHandler) ListUpstreamErrors(c *gin.Context) {
 
 	filter.View = parseOpsViewParam(c)
 	filter.Phase = "upstream"
+	// 上游错误列表需含 status<400 的 recovered 行,显式豁免客户端可见守卫。
+	filter.IncludeRecoveredUpstream = true
 	filter.Owner = "provider"
 	filter.Source = strings.TrimSpace(c.Query("error_source"))
 	filter.Query = strings.TrimSpace(c.Query("q"))
@@ -497,6 +528,8 @@ func (h *OpsHandler) ListUpstreamErrors(c *gin.Context) {
 		filter.StatusCodes = out
 	}
 
+	applyOpsErrorSortParams(c, filter)
+
 	result, err := h.opsService.GetErrorLogs(c.Request.Context(), filter)
 	if err != nil {
 		response.ErrorFrom(c, err)
diff --git a/backend/internal/handler/usage_handler.go b/backend/internal/handler/usage_handler.go
index 9d0f1d8fac..be6dc917bb 100644
--- a/backend/internal/handler/usage_handler.go
+++ b/backend/internal/handler/usage_handler.go
@@ -322,6 +322,9 @@ func (h *UsageHandler) ListErrors(c *gin.Context) {
 		filter.ErrorTypesAny = types
 	}
 
+	// 排序对齐用量明细:列白名单与方向归一在 repo 层,非法值回退 created_at DESC。
+	filter.SetSort(c.Query("sort_by"), c.Query("sort_order"))
+
 	result, err := h.opsService.ListUserErrorRequests(c.Request.Context(), subject.UserID, filter)
 	if err != nil {
 		response.ErrorFrom(c, err)
diff --git a/backend/internal/repository/ops_error_where_test.go b/backend/internal/repository/ops_error_where_test.go
index 5b9d7ab1c3..c997865ba4 100644
--- a/backend/internal/repository/ops_error_where_test.go
+++ b/backend/internal/repository/ops_error_where_test.go
@@ -85,10 +85,21 @@ func TestBuildOpsErrorLogsWhere_CyberPolicyStatusExemption(t *testing.T) {
 		t.Fatalf("default filter must still include the status >= 400 guard for non-cyber rows\nfull: %s", where)
 	}
 
-	// phase=upstream skips the status guard entirely — exemption is irrelevant there.
+	// phase=upstream WITHOUT the recovered-upstream opt-in keeps the status guard:
+	// request-error list endpoints filter by phase=upstream as a plain condition.
 	whereUpstream, _ := buildOpsErrorLogsWhere(&service.OpsErrorLogFilter{Phase: "upstream"})
-	if strings.Contains(whereUpstream, "status_code") {
-		t.Fatalf("upstream phase filter must not add any status_code clause\nfull: %s", whereUpstream)
+	if !strings.Contains(whereUpstream, "COALESCE(e.status_code, 0) >= 400") {
+		t.Fatalf("upstream phase without IncludeRecoveredUpstream must keep the status guard\nfull: %s", whereUpstream)
+	}
+	if !strings.Contains(whereUpstream, "e.error_phase = $") {
+		t.Fatalf("upstream phase filter must emit the error_phase condition\nfull: %s", whereUpstream)
+	}
+
+	// phase=upstream WITH IncludeRecoveredUpstream (ops 上游列表) skips the guard,
+	// exposing recovered (<400) upstream rows.
+	whereRecovered, _ := buildOpsErrorLogsWhere(&service.OpsErrorLogFilter{Phase: "upstream", IncludeRecoveredUpstream: true})
+	if strings.Contains(whereRecovered, "status_code") {
+		t.Fatalf("upstream phase with IncludeRecoveredUpstream must not add any status_code clause\nfull: %s", whereRecovered)
 	}
 }
 
diff --git a/backend/internal/repository/ops_repo.go b/backend/internal/repository/ops_repo.go
index 9923c08d99..2129a451c4 100644
--- a/backend/internal/repository/ops_repo.go
+++ b/backend/internal/repository/ops_repo.go
@@ -177,6 +177,37 @@ func opsInsertErrorLogArgs(input *service.OpsInsertErrorLogInput) []any {
 	}
 }
 
+// opsErrorLogsOrderBy builds the ORDER BY clause from a whitelist, mirroring
+// usageLogOrderBy semantics. Unknown SortBy falls back to created_at; e.id is
+// always appended as tiebreaker for stable pagination.
+func opsErrorLogsOrderBy(filter *service.OpsErrorLogFilter) string {
+	sortBy := ""
+	sortOrder := ""
+	if filter != nil {
+		sortBy = strings.ToLower(strings.TrimSpace(filter.SortBy))
+		sortOrder = strings.ToLower(strings.TrimSpace(filter.SortOrder))
+	}
+
+	var column string
+	switch sortBy {
+	case "model":
+		column = "COALESCE(NULLIF(TRIM(e.requested_model), ''), e.model)"
+	case "status_code":
+		// 与展示列/过滤保持同义:列表展示 COALESCE(upstream_status_code, status_code, 0),
+		// status_code 过滤也用同一表达式,故排序必须一致——否则 recovered upstream 行
+		//（status_code<400 但展示上游 5xx）排序键与显示值/分页切分不符。
+		column = "COALESCE(e.upstream_status_code, e.status_code, 0)"
+	default:
+		column = "e.created_at"
+	}
+
+	dir := "DESC"
+	if sortOrder == "asc" {
+		dir = "ASC"
+	}
+	return fmt.Sprintf("%s %s, e.id %s", column, dir, dir)
+}
+
 func (r *opsRepository) ListErrorLogs(ctx context.Context, filter *service.OpsErrorLogFilter) (*service.OpsErrorLogList, error) {
 	if r == nil || r.db == nil {
 		return nil, fmt.Errorf("nil ops repository")
@@ -233,25 +264,29 @@ SELECT
   COALESCE(a.name, ''),
   e.group_id,
   COALESCE(g.name, ''),
-  CASE WHEN e.client_ip IS NULL THEN NULL ELSE e.client_ip::text END,
+  CASE WHEN e.client_ip IS NULL THEN NULL ELSE host(e.client_ip) END,
   COALESCE(e.request_path, ''),
   e.stream,
   COALESCE(e.inbound_endpoint, ''),
   COALESCE(e.upstream_endpoint, ''),
   COALESCE(e.requested_model, ''),
   COALESCE(e.upstream_model, ''),
+  COALESCE(e.user_agent, ''),
   e.request_type,
   COALESCE(ak.name, ''),
   ak.deleted_at,
-  COALESCE(e.deleted_key_name, '')
+  COALESCE(e.deleted_key_name, ''),
+  e.deleted_key_owner_user_id,
+  COALESCE(du.email, '')
 FROM ops_error_logs e
 LEFT JOIN accounts a ON e.account_id = a.id
 LEFT JOIN groups g ON e.group_id = g.id
 LEFT JOIN users u ON e.user_id = u.id
 LEFT JOIN users u2 ON e.resolved_by_user_id = u2.id
+LEFT JOIN users du ON e.deleted_key_owner_user_id = du.id
 LEFT JOIN api_keys ak ON ak.id = e.api_key_id
 ` + where + `
-ORDER BY e.created_at DESC
+ORDER BY ` + opsErrorLogsOrderBy(filter) + `
 LIMIT $` + itoa(len(args)+1) + ` OFFSET $` + itoa(len(args)+2)
 
 	rows, err := r.db.QueryContext(ctx, selectSQL, argsWithLimit...)
@@ -279,6 +314,8 @@ LIMIT $` + itoa(len(args)+1) + ` OFFSET $` + itoa(len(args)+2)
 		var apiKeyName string
 		var apiKeyDeletedAt sql.NullTime
 		var deletedKeyName string
+		var deletedKeyOwnerID sql.NullInt64
+		var deletedKeyOwnerEmail string
 		if err := rows.Scan(
 			&item.ID,
 			&item.CreatedAt,
@@ -311,10 +348,13 @@ LIMIT $` + itoa(len(args)+1) + ` OFFSET $` + itoa(len(args)+2)
 			&item.UpstreamEndpoint,
 			&item.RequestedModel,
 			&item.UpstreamModel,
+			&item.UserAgent,
 			&requestType,
 			&apiKeyName,
 			&apiKeyDeletedAt,
 			&deletedKeyName,
+			&deletedKeyOwnerID,
+			&deletedKeyOwnerEmail,
 		); err != nil {
 			return nil, err
 		}
@@ -364,6 +404,12 @@ LIMIT $` + itoa(len(args)+1) + ` OFFSET $` + itoa(len(args)+2)
 		}
 		// 已删除：ak.deleted_at 非空（软删），或仅命中 deleted_key_name 兜底。
 		item.APIKeyDeleted = apiKeyDeletedAt.Valid || (apiKeyName == "" && deletedKeyName != "")
+		// 已删除 KEY 所有者快照:认证失败行 user_id 为空,列表用户列以此回退。
+		if deletedKeyOwnerID.Valid {
+			v := deletedKeyOwnerID.Int64
+			item.DeletedKeyOwnerUserID = &v
+			item.DeletedKeyOwnerEmail = deletedKeyOwnerEmail
+		}
 		out = append(out, &item)
 	}
 	if err := rows.Err(); err != nil {
@@ -417,7 +463,7 @@ SELECT
   COALESCE(a.name, ''),
   e.group_id,
   COALESCE(g.name, ''),
-  CASE WHEN e.client_ip IS NULL THEN NULL ELSE e.client_ip::text END,
+  CASE WHEN e.client_ip IS NULL THEN NULL ELSE host(e.client_ip) END,
   COALESCE(e.request_path, ''),
   e.stream,
   COALESCE(e.inbound_endpoint, ''),
@@ -927,12 +973,14 @@ func buildOpsErrorLogsWhere(filter *service.OpsErrorLogFilter) (string, []any) {
 	if filter != nil {
 		resolvedFilter = filter.Resolved
 	}
-	// Keep list endpoints scoped to client errors unless explicitly filtering upstream phase.
+	// Keep list endpoints scoped to client errors unless the caller explicitly opts
+	// into recovered upstream rows (Phase=="upstream" + IncludeRecoveredUpstream,
+	// ops 专用上游列表)。请求错误语义的端点即便过滤 phase=upstream 也保留该守卫。
 	// cyber_policy is exempt from the status >= 400 guard: streaming cyber hits arrive with
 	// status 200 (the SSE stream opened successfully before upstream returned response.failed),
 	// but they are always client-visible blocked requests that belong in admin + user error
 	// lists.  Without the exemption the entire streaming-path cyber sink would be invisible.
-	if phaseFilter != "upstream" {
+	if phaseFilter != "upstream" || filter == nil || !filter.IncludeRecoveredUpstream {
 		clauses = append(clauses, "(COALESCE(e.status_code, 0) >= 400 OR e.error_type = 'cyber_policy')")
 	}
 
diff --git a/backend/internal/service/ops_models.go b/backend/internal/service/ops_models.go
index 4fc6a9266e..e33dcf82a8 100644
--- a/backend/internal/service/ops_models.go
+++ b/backend/internal/service/ops_models.go
@@ -1,6 +1,9 @@
 package service
 
-import "time"
+import (
+	"strings"
+	"time"
+)
 
 type OpsSystemLog struct {
 	ID              int64          `json:"id"`
@@ -65,17 +68,22 @@ type OpsErrorLog struct {
 	RequestedModel   string `json:"requested_model"`
 	UpstreamModel    string `json:"upstream_model"`
 	RequestType      *int16 `json:"request_type"`
+	UserAgent        string `json:"user_agent"`
 
 	// 关联 api_key 名称（LEFT JOIN api_keys 取得；软删只覆盖 key 列，name 保留，故已删 key 仍有原名）。
 	APIKeyName    string `json:"api_key_name,omitempty"`
 	APIKeyDeleted bool   `json:"api_key_deleted,omitempty"`
+
+	// 已删除 KEY 所有者（INVALID_API_KEY 且该 key 曾存在时的归因快照）。
+	// 认证失败行 user_id 为空，列表用户列以此回退显示所有者。
+	DeletedKeyOwnerUserID *int64 `json:"deleted_key_owner_user_id,omitempty"`
+	DeletedKeyOwnerEmail  string `json:"deleted_key_owner_email,omitempty"`
 }
 
 type OpsErrorLogDetail struct {
 	OpsErrorLog
 
 	ErrorBody string `json:"error_body"`
-	UserAgent string `json:"user_agent"`
 
 	// Upstream context (optional)
 	UpstreamStatusCode   *int   `json:"upstream_status_code,omitempty"`
@@ -93,11 +101,10 @@ type OpsErrorLogDetail struct {
 	// vNext metric semantics
 	IsBusinessLimited bool `json:"is_business_limited"`
 
-	// Deleted key owner info (populated when INVALID_API_KEY and key was previously deleted)
-	AttemptedKeyPrefix    string `json:"attempted_key_prefix,omitempty"`
-	DeletedKeyOwnerUserID *int64 `json:"deleted_key_owner_user_id,omitempty"`
-	DeletedKeyOwnerEmail  string `json:"deleted_key_owner_email,omitempty"`
-	DeletedKeyName        string `json:"deleted_key_name,omitempty"`
+	// Deleted key owner info (populated when INVALID_API_KEY and key was previously deleted).
+	// OwnerUserID/OwnerEmail 已上移到 OpsErrorLog（列表用户列回退需要）。
+	AttemptedKeyPrefix string `json:"attempted_key_prefix,omitempty"`
+	DeletedKeyName     string `json:"deleted_key_name,omitempty"`
 
 	// Bound (non-deleted) key prefix, snapshotted at error time; mutually exclusive with AttemptedKeyPrefix.
 	APIKeyPrefix string `json:"api_key_prefix,omitempty"`
@@ -142,8 +149,14 @@ type OpsErrorLogFilter struct {
 	// ExcludeCountTokens drops count_tokens probe errors (is_count_tokens=true).
 	ExcludeCountTokens bool
 
+	// IncludeRecoveredUpstream 显式豁免 status>=400 守卫（仅在 Phase=="upstream" 时生效）：
+	// ops 专用上游错误列表需要看到 status<400 的 recovered upstream 行。
+	// 请求错误语义的端点不设此开关，phase=upstream 过滤照常生效且守卫保留。
+	IncludeRecoveredUpstream bool
+
 	// ErrorPhasesAny / ErrorTypesAny add plain ANY() filters WITHOUT touching the
-	// special-cased single `Phase` field (only Phase=="upstream" bypasses the status>=400 clause).
+	// special-cased single `Phase` field (only Phase=="upstream" with
+	// IncludeRecoveredUpstream bypasses the status>=400 clause).
 	// NOTE: these ANY filters do NOT bypass status>=400; records with error_phase='upstream'
 	// but status_code<400 (recovered upstream errors) remain excluded.
 	// Used to map user-facing coarse categories to backend conditions.
@@ -158,6 +171,19 @@ type OpsErrorLogFilter struct {
 
 	Page     int
 	PageSize int
+
+	// SortBy/SortOrder: server-side sorting aligned with the usage-log list.
+	// Repo whitelists columns (created_at/model/status_code); anything else
+	// falls back to created_at. SortOrder is "asc"/"desc" (default desc).
+	SortBy    string
+	SortOrder string
+}
+
+// SetSort normalizes raw sort_by/sort_order query values into the filter.
+// Shared by the admin and user-facing error list handlers.
+func (f *OpsErrorLogFilter) SetSort(sortBy, sortOrder string) {
+	f.SortBy = strings.TrimSpace(sortBy)
+	f.SortOrder = strings.TrimSpace(sortOrder)
 }
 
 type OpsErrorLogList struct {
diff --git a/backend/internal/service/ops_service.go b/backend/internal/service/ops_service.go
index a8c8a4bb5c..61f85ef904 100644
--- a/backend/internal/service/ops_service.go
+++ b/backend/internal/service/ops_service.go
@@ -359,10 +359,12 @@ func (s *OpsService) ListUserErrorRequests(ctx context.Context, userID int64, fi
 	filter.UserQuery = ""
 	filter.Owner = ""
 	filter.Source = ""
-	// 清空 Phase 是防御:Phase 是单值特殊字段,仅当其 == "upstream" 时 buildOpsErrorLogsWhere 才跳过 status>=400 子句。
-	// 用户端一律改走 category→ErrorPhasesAny/ErrorTypesAny(纯 ANY 过滤,不影响 status>=400 子句),
-	// 因此 recovered upstream(error_phase='upstream' 但 status<400,最终成功返回)记录对用户不可见——符合预期。
+	// 清空 Phase 是防御:用户端一律改走 category→ErrorPhasesAny/ErrorTypesAny
+	//（纯 ANY 过滤,不影响 status>=400 子句）。守卫豁免现在还需要
+	// IncludeRecoveredUpstream(用户端永不设置),recovered upstream
+	//（error_phase='upstream' 但 status<400,最终成功返回）记录对用户不可见——符合预期。
 	filter.Phase = ""
+	filter.IncludeRecoveredUpstream = false
 
 	list, err := s.opsRepo.ListErrorLogs(ctx, filter)
 	if err != nil {
diff --git a/backend/internal/service/ops_service_user_error_test.go b/backend/internal/service/ops_service_user_error_test.go
index 9027ff0788..c3b0967b67 100644
--- a/backend/internal/service/ops_service_user_error_test.go
+++ b/backend/internal/service/ops_service_user_error_test.go
@@ -184,16 +184,16 @@ func TestGetUserErrorRequestDetail_DeletedKeyOwnerAccess(t *testing.T) {
 	mk := func() *OpsErrorLogDetail {
 		return &OpsErrorLogDetail{
 			OpsErrorLog: OpsErrorLog{
-				ID:            55,
-				Phase:         "auth",
-				Type:          "api_error",
-				StatusCode:    401,
-				Message:       "Invalid API key",
-				UserID:        nil,
-				APIKeyName:    "my-old-key",
-				APIKeyDeleted: true,
+				ID:                    55,
+				Phase:                 "auth",
+				Type:                  "api_error",
+				StatusCode:            401,
+				Message:               "Invalid API key",
+				UserID:                nil,
+				APIKeyName:            "my-old-key",
+				APIKeyDeleted:         true,
+				DeletedKeyOwnerUserID: &ownerUID,
 			},
-			DeletedKeyOwnerUserID: &ownerUID,
 		}
 	}
 
diff --git a/backend/internal/service/ops_user_error.go b/backend/internal/service/ops_user_error.go
index 7dd128afa7..e3055c2392 100644
--- a/backend/internal/service/ops_user_error.go
+++ b/backend/internal/service/ops_user_error.go
@@ -3,9 +3,12 @@ package service
 import "time"
 
 // UserErrorRequest 是面向终端用户的"错误请求"精简脱敏视图（白名单）。
-// 严禁包含 client_ip / user_agent / account / api_key_prefix / upstream_endpoint /
-// user_email 等敏感或内部字段。注：message（网关标准化错误描述）与 key_name
+// 严禁包含 account / api_key_prefix / upstream_endpoint / user_email 等
+// 敏感或内部字段。注：message（网关标准化错误描述）与 key_name
 // （用户自有 API Key 名称，KeysView 中本就可见）经产品决策对该用户开放；
+// client_ip / user_agent / group_name / request_type / stream 均为该用户
+// 自己请求的属性，经产品决策（2026-07-03）开放，
+// 与用量明细已向用户展示自身 ip_address/user_agent/分组/类型 的口径对齐；
 // error_body 仅在详情接口（GetUserErrorRequestDetail）按归属校验后返回。
 type UserErrorRequest struct {
 	ID              int64     `json:"id"`
@@ -18,6 +21,11 @@ type UserErrorRequest struct {
 	Message         string    `json:"message"`
 	KeyName         string    `json:"key_name"`
 	KeyDeleted      bool      `json:"key_deleted"`
+	ClientIP        string    `json:"client_ip,omitempty"`
+	GroupName       string    `json:"group_name,omitempty"`
+	RequestType     *int16    `json:"request_type,omitempty"`
+	Stream          bool      `json:"stream"`
+	UserAgent       string    `json:"user_agent,omitempty"`
 }
 
 // UserErrorRequestList 是用户错误请求分页结果。
@@ -90,6 +98,10 @@ func ToUserErrorRequest(e *OpsErrorLog) *UserErrorRequest {
 	if model == "" {
 		model = e.Model
 	}
+	clientIP := ""
+	if e.ClientIP != nil {
+		clientIP = *e.ClientIP
+	}
 	return &UserErrorRequest{
 		ID:              e.ID,
 		CreatedAt:       e.CreatedAt,
@@ -101,6 +113,11 @@ func ToUserErrorRequest(e *OpsErrorLog) *UserErrorRequest {
 		Message:         e.Message,
 		KeyName:         e.APIKeyName,
 		KeyDeleted:      e.APIKeyDeleted,
+		ClientIP:        clientIP,
+		GroupName:       e.GroupName,
+		RequestType:     e.RequestType,
+		Stream:          e.Stream,
+		UserAgent:       e.UserAgent,
 	}
 }
 
diff --git a/backend/internal/service/ops_user_error_test.go b/backend/internal/service/ops_user_error_test.go
index 31b0c26933..9e0bc164b4 100644
--- a/backend/internal/service/ops_user_error_test.go
+++ b/backend/internal/service/ops_user_error_test.go
@@ -122,9 +122,11 @@ func TestToUserErrorRequestDetail_WhitelistAndRedacts(t *testing.T) {
 			UserEmail:        "secret@example.com",
 			ClientIP:         func() *string { s := "1.2.3.4"; return &s }(),
 			UpstreamEndpoint: "https://api.openai.com/v1/chat/completions",
+			UserAgent:        "codex_cli_rs/0.125.0",
+			GroupName:        "grp-a",
+			Stream:           true,
 		},
 		ErrorBody:          `{"error":{"message":"upstream failed","type":"server_error"}}`,
-		UserAgent:          "Mozilla/5.0 secret-agent",
 		UpstreamStatusCode: &upstreamStatus,
 	}
 
@@ -147,13 +149,27 @@ func TestToUserErrorRequestDetail_WhitelistAndRedacts(t *testing.T) {
 		t.Errorf("UpstreamStatusCode mismatch")
 	}
 
+	// client_ip / user_agent / group_name / stream 经产品决策开放（与用量明细口径对齐）
+	if out.ClientIP != "1.2.3.4" {
+		t.Errorf("want client_ip=1.2.3.4, got %q", out.ClientIP)
+	}
+	if out.UserAgent != "codex_cli_rs/0.125.0" {
+		t.Errorf("want user_agent=codex_cli_rs/0.125.0, got %q", out.UserAgent)
+	}
+	if out.GroupName != "grp-a" {
+		t.Errorf("want group_name=grp-a, got %q", out.GroupName)
+	}
+	if !out.Stream {
+		t.Errorf("want stream=true")
+	}
+
 	// 序列化后不含敏感字段
 	b, err := json.Marshal(out)
 	if err != nil {
 		t.Fatalf("json.Marshal failed: %v", err)
 	}
 	raw := string(b)
-	for _, forbidden := range []string{"user_email", "client_ip", "upstream_endpoint", "user_agent"} {
+	for _, forbidden := range []string{"user_email", "upstream_endpoint"} {
 		if strings.Contains(raw, forbidden) {
 			t.Errorf("sensitive field %q leaked in JSON output: %s", forbidden, raw)
 		}
diff --git a/frontend/src/__tests__/setup.ts b/frontend/src/__tests__/setup.ts
index b777b22e8e..9dad8c1f12 100644
--- a/frontend/src/__tests__/setup.ts
+++ b/frontend/src/__tests__/setup.ts
@@ -57,6 +57,20 @@ if (typeof globalThis.cancelIdleCallback === 'undefined') {
   }) as unknown as typeof cancelIdleCallback
 }
 
+// Mock matchMedia (jsdom 未实现;DataTable 等组件依赖它做桌面/移动分支)
+if (typeof window !== 'undefined' && typeof window.matchMedia !== 'function') {
+  window.matchMedia = ((query: string) => ({
+    matches: true, // 测试默认按桌面视口渲染表格
+    media: query,
+    onchange: null,
+    addListener: vi.fn(),
+    removeListener: vi.fn(),
+    addEventListener: vi.fn(),
+    removeEventListener: vi.fn(),
+    dispatchEvent: vi.fn(),
+  })) as unknown as typeof window.matchMedia
+}
+
 // Mock IntersectionObserver
 class MockIntersectionObserver {
   observe = vi.fn()
diff --git a/frontend/src/api/admin/ops.ts b/frontend/src/api/admin/ops.ts
index b3e53893ce..c7cbc64a4b 100644
--- a/frontend/src/api/admin/ops.ts
+++ b/frontend/src/api/admin/ops.ts
@@ -930,11 +930,16 @@ export interface OpsErrorLog {
   requested_model?: string
   upstream_model?: string
   request_type?: number | null
+  user_agent?: string
+
+  // 已删除 KEY 所有者(INVALID_API_KEY 归因快照):认证失败行 user_id 为空,
+  // 用户列以此回退显示所有者
+  deleted_key_owner_user_id?: number | null
+  deleted_key_owner_email?: string | null
 }
 
 export interface OpsErrorDetail extends OpsErrorLog {
   error_body: string
-  user_agent: string
 
   // Upstream context (optional; enriched by gateway services)
   upstream_status_code?: number | null
@@ -950,10 +955,9 @@ export interface OpsErrorDetail extends OpsErrorLog {
 
   is_business_limited: boolean
 
-  // Deleted key owner info (INVALID_API_KEY attribution)
+  // Deleted key owner info (INVALID_API_KEY attribution);
+  // owner user_id/email 已上移到 OpsErrorLog(列表用户列回退)
   attempted_key_prefix?: string | null
-  deleted_key_owner_user_id?: number | null
-  deleted_key_owner_email?: string | null
   deleted_key_name?: string | null
 
   // Bound (non-deleted) key prefix, snapshotted at error time
@@ -1098,6 +1102,8 @@ export type OpsErrorListQueryParams = {
   model?: string
 
   phase?: string
+  // 分类(用户侧粗分类码,如 auth/rate_limit/upstream),后端反查为 phase/type ANY 条件
+  category?: string
   error_owner?: string
   error_source?: string
   resolved?: string
@@ -1106,6 +1112,10 @@ export type OpsErrorListQueryParams = {
   q?: string
   status_codes?: string
   status_codes_other?: string
+
+  // 服务端排序,列白名单见后端 opsErrorLogsOrderBy(created_at/model/status_code)
+  sort_by?: string
+  sort_order?: 'asc' | 'desc'
 }
 
 // Legacy unified endpoints
diff --git a/frontend/src/api/admin/usage.ts b/frontend/src/api/admin/usage.ts
index b37996d61a..83be033c11 100644
--- a/frontend/src/api/admin/usage.ts
+++ b/frontend/src/api/admin/usage.ts
@@ -86,6 +86,10 @@ export interface AdminUsageQueryParams extends UsageQueryParams {
   billing_mode?: string
   sort_by?: string
   sort_order?: 'asc' | 'desc'
+  // 错误请求 tab 专属筛选(仅传给错误列表接口;共用同一 filters 对象)
+  error_phase?: string | null
+  error_category?: string | null
+  status_code?: number | null
 }
 
 // ==================== API Functions ====================
diff --git a/frontend/src/components/admin/usage/UsageFilters.vue b/frontend/src/components/admin/usage/UsageFilters.vue
index f16012ccff..bb63d9b8ee 100644
--- a/frontend/src/components/admin/usage/UsageFilters.vue
+++ b/frontend/src/components/admin/usage/UsageFilters.vue
@@ -121,24 +121,42 @@
           </div>
         </div>
 
-        <!-- Request Type Filter -->
-        <div class="w-full sm:w-auto sm:min-w-[180px]">
+        <!-- Request Type Filter (usage only) -->
+        <div v-if="mode !== 'errors'" class="w-full sm:w-auto sm:min-w-[180px]">
           <label class="input-label">{{ t('usage.type') }}</label>
           <Select v-model="filters.request_type" :options="requestTypeOptions" @change="emitChange" />
         </div>
 
-        <!-- Billing Type Filter -->
-        <div class="w-full sm:w-auto sm:min-w-[200px]">
+        <!-- Billing Type Filter (usage only) -->
+        <div v-if="mode !== 'errors'" class="w-full sm:w-auto sm:min-w-[200px]">
           <label class="input-label">{{ t('admin.usage.billingType') }}</label>
           <Select v-model="filters.billing_type" :options="billingTypeOptions" @change="emitChange" />
         </div>
 
-        <!-- Billing Mode Filter -->
-        <div class="w-full sm:w-auto sm:min-w-[200px]">
+        <!-- Billing Mode Filter (usage only) -->
+        <div v-if="mode !== 'errors'" class="w-full sm:w-auto sm:min-w-[200px]">
           <label class="input-label">{{ t('admin.usage.billingMode') }}</label>
           <Select v-model="filters.billing_mode" :options="billingModeOptions" @change="emitChange" />
         </div>
 
+        <!-- Error Phase Filter (errors only) -->
+        <div v-if="mode === 'errors'" class="w-full sm:w-auto sm:min-w-[180px]">
+          <label class="input-label">{{ t('admin.ops.errorLog.type') }}</label>
+          <Select v-model="filters.error_phase" :options="errorPhaseOptions" @change="emitChange" />
+        </div>
+
+        <!-- Error Category Filter (errors only) -->
+        <div v-if="mode === 'errors'" class="w-full sm:w-auto sm:min-w-[180px]">
+          <label class="input-label">{{ t('usage.errors.category') }}</label>
+          <Select v-model="filters.error_category" :options="errorCategoryOptions" @change="emitChange" />
+        </div>
+
+        <!-- Status Code Filter (errors only) -->
+        <div v-if="mode === 'errors'" class="w-full sm:w-auto sm:min-w-[180px]">
+          <label class="input-label">{{ t('admin.ops.errorLog.status') }}</label>
+          <Select v-model="filters.status_code" :options="statusCodeOptions" @change="emitChange" />
+        </div>
+
         <!-- Group Filter -->
         <div class="w-full sm:w-auto sm:min-w-[200px]">
           <label class="input-label">{{ t('admin.usage.group') }}</label>
@@ -156,12 +174,14 @@
           {{ t('common.reset') }}
         </button>
         <slot name="after-reset" />
-        <button type="button" @click="$emit('cleanup')" class="btn btn-danger">
-          {{ t('admin.usage.cleanup.button') }}
-        </button>
-        <button type="button" @click="$emit('export')" :disabled="exporting" class="btn btn-primary">
-          {{ t('usage.exportExcel') }}
-        </button>
+        <template v-if="mode !== 'errors'">
+          <button type="button" @click="$emit('cleanup')" class="btn btn-danger">
+            {{ t('admin.usage.cleanup.button') }}
+          </button>
+          <button type="button" @click="$emit('export')" :disabled="exporting" class="btn btn-primary">
+            {{ t('usage.exportExcel') }}
+          </button>
+        </template>
       </div>
     </div>
   </div>
@@ -172,6 +192,7 @@ import { ref, onMounted, onUnmounted, toRef, watch, computed } from 'vue'
 import { useI18n } from 'vue-i18n'
 import { adminAPI } from '@/api/admin'
 import Select, { type SelectOption } from '@/components/common/Select.vue'
+import { COMMON_ERROR_STATUS_CODES } from '@/utils/errorBadges'
 import type { SimpleApiKey, SimpleUser } from '@/api/admin/usage'
 
 type ModelValue = Record<string, any>
@@ -183,10 +204,13 @@ interface Props {
   endDate: string
   showActions?: boolean
   modelOptions?: string[]
+  /** errors 模式:隐藏用量专属字段/按钮,显示错误类型+状态码(错误请求 tab 用) */
+  mode?: 'usage' | 'errors'
 }
 
 const props = withDefaults(defineProps<Props>(), {
-  showActions: true
+  showActions: true,
+  mode: 'usage'
 })
 const emit = defineEmits([
   'update:modelValue',
@@ -243,6 +267,29 @@ const billingTypeOptions = ref<SelectOption[]>([
   { value: 1, label: t('admin.usage.billingTypeSubscription') }
 ])
 
+// 错误类型对应后端 phase 参数(与错误表"类型"徽章同语义)
+const errorPhaseOptions = computed<SelectOption[]>(() => [
+  { value: null, label: t('admin.usage.allTypes') },
+  { value: 'upstream', label: t('admin.ops.errorLog.typeUpstream') },
+  { value: 'request', label: t('admin.ops.errorLog.typeRequest') },
+  { value: 'auth', label: t('admin.ops.errorLog.typeAuth') },
+  { value: 'routing', label: t('admin.ops.errorLog.typeRouting') },
+  { value: 'internal', label: t('admin.ops.errorLog.typeInternal') },
+])
+
+// 分类码同用户端 /usage 错误筛选;"other" 无法反查为过滤条件,刻意不列
+const errorCategoryCodes = ['auth', 'rate_limit', 'quota', 'invalid_request', 'service_unavailable', 'upstream', 'internal', 'cyber']
+
+const errorCategoryOptions = computed<SelectOption[]>(() => [
+  { value: null, label: t('usage.errors.allCategories') },
+  ...errorCategoryCodes.map((c) => ({ value: c, label: t('usage.errors.categories.' + c) })),
+])
+
+const statusCodeOptions = computed<SelectOption[]>(() => [
+  { value: null, label: t('usage.errors.allStatuses') },
+  ...COMMON_ERROR_STATUS_CODES.map((c) => ({ value: c, label: String(c) })),
+])
+
 const billingModeOptions = ref<SelectOption[]>([
   { value: null, label: t('admin.usage.allBillingModes') },
   { value: 'token', label: t('admin.usage.billingModeToken') },
diff --git a/frontend/src/components/common/DataTable.vue b/frontend/src/components/common/DataTable.vue
index bb7315dc21..d38ebeb0e1 100644
--- a/frontend/src/components/common/DataTable.vue
+++ b/frontend/src/components/common/DataTable.vue
@@ -36,6 +36,8 @@
         v-for="(row, index) in sortedData"
         :key="resolveRowKey(row, index)"
         class="rounded-lg border border-gray-200 bg-white p-4 dark:border-dark-700 dark:bg-dark-900"
+        :class="{ 'cursor-pointer': clickableRows }"
+        @click="clickableRows && emit('rowClick', row)"
       >
         <div class="space-y-3">
           <div
@@ -166,6 +168,8 @@
             :data-index="virtualRow.index"
             :ref="measureElement"
             class="hover:bg-gray-50 dark:hover:bg-dark-800"
+            :class="{ 'cursor-pointer': clickableRows }"
+            @click="clickableRows && emit('rowClick', sortedData[virtualRow.index])"
           >
             <td
               v-for="(column, colIndex) in columns"
@@ -214,6 +218,7 @@ const isDesktopViewport = ref(
 
 const emit = defineEmits<{
   sort: [key: string, order: 'asc' | 'desc']
+  rowClick: [row: any]
 }>()
 
 // 表格容器引用
@@ -381,6 +386,8 @@ interface Props {
    * will emit 'sort' events instead of performing client-side sorting.
    */
   serverSideSort?: boolean
+  /** Emit 'rowClick' on row/card click and show pointer cursor (interactive cells should @click.stop) */
+  clickableRows?: boolean
   /** Estimated row height in px for the virtualizer (default 56) */
   estimateRowHeight?: number
   /** Number of rows to render beyond the visible area (default 5) */
diff --git a/frontend/src/components/common/IpGeoBatchToolbar.vue b/frontend/src/components/common/IpGeoBatchToolbar.vue
new file mode 100644
index 0000000000..07199f1b2a
--- /dev/null
+++ b/frontend/src/components/common/IpGeoBatchToolbar.vue
@@ -0,0 +1,59 @@
+<template>
+  <div
+    v-if="uniqueIps.length > 0"
+    class="flex flex-shrink-0 items-center justify-end gap-2 border-b border-gray-200 px-4 py-2 dark:border-dark-700"
+  >
+    <span v-if="pendingCount > 0" class="text-xs text-gray-500 dark:text-gray-400">
+      {{ t('usage.ipGeo.pending', { count: pendingCount }) }}
+    </span>
+    <button
+      type="button"
+      class="inline-flex items-center gap-1 rounded px-2 py-1 text-xs font-medium text-primary-600 transition-colors hover:bg-primary-50 disabled:cursor-not-allowed disabled:opacity-50 dark:text-primary-400 dark:hover:bg-primary-900/30"
+      :disabled="loading || pendingCount === 0"
+      @click="run"
+    >
+      {{ loading ? t('usage.ipGeo.batchFetching') : t('usage.ipGeo.batchFetch') }}
+    </button>
+  </div>
+</template>
+
+<script setup lang="ts">
+import { computed, ref } from 'vue'
+import { useI18n } from 'vue-i18n'
+import { fetchBatch, getEntry } from '@/utils/ipGeoLookup'
+
+// 当前页 IP 批量地理查询工具条:传入原始 IP 列表(可含空值),内部去重;
+// 无 IP 时自身不渲染。批量失败 emit failed,由使用方弹提示。
+const props = defineProps<{
+  ips: Array<string | null | undefined>
+}>()
+
+const emit = defineEmits<{
+  (e: 'failed'): void
+}>()
+
+const { t } = useI18n()
+
+const uniqueIps = computed(() =>
+  Array.from(new Set(props.ips.filter((ip): ip is string => Boolean(ip))))
+)
+
+const pendingCount = computed(() =>
+  uniqueIps.value.filter((ip) => {
+    const status = getEntry(ip).status
+    return status === 'idle' || status === 'error'
+  }).length
+)
+
+const loading = ref(false)
+
+const run = async () => {
+  loading.value = true
+  try {
+    const ok = await fetchBatch(uniqueIps.value)
+    if (!ok) emit('failed')
+  } finally {
+    loading.value = false
+  }
+}
+</script>
diff --git a/frontend/src/components/user/UserErrorRequestsTable.vue b/frontend/src/components/user/UserErrorRequestsTable.vue
index 42cb93ce04..2465fe2f66 100644
--- a/frontend/src/components/user/UserErrorRequestsTable.vue
+++ b/frontend/src/components/user/UserErrorRequestsTable.vue
@@ -1,91 +1,119 @@
 <template>
   <div class="flex min-h-0 flex-1 flex-col">
-    <div class="px-6 py-4 flex-shrink-0">
-      <div class="flex flex-wrap items-end gap-4">
-        <div class="min-w-[180px]">
-          <label class="input-label">{{ t('usage.errors.model') }}</label>
-          <Select
-            v-model="localModel"
-            :options="modelOptions"
-            searchable
-            creatable
-            clearable
-            :placeholder="t('usage.errors.modelPlaceholder')"
-            @change="apply"
-          />
-        </div>
-        <div class="min-w-[160px]">
-          <label class="input-label">{{ t('usage.errors.keyName') }}</label>
-          <Select
-            v-model="localApiKeyId"
-            :options="keyOptions"
-            :placeholder="t('usage.errors.allKeys')"
-            @change="apply"
-          />
-        </div>
-        <div class="min-w-[140px]">
-          <label class="input-label">{{ t('usage.errors.category') }}</label>
-          <Select
-            v-model="localCategory"
-            :options="categoryOptions"
-            :placeholder="t('usage.errors.allCategories')"
-            @change="apply"
-          />
-        </div>
-        <button class="btn btn-primary" @click="apply">
-          <Icon name="search" size="sm" />
-          {{ t('common.search') }}
-        </button>
-      </div>
-    </div>
+    <div class="card flex min-h-0 flex-1 flex-col overflow-hidden">
+      <IpGeoBatchToolbar :ips="rows.map((r) => r.client_ip)" @failed="emit('ipGeoBatchFailed')" />
 
-    <div class="min-h-0 flex-1 overflow-auto">
-      <table class="min-w-full text-sm">
-        <thead>
-          <tr>
-            <th class="px-4 py-2 text-left">{{ t('usage.errors.model') }}</th>
-            <th class="px-4 py-2 text-left">{{ t('usage.errors.keyName') }}</th>
-            <th class="px-4 py-2 text-left">{{ t('usage.errors.endpoint') }}</th>
-            <th class="px-4 py-2 text-left">{{ t('usage.errors.status') }}</th>
-            <th class="px-4 py-2 text-left">{{ t('usage.errors.category') }}</th>
-            <th class="px-4 py-2 text-left">{{ t('usage.errors.message') }}</th>
-            <th class="px-4 py-2 text-left">{{ t('usage.errors.platform') }}</th>
-            <th class="px-4 py-2 text-left">{{ t('usage.errors.time') }}</th>
-          </tr>
-        </thead>
-        <tbody>
-          <tr
-            v-for="(row, i) in rows"
-            :key="i"
-            class="border-t border-gray-100 dark:border-dark-700 cursor-pointer hover:bg-gray-50 dark:hover:bg-dark-800"
-            @click="openDetail(row.id)"
-          >
-            <td class="px-4 py-2">{{ row.model || '-' }}</td>
-            <td class="px-4 py-2">
-              <span>{{ row.key_name || '-' }}</span>
-              <span
-                v-if="row.key_deleted"
-                class="ml-1 inline-flex items-center rounded px-1 py-px text-[10px] font-medium leading-tight bg-gray-100 text-gray-500 dark:bg-dark-700 dark:text-gray-400"
-              >{{ t('usage.errors.keyDeleted') }}</span>
-            </td>
-            <td class="px-4 py-2">{{ row.inbound_endpoint || '-' }}</td>
-            <td class="px-4 py-2"><span class="badge" :class="statusClass(row.status_code)">{{ row.status_code || '-' }}</span></td>
-            <td class="px-4 py-2">{{ t('usage.errors.categories.' + row.category) }}</td>
-            <td class="px-4 py-2 max-w-[280px] truncate" :title="row.message">{{ row.message || '-' }}</td>
-            <td class="px-4 py-2">{{ row.platform || '-' }}</td>
-            <td class="px-4 py-2">{{ formatDateTime(row.created_at) }}</td>
-          </tr>
-          <tr v-if="!loading && rows.length === 0">
-            <td colspan="8" class="px-4 py-8 text-center text-gray-400">{{ t('usage.errors.empty') }}</td>
-          </tr>
-        </tbody>
-      </table>
+      <DataTable
+        :columns="columns"
+        :data="rows"
+        :loading="loading"
+        clickable-rows
+        server-side-sort
+        default-sort-key="created_at"
+        default-sort-order="desc"
+        @sort="onSort"
+        @rowClick="(row) => openDetail(row.id)"
+      >
+        <template #cell-model="{ row }">
+          <span v-if="row.model" class="text-sm font-medium text-gray-900 dark:text-white">{{ row.model }}</span>
+          <span v-else class="text-sm text-gray-400 dark:text-gray-500">-</span>
+        </template>
+
+        <template #cell-key_name="{ row }">
+          <div class="text-sm">
+            <span class="text-gray-900 dark:text-white">{{ row.key_name || '-' }}</span>
+            <span
+              v-if="row.key_deleted"
+              class="ml-1 inline-flex items-center rounded px-1 py-px text-[10px] font-medium leading-tight bg-rose-100 text-rose-600 ring-1 ring-inset ring-rose-200 dark:bg-rose-500/20 dark:text-rose-400 dark:ring-rose-500/30"
+            >{{ t('usage.errors.keyDeleted') }}</span>
+          </div>
+        </template>
+
+        <template #cell-endpoint="{ row }">
+          <div class="max-w-[320px] text-xs">
+            <div class="break-all text-gray-700 dark:text-gray-300">
+              <span class="font-medium text-gray-500 dark:text-gray-400">{{ t('usage.inbound') }}:</span>
+              <span class="ml-1">{{ row.inbound_endpoint?.trim() || '-' }}</span>
+            </div>
+          </div>
+        </template>
+
+        <template #cell-status="{ row }">
+          <span class="inline-flex items-center rounded px-2 py-0.5 text-xs font-medium" :class="statusClass(row.status_code)">
+            {{ row.status_code || '-' }}
+          </span>
+        </template>
+
+        <template #cell-category="{ row }">
+          <span class="text-sm text-gray-900 dark:text-white">{{ t('usage.errors.categories.' + row.category) }}</span>
+        </template>
+
+        <template #cell-message="{ row }">
+          <span
+            v-if="row.message"
+            class="block max-w-[280px] truncate text-sm text-gray-600 dark:text-gray-400"
+            :title="row.message"
+          >{{ row.message }}</span>
+          <span v-else class="text-sm text-gray-400 dark:text-gray-500">-</span>
+        </template>
+
+        <template #cell-group="{ row }">
+          <span
+            v-if="row.group_name"
+            class="inline-flex items-center rounded px-2 py-0.5 text-xs font-medium bg-indigo-100 text-indigo-800 dark:bg-indigo-900 dark:text-indigo-200"
+          >{{ row.group_name }}</span>
+          <span v-else class="text-sm text-gray-400 dark:text-gray-500">-</span>
+        </template>
+
+        <template #cell-type="{ row }">
+          <span
+            v-if="requestTypeBadge(row)"
+            class="inline-flex items-center rounded px-2 py-0.5 text-xs font-medium"
+            :class="requestTypeBadge(row)!.className"
+          >{{ requestTypeBadge(row)!.label }}</span>
+          <span v-else class="text-sm text-gray-400 dark:text-gray-500">-</span>
+        </template>
+
+        <template #cell-platform="{ row }">
+          <span class="text-sm text-gray-900 dark:text-white">{{ row.platform || '-' }}</span>
+        </template>
+
+        <template #cell-client_ip="{ row }">
+          <div @click.stop>
+            <div v-if="row.client_ip">
+              <span class="text-sm font-mono text-gray-600 dark:text-gray-400">{{ row.client_ip }}</span>
+              <IpGeoCell :ip="row.client_ip" />
+            </div>
+            <span v-else class="text-sm text-gray-400 dark:text-gray-500">-</span>
+          </div>
+        </template>
+
+        <template #cell-created_at="{ row }">
+          <span class="text-sm text-gray-600 dark:text-gray-400">{{ formatDateTime(row.created_at) }}</span>
+        </template>
+
+        <template #cell-user_agent="{ row }">
+          <span
+            v-if="row.user_agent"
+            class="block max-w-[320px] truncate text-sm text-gray-600 dark:text-gray-400"
+            :title="row.user_agent"
+          >{{ row.user_agent }}</span>
+          <span v-else class="text-sm text-gray-400 dark:text-gray-500">-</span>
+        </template>
+
+        <template #empty><EmptyState :message="t('usage.errors.empty')" /></template>
+      </DataTable>
     </div>
 
     <div class="flex-shrink-0">
-      <Pagination :page="page" :page-size="pageSize" :total="total"
+      <Pagination
+        v-if="total > 0"
+        :page="page"
+        :page-size="pageSize"
+        :total="total"
         @update:page="$emit('update:page', $event)"
-        @update:pageSize="$emit('update:pageSize', $event)" />
+        @update:pageSize="$emit('update:pageSize', $event)"
+      />
     </div>
 
     <UserErrorDetailModal v-model:show="showDetail" :error-id="selectedId" />
@@ -95,12 +123,22 @@
 <script setup lang="ts">
 import { ref, computed } from 'vue'
 import { useI18n } from 'vue-i18n'
+import DataTable from '@/components/common/DataTable.vue'
+import EmptyState from '@/components/common/EmptyState.vue'
 import Pagination from '@/components/common/Pagination.vue'
-import Select from '@/components/common/Select.vue'
-import Icon from '@/components/icons/Icon.vue'
 import UserErrorDetailModal from '@/components/user/UserErrorDetailModal.vue'
+import IpGeoCell from '@/components/common/IpGeoCell.vue'
+import IpGeoBatchToolbar from '@/components/common/IpGeoBatchToolbar.vue'
 import { formatDateTime } from '@/utils/format'
-import type { UserErrorRequest, ApiKey } from '@/types'
+import {
+  mapErrorSortKey,
+  numericRequestTypeKind,
+  requestTypeBadgeClass,
+  requestTypeLabelKey,
+  statusCodeBadgeClass,
+} from '@/utils/errorBadges'
+import type { UserErrorRequest } from '@/types'
+import type { Column } from '@/components/common/types'
 
 const props = defineProps<{
   rows: UserErrorRequest[]
@@ -108,46 +146,51 @@ const props = defineProps<{
   loading: boolean
   page: number
   pageSize: number
-  apiKeys?: ApiKey[]
+  /** 列设置:仅显示这些 key 的列;不传则全显(key 须与 allColumns 一致) */
+  visibleColumnKeys?: string[]
 }>()
 
 const emit = defineEmits<{
   (e: 'update:page', v: number): void
   (e: 'update:pageSize', v: number): void
-  (e: 'filter', v: { model: string; category: string; api_key_id: number | null }): void
+  (e: 'ipGeoBatchFailed'): void
+  (e: 'sort', sortBy: string, sortOrder: 'asc' | 'desc'): void
 }>()
 
+function onSort(key: string, order: 'asc' | 'desc') {
+  emit('sort', mapErrorSortKey(key), order)
+}
+
 const { t } = useI18n()
-// string | null:clearable 清空时 Select 回传 null,apply 中归一为空串
-const localModel = ref<string | null>('')
-const localCategory = ref<string>('')
-const localApiKeyId = ref<number | null>(null)
 
-const categoryCodes = ['auth', 'rate_limit', 'quota', 'invalid_request', 'service_unavailable', 'upstream', 'internal', 'cyber']
-
-const categoryOptions = computed(() => [
-  { value: '', label: t('usage.errors.allCategories') },
-  ...categoryCodes.map((c) => ({ value: c, label: t('usage.errors.categories.' + c) })),
+// 列序对齐用户端用量明细:Key → 模型 → 端点 → IP → 分组 → 类型 → 平台 → 分类
+// → 结果(状态→消息)→ 时间 → UA(用量明细 UA 同在时间之后的尾部)
+const allColumns = computed<Column[]>(() => [
+  { key: 'key_name', label: t('usage.errors.keyName') },
+  { key: 'model', label: t('usage.errors.model'), sortable: true },
+  { key: 'endpoint', label: t('usage.errors.endpoint') },
+  { key: 'client_ip', label: 'IP' },
+  { key: 'group', label: t('admin.usage.group') },
+  { key: 'type', label: t('usage.type') },
+  { key: 'platform', label: t('usage.errors.platform') },
+  { key: 'category', label: t('usage.errors.category') },
+  { key: 'status', label: t('usage.errors.status'), sortable: true },
+  { key: 'message', label: t('usage.errors.message') },
+  { key: 'created_at', label: t('usage.errors.time'), sortable: true },
+  { key: 'user_agent', label: t('usage.userAgent') },
 ])
 
-// 首项 value: null 表示不按 key 过滤；其余项取自父组件传入的 apiKeys 候选列表。
-const keyOptions = computed(() => [
-  { value: null, label: t('usage.errors.allKeys') },
-  ...(props.apiKeys ?? []).map((k) => ({ value: k.id, label: k.name })),
-])
+const columns = computed<Column[]>(() =>
+  props.visibleColumnKeys
+    ? allColumns.value.filter((c) => props.visibleColumnKeys!.includes(c.key))
+    : allColumns.value
+)
 
-// 模型候选取自当前已加载错误中出现过的模型；creatable 允许输入任意片段做后端模糊。
-const modelOptions = computed(() => {
-  const seen = new Set<string>()
-  const opts: { value: string; label: string }[] = []
-  for (const r of props.rows) {
-    if (r.model && !seen.has(r.model)) {
-      seen.add(r.model)
-      opts.push({ value: r.model, label: r.model })
-    }
-  }
-  return opts
-})
+function requestTypeBadge(row: UserErrorRequest): { label: string; className: string } | null {
+  const kind = numericRequestTypeKind(row.request_type, row.stream)
+  if (!kind) return null
+  return { label: t(requestTypeLabelKey(kind)), className: requestTypeBadgeClass(kind) }
+}
 
 const showDetail = ref(false)
 const selectedId = ref<number | null>(null)
@@ -157,17 +200,5 @@ function openDetail(id: number) {
   showDetail.value = true
 }
 
-function apply() {
-  emit('filter', {
-    model: (localModel.value ?? '').trim(),
-    category: localCategory.value || '',
-    api_key_id: localApiKeyId.value,
-  })
-}
-
-function statusClass(code: number) {
-  if (code >= 500) return 'badge-danger'
-  if (code === 429) return 'badge-warning'
-  return 'badge-gray'
-}
+const statusClass = statusCodeBadgeClass
 </script>
diff --git a/frontend/src/i18n/locales/en.ts b/frontend/src/i18n/locales/en.ts
index 831b7d7eae..7e2fd5efd0 100644
--- a/frontend/src/i18n/locales/en.ts
+++ b/frontend/src/i18n/locales/en.ts
@@ -1017,7 +1017,7 @@ export default {
       time: 'Time', model: 'Model', endpoint: 'Endpoint', status: 'Status',
       category: 'Category', platform: 'Platform', message: 'Message',
       keyName: 'Key Name', keyDeleted: 'Deleted', allKeys: 'All keys',
-      modelPlaceholder: 'Search model', allCategories: 'All categories',
+      modelPlaceholder: 'Search model', allCategories: 'All categories', allStatuses: 'All status codes',
       empty: 'No error requests', failedToLoad: 'Failed to load error requests',
       categories: {
         auth: 'Auth failed', rate_limit: 'Rate limited', quota: 'Balance/Subscription',
@@ -5088,6 +5088,7 @@ export default {
         accountId: 'Account ID',
         status: 'Status',
         message: 'Message',
+        ip: 'IP',
         latency: 'Request Duration',
         action: 'Action',
         noErrors: 'No errors in this window.',
diff --git a/frontend/src/i18n/locales/zh.ts b/frontend/src/i18n/locales/zh.ts
index 64c62f39f8..e0a4668b90 100644
--- a/frontend/src/i18n/locales/zh.ts
+++ b/frontend/src/i18n/locales/zh.ts
@@ -1021,7 +1021,7 @@ export default {
       time: '时间', model: '模型', endpoint: '端点', status: '状态码',
       category: '分类', platform: '平台', message: '错误信息',
       keyName: 'Key 名称', keyDeleted: '已删除', allKeys: '全部 Key',
-      modelPlaceholder: '搜索模型', allCategories: '全部分类',
+      modelPlaceholder: '搜索模型', allCategories: '全部分类', allStatuses: '全部状态码',
       empty: '暂无错误请求', failedToLoad: '加载错误请求失败',
       categories: {
         auth: '认证失败', rate_limit: '限流', quota: '余额/订阅',
@@ -5246,6 +5246,7 @@ export default {
         accountId: '账号 ID',
         status: '状态码',
         message: '响应内容',
+        ip: 'IP',
         latency: '请求时长',
         action: '操作',
         noErrors: '该窗口内暂无错误。',
diff --git a/frontend/src/types/index.ts b/frontend/src/types/index.ts
index 3636791ebd..a933371dd8 100644
--- a/frontend/src/types/index.ts
+++ b/frontend/src/types/index.ts
@@ -1688,6 +1688,11 @@ export interface UserErrorRequest {
   message: string
   key_name: string
   key_deleted: boolean
+  client_ip?: string
+  group_name?: string
+  request_type?: number
+  stream?: boolean
+  user_agent?: string
 }
 
 export interface UserErrorRequestDetail extends UserErrorRequest {
@@ -1705,6 +1710,9 @@ export interface UserErrorListParams {
   status_code?: number
   category?: string
   api_key_id?: number
+  // 服务端排序,列白名单见后端 opsErrorLogsOrderBy(created_at/model/status_code)
+  sort_by?: string
+  sort_order?: 'asc' | 'desc'
 }
 
 export interface UsageQueryParams {
diff --git a/frontend/src/utils/errorBadges.ts b/frontend/src/utils/errorBadges.ts
new file mode 100644
index 0000000000..bdb4657547
--- /dev/null
+++ b/frontend/src/utils/errorBadges.ts
@@ -0,0 +1,61 @@
+/**
+ * 错误请求/用量明细共享的徽章配色与列映射。
+ * 配色统一 bg-X-100/text-X-800 体系,与 UsageTable 一致。
+ */
+
+import type { UsageRequestType } from '@/types'
+
+export type UsageRequestKind = UsageRequestType
+
+/** 状态码徽章:≥500 红、429 紫、≥400 琥珀、其余灰 */
+export function statusCodeBadgeClass(code: number): string {
+  if (code >= 500) return 'bg-red-100 text-red-800 dark:bg-red-900 dark:text-red-200'
+  if (code === 429) return 'bg-purple-100 text-purple-800 dark:bg-purple-900 dark:text-purple-200'
+  if (code >= 400) return 'bg-amber-100 text-amber-800 dark:bg-amber-900 dark:text-amber-200'
+  return 'bg-gray-100 text-gray-800 dark:bg-dark-700 dark:text-gray-200'
+}
+
+/** 请求类型徽章配色(cyber 红、ws 紫、stream 蓝、sync 灰、未知琥珀) */
+export function requestTypeBadgeClass(kind: UsageRequestKind): string {
+  if (kind === 'cyber') return 'bg-red-100 text-red-800 dark:bg-red-900 dark:text-red-200'
+  if (kind === 'ws_v2') return 'bg-violet-100 text-violet-800 dark:bg-violet-900 dark:text-violet-200'
+  if (kind === 'stream') return 'bg-blue-100 text-blue-800 dark:bg-blue-900 dark:text-blue-200'
+  if (kind === 'sync') return 'bg-gray-100 text-gray-800 dark:bg-dark-700 dark:text-gray-200'
+  return 'bg-amber-100 text-amber-800 dark:bg-amber-900 dark:text-amber-200'
+}
+
+/** 请求类型 i18n 键(展示方自行 t()) */
+export function requestTypeLabelKey(kind: UsageRequestKind): string {
+  if (kind === 'cyber') return 'usage.cyber'
+  if (kind === 'ws_v2') return 'usage.ws'
+  if (kind === 'stream') return 'usage.stream'
+  if (kind === 'sync') return 'usage.sync'
+  return 'usage.unknown'
+}
+
+/**
+ * 数字 request_type(1 同步/2 流式/3 WS)→ kind;
+ * 缺失时按 stream 布尔回退,两者都缺返回 null(展示为 -)。
+ */
+export function numericRequestTypeKind(
+  requestType?: number | null,
+  stream?: boolean | null
+): UsageRequestKind | null {
+  const rt = requestType ?? (stream == null ? 0 : stream ? 2 : 1)
+  if (rt === 3) return 'ws_v2'
+  if (rt === 2) return 'stream'
+  if (rt === 1) return 'sync'
+  return null
+}
+
+/** 错误表列 key → 后端 sort_by(status 列实际按 status_code 排序) */
+export function mapErrorSortKey(key: string): string {
+  return key === 'status' ? 'status_code' : key
+}
+
+/**
+ * 错误请求筛选的常用状态码固定候选(管理端 + 用户端共用)。
+ * 用固定列表而非「当前页出现过的码」派生,避免目标状态码只在后续页/筛选外时无法选中
+ * ——后端 status_code 过滤对全量数据生效,选项不应被当前页数据限制。
+ */
+export const COMMON_ERROR_STATUS_CODES = [400, 401, 403, 404, 408, 413, 429, 499, 500, 502, 503, 504, 529]
diff --git a/frontend/src/utils/errorCategory.ts b/frontend/src/utils/errorCategory.ts
new file mode 100644
index 0000000000..58e97e14c4
--- /dev/null
+++ b/frontend/src/utils/errorCategory.ts
@@ -0,0 +1,31 @@
+/**
+ * 错误请求"分类"虚拟维度:phase + error type → 用户侧粗分类码。
+ * 镜像后端 service.MapUserErrorCategory(backend/internal/service/ops_user_error.go),
+ * 两处修改须同步。返回稳定分类码,展示文案走 i18n `usage.errors.categories.*`。
+ */
+export function mapErrorCategory(phase?: string | null, errType?: string | null): string {
+  switch ((phase || '').toLowerCase()) {
+    case 'auth':
+      return 'auth'
+    case 'routing':
+      return 'service_unavailable'
+    case 'upstream':
+    case 'network':
+      return 'upstream'
+    case 'internal':
+      return 'internal'
+    case 'request':
+      switch ((errType || '').toLowerCase()) {
+        case 'rate_limit_error':
+          return 'rate_limit'
+        case 'billing_error':
+        case 'subscription_error':
+          return 'quota'
+        case 'invalid_request_error':
+          return 'invalid_request'
+        case 'cyber_policy':
+          return 'cyber'
+      }
+  }
+  return 'other'
+}
diff --git a/frontend/src/views/admin/UsageView.vue b/frontend/src/views/admin/UsageView.vue
index fb2588cb7f..eb9a047662 100644
--- a/frontend/src/views/admin/UsageView.vue
+++ b/frontend/src/views/admin/UsageView.vue
@@ -64,7 +64,7 @@
           <TokenUsageTrend :trend-data="trendData" :loading="chartsLoading" />
         </div>
       </div>
-      <UsageFilters v-model="filters" :start-date="startDate" :end-date="endDate" :exporting="exporting" :model-options="modelNameOptions" @change="applyFilters" @refresh="refreshData" @reset="resetFilters" @cleanup="openCleanupDialog" @export="exportToExcel">
+      <UsageFilters v-model="filters" :mode="activeTab === 'errors' ? 'errors' : 'usage'" :start-date="startDate" :end-date="endDate" :exporting="exporting" :model-options="modelNameOptions" @change="applyFilters" @refresh="refreshData" @reset="resetFilters" @cleanup="openCleanupDialog" @export="exportToExcel">
         <template #after-reset>
           <div class="relative" ref="columnDropdownRef">
             <button
@@ -82,14 +82,14 @@
               class="absolute right-0 top-full z-50 mt-1 max-h-80 w-48 overflow-y-auto rounded-lg border border-gray-200 bg-white py-1 shadow-lg dark:border-dark-600 dark:bg-dark-800"
             >
               <button
-                v-for="col in toggleableColumns"
+                v-for="col in currentToggleableColumns"
                 :key="col.key"
-                @click="toggleColumn(col.key)"
+                @click="toggleCurrentColumn(col.key)"
                 class="flex w-full items-center justify-between px-4 py-2 text-left text-sm text-gray-700 hover:bg-gray-100 dark:text-gray-300 dark:hover:bg-dark-700"
               >
                 <span>{{ col.label }}</span>
                 <Icon
-                  v-if="isColumnVisible(col.key)"
+                  v-if="isCurrentColumnVisible(col.key)"
                   name="check"
                   size="sm"
                   class="text-primary-500"
@@ -126,9 +126,14 @@
         <OpsErrorLogTable
           :rows="errRows" :total="errTotal" :loading="errLoading"
           :page="errPage" :page-size="errPageSize"
+          :visible-column-keys="errVisibleColumnKeys"
+          user-clickable
+          @userClick="handleUserClick"
           @openErrorDetail="openError"
+          @sort="onErrSort"
           @update:page="onErrPage"
-          @update:pageSize="onErrPageSize" />
+          @update:pageSize="onErrPageSize"
+          @ipGeoBatchFailed="handleIpGeoBatchFailed" />
         <OpsErrorDetailModal v-model:show="showErrorModal" :error-id="selectedErrorId" :error-type="'request'" />
       </div>
     </div>
@@ -609,6 +614,74 @@ const toggleColumn = (key: string) => {
   }
 }
 
+// ---- 错误请求 tab 列设置(与用量明细同机制,独立存储) ----
+const ERR_ALWAYS_VISIBLE = ['user', 'status', 'created_at', 'actions']
+const ERR_DEFAULT_HIDDEN_COLUMNS = ['user_agent']
+const ERR_HIDDEN_COLUMNS_KEY = 'usage-error-hidden-columns'
+
+// key 集合须与 OpsErrorLogTable 内部 allColumns 一致
+const errAllColumns = computed(() => [
+  { key: 'user', label: t('admin.ops.errorLog.user') },
+  { key: 'api_key', label: t('admin.ops.errorLog.apiKey') },
+  { key: 'account', label: t('admin.ops.errorLog.account') },
+  { key: 'platform', label: t('admin.ops.errorLog.platform') },
+  { key: 'model', label: t('admin.ops.errorLog.model') },
+  { key: 'endpoint', label: t('admin.ops.errorLog.endpoint') },
+  { key: 'group', label: t('admin.ops.errorLog.group') },
+  { key: 'type', label: t('admin.ops.errorLog.type') },
+  { key: 'category', label: t('usage.errors.category') },
+  { key: 'status', label: t('admin.ops.errorLog.status') },
+  { key: 'message', label: t('admin.ops.errorLog.message') },
+  { key: 'created_at', label: t('admin.ops.errorLog.time') },
+  { key: 'user_agent', label: t('usage.userAgent') },
+  { key: 'client_ip', label: t('admin.ops.errorLog.ip') },
+  { key: 'actions', label: t('admin.ops.errorLog.action') },
+])
+
+const errHiddenColumns = reactive<Set<string>>(new Set())
+
+const errToggleableColumns = computed(() =>
+  errAllColumns.value.filter(col => !ERR_ALWAYS_VISIBLE.includes(col.key))
+)
+
+const errVisibleColumnKeys = computed(() =>
+  errAllColumns.value
+    .filter(col => ERR_ALWAYS_VISIBLE.includes(col.key) || !errHiddenColumns.has(col.key))
+    .map(col => col.key)
+)
+
+const toggleErrColumn = (key: string) => {
+  if (errHiddenColumns.has(key)) {
+    errHiddenColumns.delete(key)
+  } else {
+    errHiddenColumns.add(key)
+  }
+  try {
+    localStorage.setItem(ERR_HIDDEN_COLUMNS_KEY, JSON.stringify([...errHiddenColumns]))
+  } catch (e) {
+    console.error('Failed to save error columns:', e)
+  }
+}
+
+const loadSavedErrColumns = () => {
+  try {
+    const saved = localStorage.getItem(ERR_HIDDEN_COLUMNS_KEY)
+    const keys = saved ? (JSON.parse(saved) as string[]) : ERR_DEFAULT_HIDDEN_COLUMNS
+    keys.forEach((key) => errHiddenColumns.add(key))
+  } catch {
+    ERR_DEFAULT_HIDDEN_COLUMNS.forEach((key) => errHiddenColumns.add(key))
+  }
+}
+
+// 列设置下拉按当前 tab 分发
+const currentToggleableColumns = computed(() =>
+  activeTab.value === 'errors' ? errToggleableColumns.value : toggleableColumns.value
+)
+const isCurrentColumnVisible = (key: string) =>
+  activeTab.value === 'errors' ? !errHiddenColumns.has(key) : isColumnVisible(key)
+const toggleCurrentColumn = (key: string) =>
+  activeTab.value === 'errors' ? toggleErrColumn(key) : toggleColumn(key)
+
 const loadSavedColumns = () => {
   try {
     const saved = localStorage.getItem(HIDDEN_COLUMNS_KEY)
@@ -635,6 +708,8 @@ const errLoading = ref(false)
 const errPage = ref(1)
 const errPageSize = ref(20)
 const errTotal = ref(0)
+const errSortBy = ref('created_at')
+const errSortOrder = ref<'asc' | 'desc'>('desc')
 const showErrorModal = ref(false)
 const selectedErrorId = ref<number | null>(null)
 
@@ -656,6 +731,11 @@ const loadAdminErrors = async () => {
       account_id: filters.value.account_id ?? undefined,
       group_id: filters.value.group_id ?? undefined,
       model: filters.value.model || undefined,
+      phase: filters.value.error_phase || undefined,
+      category: filters.value.error_category || undefined,
+      status_codes: filters.value.status_code != null ? String(filters.value.status_code) : undefined,
+      sort_by: errSortBy.value,
+      sort_order: errSortOrder.value,
     })
     errRows.value = resp.items
     errTotal.value = resp.total
@@ -667,6 +747,12 @@ const loadAdminErrors = async () => {
   }
 }
 
+const onErrSort = (sortBy: string, sortOrder: 'asc' | 'desc') => {
+  errSortBy.value = sortBy
+  errSortOrder.value = sortOrder
+  errPage.value = 1
+  loadAdminErrors()
+}
 const onErrPage = (p: number) => { errPage.value = p; loadAdminErrors() }
 const onErrPageSize = (s: number) => { errPageSize.value = s; errPage.value = 1; loadAdminErrors() }
 const openError = (id: number) => { selectedErrorId.value = id; showErrorModal.value = true }
@@ -690,6 +776,7 @@ onMounted(() => {
     void loadChartData()
   }, 120)
   loadSavedColumns()
+  loadSavedErrColumns()
   document.addEventListener('click', handleColumnClickOutside)
 })
 onUnmounted(() => { abortController?.abort(); exportAbortController?.abort(); document.removeEventListener('click', handleColumnClickOutside) })
diff --git a/frontend/src/views/admin/ops/components/OpsErrorDetailsModal.vue b/frontend/src/views/admin/ops/components/OpsErrorDetailsModal.vue
index e9e4401001..0400033e89 100644
--- a/frontend/src/views/admin/ops/components/OpsErrorDetailsModal.vue
+++ b/frontend/src/views/admin/ops/components/OpsErrorDetailsModal.vue
@@ -84,6 +84,16 @@ function close() {
   emit('update:show', false)
 }
 
+const sortBy = ref('created_at')
+const sortOrder = ref<'asc' | 'desc'>('desc')
+
+function onSort(nextSortBy: string, nextSortOrder: 'asc' | 'desc') {
+  sortBy.value = nextSortBy
+  sortOrder.value = nextSortOrder
+  page.value = 1
+  void fetchErrorLogs()
+}
+
 async function fetchErrorLogs() {
   if (!props.show) return
 
@@ -93,7 +103,9 @@ async function fetchErrorLogs() {
       page: page.value,
       page_size: pageSize.value,
       time_range: props.timeRange,
-      view: viewMode.value
+      view: viewMode.value,
+      sort_by: sortBy.value,
+      sort_order: sortOrder.value
     }
 
     const platform = String(props.platform || '').trim()
@@ -253,6 +265,7 @@ watch(
             :page="page"
             :page-size="pageSize"
             @openErrorDetail="emit('openErrorDetail', $event)"
+            @sort="onSort"
 
             @update:page="page = $event"
             @update:pageSize="pageSize = $event"
diff --git a/frontend/src/views/admin/ops/components/OpsErrorLogTable.vue b/frontend/src/views/admin/ops/components/OpsErrorLogTable.vue
index becf6327a2..437f0960bd 100644
--- a/frontend/src/views/admin/ops/components/OpsErrorLogTable.vue
+++ b/frontend/src/views/admin/ops/components/OpsErrorLogTable.vue
@@ -1,272 +1,251 @@
 <template>
-  <div class="flex h-full min-h-0 flex-col bg-white dark:bg-dark-900">
-    <!-- Loading State -->
-    <div v-if="loading" class="flex flex-1 items-center justify-center py-10">
-      <div class="h-8 w-8 animate-spin rounded-full border-b-2 border-primary-600"></div>
+  <div class="flex h-full min-h-0 flex-col">
+    <div class="card flex min-h-0 flex-1 flex-col overflow-hidden">
+      <IpGeoBatchToolbar :ips="rows.map((r) => r.client_ip)" @failed="emit('ipGeoBatchFailed')" />
+
+      <DataTable
+        :columns="columns"
+        :data="rows"
+        :loading="loading"
+        clickable-rows
+        server-side-sort
+        default-sort-key="created_at"
+        default-sort-order="desc"
+        @sort="onSort"
+        @rowClick="(row) => emit('openErrorDetail', row.id)"
+      >
+        <template #cell-created_at="{ row }">
+          <span
+            class="text-sm text-gray-600 dark:text-gray-400"
+            :title="row.request_id || row.client_request_id"
+          >{{ formatDateTime(row.created_at) }}</span>
+        </template>
+
+        <template #cell-type="{ row }">
+          <span class="inline-flex items-center rounded px-2 py-0.5 text-xs font-medium" :class="getTypeBadge(row).className">
+            {{ getTypeBadge(row).label }}
+          </span>
+        </template>
+
+        <template #cell-endpoint="{ row }">
+          <div class="max-w-[320px] space-y-1 text-xs">
+            <div class="break-all text-gray-700 dark:text-gray-300">
+              <span class="font-medium text-gray-500 dark:text-gray-400">{{ t('usage.inbound') }}:</span>
+              <span class="ml-1">{{ row.inbound_endpoint?.trim() || '-' }}</span>
+            </div>
+            <div v-if="row.upstream_endpoint" class="break-all text-gray-700 dark:text-gray-300">
+              <span class="font-medium text-gray-500 dark:text-gray-400">{{ t('usage.upstream') }}:</span>
+              <span class="ml-1">{{ row.upstream_endpoint?.trim() || '-' }}</span>
+            </div>
+          </div>
+        </template>
+
+        <template #cell-platform="{ row }">
+          <span class="text-sm text-gray-900 dark:text-white">{{ row.platform || '-' }}</span>
+        </template>
+
+        <template #cell-model="{ row }">
+          <div v-if="hasModelMapping(row)" class="space-y-0.5 text-xs">
+            <div class="break-all font-medium text-gray-900 dark:text-white">{{ row.requested_model }}</div>
+            <div class="break-all text-gray-500 dark:text-gray-400"><span class="mr-0.5">↳</span>{{ row.upstream_model }}</div>
+          </div>
+          <span v-else-if="displayModel(row)" class="text-sm font-medium text-gray-900 dark:text-white">{{ displayModel(row) }}</span>
+          <span v-else class="text-sm text-gray-400 dark:text-gray-500">-</span>
+        </template>
+
+        <template #cell-group="{ row }">
+          <span
+            v-if="row.group_id"
+            class="inline-flex items-center rounded px-2 py-0.5 text-xs font-medium bg-indigo-100 text-indigo-800 dark:bg-indigo-900 dark:text-indigo-200"
+            :title="t('admin.ops.errorLog.id') + ' ' + row.group_id"
+          >
+            {{ row.group_name || '#' + row.group_id }}
+          </span>
+          <span v-else class="text-sm text-gray-400 dark:text-gray-500">-</span>
+        </template>
+
+        <template #cell-user="{ row }">
+          <div v-if="row.user_id" class="text-sm">
+            <button
+              v-if="userClickable && row.user_email"
+              class="font-medium text-primary-600 underline decoration-dashed underline-offset-2 transition-colors hover:text-primary-700 dark:text-primary-400 dark:hover:text-primary-300"
+              :title="t('admin.usage.clickToViewBalance')"
+              @click.stop="emit('userClick', row.user_id, row.user_email)"
+            >
+              {{ row.user_email }}
+            </button>
+            <span v-else class="font-medium text-gray-900 dark:text-white">{{ row.user_email || '-' }}</span>
+            <span class="ml-1 text-gray-500 dark:text-gray-400">#{{ row.user_id }}</span>
+          </div>
+          <!-- 认证失败行 user_id 为空:回退显示已删除 KEY 所有者(归因快照,与详情弹窗一致) -->
+          <div v-else-if="row.deleted_key_owner_user_id" class="text-sm">
+            <button
+              v-if="userClickable && row.deleted_key_owner_email"
+              class="font-medium text-primary-600 underline decoration-dashed underline-offset-2 transition-colors hover:text-primary-700 dark:text-primary-400 dark:hover:text-primary-300"
+              :title="t('admin.usage.clickToViewBalance')"
+              @click.stop="emit('userClick', row.deleted_key_owner_user_id, row.deleted_key_owner_email ?? undefined)"
+            >
+              {{ row.deleted_key_owner_email }}
+            </button>
+            <span v-else class="font-medium text-gray-900 dark:text-white">{{ row.deleted_key_owner_email || '-' }}</span>
+            <span class="ml-1 text-gray-500 dark:text-gray-400">#{{ row.deleted_key_owner_user_id }}</span>
+          </div>
+          <span v-else class="text-sm text-gray-400 dark:text-gray-500">-</span>
+        </template>
+
+        <template #cell-api_key="{ row }">
+          <div v-if="row.api_key_id || row.api_key_name" class="text-sm">
+            <span class="text-gray-900 dark:text-white">{{ row.api_key_name || '#' + row.api_key_id }}</span>
+            <span
+              v-if="row.api_key_deleted"
+              class="ml-1 inline-flex items-center rounded px-1 py-px text-[10px] font-medium leading-tight bg-rose-100 text-rose-600 ring-1 ring-inset ring-rose-200 dark:bg-rose-500/20 dark:text-rose-400 dark:ring-rose-500/30"
+            >{{ t('admin.ops.errorLog.keyDeletedBadge') }}</span>
+          </div>
+          <span v-else class="text-sm text-gray-400 dark:text-gray-500">-</span>
+        </template>
+
+        <template #cell-account="{ row }">
+          <span
+            v-if="row.account_id"
+            class="text-sm text-gray-900 dark:text-white"
+            :title="t('admin.ops.errorLog.accountId') + ' ' + row.account_id"
+          >{{ row.account_name || '#' + row.account_id }}</span>
+          <span v-else class="text-sm text-gray-400 dark:text-gray-500">-</span>
+        </template>
+
+        <template #cell-category="{ row }">
+          <span class="text-sm text-gray-900 dark:text-white">
+            {{ t('usage.errors.categories.' + mapErrorCategory(row.phase, row.type)) }}
+          </span>
+        </template>
+
+        <template #cell-status="{ row }">
+          <div class="flex items-center gap-1.5">
+            <span class="inline-flex items-center rounded px-2 py-0.5 text-xs font-medium" :class="getStatusClass(row.status_code)">
+              {{ row.status_code }}
+            </span>
+            <span
+              v-if="row.severity"
+              :class="['rounded px-1.5 py-0.5 text-[10px] font-medium', getSeverityClass(row.severity)]"
+            >{{ row.severity }}</span>
+            <span
+              v-if="row.request_type != null && row.request_type > 0"
+              class="inline-flex items-center rounded px-2 py-0.5 text-xs font-medium bg-gray-100 text-gray-800 dark:bg-dark-700 dark:text-gray-200"
+            >{{ formatRequestType(row.request_type) }}</span>
+          </div>
+        </template>
+
+        <template #cell-message="{ row }">
+          <span
+            v-if="row.message"
+            class="block max-w-[280px] truncate text-sm text-gray-600 dark:text-gray-400"
+            :title="row.message"
+          >{{ formatSmartMessage(row.message) || '-' }}</span>
+          <span v-else class="text-sm text-gray-400 dark:text-gray-500">-</span>
+        </template>
+
+        <template #cell-user_agent="{ row }">
+          <span
+            v-if="row.user_agent"
+            class="block max-w-[320px] truncate text-sm text-gray-600 dark:text-gray-400"
+            :title="row.user_agent"
+          >{{ row.user_agent }}</span>
+          <span v-else class="text-sm text-gray-400 dark:text-gray-500">-</span>
+        </template>
+
+        <template #cell-client_ip="{ row }">
+          <div @click.stop>
+            <div v-if="row.client_ip">
+              <span class="text-sm font-mono text-gray-600 dark:text-gray-400">{{ row.client_ip }}</span>
+              <IpGeoCell :ip="row.client_ip" />
+            </div>
+            <span v-else class="text-sm text-gray-400 dark:text-gray-500">-</span>
+          </div>
+        </template>
+
+        <template #cell-actions="{ row }">
+          <button
+            type="button"
+            class="rounded p-1 text-gray-400 transition-colors hover:bg-gray-100 hover:text-primary-600 dark:hover:bg-dark-600 dark:hover:text-primary-400"
+            :title="t('admin.ops.errorLog.details')"
+            @click.stop="emit('openErrorDetail', row.id)"
+          >
+            <svg class="h-4 w-4" fill="none" stroke="currentColor" viewBox="0 0 24 24"><path stroke-linecap="round" stroke-linejoin="round" stroke-width="2" d="M9 12h6m-6 4h6m2 5H7a2 2 0 01-2-2V5a2 2 0 012-2h5.586a1 1 0 01.707.293l5.414 5.414a1 1 0 01.293.707V19a2 2 0 01-2 2z" /></svg>
+          </button>
+        </template>
+
+        <template #empty><EmptyState :message="t('admin.ops.errorLog.noErrors')" /></template>
+      </DataTable>
     </div>
 
-    <!-- Table Container -->
-    <div v-else class="flex min-h-0 flex-1 flex-col">
-      <div class="min-h-0 flex-1 overflow-auto border-b border-gray-200 dark:border-dark-700">
-        <table class="w-full border-separate border-spacing-0">
-          <thead class="sticky top-0 z-10 bg-gray-50 dark:bg-dark-800">
-            <tr>
-              <th class="border-b border-gray-200 px-4 py-2.5 text-left text-[11px] font-bold uppercase tracking-wider text-gray-500 dark:border-dark-700 dark:text-dark-400">
-                {{ t('admin.ops.errorLog.time') }}
-              </th>
-              <th class="border-b border-gray-200 px-4 py-2.5 text-left text-[11px] font-bold uppercase tracking-wider text-gray-500 dark:border-dark-700 dark:text-dark-400">
-                {{ t('admin.ops.errorLog.type') }}
-              </th>
-              <th class="border-b border-gray-200 px-4 py-2.5 text-left text-[11px] font-bold uppercase tracking-wider text-gray-500 dark:border-dark-700 dark:text-dark-400">
-                {{ t('admin.ops.errorLog.endpoint') }}
-              </th>
-              <th class="border-b border-gray-200 px-4 py-2.5 text-left text-[11px] font-bold uppercase tracking-wider text-gray-500 dark:border-dark-700 dark:text-dark-400">
-                {{ t('admin.ops.errorLog.platform') }}
-              </th>
-              <th class="border-b border-gray-200 px-4 py-2.5 text-left text-[11px] font-bold uppercase tracking-wider text-gray-500 dark:border-dark-700 dark:text-dark-400">
-                {{ t('admin.ops.errorLog.model') }}
-              </th>
-              <th class="border-b border-gray-200 px-4 py-2.5 text-left text-[11px] font-bold uppercase tracking-wider text-gray-500 dark:border-dark-700 dark:text-dark-400">
-                {{ t('admin.ops.errorLog.group') }}
-              </th>
-              <th class="border-b border-gray-200 px-4 py-2.5 text-left text-[11px] font-bold uppercase tracking-wider text-gray-500 dark:border-dark-700 dark:text-dark-400">
-                {{ t('admin.ops.errorLog.user') }}
-              </th>
-              <th class="border-b border-gray-200 px-4 py-2.5 text-left text-[11px] font-bold uppercase tracking-wider text-gray-500 dark:border-dark-700 dark:text-dark-400">
-                {{ t('admin.ops.errorLog.apiKey') }}
-              </th>
-              <th class="border-b border-gray-200 px-4 py-2.5 text-left text-[11px] font-bold uppercase tracking-wider text-gray-500 dark:border-dark-700 dark:text-dark-400">
-                {{ t('admin.ops.errorLog.account') }}
-              </th>
-              <th class="border-b border-gray-200 px-4 py-2.5 text-left text-[11px] font-bold uppercase tracking-wider text-gray-500 dark:border-dark-700 dark:text-dark-400">
-                {{ t('admin.ops.errorLog.status') }}
-              </th>
-              <th class="border-b border-gray-200 px-4 py-2.5 text-left text-[11px] font-bold uppercase tracking-wider text-gray-500 dark:border-dark-700 dark:text-dark-400">
-                {{ t('admin.ops.errorLog.message') }}
-              </th>
-              <th class="border-b border-gray-200 px-4 py-2.5 text-right text-[11px] font-bold uppercase tracking-wider text-gray-500 dark:border-dark-700 dark:text-dark-400">
-                {{ t('admin.ops.errorLog.action') }}
-              </th>
-            </tr>
-          </thead>
-          <tbody class="divide-y divide-gray-100 dark:divide-dark-700">
-            <tr v-if="rows.length === 0">
-              <td colspan="12" class="py-12 text-center text-sm text-gray-400 dark:text-dark-500">
-                {{ t('admin.ops.errorLog.noErrors') }}
-              </td>
-            </tr>
-
-            <tr
-              v-for="log in rows"
-              :key="log.id"
-              class="group cursor-pointer transition-colors hover:bg-gray-50/80 dark:hover:bg-dark-800/50"
-              @click="emit('openErrorDetail', log.id)"
-            >
-              <!-- Time -->
-              <td class="whitespace-nowrap px-4 py-2">
-                <el-tooltip :content="log.request_id || log.client_request_id" placement="top" :show-after="500">
-                  <span class="font-mono text-xs font-medium text-gray-900 dark:text-gray-200">
-                    {{ formatDateTime(log.created_at).split(' ')[1] }}
-                  </span>
-                </el-tooltip>
-              </td>
-
-              <!-- Type -->
-              <td class="whitespace-nowrap px-4 py-2">
-                <span
-                  :class="[
-                    'inline-flex items-center rounded px-1.5 py-0.5 text-[10px] font-bold ring-1 ring-inset',
-                    getTypeBadge(log).className
-                  ]"
-                >
-                  {{ getTypeBadge(log).label }}
-                </span>
-              </td>
-
-              <!-- Endpoint -->
-              <td class="px-4 py-2">
-                <div class="max-w-[160px]">
-                  <el-tooltip v-if="log.inbound_endpoint" :content="formatEndpointTooltip(log)" placement="top" :show-after="500">
-                    <span class="truncate font-mono text-[11px] text-gray-700 dark:text-gray-300">
-                      {{ log.inbound_endpoint }}
-                    </span>
-                  </el-tooltip>
-                  <span v-else class="text-xs text-gray-400">-</span>
-                </div>
-              </td>
-
-              <!-- Platform -->
-              <td class="whitespace-nowrap px-4 py-2">
-                <span class="inline-flex items-center rounded bg-gray-100 px-1.5 py-0.5 text-[10px] font-bold uppercase text-gray-600 dark:bg-dark-700 dark:text-gray-300">
-                  {{ log.platform || '-' }}
-                </span>
-              </td>
-
-              <!-- Model -->
-              <td class="px-4 py-2">
-                <div class="max-w-[160px]">
-                  <template v-if="hasModelMapping(log)">
-                    <el-tooltip :content="modelMappingTooltip(log)" placement="top" :show-after="500">
-                      <span class="flex items-center gap-1 truncate font-mono text-[11px] text-gray-700 dark:text-gray-300">
-                        <span class="truncate">{{ log.requested_model }}</span>
-                        <span class="flex-shrink-0 text-gray-400">→</span>
-                        <span class="truncate text-primary-600 dark:text-primary-400">{{ log.upstream_model }}</span>
-                      </span>
-                    </el-tooltip>
-                  </template>
-                  <template v-else>
-                    <span v-if="displayModel(log)" class="truncate font-mono text-[11px] text-gray-700 dark:text-gray-300" :title="displayModel(log)">
-                      {{ displayModel(log) }}
-                    </span>
-                    <span v-else class="text-xs text-gray-400">-</span>
-                  </template>
-                </div>
-              </td>
-
-              <!-- Group -->
-              <td class="px-4 py-2">
-                 <el-tooltip v-if="log.group_id" :content="t('admin.ops.errorLog.id') + ' ' + log.group_id" placement="top" :show-after="500">
-                  <span class="max-w-[100px] truncate text-xs font-medium text-gray-900 dark:text-gray-200">
-                    {{ log.group_name || '-' }}
-                  </span>
-                </el-tooltip>
-                <span v-else class="text-xs text-gray-400">-</span>
-              </td>
-
-              <!-- User -->
-              <td class="px-4 py-2">
-                <el-tooltip v-if="log.user_id" :content="t('admin.ops.errorLog.userId') + ' ' + log.user_id" placement="top" :show-after="500">
-                  <span class="block max-w-[140px] truncate text-xs font-medium text-gray-900 dark:text-gray-200">
-                    {{ log.user_email || '-' }}
-                  </span>
-                </el-tooltip>
-                <span v-else class="text-xs text-gray-400">-</span>
-              </td>
-
-              <!-- API Key -->
-              <td class="px-4 py-2">
-                <div v-if="log.api_key_id || log.api_key_name" class="flex max-w-[140px] items-center gap-1">
-                  <span class="truncate text-xs font-medium text-gray-900 dark:text-gray-200" :title="log.api_key_name || ('#' + log.api_key_id)">
-                    {{ log.api_key_name || ('#' + log.api_key_id) }}
-                  </span>
-                  <span
-                    v-if="log.api_key_deleted"
-                    class="flex-shrink-0 rounded px-1 py-0.5 text-[9px] font-bold ring-1 ring-inset bg-red-50 text-red-700 ring-red-600/20 dark:bg-red-900/30 dark:text-red-400 dark:ring-red-500/30"
-                  >
-                    {{ t('admin.ops.errorLog.keyDeletedBadge') }}
-                  </span>
-                </div>
-                <span v-else class="text-xs text-gray-400">-</span>
-              </td>
-
-              <!-- Account -->
-              <td class="px-4 py-2">
-                <el-tooltip v-if="log.account_id" :content="t('admin.ops.errorLog.accountId') + ' ' + log.account_id" placement="top" :show-after="500">
-                  <span class="block max-w-[120px] truncate text-xs font-medium text-gray-900 dark:text-gray-200">
-                    {{ log.account_name || '-' }}
-                  </span>
-                </el-tooltip>
-                <span v-else class="text-xs text-gray-400">-</span>
-              </td>
-
-              <!-- Status -->
-              <td class="whitespace-nowrap px-4 py-2">
-                <div class="flex items-center gap-1.5">
-                  <span
-                    :class="[
-                      'inline-flex items-center rounded px-1.5 py-0.5 text-[10px] font-bold ring-1 ring-inset',
-                      getStatusClass(log.status_code)
-                    ]"
-                  >
-                    {{ log.status_code }}
-                  </span>
-                  <span
-                    v-if="log.severity"
-                    :class="['rounded px-1.5 py-0.5 text-[10px] font-bold', getSeverityClass(log.severity)]"
-                  >
-                    {{ log.severity }}
-                  </span>
-                  <span
-                    v-if="log.request_type != null && log.request_type > 0"
-                    class="rounded bg-gray-100 px-1.5 py-0.5 text-[10px] font-bold text-gray-600 dark:bg-dark-700 dark:text-gray-300"
-                  >
-                    {{ formatRequestType(log.request_type) }}
-                  </span>
-                </div>
-              </td>
-
-              <!-- Message (Response Content) -->
-              <td class="px-4 py-2">
-                <div class="max-w-[200px]">
-                  <p class="truncate text-[11px] font-medium text-gray-600 dark:text-gray-400" :title="log.message">
-                    {{ formatSmartMessage(log.message) || '-' }}
-                  </p>
-                </div>
-              </td>
-
-              <!-- Actions -->
-              <td class="whitespace-nowrap px-4 py-2 text-right" @click.stop>
-                <div class="flex items-center justify-end gap-3">
-                  <button type="button" class="text-primary-600 hover:text-primary-700 dark:text-primary-400 text-xs font-bold" @click="emit('openErrorDetail', log.id)">
-                    {{ t('admin.ops.errorLog.details') }}
-                  </button>
-                </div>
-              </td>
-            </tr>
-          </tbody>
-        </table>
-      </div>
-
-      <!-- Pagination -->
-      <div class="bg-gray-50/50 dark:bg-dark-800/50">
-        <Pagination
-          v-if="total > 0"
-          :total="total"
-          :page="page"
-          :page-size="pageSize"
-          @update:page="emit('update:page', $event)"
-          @update:pageSize="emit('update:pageSize', $event)"
-        />
-      </div>
+    <div class="flex-shrink-0">
+      <Pagination
+        v-if="total > 0"
+        :total="total"
+        :page="page"
+        :page-size="pageSize"
+        @update:page="emit('update:page', $event)"
+        @update:pageSize="emit('update:pageSize', $event)"
+      />
     </div>
   </div>
 </template>
 
 <script setup lang="ts">
+import { computed } from 'vue'
 import { useI18n } from 'vue-i18n'
+import DataTable from '@/components/common/DataTable.vue'
+import EmptyState from '@/components/common/EmptyState.vue'
 import Pagination from '@/components/common/Pagination.vue'
+import IpGeoCell from '@/components/common/IpGeoCell.vue'
+import IpGeoBatchToolbar from '@/components/common/IpGeoBatchToolbar.vue'
 import type { OpsErrorLog } from '@/api/admin/ops'
+import type { Column } from '@/components/common/types'
 import { getSeverityClass, formatDateTime } from '../utils/opsFormatters'
+import { mapErrorCategory } from '@/utils/errorCategory'
+import { mapErrorSortKey, statusCodeBadgeClass } from '@/utils/errorBadges'
 
 const { t } = useI18n()
 
+// 列序对齐管理端用量明细:身份(用户→Key→账号)→ 请求形态(平台→模型→端点→分组→类型)
+// → 结果(状态→消息)→ 时间→UA→IP→操作
+const allColumns = computed<Column[]>(() => [
+  { key: 'user', label: t('admin.ops.errorLog.user') },
+  { key: 'api_key', label: t('admin.ops.errorLog.apiKey') },
+  { key: 'account', label: t('admin.ops.errorLog.account') },
+  { key: 'platform', label: t('admin.ops.errorLog.platform') },
+  { key: 'model', label: t('admin.ops.errorLog.model'), sortable: true },
+  { key: 'endpoint', label: t('admin.ops.errorLog.endpoint') },
+  { key: 'group', label: t('admin.ops.errorLog.group') },
+  { key: 'type', label: t('admin.ops.errorLog.type') },
+  { key: 'category', label: t('usage.errors.category') },
+  { key: 'status', label: t('admin.ops.errorLog.status'), sortable: true },
+  { key: 'message', label: t('admin.ops.errorLog.message') },
+  { key: 'created_at', label: t('admin.ops.errorLog.time'), sortable: true },
+  { key: 'user_agent', label: t('usage.userAgent') },
+  { key: 'client_ip', label: t('admin.ops.errorLog.ip') },
+  { key: 'actions', label: t('admin.ops.errorLog.action') },
+])
+
+// 传入 visibleColumnKeys 时按其过滤(列设置);未传则全量(Ops 弹窗等使用方)
+const columns = computed<Column[]>(() =>
+  props.visibleColumnKeys
+    ? allColumns.value.filter((c) => props.visibleColumnKeys!.includes(c.key))
+    : allColumns.value
+)
+
 function isUpstreamRow(log: OpsErrorLog): boolean {
   const phase = String(log.phase || '').toLowerCase()
   const owner = String(log.error_owner || '').toLowerCase()
   return phase === 'upstream' && owner === 'provider'
 }
 
-function formatEndpointTooltip(log: OpsErrorLog): string {
-  const parts: string[] = []
-  if (log.inbound_endpoint) parts.push(`Inbound: ${log.inbound_endpoint}`)
-  if (log.upstream_endpoint) parts.push(`Upstream: ${log.upstream_endpoint}`)
-  return parts.join('\n') || ''
-}
-
 function hasModelMapping(log: OpsErrorLog): boolean {
   const requested = String(log.requested_model || '').trim()
   const upstream = String(log.upstream_model || '').trim()
   return !!requested && !!upstream && requested !== upstream
 }
 
-function modelMappingTooltip(log: OpsErrorLog): string {
-  const requested = String(log.requested_model || '').trim()
-  const upstream = String(log.upstream_model || '').trim()
-  if (!requested && !upstream) return ''
-  if (requested && upstream) return `${requested} → ${upstream}`
-  return upstream || requested
-}
-
 function displayModel(log: OpsErrorLog): string {
   const upstream = String(log.upstream_model || '').trim()
   if (upstream) return upstream
@@ -284,28 +263,29 @@ function formatRequestType(type: number | null | undefined): string {
   }
 }
 
+// 徽章配色对齐用量明细(UsageTable)的 bg-X-100/text-X-800 体系
 function getTypeBadge(log: OpsErrorLog): { label: string; className: string } {
   const phase = String(log.phase || '').toLowerCase()
   const owner = String(log.error_owner || '').toLowerCase()
 
   if (isUpstreamRow(log)) {
-    return { label: t('admin.ops.errorLog.typeUpstream'), className: 'bg-red-50 text-red-700 ring-red-600/20 dark:bg-red-900/30 dark:text-red-400 dark:ring-red-500/30' }
+    return { label: t('admin.ops.errorLog.typeUpstream'), className: 'bg-red-100 text-red-800 dark:bg-red-900 dark:text-red-200' }
   }
   if (phase === 'request' && owner === 'client') {
-    return { label: t('admin.ops.errorLog.typeRequest'), className: 'bg-amber-50 text-amber-700 ring-amber-600/20 dark:bg-amber-900/30 dark:text-amber-400 dark:ring-amber-500/30' }
+    return { label: t('admin.ops.errorLog.typeRequest'), className: 'bg-amber-100 text-amber-800 dark:bg-amber-900 dark:text-amber-200' }
   }
   if (phase === 'auth' && owner === 'client') {
-    return { label: t('admin.ops.errorLog.typeAuth'), className: 'bg-blue-50 text-blue-700 ring-blue-600/20 dark:bg-blue-900/30 dark:text-blue-400 dark:ring-blue-500/30' }
+    return { label: t('admin.ops.errorLog.typeAuth'), className: 'bg-blue-100 text-blue-800 dark:bg-blue-900 dark:text-blue-200' }
   }
   if (phase === 'routing' && owner === 'platform') {
-    return { label: t('admin.ops.errorLog.typeRouting'), className: 'bg-purple-50 text-purple-700 ring-purple-600/20 dark:bg-purple-900/30 dark:text-purple-400 dark:ring-purple-500/30' }
+    return { label: t('admin.ops.errorLog.typeRouting'), className: 'bg-purple-100 text-purple-800 dark:bg-purple-900 dark:text-purple-200' }
   }
   if (phase === 'internal' && owner === 'platform') {
-    return { label: t('admin.ops.errorLog.typeInternal'), className: 'bg-gray-100 text-gray-800 ring-gray-600/20 dark:bg-dark-700 dark:text-gray-200 dark:ring-dark-500/40' }
+    return { label: t('admin.ops.errorLog.typeInternal'), className: 'bg-gray-100 text-gray-800 dark:bg-dark-700 dark:text-gray-200' }
   }
 
-    const fallback = phase || owner || t('common.unknown')
-    return { label: fallback, className: 'bg-gray-50 text-gray-700 ring-gray-600/10 dark:bg-dark-900 dark:text-gray-300 dark:ring-dark-700' }
+  const fallback = phase || owner || t('common.unknown')
+  return { label: fallback, className: 'bg-gray-100 text-gray-800 dark:bg-dark-700 dark:text-gray-200' }
 }
 
 interface Props {
@@ -314,24 +294,30 @@ interface Props {
   loading: boolean
   page: number
   pageSize: number
+  /** 用户邮箱可点击(emit userClick),仅在有弹窗承接的使用方开启 */
+  userClickable?: boolean
+  /** 列设置:仅显示这些 key 的列;不传则全量 */
+  visibleColumnKeys?: string[]
 }
 
 interface Emits {
   (e: 'openErrorDetail', id: number): void
   (e: 'update:page', value: number): void
   (e: 'update:pageSize', value: number): void
+  (e: 'ipGeoBatchFailed'): void
+  (e: 'sort', sortBy: string, sortOrder: 'asc' | 'desc'): void
+  (e: 'userClick', userId: number, email?: string): void
 }
 
-defineProps<Props>()
+const props = defineProps<Props>()
 const emit = defineEmits<Emits>()
 
-function getStatusClass(code: number): string {
-  if (code >= 500) return 'bg-red-50 text-red-700 ring-red-600/20 dark:bg-red-900/30 dark:text-red-400 dark:ring-red-500/30'
-  if (code === 429) return 'bg-purple-50 text-purple-700 ring-purple-600/20 dark:bg-purple-900/30 dark:text-purple-400 dark:ring-purple-500/30'
-  if (code >= 400) return 'bg-amber-50 text-amber-700 ring-amber-600/20 dark:bg-amber-900/30 dark:text-amber-400 dark:ring-amber-500/30'
-  return 'bg-gray-50 text-gray-700 ring-gray-600/20 dark:bg-gray-900/30 dark:text-gray-400 dark:ring-gray-500/30'
+function onSort(key: string, order: 'asc' | 'desc') {
+  emit('sort', mapErrorSortKey(key), order)
 }
 
+const getStatusClass = statusCodeBadgeClass
+
 function formatSmartMessage(msg: string): string {
   if (!msg) return ''
 
@@ -352,6 +338,5 @@ function formatSmartMessage(msg: string): string {
   if (msg.toLowerCase().includes('rate limit')) return t('admin.ops.errorLog.commonErrors.rateLimit')
 
   return msg.length > 200 ? msg.substring(0, 200) + '...' : msg
-
 }
 </script>
diff --git a/frontend/src/views/user/UsageView.vue b/frontend/src/views/user/UsageView.vue
index b66fdaa336..bbf58532ab 100644
--- a/frontend/src/views/user/UsageView.vue
+++ b/frontend/src/views/user/UsageView.vue
@@ -68,7 +68,33 @@
 
       <div class="card p-6">
         <div class="flex flex-wrap items-end justify-between gap-4">
-          <div class="flex flex-1 flex-wrap items-end gap-4">
+          <div v-if="activeTab === 'errors'" class="flex flex-1 flex-wrap items-end gap-4">
+            <div class="w-full sm:w-auto sm:min-w-[220px]">
+              <label class="input-label">{{ t('usage.errors.keyName') }}</label>
+              <Select v-model="errorFilter.api_key_id" :options="errorKeyOptions" @change="applyErrorFilters" />
+            </div>
+            <div class="w-full sm:w-auto sm:min-w-[220px]">
+              <label class="input-label">{{ t('usage.errors.model') }}</label>
+              <Select
+                v-model="errorFilter.model"
+                :options="errorModelOptions"
+                searchable
+                creatable
+                clearable
+                :placeholder="t('usage.errors.modelPlaceholder')"
+                @change="applyErrorFilters"
+              />
+            </div>
+            <div class="w-full sm:w-auto sm:min-w-[200px]">
+              <label class="input-label">{{ t('usage.errors.category') }}</label>
+              <Select v-model="errorFilter.category" :options="errorCategoryOptions" @change="applyErrorFilters" />
+            </div>
+            <div class="w-full sm:w-auto sm:min-w-[180px]">
+              <label class="input-label">{{ t('usage.errors.status') }}</label>
+              <Select v-model="errorFilter.status_code" :options="errorStatusOptions" @change="applyErrorFilters" />
+            </div>
+          </div>
+          <div v-else class="flex flex-1 flex-wrap items-end gap-4">
             <div class="w-full sm:w-auto sm:min-w-[220px]">
               <label class="input-label">{{ t('usage.apiKeyFilter') }}</label>
               <Select v-model="filters.api_key_id" :options="apiKeyOptions" @change="applyFilters" />
@@ -96,7 +122,7 @@
           </div>
 
           <div class="flex w-full flex-wrap items-center justify-end gap-3 sm:w-auto">
-            <button type="button" @click="refreshData" :disabled="loading" class="btn btn-secondary">
+            <button type="button" @click="refreshData" :disabled="activeTab === 'errors' ? errorLoading : loading" class="btn btn-secondary">
               {{ t('common.refresh') }}
             </button>
             <button type="button" @click="resetFilters" class="btn btn-secondary">
@@ -117,18 +143,18 @@
                 class="absolute right-0 top-full z-50 mt-1 max-h-80 w-48 overflow-y-auto rounded-lg border border-gray-200 bg-white py-1 shadow-lg dark:border-dark-600 dark:bg-dark-800"
               >
                 <button
-                  v-for="col in toggleableColumns"
+                  v-for="col in currentToggleableColumns"
                   :key="col.key"
                   type="button"
-                  @click="toggleColumn(col.key)"
+                  @click="toggleCurrentColumn(col.key)"
                   class="flex w-full items-center justify-between px-4 py-2 text-left text-sm text-gray-700 hover:bg-gray-100 dark:text-gray-300 dark:hover:bg-dark-700"
                 >
                   <span>{{ col.label }}</span>
-                  <Icon v-if="isColumnVisible(col.key)" name="check" size="sm" class="text-primary-500" />
+                  <Icon v-if="isCurrentColumnVisible(col.key)" name="check" size="sm" class="text-primary-500" />
                 </button>
               </div>
             </div>
-            <button type="button" @click="exportToCSV" :disabled="exporting" class="btn btn-primary">
+            <button v-if="activeTab !== 'errors'" type="button" @click="exportToCSV" :disabled="exporting" class="btn btn-primary">
               {{ exporting ? t('usage.exporting') : t('usage.exportCsv') }}
             </button>
           </div>
@@ -175,10 +201,11 @@
         :loading="errorLoading"
         :page="errorPage"
         :page-size="errorPageSize"
-        :api-keys="apiKeys"
-        @filter="onErrorFilter"
+        :visible-column-keys="errVisibleColumnKeys"
+        @sort="onErrorSort"
         @update:page="onErrorPage"
         @update:pageSize="onErrorPageSize"
+        @ipGeoBatchFailed="handleIpGeoBatchFailed"
       />
     </div>
   </AppLayout>
@@ -219,6 +246,7 @@ import type {
   UserErrorRequest,
 } from '@/types'
 import type { Column } from '@/components/common/types'
+import { COMMON_ERROR_STATUS_CODES } from '@/utils/errorBadges'
 
 const { t } = useI18n()
 const appStore = useAppStore()
@@ -244,13 +272,53 @@ const errorRows = ref<UserErrorRequest[]>([])
 const errorLoading = ref(false)
 const errorPage = ref(1)
 const errorPageSize = ref(20)
+const errorSortBy = ref('created_at')
+const errorSortOrder = ref<'asc' | 'desc'>('desc')
 const errorTotal = ref(0)
-const errorFilter = ref<{ model: string; category: string; api_key_id: number | null }>({
+const errorFilter = ref<{ model: string | null; category: string; api_key_id: number | null; status_code: number | null }>({
   model: '',
   category: '',
   api_key_id: null,
+  status_code: null,
 })
 
+const errorKeyOptions = computed<SelectOption[]>(() => [
+  { value: null, label: t('usage.errors.allKeys') },
+  ...apiKeys.value.map((k) => ({ value: k.id, label: k.name })),
+])
+
+// 模型候选取自当前已加载错误中出现过的模型；creatable 允许输入任意片段做后端模糊。
+const errorModelOptions = computed<SelectOption[]>(() => {
+  const seen = new Set<string>()
+  const opts: SelectOption[] = []
+  for (const r of errorRows.value) {
+    if (r.model && !seen.has(r.model)) {
+      seen.add(r.model)
+      opts.push({ value: r.model, label: r.model })
+    }
+  }
+  return opts
+})
+
+const errorCategoryCodes = ['auth', 'rate_limit', 'quota', 'invalid_request', 'service_unavailable', 'upstream', 'internal', 'cyber']
+
+const errorCategoryOptions = computed<SelectOption[]>(() => [
+  { value: '', label: t('usage.errors.allCategories') },
+  ...errorCategoryCodes.map((c) => ({ value: c, label: t('usage.errors.categories.' + c) })),
+])
+
+// 状态码候选用固定常用列表(与管理端 UsageFilters 共用常量),不受当前页数据限制:
+// 后端 status_code 过滤对全量生效,若只列当前页出现过的码,用户就选不到仅在后续页的码。
+const errorStatusOptions = computed<SelectOption[]>(() => [
+  { value: null, label: t('usage.errors.allStatuses') },
+  ...COMMON_ERROR_STATUS_CODES.map((c) => ({ value: c, label: String(c) })),
+])
+
+const applyErrorFilters = () => {
+  errorPage.value = 1
+  void loadErrors()
+}
+
 let abortController: AbortController | null = null
 let chartReqSeq = 0
 let statsReqSeq = 0
@@ -486,6 +554,10 @@ const resetFilters = () => {
   }
   granularity.value = getGranularityForRange(range.start, range.end)
   applyFilters()
+  if (activeTab.value === 'errors') {
+    errorFilter.value = { model: '', category: '', api_key_id: null, status_code: null }
+    applyErrorFilters()
+  }
 }
 
 const onDateRangeChange = (range: { startDate: string; endDate: string; preset: string | null }) => {
@@ -663,6 +735,63 @@ const loadSavedColumns = () => {
   }
 }
 
+// 错误请求 tab 独立列设置(机制同用量列设置,存储互不影响)
+const ERR_ALWAYS_VISIBLE = ['status', 'created_at']
+const ERR_DEFAULT_HIDDEN_COLUMNS = ['user_agent']
+const ERR_HIDDEN_COLUMNS_KEY = 'user-usage-error-hidden-columns'
+
+// key 须与 UserErrorRequestsTable 的 allColumns 一致
+const errAllColumns = computed<Column[]>(() => [
+  { key: 'key_name', label: t('usage.errors.keyName') },
+  { key: 'model', label: t('usage.errors.model') },
+  { key: 'endpoint', label: t('usage.errors.endpoint') },
+  { key: 'client_ip', label: 'IP' },
+  { key: 'group', label: t('admin.usage.group') },
+  { key: 'type', label: t('usage.type') },
+  { key: 'platform', label: t('usage.errors.platform') },
+  { key: 'category', label: t('usage.errors.category') },
+  { key: 'status', label: t('usage.errors.status') },
+  { key: 'message', label: t('usage.errors.message') },
+  { key: 'created_at', label: t('usage.errors.time') },
+  { key: 'user_agent', label: t('usage.userAgent') },
+])
+
+const errHiddenColumns = reactive<Set<string>>(new Set())
+const errToggleableColumns = computed(() =>
+  errAllColumns.value.filter((col) => !ERR_ALWAYS_VISIBLE.includes(col.key))
+)
+const errVisibleColumnKeys = computed(() =>
+  errAllColumns.value
+    .filter((col) => ERR_ALWAYS_VISIBLE.includes(col.key) || !errHiddenColumns.has(col.key))
+    .map((col) => col.key)
+)
+const isErrColumnVisible = (key: string) => !errHiddenColumns.has(key)
+const toggleErrColumn = (key: string) => {
+  if (errHiddenColumns.has(key)) errHiddenColumns.delete(key)
+  else errHiddenColumns.add(key)
+  localStorage.setItem(ERR_HIDDEN_COLUMNS_KEY, JSON.stringify([...errHiddenColumns]))
+}
+const loadSavedErrColumns = () => {
+  try {
+    const saved = localStorage.getItem(ERR_HIDDEN_COLUMNS_KEY)
+    const values = saved ? (JSON.parse(saved) as string[]) : ERR_DEFAULT_HIDDEN_COLUMNS
+    values.forEach((key) => errHiddenColumns.add(key))
+  } catch {
+    ERR_DEFAULT_HIDDEN_COLUMNS.forEach((key) => errHiddenColumns.add(key))
+  }
+}
+
+// 列设置下拉按当前 tab 分发
+const currentToggleableColumns = computed(() =>
+  activeTab.value === 'errors' ? errToggleableColumns.value : toggleableColumns.value
+)
+const isCurrentColumnVisible = (key: string) =>
+  activeTab.value === 'errors' ? isErrColumnVisible(key) : isColumnVisible(key)
+const toggleCurrentColumn = (key: string) => {
+  if (activeTab.value === 'errors') toggleErrColumn(key)
+  else toggleColumn(key)
+}
+
 const showColumnDropdown = ref(false)
 const columnDropdownRef = ref<HTMLElement | null>(null)
 const handleColumnClickOutside = (event: MouseEvent) => {
@@ -702,9 +831,12 @@ const loadErrors = async () => {
       page_size: errorPageSize.value,
       start_date: startDate.value,
       end_date: endDate.value,
-      model: errorFilter.value.model || undefined,
+      model: (errorFilter.value.model ?? '').trim() || undefined,
       category: errorFilter.value.category || undefined,
       api_key_id: errorFilter.value.api_key_id ?? undefined,
+      status_code: errorFilter.value.status_code ?? undefined,
+      sort_by: errorSortBy.value,
+      sort_order: errorSortOrder.value,
     })
     errorRows.value = resp.items
     errorTotal.value = resp.total
@@ -716,8 +848,9 @@ const loadErrors = async () => {
   }
 }
 
-const onErrorFilter = (filter: { model: string; category: string; api_key_id: number | null }) => {
-  errorFilter.value = filter
+const onErrorSort = (sortBy: string, sortOrder: 'asc' | 'desc') => {
+  errorSortBy.value = sortBy
+  errorSortOrder.value = sortOrder
   errorPage.value = 1
   void loadErrors()
 }
@@ -740,6 +873,7 @@ const switchToErrors = () => {
 
 onMounted(() => {
   loadSavedColumns()
+  loadSavedErrColumns()
   document.addEventListener('click', handleColumnClickOutside)
   void loadFilterOptions()
   refreshData()

From df2cedeead45fc220cfb06b11a340fc1b24612d7 Mon Sep 17 00:00:00 2001
From: zhaoteng45 <23356620+zhaoteng45@users.noreply.github.com>
Date: Fri, 3 Jul 2026 17:22:49 +0000
Subject: [PATCH 10/28] fix: normalize antigravity gemini 3.1 pro routing

---
 backend/internal/domain/constants.go          | 11 ++-
 backend/internal/domain/constants_test.go     | 22 +++++
 backend/internal/service/account.go           | 56 +++++++++++++
 .../internal/service/account_wildcard_test.go | 82 +++++++++++++++++++
 .../__tests__/useModelWhitelist.spec.ts       |  6 ++
 frontend/src/composables/useModelWhitelist.ts |  9 +-
 6 files changed, 180 insertions(+), 6 deletions(-)

diff --git a/backend/internal/domain/constants.go b/backend/internal/domain/constants.go
index 0d3bf88705..044e07d7a2 100644
--- a/backend/internal/domain/constants.go
+++ b/backend/internal/domain/constants.go
@@ -68,6 +68,9 @@ const (
 	SubscriptionStatusSuspended = "suspended"
 )
 
+// AntigravityGemini31ProAgentModel is the upstream route for Gemini 3.1 Pro High.
+const AntigravityGemini31ProAgentModel = "gemini-pro-agent"
+
 // DefaultAntigravityModelMapping 是 Antigravity 平台的默认模型映射
 // 当账号未配置 model_mapping 时使用此默认值
 // 与前端 useModelWhitelist.ts 中的 antigravityDefaultMappings 保持一致
@@ -103,10 +106,12 @@ var DefaultAntigravityModelMapping = map[string]string{
 	"gemini-3-flash-preview": "gemini-3-flash",
 	"gemini-3-pro-preview":   "gemini-3-pro-high",
 	// Gemini 3.1 白名单
-	"gemini-3.1-pro-high": "gemini-3.1-pro-high",
-	"gemini-3.1-pro-low":  "gemini-3.1-pro-low",
+	AntigravityGemini31ProAgentModel: AntigravityGemini31ProAgentModel,
+	"gemini-3.1-pro":                 AntigravityGemini31ProAgentModel,
+	"gemini-3.1-pro-high":            AntigravityGemini31ProAgentModel,
+	"gemini-3.1-pro-low":             "gemini-3.1-pro-low",
 	// Gemini 3.1 preview 映射
-	"gemini-3.1-pro-preview": "gemini-3.1-pro-high",
+	"gemini-3.1-pro-preview": AntigravityGemini31ProAgentModel,
 	// Gemini 3.1 image 白名单
 	"gemini-3.1-flash-image": "gemini-3.1-flash-image",
 	// Gemini 3.1 image preview 映射
diff --git a/backend/internal/domain/constants_test.go b/backend/internal/domain/constants_test.go
index 0b24aea915..0fb9054f7e 100644
--- a/backend/internal/domain/constants_test.go
+++ b/backend/internal/domain/constants_test.go
@@ -43,6 +43,28 @@ func TestDefaultAntigravityModelMapping_ContainsNewClaudeModels(t *testing.T) {
 	}
 }
 
+func TestDefaultAntigravityModelMapping_Gemini31ProAliases(t *testing.T) {
+	t.Parallel()
+
+	cases := map[string]string{
+		AntigravityGemini31ProAgentModel: AntigravityGemini31ProAgentModel,
+		"gemini-3.1-pro":                 AntigravityGemini31ProAgentModel,
+		"gemini-3.1-pro-high":            AntigravityGemini31ProAgentModel,
+		"gemini-3.1-pro-preview":         AntigravityGemini31ProAgentModel,
+		"gemini-3.1-pro-low":             "gemini-3.1-pro-low",
+	}
+
+	for from, want := range cases {
+		got, ok := DefaultAntigravityModelMapping[from]
+		if !ok {
+			t.Fatalf("expected mapping for %q to exist", from)
+		}
+		if got != want {
+			t.Fatalf("unexpected mapping for %q: got %q want %q", from, got, want)
+		}
+	}
+}
+
 func TestDefaultBedrockModelMapping_ContainsNewClaudeModels(t *testing.T) {
 	t.Parallel()
 
diff --git a/backend/internal/service/account.go b/backend/internal/service/account.go
index 7d5e70a15b..26af0a7281 100644
--- a/backend/internal/service/account.go
+++ b/backend/internal/service/account.go
@@ -580,6 +580,7 @@ func (a *Account) resolveModelMapping(rawMapping map[string]any) map[string]stri
 				"gemini-3.1-pro-high",
 				"gemini-3.1-pro-low",
 			})
+			applyAntigravityGemini31ProAliases(result)
 		}
 		return result
 	}
@@ -646,6 +647,61 @@ func ensureAntigravityDefaultPassthroughs(mapping map[string]string, models []st
 	}
 }
 
+func applyAntigravityGemini31ProAliases(mapping map[string]string) {
+	target := strings.TrimSpace(mapping[domain.AntigravityGemini31ProAgentModel])
+	if target == "" {
+		return
+	}
+
+	aliases := []struct {
+		model         string
+		legacyTargets map[string]struct{}
+	}{
+		{
+			model: "gemini-3.1-pro",
+			legacyTargets: map[string]struct{}{
+				"gemini-3.1-pro": {},
+			},
+		},
+		{
+			model: "gemini-3.1-pro-high",
+			legacyTargets: map[string]struct{}{
+				"gemini-3.1-pro-high": {},
+			},
+		},
+		{
+			model: "gemini-3.1-pro-preview",
+			legacyTargets: map[string]struct{}{
+				"gemini-3.1-pro-preview": {},
+				"gemini-3.1-pro-high":    {},
+			},
+		},
+	}
+
+	for _, alias := range aliases {
+		current, exists := mapping[alias.model]
+		if exists {
+			if _, legacy := alias.legacyTargets[current]; legacy {
+				mapping[alias.model] = target
+			}
+			continue
+		}
+		if mappingHasWildcardForModel(mapping, alias.model) {
+			continue
+		}
+		mapping[alias.model] = target
+	}
+}
+
+func mappingHasWildcardForModel(mapping map[string]string, model string) bool {
+	for pattern := range mapping {
+		if matchWildcard(pattern, model) {
+			return true
+		}
+	}
+	return false
+}
+
 func normalizeRequestedModelForLookup(platform, requestedModel string) string {
 	trimmed := strings.TrimSpace(requestedModel)
 	if trimmed == "" {
diff --git a/backend/internal/service/account_wildcard_test.go b/backend/internal/service/account_wildcard_test.go
index d903b940a5..6ce804bb16 100644
--- a/backend/internal/service/account_wildcard_test.go
+++ b/backend/internal/service/account_wildcard_test.go
@@ -4,6 +4,8 @@ package service
 
 import (
 	"testing"
+
+	"github.com/Wei-Shaw/sub2api/internal/domain"
 )
 
 func TestMatchWildcard(t *testing.T) {
@@ -320,6 +322,86 @@ func TestAccountGetMappedModel(t *testing.T) {
 	}
 }
 
+func TestAccountGetModelMapping_AntigravityNormalizesGemini31ProAliases(t *testing.T) {
+	t.Parallel()
+
+	account := &Account{
+		Platform: PlatformAntigravity,
+		Credentials: map[string]any{
+			"model_mapping": map[string]any{
+				domain.AntigravityGemini31ProAgentModel: domain.AntigravityGemini31ProAgentModel,
+				"gemini-3.1-pro-high":                   "gemini-3.1-pro-high",
+				"gemini-3.1-pro-preview":                "gemini-3.1-pro-high",
+			},
+		},
+	}
+
+	mapping := account.GetModelMapping()
+
+	if got := mapping["gemini-3.1-pro"]; got != domain.AntigravityGemini31ProAgentModel {
+		t.Fatalf("expected gemini-3.1-pro to map to %q, got %q", domain.AntigravityGemini31ProAgentModel, got)
+	}
+	if got := mapping["gemini-3.1-pro-high"]; got != domain.AntigravityGemini31ProAgentModel {
+		t.Fatalf("expected gemini-3.1-pro-high to map to %q, got %q", domain.AntigravityGemini31ProAgentModel, got)
+	}
+	if got := mapping["gemini-3.1-pro-preview"]; got != domain.AntigravityGemini31ProAgentModel {
+		t.Fatalf("expected gemini-3.1-pro-preview to map to %q, got %q", domain.AntigravityGemini31ProAgentModel, got)
+	}
+}
+
+func TestAccountGetModelMapping_AntigravityPreservesGemini31ProOverrides(t *testing.T) {
+	t.Parallel()
+
+	account := &Account{
+		Platform: PlatformAntigravity,
+		Credentials: map[string]any{
+			"model_mapping": map[string]any{
+				domain.AntigravityGemini31ProAgentModel: domain.AntigravityGemini31ProAgentModel,
+				"gemini-3.1-pro-high":                   "custom-high",
+				"gemini-3.1-pro-preview":                "custom-preview",
+			},
+		},
+	}
+
+	mapping := account.GetModelMapping()
+
+	if got := mapping["gemini-3.1-pro-high"]; got != "custom-high" {
+		t.Fatalf("expected gemini-3.1-pro-high override to be preserved, got %q", got)
+	}
+	if got := mapping["gemini-3.1-pro-preview"]; got != "custom-preview" {
+		t.Fatalf("expected gemini-3.1-pro-preview override to be preserved, got %q", got)
+	}
+	if got := mapping["gemini-3.1-pro"]; got != domain.AntigravityGemini31ProAgentModel {
+		t.Fatalf("expected gemini-3.1-pro alias to default to %q, got %q", domain.AntigravityGemini31ProAgentModel, got)
+	}
+}
+
+func TestAccountGetModelMapping_AntigravityGemini31ProAliasesRespectWildcard(t *testing.T) {
+	t.Parallel()
+
+	account := &Account{
+		Platform: PlatformAntigravity,
+		Credentials: map[string]any{
+			"model_mapping": map[string]any{
+				domain.AntigravityGemini31ProAgentModel: domain.AntigravityGemini31ProAgentModel,
+				"gemini-3.1-*":                          "custom-wildcard",
+			},
+		},
+	}
+
+	mapping := account.GetModelMapping()
+
+	if got := mapping["gemini-3.1-pro"]; got != "" {
+		t.Fatalf("expected gemini-3.1-pro exact alias to stay unset when wildcard exists, got %q", got)
+	}
+	if got := mapping["gemini-3.1-pro-high"]; got != "" {
+		t.Fatalf("expected gemini-3.1-pro-high exact alias to stay unset when wildcard exists, got %q", got)
+	}
+	if got := mapping["gemini-3.1-pro-preview"]; got != "" {
+		t.Fatalf("expected gemini-3.1-pro-preview exact alias to stay unset when wildcard exists, got %q", got)
+	}
+}
+
 func TestAccountResolveMappedModel(t *testing.T) {
 	tests := []struct {
 		name           string
diff --git a/frontend/src/composables/__tests__/useModelWhitelist.spec.ts b/frontend/src/composables/__tests__/useModelWhitelist.spec.ts
index 7f3157dfff..d7e70e309d 100644
--- a/frontend/src/composables/__tests__/useModelWhitelist.spec.ts
+++ b/frontend/src/composables/__tests__/useModelWhitelist.spec.ts
@@ -58,6 +58,12 @@ describe('useModelWhitelist', () => {
     expect(models.indexOf('gemini-2.5-flash-image')).toBeLessThan(models.indexOf('gemini-2.5-flash-lite'))
   })
 
+  it('antigravity 模型列表包含 Gemini 3.1 Pro 通用别名', () => {
+    const models = getModelsByPlatform('antigravity')
+
+    expect(models).toContain('gemini-3.1-pro')
+  })
+
   it('whitelist 模式会忽略通配符条目', () => {
     const mapping = buildModelMappingObject('whitelist', ['claude-*', 'gemini-3.1-flash-image'], [])
     expect(mapping).toEqual({
diff --git a/frontend/src/composables/useModelWhitelist.ts b/frontend/src/composables/useModelWhitelist.ts
index 43449b09ee..0fea5d93f8 100644
--- a/frontend/src/composables/useModelWhitelist.ts
+++ b/frontend/src/composables/useModelWhitelist.ts
@@ -74,6 +74,7 @@ const antigravityModels = [
   'gemini-3-pro-high',
   'gemini-3-pro-low',
   // Gemini 3.1 系列
+  'gemini-3.1-pro',
   'gemini-3.1-pro-high',
   'gemini-3.1-pro-low',
   'gemini-3-pro-image',
@@ -302,11 +303,13 @@ const antigravityPresetMappings = [
   { label: 'Sonnet3.5→4.6', from: 'claude-3-5-sonnet-20241022', to: 'claude-sonnet-4-6', color: 'bg-teal-100 text-teal-700 hover:bg-teal-200 dark:bg-teal-900/30 dark:text-teal-400' },
   { label: 'Opus4.5→4.6', from: 'claude-opus-4-5-20251101', to: 'claude-opus-4-6-thinking', color: 'bg-violet-100 text-violet-700 hover:bg-violet-200 dark:bg-violet-900/30 dark:text-violet-400' },
   // Gemini 3→3.1 映射
-  { label: '3-Pro-Preview→3.1-Pro-High', from: 'gemini-3-pro-preview', to: 'gemini-3.1-pro-high', color: 'bg-amber-100 text-amber-700 hover:bg-amber-200 dark:bg-amber-900/30 dark:text-amber-400' },
-  { label: '3-Pro-High→3.1-Pro-High', from: 'gemini-3-pro-high', to: 'gemini-3.1-pro-high', color: 'bg-orange-100 text-orange-700 hover:bg-orange-200 dark:bg-orange-900/30 dark:text-orange-400' },
+  { label: '3-Pro-Preview→Pro-Agent', from: 'gemini-3-pro-preview', to: 'gemini-pro-agent', color: 'bg-amber-100 text-amber-700 hover:bg-amber-200 dark:bg-amber-900/30 dark:text-amber-400' },
+  { label: '3-Pro-High→Pro-Agent', from: 'gemini-3-pro-high', to: 'gemini-pro-agent', color: 'bg-orange-100 text-orange-700 hover:bg-orange-200 dark:bg-orange-900/30 dark:text-orange-400' },
   { label: '3-Pro-Low→3.1-Pro-Low', from: 'gemini-3-pro-low', to: 'gemini-3.1-pro-low', color: 'bg-yellow-100 text-yellow-700 hover:bg-yellow-200 dark:bg-yellow-900/30 dark:text-yellow-400' },
-  { label: '3.1-Pro-High passthrough', from: 'gemini-3.1-pro-high', to: 'gemini-3.1-pro-high', color: 'bg-orange-100 text-orange-700 hover:bg-orange-200 dark:bg-orange-900/30 dark:text-orange-400' },
+  { label: '3.1-Pro→Pro-Agent', from: 'gemini-3.1-pro', to: 'gemini-pro-agent', color: 'bg-orange-100 text-orange-700 hover:bg-orange-200 dark:bg-orange-900/30 dark:text-orange-400' },
+  { label: '3.1-Pro-High→Pro-Agent', from: 'gemini-3.1-pro-high', to: 'gemini-pro-agent', color: 'bg-orange-100 text-orange-700 hover:bg-orange-200 dark:bg-orange-900/30 dark:text-orange-400' },
   { label: '3.1-Pro-Low passthrough', from: 'gemini-3.1-pro-low', to: 'gemini-3.1-pro-low', color: 'bg-yellow-100 text-yellow-700 hover:bg-yellow-200 dark:bg-yellow-900/30 dark:text-yellow-400' },
+  { label: '3.1-Pro-Preview→Pro-Agent', from: 'gemini-3.1-pro-preview', to: 'gemini-pro-agent', color: 'bg-amber-100 text-amber-700 hover:bg-amber-200 dark:bg-amber-900/30 dark:text-amber-400' },
   // Gemini 通配符映射
   { label: 'Gemini 3→Flash', from: 'gemini-3*', to: 'gemini-3-flash', color: 'bg-yellow-100 text-yellow-700 hover:bg-yellow-200 dark:bg-yellow-900/30 dark:text-yellow-400' },
   { label: 'Gemini 2.5→Flash', from: 'gemini-2.5*', to: 'gemini-2.5-flash', color: 'bg-orange-100 text-orange-700 hover:bg-orange-200 dark:bg-orange-900/30 dark:text-orange-400' },

From e5dc1f597271fd1424df94fe4cb16e83dae1fbf0 Mon Sep 17 00:00:00 2001
From: jjaw <jjaw@jjaw.cn>
Date: Sat, 4 Jul 2026 04:30:15 +0800
Subject: [PATCH 11/28] =?UTF-8?q?=E6=B7=BB=E5=8A=A0=20token=5Fexpired=20?=
 =?UTF-8?q?=E5=88=B7=E6=96=B0=E4=B8=8D=E5=8F=AF=E9=87=8D=E8=AF=95=E5=88=A4?=
 =?UTF-8?q?=E6=96=AD?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 backend/internal/service/token_refresh_service.go      | 1 +
 backend/internal/service/token_refresh_service_test.go | 1 +
 2 files changed, 2 insertions(+)

diff --git a/backend/internal/service/token_refresh_service.go b/backend/internal/service/token_refresh_service.go
index 08761f8220..9ca369a52a 100644
--- a/backend/internal/service/token_refresh_service.go
+++ b/backend/internal/service/token_refresh_service.go
@@ -446,6 +446,7 @@ func isNonRetryableRefreshError(err error) bool {
 	nonRetryable := []string{
 		"invalid_grant",             // refresh_token 已失效
 		"invalid_refresh_token",     // refresh_token 无效, team 账号工作区被删除会出现
+		"token_expired",             // OpenAI refresh_token 已过期，需要重新授权
 		"app_session_terminated",    // refresh_token team 账号工作区被删除
 		"refresh_token_reused",      // OpenAI refresh_token 已被使用，必须重新授权
 		"refresh_token_invalidated", // OpenAI session ended; refresh token invalidated
diff --git a/backend/internal/service/token_refresh_service_test.go b/backend/internal/service/token_refresh_service_test.go
index d2315c1db5..6f227352f9 100644
--- a/backend/internal/service/token_refresh_service_test.go
+++ b/backend/internal/service/token_refresh_service_test.go
@@ -541,6 +541,7 @@ func TestIsNonRetryableRefreshError(t *testing.T) {
 		{name: "invalid_grant", err: errors.New("invalid_grant"), expected: true},
 		{name: "invalid_client", err: errors.New("invalid_client"), expected: true},
 		{name: "invalid_refresh_token", err: errors.New(`OPENAI_OAUTH_TOKEN_REFRESH_FAILED: token refresh failed: status 401, body: {"error":{"code":"invalid_refresh_token"}}`), expected: true},
+		{name: "token_expired", err: errors.New(`OPENAI_OAUTH_TOKEN_REFRESH_FAILED: token refresh failed: status 401, body: {"error":{"code":"token_expired"}}`), expected: true},
 		{name: "refresh_token_reused", err: errors.New(`OPENAI_OAUTH_TOKEN_REFRESH_FAILED: token refresh failed: status 401, body: {"error":{"code":"refresh_token_reused"}}`), expected: true},
 		{name: "app_session_terminated", err: errors.New(`OPENAI_OAUTH_TOKEN_REFRESH_FAILED: token refresh failed: status 401, body: {"error": {"code": "app_session_terminated"}}`), expected: true},
 		{name: "unauthorized_client", err: errors.New("unauthorized_client"), expected: true},

From b3f796972e72536698c86ff87dd7d7155105e199 Mon Sep 17 00:00:00 2001
From: shaw <dev@sub2api.org>
Date: Sat, 4 Jul 2026 08:28:30 +0800
Subject: [PATCH 12/28] feat(anthropic): treat 7d_oi (Fable) window 429 as
 model-level rate limit

Anthropic OAuth/setup-token accounts now expose a Fable-only 7d window in
429 responses (anthropic-ratelimit-unified-7d_oi-*). Previously, a 429
triggered solely by this window fell through to the legacy path and marked
the whole account rate limited (picking the sooner 5h/7d reset) even though
5h/7d were still "allowed".

Backend:
- HandleUpstreamError: when only 7d_oi is rejected/exceeded, set a
  model-level rate limit (scope "claude-fable-5") and skip account-level
  SetRateLimited / session-window rewrite / temp-unsched entirely.
  Real 5h/7d exhaustion keeps the exact legacy account-level behavior.
- modelRateLimitKeysForRequest: append the Fable family scope for any
  mapped Anthropic model containing "fable", so claude-fable-5[1m] and
  other variants are excluded from scheduling together.
- Passive sampling: collect 7d_oi utilization/reset into Extra
  (passive_usage_7d_oi_*), including from the 429 response itself so the
  usage bar doesn't freeze at a stale value during the limit window;
  cleared on 5h window reset like the existing 7d keys.
- Usage API: new seven_day_fable window in UsageInfo. GetPassiveUsage
  builds it from passive samples; active queries parse the upstream
  seven_day_overage_included field (claim name from the
  representative-claim header) and fall back to passive data when the
  upstream usage API doesn't return it, so the active-query button keeps
  the bar populated. seven_day_sonnet handling is preserved unchanged.

Frontend:
- /admin/accounts usage cell renders a "7d F" (amber) progress bar next to
  the existing 5h / 7d / 7d S bars for Anthropic OAuth/setup-token accounts.
---
 .../internal/service/account_usage_service.go |  91 +++++++---
 .../account_usage_service_fable_test.go       | 119 +++++++++++++
 backend/internal/service/model_rate_limit.go  |  12 ++
 .../internal/service/model_rate_limit_test.go |  44 +++++
 backend/internal/service/ratelimit_service.go | 152 +++++++++++++---
 .../ratelimit_service_anthropic_test.go       | 137 ++++++++++++++
 ...mit_service_anthropic_window_limit_test.go | 167 +++++++++++++++++-
 .../service/ratelimit_session_window_test.go  |  53 ++++++
 .../components/account/AccountUsageCell.vue   |   9 +
 .../__tests__/AccountUsageCell.spec.ts        |  98 ++++++++++
 frontend/src/types/index.ts                   |   1 +
 11 files changed, 835 insertions(+), 48 deletions(-)
 create mode 100644 backend/internal/service/account_usage_service_fable_test.go

diff --git a/backend/internal/service/account_usage_service.go b/backend/internal/service/account_usage_service.go
index 3c50baaec2..d9f6200359 100644
--- a/backend/internal/service/account_usage_service.go
+++ b/backend/internal/service/account_usage_service.go
@@ -184,6 +184,7 @@ type UsageInfo struct {
 	FiveHour           *UsageProgress `json:"five_hour"`                      // 5小时窗口
 	SevenDay           *UsageProgress `json:"seven_day,omitempty"`            // 7天窗口
 	SevenDaySonnet     *UsageProgress `json:"seven_day_sonnet,omitempty"`     // 7天Sonnet窗口
+	SevenDayFable      *UsageProgress `json:"seven_day_fable,omitempty"`      // 7天Fable窗口（响应头 7d_oi）
 	GeminiSharedDaily  *UsageProgress `json:"gemini_shared_daily,omitempty"`  // Gemini shared pool RPD (Google One / Code Assist)
 	GeminiProDaily     *UsageProgress `json:"gemini_pro_daily,omitempty"`     // Gemini Pro 日配额
 	GeminiFlashDaily   *UsageProgress `json:"gemini_flash_daily,omitempty"`   // Gemini Flash 日配额
@@ -236,6 +237,12 @@ type UsageInfo struct {
 	Error string `json:"error,omitempty"`
 }
 
+// ClaudeUsageWindow Anthropic /api/oauth/usage 返回的单个用量窗口
+type ClaudeUsageWindow struct {
+	Utilization float64 `json:"utilization"`
+	ResetsAt    string  `json:"resets_at"`
+}
+
 // ClaudeUsageResponse Anthropic API返回的usage结构
 type ClaudeUsageResponse struct {
 	FiveHour struct {
@@ -250,6 +257,10 @@ type ClaudeUsageResponse struct {
 		Utilization float64 `json:"utilization"`
 		ResetsAt    string  `json:"resets_at"`
 	} `json:"seven_day_sonnet"`
+	// Fable 专属 7d 窗口（对应响应头 7d_oi，claim 名为 seven_day_overage_included，
+	// 见 anthropic-ratelimit-unified-representative-claim 头）。上游 usage API
+	// 若不下发该字段，GetUsage 会用被动采样数据回填。
+	SevenDayOverageIncluded ClaudeUsageWindow `json:"seven_day_overage_included"`
 }
 
 // ClaudeUsageFetchOptions 包含获取 Claude 用量数据所需的所有选项
@@ -429,6 +440,12 @@ func (s *AccountUsageService) GetUsage(ctx context.Context, accountID int64, for
 		// 5. 将主动查询结果同步到被动缓存，下次 passive 加载即为最新值
 		s.syncActiveToPassive(ctx, account.ID, usage)
 
+		// 6. 上游 usage API 目前不一定下发 Fable 7d 窗口；缺失时回填被动采样
+		// （7d_oi 响应头）的数据，避免主动查询后 7d F 进度条丢失。
+		if usage.SevenDayFable == nil {
+			usage.SevenDayFable = buildPassiveUsageWindow(account.Extra, "passive_usage_7d_oi_utilization", "passive_usage_7d_oi_reset")
+		}
+
 		s.tryClearRecoverableAccountError(ctx, account)
 		return usage, nil
 	}
@@ -471,25 +488,10 @@ func (s *AccountUsageService) GetPassiveUsage(ctx context.Context, accountID int
 	}
 
 	// 构建 7d 窗口（从被动采样数据）
-	util7d := parseExtraFloat64(account.Extra["passive_usage_7d_utilization"])
-	reset7dRaw := parseExtraFloat64(account.Extra["passive_usage_7d_reset"])
-	if util7d > 0 || reset7dRaw > 0 {
-		var resetAt *time.Time
-		var remaining int
-		if reset7dRaw > 0 {
-			t := time.Unix(int64(reset7dRaw), 0)
-			resetAt = &t
-			remaining = int(time.Until(t).Seconds())
-			if remaining < 0 {
-				remaining = 0
-			}
-		}
-		info.SevenDay = &UsageProgress{
-			Utilization:      util7d * 100,
-			ResetsAt:         resetAt,
-			RemainingSeconds: remaining,
-		}
-	}
+	info.SevenDay = buildPassiveUsageWindow(account.Extra, "passive_usage_7d_utilization", "passive_usage_7d_reset")
+
+	// 构建 7d Fable 窗口（从被动采样的 7d_oi 响应头数据）
+	info.SevenDayFable = buildPassiveUsageWindow(account.Extra, "passive_usage_7d_oi_utilization", "passive_usage_7d_oi_reset")
 
 	// 添加窗口统计
 	s.addWindowStats(ctx, account, info)
@@ -497,6 +499,31 @@ func (s *AccountUsageService) GetPassiveUsage(ctx context.Context, accountID int
 	return info, nil
 }
 
+// buildPassiveUsageWindow 从 Extra 中的被动采样数据（utilization 为 0-1 小数、reset 为 Unix 秒）
+// 构建用量窗口，无数据时返回 nil。
+func buildPassiveUsageWindow(extra map[string]any, utilKey, resetKey string) *UsageProgress {
+	util := parseExtraFloat64(extra[utilKey])
+	resetRaw := parseExtraFloat64(extra[resetKey])
+	if util <= 0 && resetRaw <= 0 {
+		return nil
+	}
+	var resetAt *time.Time
+	var remaining int
+	if resetRaw > 0 {
+		t := time.Unix(int64(resetRaw), 0)
+		resetAt = &t
+		remaining = int(time.Until(t).Seconds())
+		if remaining < 0 {
+			remaining = 0
+		}
+	}
+	return &UsageProgress{
+		Utilization:      util * 100,
+		ResetsAt:         resetAt,
+		RemainingSeconds: remaining,
+	}
+}
+
 // syncActiveToPassive 将主动查询的最新数据回写到 Extra 被动缓存，
 // 这样下次被动加载时能看到最新值。
 func (s *AccountUsageService) syncActiveToPassive(ctx context.Context, accountID int64, usage *UsageInfo) {
@@ -511,6 +538,12 @@ func (s *AccountUsageService) syncActiveToPassive(ctx context.Context, accountID
 			extraUpdates["passive_usage_7d_reset"] = usage.SevenDay.ResetsAt.Unix()
 		}
 	}
+	if usage.SevenDayFable != nil {
+		extraUpdates["passive_usage_7d_oi_utilization"] = usage.SevenDayFable.Utilization / 100
+		if usage.SevenDayFable.ResetsAt != nil {
+			extraUpdates["passive_usage_7d_oi_reset"] = usage.SevenDayFable.ResetsAt.Unix()
+		}
+	}
 
 	if len(extraUpdates) > 0 {
 		extraUpdates["passive_usage_sampled_at"] = time.Now().UTC().Format(time.RFC3339)
@@ -1010,8 +1043,8 @@ func enrichUsageWithAccountError(info *UsageInfo, account *Account) {
 // 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) {
 	// 修复：即使 FiveHour 为 nil，也要尝试获取统计数据
-	// 因为 SevenDay/SevenDaySonnet 可能需要
-	if usage.FiveHour == nil && usage.SevenDay == nil && usage.SevenDaySonnet == nil {
+	// 因为 SevenDay/SevenDaySonnet/SevenDayFable 可能需要
+	if usage.FiveHour == nil && usage.SevenDay == nil && usage.SevenDaySonnet == nil && usage.SevenDayFable == nil {
 		return
 	}
 
@@ -1347,6 +1380,22 @@ func (s *AccountUsageService) buildUsageInfo(resp *ClaudeUsageResponse, updatedA
 		}
 	}
 
+	// 7天Fable窗口（响应头 7d_oi 对应的窗口）
+	if fable := resp.SevenDayOverageIncluded; fable.ResetsAt != "" {
+		if fableReset, err := parseTime(fable.ResetsAt); err == nil {
+			info.SevenDayFable = &UsageProgress{
+				Utilization:      fable.Utilization,
+				ResetsAt:         &fableReset,
+				RemainingSeconds: int(time.Until(fableReset).Seconds()),
+			}
+		} else {
+			log.Printf("Failed to parse SevenDayFable.ResetsAt: %s, error: %v", fable.ResetsAt, err)
+			info.SevenDayFable = &UsageProgress{
+				Utilization: fable.Utilization,
+			}
+		}
+	}
+
 	return info
 }
 
diff --git a/backend/internal/service/account_usage_service_fable_test.go b/backend/internal/service/account_usage_service_fable_test.go
new file mode 100644
index 0000000000..60f2e9c79f
--- /dev/null
+++ b/backend/internal/service/account_usage_service_fable_test.go
@@ -0,0 +1,119 @@
+package service
+
+import (
+	"encoding/json"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/require"
+)
+
+func TestClaudeUsageResponse_FableWindowDecoding(t *testing.T) {
+	t.Run("seven_day_overage_included", func(t *testing.T) {
+		raw := `{
+  "five_hour": {"utilization": 12.0, "resets_at": "2026-07-03T10:00:00Z"},
+  "seven_day": {"utilization": 34.0, "resets_at": "2026-07-08T00:00:00Z"},
+  "seven_day_overage_included": {"utilization": 56.0, "resets_at": "2026-07-08T03:00:00Z"}
+}`
+		var resp ClaudeUsageResponse
+		require.NoError(t, json.Unmarshal([]byte(raw), &resp))
+		require.Equal(t, 56.0, resp.SevenDayOverageIncluded.Utilization)
+		require.Equal(t, "2026-07-08T03:00:00Z", resp.SevenDayOverageIncluded.ResetsAt)
+	})
+
+	t.Run("absent", func(t *testing.T) {
+		raw := `{"five_hour": {"utilization": 12.0, "resets_at": "2026-07-03T10:00:00Z"}}`
+		var resp ClaudeUsageResponse
+		require.NoError(t, json.Unmarshal([]byte(raw), &resp))
+		require.Zero(t, resp.SevenDayOverageIncluded.Utilization)
+		require.Empty(t, resp.SevenDayOverageIncluded.ResetsAt)
+	})
+}
+
+func TestBuildUsageInfo_SevenDayFable(t *testing.T) {
+	svc := &AccountUsageService{}
+	now := time.Now()
+
+	resetAt := now.Add(72 * time.Hour).UTC().Truncate(time.Second)
+	var resp ClaudeUsageResponse
+	resp.FiveHour.Utilization = 10
+	resp.SevenDayOverageIncluded = ClaudeUsageWindow{
+		Utilization: 88,
+		ResetsAt:    resetAt.Format(time.RFC3339),
+	}
+
+	info := svc.buildUsageInfo(&resp, &now)
+	require.NotNil(t, info.SevenDayFable)
+	require.Equal(t, 88.0, info.SevenDayFable.Utilization)
+	require.NotNil(t, info.SevenDayFable.ResetsAt)
+	require.True(t, info.SevenDayFable.ResetsAt.Equal(resetAt))
+	require.Greater(t, info.SevenDayFable.RemainingSeconds, 0)
+
+	// 无 Fable 数据时不应创建窗口
+	var empty ClaudeUsageResponse
+	empty.FiveHour.Utilization = 10
+	info = svc.buildUsageInfo(&empty, &now)
+	require.Nil(t, info.SevenDayFable)
+}
+
+func TestBuildPassiveUsageWindow(t *testing.T) {
+	future := time.Now().Add(48 * time.Hour).Unix()
+
+	t.Run("utilization and reset", func(t *testing.T) {
+		window := buildPassiveUsageWindow(map[string]any{
+			"passive_usage_7d_oi_utilization": 0.87,
+			"passive_usage_7d_oi_reset":       float64(future),
+		}, "passive_usage_7d_oi_utilization", "passive_usage_7d_oi_reset")
+		require.NotNil(t, window)
+		require.InDelta(t, 87.0, window.Utilization, 1e-9)
+		require.NotNil(t, window.ResetsAt)
+		require.Equal(t, future, window.ResetsAt.Unix())
+		require.Greater(t, window.RemainingSeconds, 0)
+	})
+
+	t.Run("no data returns nil", func(t *testing.T) {
+		require.Nil(t, buildPassiveUsageWindow(nil, "u", "r"))
+		require.Nil(t, buildPassiveUsageWindow(map[string]any{}, "u", "r"))
+	})
+
+	t.Run("expired reset clamps remaining to zero", func(t *testing.T) {
+		past := time.Now().Add(-time.Hour).Unix()
+		window := buildPassiveUsageWindow(map[string]any{
+			"u": 0.5,
+			"r": float64(past),
+		}, "u", "r")
+		require.NotNil(t, window)
+		require.Equal(t, 0, window.RemainingSeconds)
+	})
+
+	t.Run("utilization only", func(t *testing.T) {
+		window := buildPassiveUsageWindow(map[string]any{"u": 0.25}, "u", "r")
+		require.NotNil(t, window)
+		require.InDelta(t, 25.0, window.Utilization, 1e-9)
+		require.Nil(t, window.ResetsAt)
+	})
+}
+
+func TestSyncActiveToPassive_WritesFableExtras(t *testing.T) {
+	repo := &accountUsageCodexProbeRepo{updateExtraCh: make(chan map[string]any, 1)}
+	svc := &AccountUsageService{accountRepo: repo}
+
+	resetAt := time.Now().Add(72 * time.Hour).Truncate(time.Second)
+	usage := &UsageInfo{
+		SevenDayFable: &UsageProgress{
+			Utilization: 87,
+			ResetsAt:    &resetAt,
+		},
+	}
+
+	svc.syncActiveToPassive(t.Context(), 1, usage)
+
+	select {
+	case updates := <-repo.updateExtraCh:
+		require.InDelta(t, 0.87, updates["passive_usage_7d_oi_utilization"], 1e-9)
+		require.Equal(t, resetAt.Unix(), updates["passive_usage_7d_oi_reset"])
+		require.Contains(t, updates, "passive_usage_sampled_at")
+	default:
+		t.Fatal("expected UpdateExtra to be called with fable extras")
+	}
+}
diff --git a/backend/internal/service/model_rate_limit.go b/backend/internal/service/model_rate_limit.go
index 1568a00f78..195f5a1dfd 100644
--- a/backend/internal/service/model_rate_limit.go
+++ b/backend/internal/service/model_rate_limit.go
@@ -12,6 +12,9 @@ const (
 	modelRateLimitsKey                 = "model_rate_limits"
 	antigravityGeminiModelRateLimitKey = "antigravity:gemini"
 	openAIImageGenerationRateLimitKey  = "openai:image_generation"
+	// anthropicFableRateLimitKey 是 Anthropic 7d_oi（Fable 专属 7d 窗口）限流的
+	// 家族级 scope：命中后所有 Fable 变体（含 [1m] 等后缀）都不再调度到该账号。
+	anthropicFableRateLimitKey = "claude-fable-5"
 )
 
 // isRateLimitActiveForKey 检查指定 key 的限流是否生效
@@ -82,10 +85,19 @@ func (a *Account) modelRateLimitKeysForRequest(ctx context.Context, requestedMod
 		if openAIImageGenerationRateLimitApplies(ctx, requestedModel, modelKey) && modelKey != openAIImageGenerationRateLimitKey {
 			keys = append(keys, openAIImageGenerationRateLimitKey)
 		}
+	case PlatformAnthropic:
+		if isAnthropicFableModel(modelKey) && modelKey != anthropicFableRateLimitKey {
+			keys = append(keys, anthropicFableRateLimitKey)
+		}
 	}
 	return keys
 }
 
+// isAnthropicFableModel 判断是否为 Fable 模型家族（claude-fable-5、claude-fable-5[1m] 等变体）
+func isAnthropicFableModel(model string) bool {
+	return strings.Contains(strings.ToLower(model), "fable")
+}
+
 func openAIImageGenerationRateLimitApplies(ctx context.Context, requestedModel, modelKey string) bool {
 	if isOpenAIImageGenerationModel(requestedModel) || isOpenAIImageGenerationModel(modelKey) {
 		return true
diff --git a/backend/internal/service/model_rate_limit_test.go b/backend/internal/service/model_rate_limit_test.go
index c62e4c4b9b..430d91baf6 100644
--- a/backend/internal/service/model_rate_limit_test.go
+++ b/backend/internal/service/model_rate_limit_test.go
@@ -499,3 +499,47 @@ func TestGetRateLimitRemainingTime(t *testing.T) {
 		})
 	}
 }
+
+func TestIsModelRateLimited_AnthropicFableFamilyKey(t *testing.T) {
+	now := time.Now()
+	future := now.Add(48 * time.Hour).Format(time.RFC3339)
+
+	account := &Account{
+		Platform: PlatformAnthropic,
+		Extra: map[string]any{
+			modelRateLimitsKey: map[string]any{
+				anthropicFableRateLimitKey: map[string]any{
+					"rate_limit_reset_at": future,
+				},
+			},
+		},
+	}
+
+	tests := []struct {
+		requestedModel string
+		expected       bool
+	}{
+		{"claude-fable-5", true},
+		{"claude-fable-5[1m]", true},      // 家族 key 覆盖变体
+		{"Claude-Fable-5-20260601", true}, // 大小写不敏感
+		{"claude-sonnet-4-6", false},      // 其他模型不受影响
+		{"claude-opus-4-8", false},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.requestedModel, func(t *testing.T) {
+			got := account.isModelRateLimitedWithContext(context.Background(), tc.requestedModel)
+			require.Equal(t, tc.expected, got)
+			remaining := account.GetModelRateLimitRemainingTimeWithContext(context.Background(), tc.requestedModel)
+			require.Equal(t, tc.expected, remaining > 0)
+		})
+	}
+}
+
+func TestIsAnthropicFableModel(t *testing.T) {
+	require.True(t, isAnthropicFableModel("claude-fable-5"))
+	require.True(t, isAnthropicFableModel("claude-fable-5[1m]"))
+	require.True(t, isAnthropicFableModel("Claude-Fable-5"))
+	require.False(t, isAnthropicFableModel("claude-sonnet-4-6"))
+	require.False(t, isAnthropicFableModel(""))
+}
diff --git a/backend/internal/service/ratelimit_service.go b/backend/internal/service/ratelimit_service.go
index 9e968853f4..2b9c85a2fa 100644
--- a/backend/internal/service/ratelimit_service.go
+++ b/backend/internal/service/ratelimit_service.go
@@ -186,9 +186,14 @@ func (s *RateLimitService) HandleUpstreamError(ctx context.Context, account *Acc
 	// otherwise a broad "rate limit" keyword rule can shorten a multi-hour
 	// cooldown to a local temporary pause.
 	if statusCode == http.StatusTooManyRequests && account.Platform == PlatformAnthropic {
+		// 7d_oi 是 Fable 模型专属的 7d 窗口：只标记模型级限流，账号对其他模型仍可调度。
+		fableLimited := s.persistAnthropicFableWindowLimit(ctx, account, headers)
 		if s.persistAnthropicExhaustedWindowLimit(ctx, account, headers) {
 			return false
 		}
+		if fableLimited {
+			return false
+		}
 	}
 
 	// 先尝试临时不可调度规则（401除外）
@@ -1140,11 +1145,25 @@ func selectAnthropicExhaustedWindow(headers http.Header, now time.Time) *anthrop
 }
 
 func isAnthropic5hRejected(headers http.Header) bool {
-	return strings.EqualFold(strings.TrimSpace(headers.Get("anthropic-ratelimit-unified-5h-status")), "rejected")
+	return isAnthropicWindowRejected(headers, "5h")
+}
+
+func isAnthropicWindowRejected(headers http.Header, window string) bool {
+	return strings.EqualFold(strings.TrimSpace(headers.Get("anthropic-ratelimit-unified-"+window+"-status")), "rejected")
 }
 
 func parseAnthropicWindowReset(headers http.Header, window string, now time.Time) (time.Time, bool) {
-	raw := strings.TrimSpace(headers.Get("anthropic-ratelimit-unified-" + window + "-reset"))
+	maxAge := 8 * 24 * time.Hour
+	if window == "5h" {
+		maxAge = 6 * time.Hour
+	}
+	return parseAnthropicResetTimestamp(headers.Get("anthropic-ratelimit-unified-"+window+"-reset"), now, maxAge)
+}
+
+// parseAnthropicResetTimestamp 解析 Anthropic reset 头的 Unix 时间戳（自动识别毫秒），
+// 并校验落在 (now, now+maxAge] 的合理区间内。
+func parseAnthropicResetTimestamp(raw string, now time.Time, maxAge time.Duration) (time.Time, bool) {
+	raw = strings.TrimSpace(raw)
 	if raw == "" {
 		return time.Time{}, false
 	}
@@ -1156,15 +1175,7 @@ func parseAnthropicWindowReset(headers http.Header, window string, now time.Time
 		ts = ts / 1000
 	}
 	resetAt := time.Unix(ts, 0)
-	if !resetAt.After(now) {
-		return time.Time{}, false
-	}
-
-	maxAge := 8 * 24 * time.Hour
-	if window == "5h" {
-		maxAge = 6 * time.Hour
-	}
-	if resetAt.After(now.Add(maxAge)) {
+	if !resetAt.After(now) || resetAt.After(now.Add(maxAge)) {
 		return time.Time{}, false
 	}
 	return resetAt, true
@@ -1218,6 +1229,76 @@ func (s *RateLimitService) persistAnthropicExhaustedWindowLimit(ctx context.Cont
 	return true
 }
 
+const anthropicFableWindowReason = "anthropic_7d_oi_window_exhausted"
+
+// selectAnthropicFableWindowLimit parses the Anthropic 7d_oi per-model window
+// headers (the Fable-only 7d window, e.g. anthropic-ratelimit-unified-7d_oi-*).
+// Unlike 5h/7d, exhaustion of this window only limits the Fable model family —
+// the account must stay schedulable for other models.
+//
+// The 7d_oi surpassed-threshold header carries a float ("1.0") rather than
+// "true", so exhaustion is detected via status=rejected or utilization >= 1.0.
+// When the 7d_oi reset header is missing, the aggregated
+// anthropic-ratelimit-unified-reset is used (it mirrors the binding claim's
+// reset when 7d_oi is the representative claim).
+func selectAnthropicFableWindowLimit(headers http.Header, now time.Time) *anthropicWindowLimit {
+	if !isAnthropicWindowRejected(headers, "7d_oi") && !isAnthropicWindowExceeded(headers, "7d_oi") {
+		return nil
+	}
+	resetAt, ok := parseAnthropicWindowReset(headers, "7d_oi", now)
+	if !ok {
+		resetAt, ok = parseAnthropicAggregateReset(headers, now)
+	}
+	if !ok {
+		return nil
+	}
+	return &anthropicWindowLimit{
+		window:  "7d_oi",
+		resetAt: resetAt,
+		reason:  anthropicFableWindowReason,
+	}
+}
+
+// parseAnthropicAggregateReset parses the aggregated
+// anthropic-ratelimit-unified-reset header with the same sanity checks as the
+// per-window variant (7d scale).
+func parseAnthropicAggregateReset(headers http.Header, now time.Time) (time.Time, bool) {
+	return parseAnthropicResetTimestamp(headers.Get("anthropic-ratelimit-unified-reset"), now, 8*24*time.Hour)
+}
+
+// persistAnthropicFableWindowLimit marks the Fable model family as rate limited
+// when the 7d_oi window is exhausted. Returns true when the 7d_oi window was the
+// (or a) trigger of this 429, so the caller must not fall through to logic that
+// would mark the whole account as rate limited.
+func (s *RateLimitService) persistAnthropicFableWindowLimit(ctx context.Context, account *Account, headers http.Header) bool {
+	if s == nil || s.accountRepo == nil || account == nil {
+		return false
+	}
+	now := time.Now()
+	limit := selectAnthropicFableWindowLimit(headers, now)
+	if limit == nil {
+		return false
+	}
+	// 429 响应头本身携带最新的窗口用量（7d_oi utilization=1.0）。限流期内
+	// Fable 请求不再调度到该账号，若不在此处采样，7d F 进度条会冻结在
+	// 限流前的旧值直到窗口重置。
+	s.samplePassiveUsageFromHeaders(ctx, account, headers)
+	if err := s.accountRepo.SetModelRateLimit(ctx, account.ID, anthropicFableRateLimitKey, limit.resetAt, limit.reason); err != nil {
+		slog.Warn("anthropic_fable_window_rate_limit_set_failed",
+			"account_id", account.ID,
+			"scope", anthropicFableRateLimitKey,
+			"reset_at", limit.resetAt,
+			"error", err)
+		return true
+	}
+	slog.Info("anthropic_fable_window_model_rate_limited",
+		"account_id", account.ID,
+		"scope", anthropicFableRateLimitKey,
+		"reset_at", limit.resetAt,
+		"reset_in", time.Until(limit.resetAt).Truncate(time.Second))
+	return true
+}
+
 // calculateAnthropic429ResetTime parses Anthropic's per-window rate-limit headers
 // to determine which window (5h or 7d) actually triggered the 429.
 //
@@ -1541,10 +1622,12 @@ func (s *RateLimitService) UpdateSessionWindow(ctx context.Context, account *Acc
 	// 窗口重置时清除旧的 utilization 和被动采样数据，避免残留上个窗口的数据
 	if windowEnd != nil && needInitWindow {
 		_ = s.accountRepo.UpdateExtra(ctx, account.ID, map[string]any{
-			"session_window_utilization":   nil,
-			"passive_usage_7d_utilization": nil,
-			"passive_usage_7d_reset":       nil,
-			"passive_usage_sampled_at":     nil,
+			"session_window_utilization":      nil,
+			"passive_usage_7d_utilization":    nil,
+			"passive_usage_7d_reset":          nil,
+			"passive_usage_7d_oi_utilization": nil,
+			"passive_usage_7d_oi_reset":       nil,
+			"passive_usage_sampled_at":        nil,
 		})
 	}
 
@@ -1552,8 +1635,21 @@ func (s *RateLimitService) UpdateSessionWindow(ctx context.Context, account *Acc
 		slog.Warn("session_window_update_failed", "account_id", account.ID, "error", err)
 	}
 
-	// 被动采样：从响应头收集 5h + 7d utilization，合并为一次 DB 写入
-	extraUpdates := make(map[string]any, 4)
+	// 被动采样：从响应头收集 5h + 7d + 7d_oi utilization，合并为一次 DB 写入
+	s.samplePassiveUsageFromHeaders(ctx, account, headers)
+
+	// 如果状态为allowed且之前有限流，说明窗口已重置，清除限流状态
+	if status == "allowed" && account.IsRateLimited() {
+		if err := s.ClearRateLimit(ctx, account.ID); err != nil {
+			slog.Warn("rate_limit_clear_failed", "account_id", account.ID, "error", err)
+		}
+	}
+}
+
+// samplePassiveUsageFromHeaders 从 Anthropic 响应头收集 5h/7d/7d_oi 的
+// utilization 与 reset 被动采样数据，合并为一次 Extra 写入。无数据时不写。
+func (s *RateLimitService) samplePassiveUsageFromHeaders(ctx context.Context, account *Account, headers http.Header) {
+	extraUpdates := make(map[string]any, 6)
 	// 5h utilization（0-1 小数），供 estimateSetupTokenUsage 使用
 	if utilStr := headers.Get("anthropic-ratelimit-unified-5h-utilization"); utilStr != "" {
 		if util, err := strconv.ParseFloat(utilStr, 64); err == nil {
@@ -1575,19 +1671,27 @@ func (s *RateLimitService) UpdateSessionWindow(ctx context.Context, account *Acc
 			extraUpdates["passive_usage_7d_reset"] = ts
 		}
 	}
+	// 7d_oi (Fable 专属 7d 窗口) utilization（0-1 小数）
+	if utilStr := headers.Get("anthropic-ratelimit-unified-7d_oi-utilization"); utilStr != "" {
+		if util, err := strconv.ParseFloat(utilStr, 64); err == nil {
+			extraUpdates["passive_usage_7d_oi_utilization"] = util
+		}
+	}
+	// 7d_oi reset timestamp
+	if resetStr := headers.Get("anthropic-ratelimit-unified-7d_oi-reset"); resetStr != "" {
+		if ts, err := strconv.ParseInt(resetStr, 10, 64); err == nil {
+			if ts > 1e11 {
+				ts = ts / 1000
+			}
+			extraUpdates["passive_usage_7d_oi_reset"] = ts
+		}
+	}
 	if len(extraUpdates) > 0 {
 		extraUpdates["passive_usage_sampled_at"] = time.Now().UTC().Format(time.RFC3339)
 		if err := s.accountRepo.UpdateExtra(ctx, account.ID, extraUpdates); err != nil {
 			slog.Warn("passive_usage_update_failed", "account_id", account.ID, "error", err)
 		}
 	}
-
-	// 如果状态为allowed且之前有限流，说明窗口已重置，清除限流状态
-	if status == "allowed" && account.IsRateLimited() {
-		if err := s.ClearRateLimit(ctx, account.ID); err != nil {
-			slog.Warn("rate_limit_clear_failed", "account_id", account.ID, "error", err)
-		}
-	}
 }
 
 // ClearRateLimit 清除账号的限流状态
diff --git a/backend/internal/service/ratelimit_service_anthropic_test.go b/backend/internal/service/ratelimit_service_anthropic_test.go
index eaeaf30e60..0e75b4f914 100644
--- a/backend/internal/service/ratelimit_service_anthropic_test.go
+++ b/backend/internal/service/ratelimit_service_anthropic_test.go
@@ -2,6 +2,7 @@ package service
 
 import (
 	"net/http"
+	"strconv"
 	"testing"
 	"time"
 )
@@ -181,6 +182,142 @@ func TestIsAnthropicWindowExceeded(t *testing.T) {
 	}
 }
 
+func TestSelectAnthropicFableWindowLimit_RejectedStatus(t *testing.T) {
+	now := time.Now()
+	reset := now.Add(80 * time.Hour).Truncate(time.Second)
+
+	headers := http.Header{}
+	headers.Set("anthropic-ratelimit-unified-7d_oi-status", "rejected")
+	headers.Set("anthropic-ratelimit-unified-7d_oi-utilization", "1.0")
+	headers.Set("anthropic-ratelimit-unified-7d_oi-surpassed-threshold", "1.0")
+	headers.Set("anthropic-ratelimit-unified-7d_oi-reset", strconv.FormatInt(reset.Unix(), 10))
+
+	limit := selectAnthropicFableWindowLimit(headers, now)
+	if limit == nil {
+		t.Fatal("expected non-nil limit")
+	}
+	if !limit.resetAt.Equal(reset) {
+		t.Errorf("expected resetAt=%v, got %v", reset, limit.resetAt)
+	}
+	if limit.reason != anthropicFableWindowReason {
+		t.Errorf("expected reason=%q, got %q", anthropicFableWindowReason, limit.reason)
+	}
+}
+
+func TestSelectAnthropicFableWindowLimit_UtilizationOnly(t *testing.T) {
+	// 无 status 头时，utilization >= 1.0 也应视为超限
+	now := time.Now()
+	reset := now.Add(3 * 24 * time.Hour).Truncate(time.Second)
+
+	headers := http.Header{}
+	headers.Set("anthropic-ratelimit-unified-7d_oi-utilization", "1.0")
+	headers.Set("anthropic-ratelimit-unified-7d_oi-reset", strconv.FormatInt(reset.Unix(), 10))
+
+	limit := selectAnthropicFableWindowLimit(headers, now)
+	if limit == nil {
+		t.Fatal("expected non-nil limit")
+	}
+	if !limit.resetAt.Equal(reset) {
+		t.Errorf("expected resetAt=%v, got %v", reset, limit.resetAt)
+	}
+}
+
+func TestSelectAnthropicFableWindowLimit_AllowedReturnsNil(t *testing.T) {
+	now := time.Now()
+	headers := http.Header{}
+	headers.Set("anthropic-ratelimit-unified-7d_oi-status", "allowed")
+	headers.Set("anthropic-ratelimit-unified-7d_oi-utilization", "0.56")
+	headers.Set("anthropic-ratelimit-unified-7d_oi-reset", strconv.FormatInt(now.Add(80*time.Hour).Unix(), 10))
+
+	if limit := selectAnthropicFableWindowLimit(headers, now); limit != nil {
+		t.Errorf("expected nil limit for allowed window, got %+v", limit)
+	}
+}
+
+func TestSelectAnthropicFableWindowLimit_NoHeadersReturnsNil(t *testing.T) {
+	if limit := selectAnthropicFableWindowLimit(http.Header{}, time.Now()); limit != nil {
+		t.Errorf("expected nil limit for empty headers, got %+v", limit)
+	}
+}
+
+func TestSelectAnthropicFableWindowLimit_FallsBackToAggregateReset(t *testing.T) {
+	// 7d_oi-reset 缺失时回退聚合 anthropic-ratelimit-unified-reset
+	now := time.Now()
+	reset := now.Add(80 * time.Hour).Truncate(time.Second)
+
+	headers := http.Header{}
+	headers.Set("anthropic-ratelimit-unified-7d_oi-status", "rejected")
+	headers.Set("anthropic-ratelimit-unified-reset", strconv.FormatInt(reset.Unix(), 10))
+
+	limit := selectAnthropicFableWindowLimit(headers, now)
+	if limit == nil {
+		t.Fatal("expected non-nil limit via aggregate reset fallback")
+	}
+	if !limit.resetAt.Equal(reset) {
+		t.Errorf("expected resetAt=%v, got %v", reset, limit.resetAt)
+	}
+}
+
+func TestSelectAnthropicFableWindowLimit_RejectedWithoutAnyResetReturnsNil(t *testing.T) {
+	headers := http.Header{}
+	headers.Set("anthropic-ratelimit-unified-7d_oi-status", "rejected")
+
+	if limit := selectAnthropicFableWindowLimit(headers, time.Now()); limit != nil {
+		t.Errorf("expected nil limit when no reset time available, got %+v", limit)
+	}
+}
+
+func TestParseAnthropicAggregateReset(t *testing.T) {
+	now := time.Now()
+	future := now.Add(80 * time.Hour).Truncate(time.Second)
+
+	tests := []struct {
+		name   string
+		value  string
+		want   time.Time
+		wantOK bool
+	}{
+		{"valid seconds", strconv.FormatInt(future.Unix(), 10), future, true},
+		{"valid milliseconds", strconv.FormatInt(future.UnixMilli(), 10), future, true},
+		{"empty", "", time.Time{}, false},
+		{"garbage", "abc", time.Time{}, false},
+		{"in the past", strconv.FormatInt(now.Add(-time.Hour).Unix(), 10), time.Time{}, false},
+		{"too far in the future", strconv.FormatInt(now.Add(30*24*time.Hour).Unix(), 10), time.Time{}, false},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			headers := http.Header{}
+			if tc.value != "" {
+				headers.Set("anthropic-ratelimit-unified-reset", tc.value)
+			}
+			got, ok := parseAnthropicAggregateReset(headers, now)
+			if ok != tc.wantOK {
+				t.Fatalf("expected ok=%v, got %v", tc.wantOK, ok)
+			}
+			if ok && !got.Equal(tc.want) {
+				t.Errorf("expected %v, got %v", tc.want, got)
+			}
+		})
+	}
+}
+
+func TestIsAnthropicWindowRejected(t *testing.T) {
+	headers := http.Header{}
+	headers.Set("anthropic-ratelimit-unified-7d_oi-status", "Rejected")
+	headers.Set("anthropic-ratelimit-unified-5h-status", "allowed")
+
+	if !isAnthropicWindowRejected(headers, "7d_oi") {
+		t.Error("expected 7d_oi to be rejected (case insensitive)")
+	}
+	if isAnthropicWindowRejected(headers, "5h") {
+		t.Error("expected 5h not rejected")
+	}
+	if isAnthropicWindowRejected(headers, "7d") {
+		t.Error("expected missing 7d status not rejected")
+	}
+}
+
 // assertAnthropicResult is a test helper that verifies the result is non-nil and
 // has the expected resetAt unix timestamp.
 func assertAnthropicResult(t *testing.T, result *anthropic429Result, wantUnix int64) {
diff --git a/backend/internal/service/ratelimit_service_anthropic_window_limit_test.go b/backend/internal/service/ratelimit_service_anthropic_window_limit_test.go
index 6168b250a6..a13e8fe333 100644
--- a/backend/internal/service/ratelimit_service_anthropic_window_limit_test.go
+++ b/backend/internal/service/ratelimit_service_anthropic_window_limit_test.go
@@ -14,9 +14,14 @@ import (
 
 type anthropicWindowLimitRepo struct {
 	mockAccountRepoForGemini
-	rateLimitCalls     int
-	tempUnschedCalls   int
-	lastRateLimitReset time.Time
+	rateLimitCalls          int
+	tempUnschedCalls        int
+	lastRateLimitReset      time.Time
+	modelRateLimitCalls     int
+	lastModelRateLimitScope string
+	lastModelRateLimitReset time.Time
+	sessionWindowCalls      int
+	lastExtraUpdates        map[string]any
 }
 
 func (r *anthropicWindowLimitRepo) SetRateLimited(_ context.Context, _ int64, resetAt time.Time) error {
@@ -30,6 +35,23 @@ func (r *anthropicWindowLimitRepo) SetTempUnschedulable(_ context.Context, _ int
 	return nil
 }
 
+func (r *anthropicWindowLimitRepo) SetModelRateLimit(_ context.Context, _ int64, scope string, resetAt time.Time, _ ...string) error {
+	r.modelRateLimitCalls++
+	r.lastModelRateLimitScope = scope
+	r.lastModelRateLimitReset = resetAt
+	return nil
+}
+
+func (r *anthropicWindowLimitRepo) UpdateSessionWindow(_ context.Context, _ int64, _, _ *time.Time, _ string) error {
+	r.sessionWindowCalls++
+	return nil
+}
+
+func (r *anthropicWindowLimitRepo) UpdateExtra(_ context.Context, _ int64, updates map[string]any) error {
+	r.lastExtraUpdates = updates
+	return nil
+}
+
 func TestHandleUpstreamError_AnthropicWindowLimitPreemptsTempUnschedRule(t *testing.T) {
 	resetAt := time.Now().Add(3 * time.Hour).Truncate(time.Second)
 	headers := http.Header{}
@@ -66,3 +88,142 @@ func TestHandleUpstreamError_AnthropicWindowLimitPreemptsTempUnschedRule(t *test
 	require.Equal(t, 1, repo.rateLimitCalls)
 	require.Equal(t, resetAt, repo.lastRateLimitReset)
 }
+
+// fable429Headers 构造 7d_oi（Fable 专属 7d 窗口）触发 429 的完整响应头，
+// 数值取自真实抓包（5h/7d 均 allowed，仅 7d_oi rejected）。
+func fable429Headers(reset5h, resetOI time.Time) http.Header {
+	headers := http.Header{}
+	headers.Set("anthropic-ratelimit-unified-5h-reset", strconv.FormatInt(reset5h.Unix(), 10))
+	headers.Set("anthropic-ratelimit-unified-5h-status", "allowed")
+	headers.Set("anthropic-ratelimit-unified-5h-utilization", "0.41")
+	headers.Set("anthropic-ratelimit-unified-7d-reset", strconv.FormatInt(resetOI.Unix(), 10))
+	headers.Set("anthropic-ratelimit-unified-7d-status", "allowed")
+	headers.Set("anthropic-ratelimit-unified-7d-utilization", "0.56")
+	headers.Set("anthropic-ratelimit-unified-7d_oi-reset", strconv.FormatInt(resetOI.Unix(), 10))
+	headers.Set("anthropic-ratelimit-unified-7d_oi-status", "rejected")
+	headers.Set("anthropic-ratelimit-unified-7d_oi-surpassed-threshold", "1.0")
+	headers.Set("anthropic-ratelimit-unified-7d_oi-utilization", "1.0")
+	headers.Set("anthropic-ratelimit-unified-fallback-percentage", "0.5")
+	headers.Set("anthropic-ratelimit-unified-overage-disabled-reason", "org_level_disabled")
+	headers.Set("anthropic-ratelimit-unified-overage-status", "rejected")
+	headers.Set("anthropic-ratelimit-unified-representative-claim", "seven_day_overage_included")
+	headers.Set("anthropic-ratelimit-unified-reset", strconv.FormatInt(resetOI.Unix(), 10))
+	headers.Set("anthropic-ratelimit-unified-status", "rejected")
+	return headers
+}
+
+func TestHandleUpstreamError_Anthropic7dOiOnlyMarksModelRateLimit(t *testing.T) {
+	now := time.Now()
+	reset5h := now.Add(2 * time.Hour).Truncate(time.Second)
+	resetOI := now.Add(80 * time.Hour).Truncate(time.Second)
+	headers := fable429Headers(reset5h, resetOI)
+
+	repo := &anthropicWindowLimitRepo{}
+	svc := NewRateLimitService(repo, nil, nil, nil, nil)
+	account := &Account{
+		ID:       42,
+		Type:     AccountTypeOAuth,
+		Platform: PlatformAnthropic,
+		Credentials: map[string]any{
+			"temp_unschedulable_enabled": true,
+			"temp_unschedulable_rules": []any{
+				map[string]any{
+					"error_code":       float64(http.StatusTooManyRequests),
+					"keywords":         []any{"rate limit"},
+					"duration_minutes": float64(10),
+				},
+			},
+		},
+	}
+
+	shouldDisable := svc.HandleUpstreamError(
+		context.Background(),
+		account,
+		http.StatusTooManyRequests,
+		headers,
+		[]byte(`{"type":"error","error":{"type":"rate_limit_error","message":"This request would exceed your account's rate limit. Please try again later."}}`),
+		"claude-fable-5",
+	)
+
+	require.False(t, shouldDisable)
+	require.Zero(t, repo.rateLimitCalls, "7d_oi (Fable-only) window must not mark the whole account rate limited")
+	require.Zero(t, repo.tempUnschedCalls, "7d_oi window must not trigger local temp-unsched rules")
+	require.Zero(t, repo.sessionWindowCalls, "7d_oi window must not rewrite the 5h session window as rejected")
+	require.Equal(t, 1, repo.modelRateLimitCalls)
+	require.Equal(t, anthropicFableRateLimitKey, repo.lastModelRateLimitScope)
+	require.Equal(t, resetOI, repo.lastModelRateLimitReset)
+
+	// 429 响应头也要被动采样，避免 7d F 进度条在限流期内冻结在旧值
+	require.NotNil(t, repo.lastExtraUpdates)
+	require.Equal(t, 1.0, repo.lastExtraUpdates["passive_usage_7d_oi_utilization"])
+	require.Equal(t, resetOI.Unix(), repo.lastExtraUpdates["passive_usage_7d_oi_reset"])
+	require.Equal(t, 0.41, repo.lastExtraUpdates["session_window_utilization"])
+}
+
+func TestHandleUpstreamError_Anthropic5hWindowStillWinsOver7dOi(t *testing.T) {
+	// 5h 窗口 rejected 时必须仍按账号级限流处理（用 5h reset），同时记录 Fable 模型限流。
+	now := time.Now()
+	reset5h := now.Add(2 * time.Hour).Truncate(time.Second)
+	resetOI := now.Add(80 * time.Hour).Truncate(time.Second)
+	headers := fable429Headers(reset5h, resetOI)
+	headers.Set("anthropic-ratelimit-unified-5h-status", "rejected")
+	headers.Set("anthropic-ratelimit-unified-5h-utilization", "1.0")
+
+	repo := &anthropicWindowLimitRepo{}
+	svc := NewRateLimitService(repo, nil, nil, nil, nil)
+	account := &Account{ID: 42, Type: AccountTypeOAuth, Platform: PlatformAnthropic}
+
+	svc.HandleUpstreamError(context.Background(), account, http.StatusTooManyRequests, headers, nil, "claude-fable-5")
+
+	require.Equal(t, 1, repo.rateLimitCalls, "exhausted 5h window must still rate limit the account")
+	require.Equal(t, reset5h, repo.lastRateLimitReset)
+	require.Equal(t, 1, repo.modelRateLimitCalls)
+	require.Equal(t, anthropicFableRateLimitKey, repo.lastModelRateLimitScope)
+}
+
+func TestHandleUpstreamError_AnthropicAccountWindowStillWinsOver7dOi(t *testing.T) {
+	// 7d 窗口真超限时必须仍按账号级限流处理，同时记录 Fable 模型限流。
+	now := time.Now()
+	reset5h := now.Add(2 * time.Hour).Truncate(time.Second)
+	resetOI := now.Add(80 * time.Hour).Truncate(time.Second)
+	headers := fable429Headers(reset5h, resetOI)
+	headers.Set("anthropic-ratelimit-unified-7d-status", "rejected")
+	headers.Set("anthropic-ratelimit-unified-7d-utilization", "1.02")
+
+	repo := &anthropicWindowLimitRepo{}
+	svc := NewRateLimitService(repo, nil, nil, nil, nil)
+	account := &Account{ID: 42, Type: AccountTypeOAuth, Platform: PlatformAnthropic}
+
+	svc.HandleUpstreamError(context.Background(), account, http.StatusTooManyRequests, headers, nil, "claude-fable-5")
+
+	require.Equal(t, 1, repo.rateLimitCalls, "exhausted 7d window must still rate limit the account")
+	require.Equal(t, resetOI, repo.lastRateLimitReset)
+	require.Equal(t, 1, repo.modelRateLimitCalls, "Fable model rate limit should also be recorded")
+	require.Equal(t, anthropicFableRateLimitKey, repo.lastModelRateLimitScope)
+}
+
+func TestHandleUpstreamError_Anthropic429Without7dOiKeepsLegacyBehavior(t *testing.T) {
+	// 无 7d_oi 头、5h/7d 均未超限的 429：保持旧行为（按较早 reset 标记账号限流）。
+	now := time.Now()
+	reset5h := now.Add(2 * time.Hour).Truncate(time.Second)
+	reset7d := now.Add(80 * time.Hour).Truncate(time.Second)
+
+	headers := http.Header{}
+	headers.Set("anthropic-ratelimit-unified-5h-reset", strconv.FormatInt(reset5h.Unix(), 10))
+	headers.Set("anthropic-ratelimit-unified-5h-status", "allowed")
+	headers.Set("anthropic-ratelimit-unified-5h-utilization", "0.41")
+	headers.Set("anthropic-ratelimit-unified-7d-reset", strconv.FormatInt(reset7d.Unix(), 10))
+	headers.Set("anthropic-ratelimit-unified-7d-status", "allowed")
+	headers.Set("anthropic-ratelimit-unified-7d-utilization", "0.56")
+
+	repo := &anthropicWindowLimitRepo{}
+	svc := NewRateLimitService(repo, nil, nil, nil, nil)
+	account := &Account{ID: 42, Type: AccountTypeOAuth, Platform: PlatformAnthropic}
+
+	svc.HandleUpstreamError(context.Background(), account, http.StatusTooManyRequests, headers, nil, "claude-fable-5")
+
+	require.Zero(t, repo.modelRateLimitCalls, "no 7d_oi signal → no model rate limit")
+	require.Equal(t, 1, repo.rateLimitCalls)
+	require.Equal(t, reset5h, repo.lastRateLimitReset, "legacy path picks the sooner reset")
+	require.Equal(t, 1, repo.sessionWindowCalls)
+}
diff --git a/backend/internal/service/ratelimit_session_window_test.go b/backend/internal/service/ratelimit_session_window_test.go
index 9337ae5f8f..cb19227e54 100644
--- a/backend/internal/service/ratelimit_session_window_test.go
+++ b/backend/internal/service/ratelimit_session_window_test.go
@@ -367,6 +367,59 @@ func TestUpdateSessionWindow_NoClearUtilizationOnCorrection(t *testing.T) {
 	}
 }
 
+func TestUpdateSessionWindow_SamplesFable7dOiHeaders(t *testing.T) {
+	// 被动采样应收集 7d_oi（Fable 专属 7d 窗口）的 utilization 和 reset。
+	existingEnd := time.Now().Add(3 * time.Hour)
+	resetOIUnix := time.Now().Add(80 * time.Hour).Unix()
+
+	repo := &sessionWindowMockRepo{}
+	svc := newRateLimitServiceForTest(repo)
+
+	account := &Account{ID: 90, SessionWindowEnd: &existingEnd} // needInitWindow=false
+	headers := http.Header{}
+	headers.Set("anthropic-ratelimit-unified-5h-status", "allowed")
+	headers.Set("anthropic-ratelimit-unified-7d_oi-utilization", "0.87")
+	headers.Set("anthropic-ratelimit-unified-7d_oi-reset", fmt.Sprintf("%d", resetOIUnix))
+
+	svc.UpdateSessionWindow(context.Background(), account, headers)
+
+	if len(repo.updateExtraCalls) != 1 {
+		t.Fatalf("expected 1 UpdateExtra call, got %d", len(repo.updateExtraCalls))
+	}
+	updates := repo.updateExtraCalls[0].Updates
+	if val, ok := updates["passive_usage_7d_oi_utilization"].(float64); !ok || val != 0.87 {
+		t.Errorf("expected passive_usage_7d_oi_utilization=0.87, got %v", updates["passive_usage_7d_oi_utilization"])
+	}
+	if val, ok := updates["passive_usage_7d_oi_reset"].(int64); !ok || val != resetOIUnix {
+		t.Errorf("expected passive_usage_7d_oi_reset=%d, got %v", resetOIUnix, updates["passive_usage_7d_oi_reset"])
+	}
+}
+
+func TestUpdateSessionWindow_ClearsFable7dOiOnWindowReset(t *testing.T) {
+	// 5h 窗口重置时应连同清除 7d_oi 被动采样数据，与 7d 行为一致。
+	resetUnix := time.Now().Add(3 * time.Hour).Unix()
+
+	repo := &sessionWindowMockRepo{}
+	svc := newRateLimitServiceForTest(repo)
+
+	account := &Account{ID: 91} // no existing window → needInitWindow=true
+	headers := http.Header{}
+	headers.Set("anthropic-ratelimit-unified-5h-status", "allowed")
+	headers.Set("anthropic-ratelimit-unified-5h-reset", fmt.Sprintf("%d", resetUnix))
+
+	svc.UpdateSessionWindow(context.Background(), account, headers)
+
+	if len(repo.updateExtraCalls) != 1 {
+		t.Fatalf("expected 1 UpdateExtra (clear) call, got %d", len(repo.updateExtraCalls))
+	}
+	clearUpdates := repo.updateExtraCalls[0].Updates
+	for _, key := range []string{"passive_usage_7d_oi_utilization", "passive_usage_7d_oi_reset"} {
+		if val, present := clearUpdates[key]; !present || val != nil {
+			t.Errorf("expected %s cleared to nil on window reset, got present=%v val=%v", key, present, val)
+		}
+	}
+}
+
 func TestUpdateSessionWindow_NoStatusHeader(t *testing.T) {
 	// Should return immediately if no status header.
 	repo := &sessionWindowMockRepo{}
diff --git a/frontend/src/components/account/AccountUsageCell.vue b/frontend/src/components/account/AccountUsageCell.vue
index 67d7b165dc..a4cb70a9c3 100644
--- a/frontend/src/components/account/AccountUsageCell.vue
+++ b/frontend/src/components/account/AccountUsageCell.vue
@@ -68,6 +68,15 @@
           color="purple"
         />
 
+        <!-- 7d Fable Window (7d_oi) -->
+        <UsageProgressBar
+          v-if="usageInfo.seven_day_fable"
+          label="7d F"
+          :utilization="usageInfo.seven_day_fable.utilization"
+          :resets-at="usageInfo.seven_day_fable.resets_at"
+          color="amber"
+        />
+
         <!-- Passive sampling label + active query button -->
         <div class="flex items-center gap-1.5 mt-0.5">
           <span
diff --git a/frontend/src/components/account/__tests__/AccountUsageCell.spec.ts b/frontend/src/components/account/__tests__/AccountUsageCell.spec.ts
index fa4104f6dc..55efc197a7 100644
--- a/frontend/src/components/account/__tests__/AccountUsageCell.spec.ts
+++ b/frontend/src/components/account/__tests__/AccountUsageCell.spec.ts
@@ -655,4 +655,102 @@ describe('AccountUsageCell', () => {
 		expect(wrapper.text()).toContain('A $0.00')
 		expect(wrapper.text()).toContain('U $0.00')
   })
+
+  it('Anthropic OAuth 会渲染 7d F (Fable) 进度条，且 7d S 逻辑保留', async () => {
+    getUsage.mockResolvedValue({
+      source: 'passive',
+      five_hour: {
+        utilization: 41,
+        resets_at: '2026-07-03T10:00:00Z',
+        remaining_seconds: 3600
+      },
+      seven_day: {
+        utilization: 56,
+        resets_at: '2026-07-06T22:00:00Z',
+        remaining_seconds: 300000
+      },
+      seven_day_sonnet: {
+        utilization: 30,
+        resets_at: '2026-07-06T22:00:00Z',
+        remaining_seconds: 300000
+      },
+      seven_day_fable: {
+        utilization: 100,
+        resets_at: '2026-07-06T22:00:00Z',
+        remaining_seconds: 300000
+      }
+    })
+
+    const wrapper = mount(AccountUsageCell, {
+      props: {
+        account: makeAccount({
+          id: 3001,
+          platform: 'anthropic',
+          type: 'oauth',
+          extra: {}
+        })
+      },
+      global: {
+        stubs: {
+          UsageProgressBar: {
+            props: ['label', 'utilization', 'resetsAt', 'color'],
+            template: '<div class="usage-bar">{{ label }}|{{ utilization }}</div>'
+          },
+          AccountQuotaInfo: true,
+          GrokQuotaProbeCell: true
+        }
+      }
+    })
+
+    await flushPromises()
+
+    expect(wrapper.text()).toContain('5h|41')
+    expect(wrapper.text()).toContain('7d|56')
+    expect(wrapper.text()).toContain('7d S|30')
+    expect(wrapper.text()).toContain('7d F|100')
+  })
+
+  it('Anthropic OAuth 无 Fable 数据时不渲染 7d F 进度条', async () => {
+    getUsage.mockResolvedValue({
+      source: 'passive',
+      five_hour: {
+        utilization: 41,
+        resets_at: '2026-07-03T10:00:00Z',
+        remaining_seconds: 3600
+      },
+      seven_day: {
+        utilization: 56,
+        resets_at: '2026-07-06T22:00:00Z',
+        remaining_seconds: 300000
+      }
+    })
+
+    const wrapper = mount(AccountUsageCell, {
+      props: {
+        account: makeAccount({
+          id: 3002,
+          platform: 'anthropic',
+          type: 'oauth',
+          extra: {}
+        })
+      },
+      global: {
+        stubs: {
+          UsageProgressBar: {
+            props: ['label', 'utilization', 'resetsAt', 'color'],
+            template: '<div class="usage-bar">{{ label }}|{{ utilization }}</div>'
+          },
+          AccountQuotaInfo: true,
+          GrokQuotaProbeCell: true
+        }
+      }
+    })
+
+    await flushPromises()
+
+    expect(wrapper.text()).toContain('5h|41')
+    expect(wrapper.text()).toContain('7d|56')
+    expect(wrapper.text()).not.toContain('7d S')
+    expect(wrapper.text()).not.toContain('7d F')
+  })
 })
diff --git a/frontend/src/types/index.ts b/frontend/src/types/index.ts
index 3636791ebd..f304a85341 100644
--- a/frontend/src/types/index.ts
+++ b/frontend/src/types/index.ts
@@ -984,6 +984,7 @@ export interface AccountUsageInfo {
   five_hour: UsageProgress | null
   seven_day: UsageProgress | null
   seven_day_sonnet: UsageProgress | null
+  seven_day_fable?: UsageProgress | null
   gemini_shared_daily?: UsageProgress | null
   gemini_pro_daily?: UsageProgress | null
   gemini_flash_daily?: UsageProgress | null

From be297b90ce30d4a8fda03d00a36c937b2161980e Mon Sep 17 00:00:00 2001
From: shaw <dev@sub2api.org>
Date: Sat, 4 Jul 2026 09:49:37 +0800
Subject: [PATCH 13/28] refactor(frontend): merge Codex image bridge and tool
 policy into one four-state control

Replace the two adjacent account-level controls (3-state injection bridge
+ 2-state explicit tool policy) in EditAccountModal with a single
four-state selector: follow channel / force inject / no injection /
block all. The four states map onto the existing extra keys
(codex_image_generation_bridge, codex_image_generation_explicit_tool_policy)
with no backend change; strip already disables the bridge upstream, so
the effective decision space is exactly these four states.

- load: strip takes precedence and reads as "block all"; legacy
  bridge_enabled key still recognized and cleaned up on save
- save: block writes policy=strip and drops the bridge override;
  inherit clears both keys
- semantic accent colors per state (sky/emerald/amber/rose) with dark
  mode support; state badge follows the same palette
- replace account-scope i18n keys in zh/en (channel-scope keys untouched)
- rewrite specs to cover all four states (25/25 passing)
---
 .../components/account/EditAccountModal.vue   | 183 +++++++-----------
 .../__tests__/EditAccountModal.spec.ts        |  35 +++-
 frontend/src/i18n/locales/en.ts               |  34 ++--
 frontend/src/i18n/locales/zh.ts               |  33 ++--
 4 files changed, 126 insertions(+), 159 deletions(-)

diff --git a/frontend/src/components/account/EditAccountModal.vue b/frontend/src/components/account/EditAccountModal.vue
index a596183207..3670f233e3 100644
--- a/frontend/src/components/account/EditAccountModal.vue
+++ b/frontend/src/components/account/EditAccountModal.vue
@@ -1370,7 +1370,7 @@
         </div>
       </div>
 
-      <!-- OpenAI Codex 图片生成桥接账号级覆盖 -->
+      <!-- OpenAI Codex 图片工具统一策略（自动注入 + 客户端显式携带） -->
       <div
         v-if="account?.platform === 'openai' && (account?.type === 'oauth' || account?.type === 'setup-token' || account?.type === 'apikey')"
         class="border-t border-gray-200 pt-4 dark:border-dark-600"
@@ -1382,91 +1382,39 @@
             </div>
             <div class="min-w-0 flex-1">
               <div class="flex flex-wrap items-center gap-2">
-                <label class="input-label mb-0">{{ t('admin.accounts.openai.codexImageGenerationBridge') }}</label>
+                <label class="input-label mb-0">{{ t('admin.accounts.openai.codexImageTool') }}</label>
                 <span
                   class="rounded-full px-2 py-0.5 text-[11px] font-medium"
-                  :class="codexImageGenerationBridgeBadgeClass"
+                  :class="codexImageToolBadgeClass"
                 >
-                  {{ codexImageGenerationBridgeBadgeLabel }}
+                  {{ codexImageToolBadgeLabel }}
                 </span>
               </div>
               <p class="mt-1 text-xs leading-5 text-slate-600 dark:text-slate-300">
-                {{ t('admin.accounts.openai.codexImageGenerationBridgeDesc') }}
+                {{ t('admin.accounts.openai.codexImageToolDesc') }}
               </p>
             </div>
           </div>
           <div class="border-t border-sky-100 bg-white/70 p-2 dark:border-sky-900/50 dark:bg-dark-800/70">
-            <div class="grid grid-cols-1 gap-2 sm:grid-cols-3">
+            <div class="grid grid-cols-1 gap-2 sm:grid-cols-2">
               <button
-                v-for="option in codexImageGenerationBridgeOptions"
+                v-for="option in codexImageToolOptions"
                 :key="option.value"
                 type="button"
-                :data-testid="`codex-image-bridge-${option.value}`"
-                @click="codexImageGenerationBridgeMode = option.value"
+                :data-testid="`codex-image-tool-${option.value}`"
+                @click="codexImageToolMode = option.value"
                 :class="[
-                  'group flex min-h-[68px] items-start gap-2 rounded-md border px-3 py-2 text-left transition-all',
-                  codexImageGenerationBridgeMode === option.value
-                    ? 'border-sky-300 bg-sky-50 text-sky-900 shadow-sm ring-1 ring-sky-200 dark:border-sky-700 dark:bg-sky-900/25 dark:text-sky-100 dark:ring-sky-800'
+                  'group flex min-h-[62px] items-start gap-2 rounded-md border px-3 py-2 text-left transition-all',
+                  codexImageToolMode === option.value
+                    ? option.selectedCardClass
                     : 'border-transparent bg-transparent text-slate-600 hover:border-gray-200 hover:bg-gray-50 dark:text-slate-300 dark:hover:border-dark-500 dark:hover:bg-dark-700'
                 ]"
               >
                 <span
                   :class="[
                     'mt-0.5 flex h-5 w-5 shrink-0 items-center justify-center rounded-full border transition-colors',
-                    codexImageGenerationBridgeMode === option.value
-                      ? 'border-sky-500 bg-sky-500 text-white'
-                      : 'border-gray-300 text-transparent group-hover:border-gray-400 dark:border-dark-500'
-                  ]"
-                >
-                  <Icon name="check" size="xs" :stroke-width="2" />
-                </span>
-                <span class="min-w-0">
-                  <span class="block text-sm font-medium">{{ option.label }}</span>
-                  <span class="mt-0.5 block text-xs leading-4 text-slate-500 dark:text-slate-400">{{ option.description }}</span>
-                </span>
-              </button>
-            </div>
-          </div>
-        </div>
-      </div>
-
-      <!-- OpenAI Codex 显式 image_generation tool 策略 -->
-      <div
-        v-if="account?.platform === 'openai' && (account?.type === 'oauth' || account?.type === 'setup-token' || account?.type === 'apikey')"
-        class="border-t border-gray-200 pt-4 dark:border-dark-600"
-      >
-        <div class="overflow-hidden rounded-lg border border-gray-200 bg-white shadow-sm dark:border-dark-600 dark:bg-dark-800">
-          <div class="flex items-start gap-3 px-4 py-3">
-            <div class="mt-0.5 flex h-9 w-9 shrink-0 items-center justify-center rounded-md bg-gray-50 text-slate-600 ring-1 ring-gray-200 dark:bg-dark-700 dark:text-slate-300 dark:ring-dark-600">
-              <Icon name="filter" size="sm" />
-            </div>
-            <div class="min-w-0 flex-1">
-              <label class="input-label mb-0">{{ t('admin.accounts.openai.codexImageToolPolicy') }}</label>
-              <p class="mt-1 text-xs leading-5 text-slate-600 dark:text-slate-300">
-                {{ t('admin.accounts.openai.codexImageToolPolicyDesc') }}
-              </p>
-            </div>
-          </div>
-          <div class="border-t border-gray-200 bg-gray-50/70 p-2 dark:border-dark-600 dark:bg-dark-700/50">
-            <div class="grid grid-cols-1 gap-2 sm:grid-cols-2">
-              <button
-                v-for="option in codexImageToolPolicyOptions"
-                :key="option.value"
-                type="button"
-                :data-testid="`codex-image-tool-policy-${option.value}`"
-                @click="codexImageToolPolicyMode = option.value"
-                :class="[
-                  'group flex min-h-[62px] items-start gap-2 rounded-md border px-3 py-2 text-left transition-all',
-                  codexImageToolPolicyMode === option.value
-                    ? 'border-slate-300 bg-white text-slate-900 shadow-sm ring-1 ring-slate-200 dark:border-slate-600 dark:bg-dark-800 dark:text-slate-100 dark:ring-dark-500'
-                    : 'border-transparent bg-transparent text-slate-600 hover:border-gray-200 hover:bg-white dark:text-slate-300 dark:hover:border-dark-500 dark:hover:bg-dark-800'
-                ]"
-              >
-                <span
-                  :class="[
-                    'mt-0.5 flex h-5 w-5 shrink-0 items-center justify-center rounded-full border transition-colors',
-                    codexImageToolPolicyMode === option.value
-                      ? 'border-slate-600 bg-slate-700 text-white dark:border-slate-400 dark:bg-slate-500'
+                    codexImageToolMode === option.value
+                      ? option.selectedDotClass
                       : 'border-gray-300 text-transparent group-hover:border-gray-400 dark:border-dark-500'
                   ]"
                 >
@@ -2679,10 +2627,8 @@ const openaiOAuthResponsesWebSocketV2Mode = ref<OpenAIWSMode>(OPENAI_WS_MODE_OFF
 const openaiAPIKeyResponsesWebSocketV2Mode = ref<OpenAIWSMode>(OPENAI_WS_MODE_OFF)
 const codexCLIOnlyEnabled = ref(false)
 const codexCLIOnlyAppServerEnabled = ref(false)
-type CodexImageGenerationBridgeMode = 'inherit' | 'enabled' | 'disabled'
-const codexImageGenerationBridgeMode = ref<CodexImageGenerationBridgeMode>('inherit')
-type CodexImageToolPolicyMode = 'allow' | 'strip'
-const codexImageToolPolicyMode = ref<CodexImageToolPolicyMode>('allow')
+type CodexImageToolMode = 'inherit' | 'enabled' | 'disabled' | 'block'
+const codexImageToolMode = ref<CodexImageToolMode>('inherit')
 type AnthropicAPIKeyAuthScheme = 'x_api_key' | 'authorization_bearer'
 const anthropicPassthroughEnabled = ref(false)
 const anthropicAPIKeyAuthScheme = ref<AnthropicAPIKeyAuthScheme>('x_api_key')
@@ -2736,58 +2682,61 @@ const openaiResponsesWebSocketV2Mode = computed({
 const openAIWSModeConcurrencyHintKey = computed(() =>
   resolveOpenAIWSModeConcurrencyHintKey(openaiResponsesWebSocketV2Mode.value)
 )
-const codexImageGenerationBridgeOptions = computed<Array<{
-  value: CodexImageGenerationBridgeMode
+const codexImageToolOptions = computed<Array<{
+  value: CodexImageToolMode
   label: string
   description: string
+  selectedCardClass: string
+  selectedDotClass: string
 }>>(() => [
   {
     value: 'inherit',
-    label: t('admin.accounts.openai.codexImageGenerationBridgeInherit'),
-    description: t('admin.accounts.openai.codexImageGenerationBridgeInheritDesc')
+    label: t('admin.accounts.openai.codexImageToolInherit'),
+    description: t('admin.accounts.openai.codexImageToolInheritDesc'),
+    selectedCardClass: 'border-sky-300 bg-sky-50 text-sky-900 shadow-sm ring-1 ring-sky-200 dark:border-sky-700 dark:bg-sky-900/25 dark:text-sky-100 dark:ring-sky-800',
+    selectedDotClass: 'border-sky-500 bg-sky-500 text-white'
   },
   {
     value: 'enabled',
-    label: t('admin.accounts.openai.codexImageGenerationBridgeEnabled'),
-    description: t('admin.accounts.openai.codexImageGenerationBridgeEnabledDesc')
+    label: t('admin.accounts.openai.codexImageToolEnabled'),
+    description: t('admin.accounts.openai.codexImageToolEnabledDesc'),
+    selectedCardClass: 'border-emerald-300 bg-emerald-50 text-emerald-900 shadow-sm ring-1 ring-emerald-200 dark:border-emerald-700 dark:bg-emerald-900/25 dark:text-emerald-100 dark:ring-emerald-800',
+    selectedDotClass: 'border-emerald-500 bg-emerald-500 text-white'
   },
   {
     value: 'disabled',
-    label: t('admin.accounts.openai.codexImageGenerationBridgeDisabled'),
-    description: t('admin.accounts.openai.codexImageGenerationBridgeDisabledDesc')
-  }
-])
-const codexImageToolPolicyOptions = computed<Array<{
-  value: CodexImageToolPolicyMode
-  label: string
-  description: string
-}>>(() => [
-  {
-    value: 'allow',
-    label: t('admin.accounts.openai.codexImageToolPolicyAllow'),
-    description: t('admin.accounts.openai.codexImageToolPolicyAllowDesc')
+    label: t('admin.accounts.openai.codexImageToolDisabled'),
+    description: t('admin.accounts.openai.codexImageToolDisabledDesc'),
+    selectedCardClass: 'border-amber-300 bg-amber-50 text-amber-900 shadow-sm ring-1 ring-amber-200 dark:border-amber-700 dark:bg-amber-900/25 dark:text-amber-100 dark:ring-amber-800',
+    selectedDotClass: 'border-amber-500 bg-amber-500 text-white'
   },
   {
-    value: 'strip',
-    label: t('admin.accounts.openai.codexImageToolPolicyStrip'),
-    description: t('admin.accounts.openai.codexImageToolPolicyStripDesc')
+    value: 'block',
+    label: t('admin.accounts.openai.codexImageToolBlock'),
+    description: t('admin.accounts.openai.codexImageToolBlockDesc'),
+    selectedCardClass: 'border-rose-300 bg-rose-50 text-rose-900 shadow-sm ring-1 ring-rose-200 dark:border-rose-700 dark:bg-rose-900/25 dark:text-rose-100 dark:ring-rose-800',
+    selectedDotClass: 'border-rose-500 bg-rose-500 text-white'
   }
 ])
-const codexImageGenerationBridgeBadgeLabel = computed(() => {
-  switch (codexImageGenerationBridgeMode.value) {
+const codexImageToolBadgeLabel = computed(() => {
+  switch (codexImageToolMode.value) {
     case 'enabled':
-      return t('admin.accounts.openai.codexImageGenerationBridgeBadgeEnabled')
+      return t('admin.accounts.openai.codexImageToolBadgeEnabled')
     case 'disabled':
-      return t('admin.accounts.openai.codexImageGenerationBridgeBadgeDisabled')
+      return t('admin.accounts.openai.codexImageToolBadgeDisabled')
+    case 'block':
+      return t('admin.accounts.openai.codexImageToolBadgeBlock')
     default:
-      return t('admin.accounts.openai.codexImageGenerationBridgeBadgeInherit')
+      return t('admin.accounts.openai.codexImageToolBadgeInherit')
   }
 })
-const codexImageGenerationBridgeBadgeClass = computed(() => {
-  switch (codexImageGenerationBridgeMode.value) {
+const codexImageToolBadgeClass = computed(() => {
+  switch (codexImageToolMode.value) {
     case 'enabled':
       return 'bg-emerald-100 text-emerald-700 dark:bg-emerald-900/40 dark:text-emerald-300'
     case 'disabled':
+      return 'bg-amber-100 text-amber-700 dark:bg-amber-900/40 dark:text-amber-300'
+    case 'block':
       return 'bg-rose-100 text-rose-700 dark:bg-rose-900/40 dark:text-rose-300'
     default:
       return 'bg-slate-100 text-slate-600 dark:bg-dark-600 dark:text-slate-300'
@@ -3105,8 +3054,7 @@ const syncFormFromAccount = (newAccount: Account | null) => {
   openaiAPIKeyResponsesWebSocketV2Mode.value = OPENAI_WS_MODE_OFF
   codexCLIOnlyEnabled.value = false
   codexCLIOnlyAppServerEnabled.value = false
-  codexImageGenerationBridgeMode.value = 'inherit'
-  codexImageToolPolicyMode.value = 'allow'
+  codexImageToolMode.value = 'inherit'
   anthropicPassthroughEnabled.value = false
   anthropicAPIKeyAuthScheme.value = 'x_api_key'
   webSearchEmulationMode.value = 'default'
@@ -3125,14 +3073,13 @@ const syncFormFromAccount = (newAccount: Account | null) => {
     const codexImageGenerationBridgeValue = typeof extra?.codex_image_generation_bridge === 'boolean'
       ? extra.codex_image_generation_bridge
       : extra?.codex_image_generation_bridge_enabled
-    if (codexImageGenerationBridgeValue === true) {
-      codexImageGenerationBridgeMode.value = 'enabled'
+    if (extra?.codex_image_generation_explicit_tool_policy === 'strip') {
+      codexImageToolMode.value = 'block'
+    } else if (codexImageGenerationBridgeValue === true) {
+      codexImageToolMode.value = 'enabled'
     } else if (codexImageGenerationBridgeValue === false) {
-      codexImageGenerationBridgeMode.value = 'disabled'
+      codexImageToolMode.value = 'disabled'
     }
-    codexImageToolPolicyMode.value = extra?.codex_image_generation_explicit_tool_policy === 'strip'
-      ? 'strip'
-      : 'allow'
     openaiOAuthResponsesWebSocketV2Mode.value = resolveOpenAIWSModeFromExtra(extra, {
       modeKey: 'openai_oauth_responses_websockets_v2_mode',
       enabledKey: 'openai_oauth_responses_websockets_v2_enabled',
@@ -4277,15 +4224,19 @@ const handleSubmit = async () => {
 		}
 
 		delete newExtra.codex_image_generation_bridge_enabled
-      if (codexImageGenerationBridgeMode.value === 'inherit') {
-        delete newExtra.codex_image_generation_bridge
-      } else {
-        newExtra.codex_image_generation_bridge = codexImageGenerationBridgeMode.value === 'enabled'
-      }
-      if (codexImageToolPolicyMode.value === 'strip') {
-        newExtra.codex_image_generation_explicit_tool_policy = 'strip'
-      } else {
-        delete newExtra.codex_image_generation_explicit_tool_policy
+      switch (codexImageToolMode.value) {
+        case 'enabled':
+        case 'disabled':
+          newExtra.codex_image_generation_bridge = codexImageToolMode.value === 'enabled'
+          delete newExtra.codex_image_generation_explicit_tool_policy
+          break
+        case 'block':
+          newExtra.codex_image_generation_explicit_tool_policy = 'strip'
+          delete newExtra.codex_image_generation_bridge
+          break
+        default:
+          delete newExtra.codex_image_generation_bridge
+          delete newExtra.codex_image_generation_explicit_tool_policy
       }
 
       if (props.account.type === 'oauth' || props.account.type === 'setup-token') {
diff --git a/frontend/src/components/account/__tests__/EditAccountModal.spec.ts b/frontend/src/components/account/__tests__/EditAccountModal.spec.ts
index 01c380bc67..c148d04a6f 100644
--- a/frontend/src/components/account/__tests__/EditAccountModal.spec.ts
+++ b/frontend/src/components/account/__tests__/EditAccountModal.spec.ts
@@ -611,7 +611,7 @@ describe('EditAccountModal', () => {
     expect(updateAccountMock.mock.calls[0]?.[1]?.extra?.openai_responses_supported).toBe(true)
   })
 
-  it('submits account-level Codex image generation bridge override', async () => {
+  it('submits Codex image tool force-inject mode as bridge override', async () => {
     const account = buildAccount()
     account.extra = {
       codex_image_generation_bridge: false,
@@ -624,15 +624,16 @@ describe('EditAccountModal', () => {
 
     const wrapper = mountModal(account)
 
-    await wrapper.get('button[data-testid="codex-image-bridge-enabled"]').trigger('click')
+    await wrapper.get('button[data-testid="codex-image-tool-enabled"]').trigger('click')
     await wrapper.get('form#edit-account-form').trigger('submit.prevent')
 
     expect(updateAccountMock).toHaveBeenCalledTimes(1)
     expect(updateAccountMock.mock.calls[0]?.[1]?.extra?.codex_image_generation_bridge).toBe(true)
     expect(updateAccountMock.mock.calls[0]?.[1]?.extra).not.toHaveProperty('codex_image_generation_bridge_enabled')
+    expect(updateAccountMock.mock.calls[0]?.[1]?.extra).not.toHaveProperty('codex_image_generation_explicit_tool_policy')
   })
 
-  it('submits account-level Codex image_generation tool strip policy', async () => {
+  it('submits Codex image tool no-injection mode without strip policy', async () => {
     const account = buildAccount()
     updateAccountMock.mockReset()
     checkMixedChannelRiskMock.mockReset()
@@ -641,14 +642,35 @@ describe('EditAccountModal', () => {
 
     const wrapper = mountModal(account)
 
-    await wrapper.get('button[data-testid="codex-image-tool-policy-strip"]').trigger('click')
+    await wrapper.get('button[data-testid="codex-image-tool-disabled"]').trigger('click')
+    await wrapper.get('form#edit-account-form').trigger('submit.prevent')
+
+    expect(updateAccountMock).toHaveBeenCalledTimes(1)
+    expect(updateAccountMock.mock.calls[0]?.[1]?.extra?.codex_image_generation_bridge).toBe(false)
+    expect(updateAccountMock.mock.calls[0]?.[1]?.extra).not.toHaveProperty('codex_image_generation_explicit_tool_policy')
+  })
+
+  it('submits Codex image tool block mode as strip policy and clears bridge override', async () => {
+    const account = buildAccount()
+    account.extra = {
+      codex_image_generation_bridge: true
+    }
+    updateAccountMock.mockReset()
+    checkMixedChannelRiskMock.mockReset()
+    checkMixedChannelRiskMock.mockResolvedValue({ has_risk: false })
+    updateAccountMock.mockResolvedValue(account)
+
+    const wrapper = mountModal(account)
+
+    await wrapper.get('button[data-testid="codex-image-tool-block"]').trigger('click')
     await wrapper.get('form#edit-account-form').trigger('submit.prevent')
 
     expect(updateAccountMock).toHaveBeenCalledTimes(1)
     expect(updateAccountMock.mock.calls[0]?.[1]?.extra?.codex_image_generation_explicit_tool_policy).toBe('strip')
+    expect(updateAccountMock.mock.calls[0]?.[1]?.extra).not.toHaveProperty('codex_image_generation_bridge')
   })
 
-  it('keeps account-level Codex image_generation tool policy unset when allowing', async () => {
+  it('loads strip policy as block mode and clears both keys when reset to inherit', async () => {
     const account = buildAccount()
     account.extra = {
       codex_image_generation_explicit_tool_policy: 'strip'
@@ -660,11 +682,12 @@ describe('EditAccountModal', () => {
 
     const wrapper = mountModal(account)
 
-    await wrapper.get('button[data-testid="codex-image-tool-policy-allow"]').trigger('click')
+    await wrapper.get('button[data-testid="codex-image-tool-inherit"]').trigger('click')
     await wrapper.get('form#edit-account-form').trigger('submit.prevent')
 
     expect(updateAccountMock).toHaveBeenCalledTimes(1)
     expect(updateAccountMock.mock.calls[0]?.[1]?.extra).not.toHaveProperty('codex_image_generation_explicit_tool_policy')
+    expect(updateAccountMock.mock.calls[0]?.[1]?.extra).not.toHaveProperty('codex_image_generation_bridge')
   })
 
   it('setup-token account can select and submit OAuth WS mode', async () => {
diff --git a/frontend/src/i18n/locales/en.ts b/frontend/src/i18n/locales/en.ts
index 5eed782237..ac831278bd 100644
--- a/frontend/src/i18n/locales/en.ts
+++ b/frontend/src/i18n/locales/en.ts
@@ -3557,25 +3557,21 @@ export default {
         codexCLIOnlyAppServer: 'Allow Codex app-server clients',
         codexCLIOnlyAppServerDesc:
           "Effective only when the switch above is on. When enabled, this account also allows third-party clients that embed the Codex engine over the app-server protocol (e.g. Claude Code's codex plugin); they still pass the global engine-fingerprint gate. OR-combined with the global app-server toggle.",
-        codexImageGenerationBridge: 'Codex image-generation bridge',
-        codexImageGenerationBridgeDesc:
-          'Account policy takes precedence over channel and global settings. Only controls whether Codex requests through the /responses text endpoint receive the image_generation tool; standalone image-generation endpoints are unaffected.',
-        codexImageGenerationBridgeInherit: 'Follow channel',
-        codexImageGenerationBridgeInheritDesc: 'Do not write an account override; use the channel or global policy.',
-        codexImageGenerationBridgeEnabled: 'Force on',
-        codexImageGenerationBridgeEnabledDesc: 'Allow image tool injection for Codex /responses requests.',
-        codexImageGenerationBridgeDisabled: 'Force off',
-        codexImageGenerationBridgeDisabledDesc: 'Block image tool injection for Codex /responses requests.',
-        codexImageGenerationBridgeBadgeInherit: 'Channel policy',
-        codexImageGenerationBridgeBadgeEnabled: 'Account on',
-        codexImageGenerationBridgeBadgeDisabled: 'Account off',
-        codexImageToolPolicy: 'Codex image tool forwarding',
-        codexImageToolPolicyDesc:
-          'Only applies to client-provided image_generation tools in Codex /responses text requests; standalone image-generation endpoints are unaffected.',
-        codexImageToolPolicyAllow: 'Allow',
-        codexImageToolPolicyAllowDesc: 'Default behavior. Keep client-provided image_generation tools.',
-        codexImageToolPolicyStrip: 'Remove',
-        codexImageToolPolicyStripDesc: 'Delete image_generation tools and matching tool_choice before forwarding.',
+        codexImageTool: 'Codex image tool',
+        codexImageToolDesc:
+          'One policy for the image_generation tool on Codex /responses text requests: whether it is auto-injected, and whether client-provided tools pass through. Account policy takes precedence over channel and global settings; standalone image-generation endpoints are unaffected.',
+        codexImageToolInherit: 'Follow channel',
+        codexImageToolInheritDesc: 'No account override; injection follows the channel or global policy, and client-provided image tools pass through.',
+        codexImageToolEnabled: 'Force inject',
+        codexImageToolEnabledDesc: 'Always inject the image tool for Codex /responses requests.',
+        codexImageToolDisabled: 'No injection',
+        codexImageToolDisabledDesc: 'Never auto-inject; client-provided image tools still pass through.',
+        codexImageToolBlock: 'Block all',
+        codexImageToolBlockDesc: 'No injection, and client-provided image tools plus matching tool_choice are removed.',
+        codexImageToolBadgeInherit: 'Channel policy',
+        codexImageToolBadgeEnabled: 'Force inject',
+        codexImageToolBadgeDisabled: 'No injection',
+        codexImageToolBadgeBlock: 'Blocked',
         compactMode: 'Compact mode',
         compactModeDesc:
           'Controls how this account participates in /responses/compact routing. Auto follows probe results, Force On always allows, Force Off always excludes.',
diff --git a/frontend/src/i18n/locales/zh.ts b/frontend/src/i18n/locales/zh.ts
index 9ec540a951..dc090a458f 100644
--- a/frontend/src/i18n/locales/zh.ts
+++ b/frontend/src/i18n/locales/zh.ts
@@ -3730,24 +3730,21 @@ export default {
         codexCLIOnlyDesc: '仅对 OpenAI OAuth 生效。开启后仅允许 Codex 官方客户端家族访问；关闭后完全绕过并保持原逻辑。',
         codexCLIOnlyAppServer: '允许 Codex app-server 客户端',
         codexCLIOnlyAppServerDesc: '仅在上方开关开启时生效。开启后本账号额外放行内嵌 Codex 引擎、经 app-server 协议接入的第三方客户端（如 Claude Code 的 codex 插件），仍需通过全局引擎指纹门；与全局 app-server 开关取 OR（任一开即放行）。',
-        codexImageGenerationBridge: 'Codex 图片生成桥接',
-        codexImageGenerationBridgeDesc:
-          '账号级策略优先于渠道和全局配置。仅控制 Codex 走 /responses 文本端点时是否注入 image_generation 工具；不影响独立图片生成接口。',
-        codexImageGenerationBridgeInherit: '跟随渠道',
-        codexImageGenerationBridgeInheritDesc: '不写入账号覆盖，继续使用渠道或全局策略。',
-        codexImageGenerationBridgeEnabled: '强制开启',
-        codexImageGenerationBridgeEnabledDesc: '允许 Codex /responses 请求获得图片工具注入。',
-        codexImageGenerationBridgeDisabled: '强制关闭',
-        codexImageGenerationBridgeDisabledDesc: '阻断 Codex /responses 的图片工具注入。',
-        codexImageGenerationBridgeBadgeInherit: '渠道策略',
-        codexImageGenerationBridgeBadgeEnabled: '账号开启',
-        codexImageGenerationBridgeBadgeDisabled: '账号关闭',
-        codexImageToolPolicy: 'Codex 图片工具传递策略',
-        codexImageToolPolicyDesc: '仅作用于 Codex /responses 文本端点中客户端显式携带的 image_generation tool；不影响独立图片生成接口。',
-        codexImageToolPolicyAllow: '放行',
-        codexImageToolPolicyAllowDesc: '默认行为，保留客户端显式发送的 image_generation tool。',
-        codexImageToolPolicyStrip: '移除',
-        codexImageToolPolicyStripDesc: '转发前删除 image_generation tool 和指向它的 tool_choice。',
+        codexImageTool: 'Codex 图片工具',
+        codexImageToolDesc:
+          '统一控制 Codex /responses 文本请求的 image_generation 图片工具：是否自动注入，以及客户端自带该工具时是否放行。账号级策略优先于渠道和全局配置，不影响独立图片生成接口。',
+        codexImageToolInherit: '跟随渠道',
+        codexImageToolInheritDesc: '不写入账号覆盖，是否注入由渠道或全局策略决定；客户端自带的图片工具照常放行。',
+        codexImageToolEnabled: '强制注入',
+        codexImageToolEnabledDesc: '始终为 Codex /responses 请求注入图片工具。',
+        codexImageToolDisabled: '关闭注入',
+        codexImageToolDisabledDesc: '不自动注入；客户端自带的图片工具仍会放行。',
+        codexImageToolBlock: '完全阻断',
+        codexImageToolBlockDesc: '不注入，并移除客户端自带的图片工具及指向它的 tool_choice。',
+        codexImageToolBadgeInherit: '渠道策略',
+        codexImageToolBadgeEnabled: '强制注入',
+        codexImageToolBadgeDisabled: '关闭注入',
+        codexImageToolBadgeBlock: '完全阻断',
         compactMode: 'Compact 模式',
         compactModeDesc:
           '控制本账号在 /responses/compact 调度中的参与方式。Auto 跟随探测结果，Force On 强制允许，Force Off 强制排除。',

From 6bd248fd1fa4e45af448ad75e737ecd04f6c3203 Mon Sep 17 00:00:00 2001
From: wucm667 <stevenwucongmin@gmail.com>
Date: Fri, 3 Jul 2026 18:01:11 +0800
Subject: [PATCH 14/28] fix(admin): avoid merging Codex access-only imports

---
 .../handler/admin/account_codex_import.go     |  71 ++-
 .../admin/account_codex_import_test.go        | 492 ++++++++++++++++--
 2 files changed, 521 insertions(+), 42 deletions(-)

diff --git a/backend/internal/handler/admin/account_codex_import.go b/backend/internal/handler/admin/account_codex_import.go
index 6ec9495d84..01a5fbfa1c 100644
--- a/backend/internal/handler/admin/account_codex_import.go
+++ b/backend/internal/handler/admin/account_codex_import.go
@@ -253,6 +253,17 @@ func (h *AccountHandler) importCodexSessions(ctx context.Context, req CodexSessi
 					Message: "已有账号未记录 chatgpt_user_id，已按共享的 chatgpt_account_id 匹配并回填，请确认两者属于同一用户",
 				})
 			}
+			preserveExistingRefresh := item.RefreshToken == "" &&
+				codexCredentialString(existing.Credentials, "refresh_token") != ""
+			if preserveExistingRefresh {
+				result.Warnings = append(result.Warnings, CodexSessionImportMessage{
+					Index:   entry.Index,
+					Name:    accountName,
+					Message: "已有账号包含 refresh_token，本次 accessToken-only 导入已保留自动续期凭据",
+				})
+				effectiveExpiresAt = nil
+				autoPauseOnExpired = nil
+			}
 			mergedCredentials := mergeCodexImportCredentials(existing.Credentials, credentials, item)
 			mergedExtra := mergeCodexImportMap(existing.Extra, extra)
 			updateInput := &service.UpdateAccountInput{
@@ -592,7 +603,7 @@ func normalizeCodexImportEntry(entry codexImportEntry) (*codexImportAccount, err
 
 	fingerprint := codexTokenFingerprint(item.AccessToken)
 	item.Extra["access_token_sha256"] = fingerprint
-	item.IdentityKeys = buildCodexIdentityKeys(item.AccountID, item.UserID, item.Email, item.AccessToken)
+	item.IdentityKeys = buildCodexImportIdentityKeys(item.AccountID, item.UserID, item.Email, item.AccessToken, item.RefreshToken)
 	item.Name = buildCodexImportAccountName(item, entry.Index)
 
 	return item, nil
@@ -815,13 +826,25 @@ func sanitizeCodexImportCredentialExtras(input map[string]any) map[string]any {
 	return out
 }
 
-// buildCodexIdentityKeys 按身份强度排序生成匹配键：chatgpt_account_id 在同一
-// ChatGPT 团队内是共享的，因此 account: 键排在最后，且命中时还需通过
-// codexIdentityConflicts 的跨用户校验才生效。
-func buildCodexIdentityKeys(accountID, userID, email, accessToken string) []string {
+// buildCodexImportIdentityKeys 生成导入条目的匹配键。refresh_token 缺失时
+// Codex session 只能作为 accessToken-only 凭据使用，此时以 access token
+// 指纹作为唯一稳定身份，避免同 workspace 下共享的 account/user 标识误合并。
+func buildCodexImportIdentityKeys(accountID, userID, email, accessToken, refreshToken string) []string {
+	accessToken = strings.TrimSpace(accessToken)
+	refreshToken = strings.TrimSpace(refreshToken)
+	if refreshToken == "" && accessToken != "" {
+		return []string{"access:" + codexTokenFingerprint(accessToken)}
+	}
+	return buildCodexStoredIdentityKeys(accountID, userID, email, accessToken)
+}
+
+// buildCodexStoredIdentityKeys 生成存量账号索引键，保留 user/account 维度，
+// 让 accessToken-only 账号后续升级为完整 OAuth 时仍能命中并更新原账号。
+func buildCodexStoredIdentityKeys(accountID, userID, email, accessToken string) []string {
 	keys := make([]string, 0, 3)
 	accountID = strings.TrimSpace(accountID)
 	userID = strings.TrimSpace(userID)
+	accessToken = strings.TrimSpace(accessToken)
 	if userID != "" {
 		keys = append(keys, "user:"+userID)
 	}
@@ -830,7 +853,7 @@ func buildCodexIdentityKeys(accountID, userID, email, accessToken string) []stri
 			keys = append(keys, "email:"+email)
 		}
 	}
-	if accessToken = strings.TrimSpace(accessToken); accessToken != "" {
+	if accessToken != "" {
 		keys = append(keys, "access:"+codexTokenFingerprint(accessToken))
 	}
 	if accountID != "" {
@@ -854,7 +877,8 @@ func (i *codexAccountIndex) Add(account service.Account) {
 	if i.accountsByKey == nil {
 		i.accountsByKey = map[string][]service.Account{}
 	}
-	keys := buildCodexIdentityKeys(
+	i.remove(account.ID)
+	keys := buildCodexStoredIdentityKeys(
 		codexCredentialString(account.Credentials, "chatgpt_account_id"),
 		codexCredentialString(account.Credentials, "chatgpt_user_id"),
 		codexCredentialString(account.Credentials, "email"),
@@ -865,6 +889,22 @@ func (i *codexAccountIndex) Add(account service.Account) {
 	}
 }
 
+func (i *codexAccountIndex) remove(accountID int64) {
+	for key, accounts := range i.accountsByKey {
+		kept := accounts[:0]
+		for _, account := range accounts {
+			if account.ID != accountID {
+				kept = append(kept, account)
+			}
+		}
+		if len(kept) == 0 {
+			delete(i.accountsByKey, key)
+			continue
+		}
+		i.accountsByKey[key] = kept
+	}
+}
+
 // upsertCodexAccount 保留同一键下的全部候选账号（共享的 account: 键可对应
 // 团队内多个账号），同一账号重复 Add 时原位替换为最新状态。
 func upsertCodexAccount(accounts []service.Account, account service.Account) []service.Account {
@@ -894,9 +934,9 @@ func (i *codexAccountIndex) Find(keys []string, userID string) (*service.Account
 }
 
 // codexIdentityConflicts 判断 account: 键的命中是否把同一 ChatGPT 团队的两个
-// 不同成员误连到一起：双方都携带 user id 且不相等时视为冲突。任一侧缺少
-// user id 时保留匹配，使早期未记录 chatgpt_user_id 的存量账号仍能被更新
-// （并借助凭据合并回填 user id），而不是产生重复账号。
+// 不同成员误连到一起：双方都携带 user id 且不相等时视为冲突。存量索引侧
+// 仍保留 account 键，任一侧缺少 user id 时允许匹配，使含 refresh_token
+// 的常规导入和 accessToken-only 账号升级为完整 OAuth 时仍能更新原账号。
 func codexIdentityConflicts(key, userID, storedUserID string) bool {
 	if !strings.HasPrefix(key, "account:") {
 		return false
@@ -948,8 +988,15 @@ func mergeCodexImportCredentials(existing, incoming map[string]any, item *codexI
 		return out
 	}
 	if strings.TrimSpace(item.RefreshToken) == "" {
-		delete(out, "refresh_token")
-		delete(out, "client_id")
+		if codexCredentialString(existing, "refresh_token") == "" {
+			delete(out, "refresh_token")
+			delete(out, "client_id")
+		} else {
+			out["refresh_token"] = existing["refresh_token"]
+			if clientID, ok := existing["client_id"]; ok {
+				out["client_id"] = clientID
+			}
+		}
 	}
 	if strings.TrimSpace(item.IDToken) == "" {
 		delete(out, "id_token")
diff --git a/backend/internal/handler/admin/account_codex_import_test.go b/backend/internal/handler/admin/account_codex_import_test.go
index f4ee5bd7db..a52463aa86 100644
--- a/backend/internal/handler/admin/account_codex_import_test.go
+++ b/backend/internal/handler/admin/account_codex_import_test.go
@@ -1,6 +1,7 @@
 package admin
 
 import (
+	"context"
 	"encoding/base64"
 	"encoding/json"
 	"fmt"
@@ -144,7 +145,7 @@ func TestNormalizeCodexSessionJSONExtractsCredentialsAndIgnoresSessionToken(t *t
 	}
 }
 
-func TestMergeCodexImportCredentialsClearsStaleRefreshFieldsWhenIncomingHasNoRefreshToken(t *testing.T) {
+func TestMergeCodexImportCredentialsPreservesExistingRefreshFieldsWhenIncomingHasNoRefreshToken(t *testing.T) {
 	existing := map[string]any{
 		"access_token":       "old-access-token",
 		"refresh_token":      "old-refresh-token",
@@ -171,11 +172,11 @@ func TestMergeCodexImportCredentialsClearsStaleRefreshFieldsWhenIncomingHasNoRef
 	if merged["chatgpt_account_id"] != "acct-new" {
 		t.Fatalf("chatgpt_account_id = %v, want acct-new", merged["chatgpt_account_id"])
 	}
-	if _, ok := merged["refresh_token"]; ok {
-		t.Fatalf("refresh_token should be cleared")
+	if merged["refresh_token"] != "old-refresh-token" {
+		t.Fatalf("refresh_token = %v, want old-refresh-token", merged["refresh_token"])
 	}
-	if _, ok := merged["client_id"]; ok {
-		t.Fatalf("client_id should be cleared")
+	if merged["client_id"] != "old-client-id" {
+		t.Fatalf("client_id = %v, want old-client-id", merged["client_id"])
 	}
 	if _, ok := merged["id_token"]; ok {
 		t.Fatalf("id_token should be cleared")
@@ -301,9 +302,9 @@ func TestResolveCodexImportExpiryForNoRefreshTokenUsesEarlierRequestExpiry(t *te
 }
 
 func TestCodexIdentityKeysPreferStrongIdentifiers(t *testing.T) {
-	keys := buildCodexIdentityKeys("acct-1", "user-1", "same@example.com", "token")
+	keys := buildCodexImportIdentityKeys("acct-1", "user-1", "same@example.com", "token", "refresh")
 	if len(keys) == 0 || keys[0] != "user:user-1" {
-		t.Fatalf("user key should have highest priority: %v", keys)
+		t.Fatalf("user key should have highest priority when refresh token exists: %v", keys)
 	}
 	if keys[len(keys)-1] != "account:acct-1" {
 		t.Fatalf("shared account key should be the last fallback: %v", keys)
@@ -314,7 +315,7 @@ func TestCodexIdentityKeysPreferStrongIdentifiers(t *testing.T) {
 		}
 	}
 
-	keys = buildCodexIdentityKeys("", "", "same@example.com", "token")
+	keys = buildCodexImportIdentityKeys("", "", "same@example.com", "token", "refresh")
 	hasEmail := false
 	for _, key := range keys {
 		if key == "email:same@example.com" {
@@ -324,6 +325,11 @@ func TestCodexIdentityKeysPreferStrongIdentifiers(t *testing.T) {
 	if !hasEmail {
 		t.Fatalf("weak identity should include email fallback: %v", keys)
 	}
+
+	keys = buildCodexImportIdentityKeys("acct-1", "user-1", "same@example.com", "token", "")
+	if len(keys) != 1 || !strings.HasPrefix(keys[0], "access:") {
+		t.Fatalf("accessToken-only identity should use only access fingerprint: %v", keys)
+	}
 }
 
 func TestCodexAccountIndexDoesNotMatchDifferentUsersInSameChatGPTAccount(t *testing.T) {
@@ -333,35 +339,37 @@ func TestCodexAccountIndexDoesNotMatchDifferentUsersInSameChatGPTAccount(t *test
 			"chatgpt_account_id": "team-1",
 			"chatgpt_user_id":    "user-1",
 			"access_token":       "token-1",
+			"refresh_token":      "refresh-1",
 		},
 	}
 	index := buildCodexAccountIndex([]service.Account{existing})
 
-	keys := buildCodexIdentityKeys("team-1", "user-2", "", "token-2")
+	keys := buildCodexImportIdentityKeys("team-1", "user-2", "", "token-2", "refresh-2")
 	if got, _ := index.Find(keys, "user-2"); got != nil {
 		t.Fatalf("Find matched account ID %d for a different chatgpt_user_id in the same team", got.ID)
 	}
 
-	keys = buildCodexIdentityKeys("team-1", "user-1", "", "token-2")
+	keys = buildCodexImportIdentityKeys("team-1", "user-1", "", "token-2", "refresh-2")
 	got, _ := index.Find(keys, "user-1")
 	if got == nil || got.ID != existing.ID {
 		t.Fatalf("Find by same chatgpt_user_id = %v, want account ID %d", got, existing.ID)
 	}
 }
 
-func TestCodexAccountIndexFallsBackToAccountKeyWhenUserIDMissing(t *testing.T) {
-	// 存量账号缺少 chatgpt_user_id：携带 user id 的重新导入应命中并更新（回填），
-	// 而不是创建重复账号。
+func TestCodexAccountIndexFallsBackToAccountKeyWhenRefreshTokenExistsAndUserIDMissing(t *testing.T) {
+	// 含 refresh_token 的常规导入沿用 a5638a4e 的兼容逻辑：存量账号缺少
+	// chatgpt_user_id 时，携带 user id 的重新导入仍可命中并回填。
 	legacy := service.Account{
 		ID: 20,
 		Credentials: map[string]any{
 			"chatgpt_account_id": "team-1",
 			"access_token":       "token-old",
+			"refresh_token":      "refresh-old",
 		},
 	}
 	index := buildCodexAccountIndex([]service.Account{legacy})
 
-	keys := buildCodexIdentityKeys("team-1", "user-1", "", "token-new")
+	keys := buildCodexImportIdentityKeys("team-1", "user-1", "", "token-new", "refresh-new")
 	got, matchedKey := index.Find(keys, "user-1")
 	if got == nil || got.ID != legacy.ID {
 		t.Fatalf("Find legacy account without stored user id = %v, want account ID %d", got, legacy.ID)
@@ -370,30 +378,59 @@ func TestCodexAccountIndexFallsBackToAccountKeyWhenUserIDMissing(t *testing.T) {
 		t.Fatalf("matched key = %q, want account:team-1", matchedKey)
 	}
 
-	// 反向：导入条目无法解析出 user id 时，仍应通过 account 键命中已有账号。
+	// 反向：含 refresh_token 的导入条目无法解析出 user id 时，仍应通过
+	// account 键命中已有账号，保持常规导入去重行为。
 	full := service.Account{
 		ID: 21,
 		Credentials: map[string]any{
 			"chatgpt_account_id": "team-2",
 			"chatgpt_user_id":    "user-9",
 			"access_token":       "token-old",
+			"refresh_token":      "refresh-old",
 		},
 	}
 	index = buildCodexAccountIndex([]service.Account{full})
 
-	keys = buildCodexIdentityKeys("team-2", "", "", "token-opaque")
+	keys = buildCodexImportIdentityKeys("team-2", "", "", "token-opaque", "refresh-new")
 	got, _ = index.Find(keys, "")
 	if got == nil || got.ID != full.ID {
 		t.Fatalf("Find by account key without entry user id = %v, want account ID %d", got, full.ID)
 	}
 }
 
+func TestCodexAccountIndexAccessTokenOnlyUsesTokenFingerprint(t *testing.T) {
+	existing := service.Account{
+		ID: 22,
+		Credentials: map[string]any{
+			"chatgpt_account_id": "team-1",
+			"chatgpt_user_id":    "user-1",
+			"access_token":       "token-old",
+		},
+	}
+	index := buildCodexAccountIndex([]service.Account{existing})
+
+	keys := buildCodexImportIdentityKeys("team-1", "user-1", "", "token-new", "")
+	if got, matchedKey := index.Find(keys, "user-1"); got != nil {
+		t.Fatalf("accessToken-only import matched by %q despite different token: account ID %d", matchedKey, got.ID)
+	}
+
+	keys = buildCodexImportIdentityKeys("team-1", "user-1", "", "token-old", "")
+	got, matchedKey := index.Find(keys, "user-1")
+	if got == nil || got.ID != existing.ID {
+		t.Fatalf("Find accessToken-only duplicate by fingerprint = %v, want account ID %d", got, existing.ID)
+	}
+	if !strings.HasPrefix(matchedKey, "access:") {
+		t.Fatalf("matched key = %q, want access fingerprint", matchedKey)
+	}
+}
+
 func TestCodexAccountIndexKeepsAllCandidatesForSharedAccountKey(t *testing.T) {
 	legacy := service.Account{
 		ID: 30,
 		Credentials: map[string]any{
 			"chatgpt_account_id": "team-1",
 			"access_token":       "token-legacy",
+			"refresh_token":      "refresh-legacy",
 		},
 	}
 	member := service.Account{
@@ -402,10 +439,11 @@ func TestCodexAccountIndexKeepsAllCandidatesForSharedAccountKey(t *testing.T) {
 			"chatgpt_account_id": "team-1",
 			"chatgpt_user_id":    "user-2",
 			"access_token":       "token-member",
+			"refresh_token":      "refresh-member",
 		},
 	}
 
-	// 无论索引构建顺序如何，携带新 user id 的条目都应跳过 user-2 的账号、
+	// 无论索引构建顺序如何，携带新 user id 的条目都应跳过 user-2 的账号，
 	// 命中缺少 user id 的存量账号，而不是因单一候选被遮蔽而落空。
 	for _, accounts := range [][]service.Account{
 		{member, legacy},
@@ -413,7 +451,7 @@ func TestCodexAccountIndexKeepsAllCandidatesForSharedAccountKey(t *testing.T) {
 	} {
 		index := buildCodexAccountIndex(accounts)
 
-		keys := buildCodexIdentityKeys("team-1", "user-1", "", "token-new")
+		keys := buildCodexImportIdentityKeys("team-1", "user-1", "", "token-new", "refresh-new")
 		got, matchedKey := index.Find(keys, "user-1")
 		if got == nil || got.ID != legacy.ID {
 			t.Fatalf("Find with shared account key = %v, want legacy account ID %d", got, legacy.ID)
@@ -422,7 +460,7 @@ func TestCodexAccountIndexKeepsAllCandidatesForSharedAccountKey(t *testing.T) {
 			t.Fatalf("matched key = %q, want account:team-1", matchedKey)
 		}
 
-		keys = buildCodexIdentityKeys("team-1", "user-2", "", "token-new")
+		keys = buildCodexImportIdentityKeys("team-1", "user-2", "", "token-new", "refresh-new")
 		got, matchedKey = index.Find(keys, "user-2")
 		if got == nil || got.ID != member.ID {
 			t.Fatalf("Find by user key = %v, want member account ID %d", got, member.ID)
@@ -449,18 +487,19 @@ func TestCodexAccountIndexUpsertReplacesSameAccount(t *testing.T) {
 			"chatgpt_account_id": "team-1",
 			"chatgpt_user_id":    "user-1",
 			"access_token":       "token-new",
+			"refresh_token":      "refresh-new",
 		},
 	}
 	index.Add(backfilled)
 
 	// 回填后同一账号在 account 键下应被原位替换而非残留旧副本：
 	// 其他成员的条目不应再通过旧副本（无 user id）命中该账号。
-	keys := buildCodexIdentityKeys("team-1", "user-2", "", "token-other")
-	if got, _ := index.Find(keys, "user-2"); got != nil {
-		t.Fatalf("stale candidate matched after upsert: account ID %d", got.ID)
+	keys := buildCodexImportIdentityKeys("team-1", "user-2", "", "token-other", "refresh-other")
+	if got, matchedKey := index.Find(keys, "user-2"); got != nil {
+		t.Fatalf("stale candidate matched after upsert by %q: account ID %d", matchedKey, got.ID)
 	}
 
-	keys = buildCodexIdentityKeys("team-1", "user-1", "", "token-other")
+	keys = buildCodexImportIdentityKeys("team-1", "user-1", "", "token-other", "refresh-other")
 	got, _ := index.Find(keys, "user-1")
 	if got == nil || got.ID != backfilled.ID {
 		t.Fatalf("Find after upsert = %v, want account ID %d", got, backfilled.ID)
@@ -472,28 +511,421 @@ func TestCodexAccountIndexUpsertReplacesSameAccount(t *testing.T) {
 
 func TestCodexIdentitySeenDistinguishesTeamMembers(t *testing.T) {
 	seen := map[string]codexSeenIdentity{}
-	member1 := buildCodexIdentityKeys("team-1", "user-1", "", "token-1")
+	member1 := buildCodexImportIdentityKeys("team-1", "user-1", "", "token-1", "refresh-1")
 	markCodexIdentitySeen(seen, member1, 1, "user-1")
 
-	member2 := buildCodexIdentityKeys("team-1", "user-2", "", "token-2")
+	member2 := buildCodexImportIdentityKeys("team-1", "user-2", "", "token-2", "refresh-2")
 	if index, ok := firstSeenCodexIdentity(seen, member2, "user-2"); ok {
 		t.Fatalf("different team member treated as duplicate of entry %d", index)
 	}
 
-	again := buildCodexIdentityKeys("team-1", "user-1", "", "token-3")
+	again := buildCodexImportIdentityKeys("team-1", "user-1", "", "token-3", "refresh-3")
 	index, ok := firstSeenCodexIdentity(seen, again, "user-1")
 	if !ok || index != 1 {
 		t.Fatalf("same user re-entry dedup = (%d, %v), want (1, true)", index, ok)
 	}
 
-	// 无 user id 的条目与已见同 account 条目视为重复（保守跳过，与既有行为一致）。
-	opaque := buildCodexIdentityKeys("team-1", "", "", "token-4")
+	// 无 user id 的条目不应因共享 account id 与已见团队成员互相去重；
+	// 只有相同 access token 指纹才视为重复。
+	opaque := buildCodexImportIdentityKeys("team-1", "", "", "token-4", "")
 	index, ok = firstSeenCodexIdentity(seen, opaque, "")
-	if !ok || index != 1 {
-		t.Fatalf("entry without user id dedup = (%d, %v), want (1, true)", index, ok)
+	if ok {
+		t.Fatalf("entry without user id dedup = (%d, %v), want no match", index, ok)
 	}
 }
 
+func TestNormalizeCodexImportUsesJWTSubForAccessTokenOnlyIdentity(t *testing.T) {
+	accessToken := buildCodexImportTestJWT(t, time.Now().Add(time.Hour), map[string]any{
+		"sub": "user-from-access-token",
+		"https://api.openai.com/auth": map[string]any{
+			"chatgpt_account_id": "workspace-1",
+		},
+	})
+
+	item, err := normalizeCodexImportEntry(codexImportEntry{Index: 1, Value: accessToken})
+	if err != nil {
+		t.Fatalf("normalizeCodexImportEntry error = %v", err)
+	}
+	if item.UserID != "user-from-access-token" {
+		t.Fatalf("UserID = %q, want JWT sub", item.UserID)
+	}
+	if len(item.IdentityKeys) != 1 || !strings.HasPrefix(item.IdentityKeys[0], "access:") {
+		t.Fatalf("IdentityKeys = %v, want access fingerprint only for accessToken-only import", item.IdentityKeys)
+	}
+	if got := item.Credentials["chatgpt_user_id"]; got != "user-from-access-token" {
+		t.Fatalf("credential chatgpt_user_id = %v, want JWT sub", got)
+	}
+}
+
+func TestImportCodexSessionsAccessTokenOnlySameWorkspaceDifferentUsersCreatesTwoAccounts(t *testing.T) {
+	svc := newCodexImportMemoryAdminService(nil)
+	handler := NewAccountHandler(svc, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil)
+	req := CodexSessionImportRequest{SkipDefaultGroupBind: boolPtr(true)}
+	entries := []codexImportEntry{
+		{Index: 1, Value: buildCodexAccessOnlyImportValue(t, "workspace-1", "user-1")},
+		{Index: 2, Value: buildCodexAccessOnlyImportValue(t, "workspace-1", "user-2")},
+	}
+
+	result, err := handler.importCodexSessions(context.Background(), req, entries)
+	if err != nil {
+		t.Fatalf("importCodexSessions error = %v", err)
+	}
+	if result.Created != 2 || result.Updated != 0 || result.Skipped != 0 || result.Failed != 0 {
+		t.Fatalf("result = %+v, want two created accounts", result)
+	}
+	if len(svc.createdAccounts) != 2 {
+		t.Fatalf("created accounts = %d, want 2", len(svc.createdAccounts))
+	}
+	if svc.createdAccounts[0].Credentials["chatgpt_user_id"] == svc.createdAccounts[1].Credentials["chatgpt_user_id"] {
+		t.Fatalf("created accounts share user id: %v", svc.createdAccounts)
+	}
+}
+
+func TestImportCodexSessionsAccessTokenOnlySameWorkspaceAndUserDifferentTokensCreatesTwoAccounts(t *testing.T) {
+	svc := newCodexImportMemoryAdminService(nil)
+	handler := NewAccountHandler(svc, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil)
+	req := CodexSessionImportRequest{SkipDefaultGroupBind: boolPtr(true)}
+	entries := []codexImportEntry{
+		{Index: 1, Value: map[string]any{
+			"access_token": buildCodexImportTestJWT(t, time.Now().Add(time.Hour), map[string]any{
+				"sub": "shared-user",
+				"jti": "token-1",
+				"https://api.openai.com/auth": map[string]any{
+					"chatgpt_account_id": "workspace-1",
+				},
+			}),
+		}},
+		{Index: 2, Value: map[string]any{
+			"access_token": buildCodexImportTestJWT(t, time.Now().Add(time.Hour), map[string]any{
+				"sub": "shared-user",
+				"jti": "token-2",
+				"https://api.openai.com/auth": map[string]any{
+					"chatgpt_account_id": "workspace-1",
+				},
+			}),
+		}},
+	}
+
+	result, err := handler.importCodexSessions(context.Background(), req, entries)
+	if err != nil {
+		t.Fatalf("importCodexSessions error = %v", err)
+	}
+	if result.Created != 2 || result.Updated != 0 || result.Skipped != 0 || result.Failed != 0 {
+		t.Fatalf("result = %+v, want two created accounts", result)
+	}
+	if len(svc.createdAccounts) != 2 {
+		t.Fatalf("created accounts = %d, want 2", len(svc.createdAccounts))
+	}
+}
+
+func TestImportCodexSessionsAccessTokenOnlySameUserUpdatesExisting(t *testing.T) {
+	existingToken := buildCodexAccessToken(t, "workspace-1", "user-1", time.Now().Add(time.Hour))
+	svc := newCodexImportMemoryAdminService([]service.Account{{
+		ID:       10,
+		Name:     "existing",
+		Platform: service.PlatformOpenAI,
+		Type:     service.AccountTypeOAuth,
+		Credentials: map[string]any{
+			"chatgpt_account_id": "workspace-1",
+			"chatgpt_user_id":    "user-1",
+			"access_token":       existingToken,
+		},
+	}})
+	handler := NewAccountHandler(svc, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil)
+	req := CodexSessionImportRequest{SkipDefaultGroupBind: boolPtr(true)}
+	entries := []codexImportEntry{
+		{Index: 1, Value: map[string]any{"access_token": existingToken}},
+	}
+
+	result, err := handler.importCodexSessions(context.Background(), req, entries)
+	if err != nil {
+		t.Fatalf("importCodexSessions error = %v", err)
+	}
+	if result.Created != 0 || result.Updated != 1 || result.Failed != 0 {
+		t.Fatalf("result = %+v, want one updated account", result)
+	}
+	if len(svc.createdAccounts) != 0 {
+		t.Fatalf("created accounts = %d, want 0", len(svc.createdAccounts))
+	}
+	if len(svc.updatedAccounts) != 1 || svc.updatedAccounts[0].id != 10 {
+		t.Fatalf("updated accounts = %+v, want account 10", svc.updatedAccounts)
+	}
+}
+
+func TestImportCodexSessionsUpgradesAccessTokenOnlyAccountWithRefreshToken(t *testing.T) {
+	oldToken := buildCodexAccessTokenWithJTI(t, "workspace-1", "user-1", "old-token", time.Now().Add(time.Hour))
+	newToken := buildCodexAccessTokenWithJTI(t, "workspace-1", "user-1", "new-token", time.Now().Add(time.Hour))
+	svc := newCodexImportMemoryAdminService([]service.Account{{
+		ID:       12,
+		Name:     "existing",
+		Platform: service.PlatformOpenAI,
+		Type:     service.AccountTypeOAuth,
+		Credentials: map[string]any{
+			"chatgpt_account_id": "workspace-1",
+			"chatgpt_user_id":    "user-1",
+			"access_token":       oldToken,
+		},
+	}})
+	handler := NewAccountHandler(svc, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil)
+	req := CodexSessionImportRequest{SkipDefaultGroupBind: boolPtr(true)}
+	entries := []codexImportEntry{
+		{Index: 1, Value: map[string]any{
+			"access_token":  newToken,
+			"refresh_token": "refresh-new",
+		}},
+	}
+
+	result, err := handler.importCodexSessions(context.Background(), req, entries)
+	if err != nil {
+		t.Fatalf("importCodexSessions error = %v", err)
+	}
+	if result.Created != 0 || result.Updated != 1 || result.Failed != 0 {
+		t.Fatalf("result = %+v, want one updated account", result)
+	}
+	if len(svc.updatedAccounts) != 1 || svc.updatedAccounts[0].id != 12 {
+		t.Fatalf("updated accounts = %+v, want account 12", svc.updatedAccounts)
+	}
+	if got := svc.updatedAccounts[0].input.Credentials["refresh_token"]; got != "refresh-new" {
+		t.Fatalf("updated refresh_token = %v, want refresh-new", got)
+	}
+}
+
+func TestImportCodexSessionsAccessTokenOnlyPreservesExistingRefreshToken(t *testing.T) {
+	existingToken := buildCodexAccessToken(t, "workspace-1", "user-1", time.Now().Add(time.Hour))
+	svc := newCodexImportMemoryAdminService([]service.Account{{
+		ID:       13,
+		Name:     "existing",
+		Platform: service.PlatformOpenAI,
+		Type:     service.AccountTypeOAuth,
+		Credentials: map[string]any{
+			"chatgpt_account_id": "workspace-1",
+			"chatgpt_user_id":    "user-1",
+			"access_token":       existingToken,
+			"refresh_token":      "refresh-old",
+			"client_id":          "client-old",
+		},
+	}})
+	handler := NewAccountHandler(svc, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil)
+	req := CodexSessionImportRequest{SkipDefaultGroupBind: boolPtr(true)}
+	entries := []codexImportEntry{
+		{Index: 1, Value: map[string]any{"access_token": existingToken}},
+	}
+
+	result, err := handler.importCodexSessions(context.Background(), req, entries)
+	if err != nil {
+		t.Fatalf("importCodexSessions error = %v", err)
+	}
+	if result.Created != 0 || result.Updated != 1 || result.Failed != 0 {
+		t.Fatalf("result = %+v, want one updated account", result)
+	}
+	update := svc.updatedAccounts[0].input
+	if got := update.Credentials["refresh_token"]; got != "refresh-old" {
+		t.Fatalf("refresh_token = %v, want refresh-old", got)
+	}
+	if got := update.Credentials["client_id"]; got != "client-old" {
+		t.Fatalf("client_id = %v, want client-old", got)
+	}
+	if update.ExpiresAt != nil {
+		t.Fatalf("ExpiresAt = %v, want nil to preserve OAuth account expiry", *update.ExpiresAt)
+	}
+	if update.AutoPauseOnExpired != nil {
+		t.Fatalf("AutoPauseOnExpired = %v, want nil to preserve OAuth account scheduling", *update.AutoPauseOnExpired)
+	}
+}
+
+func TestImportCodexSessionsBatchOldAccessTokenDoesNotRollbackRefreshToken(t *testing.T) {
+	oldToken := buildCodexAccessTokenWithJTI(t, "workspace-1", "user-1", "old-token", time.Now().Add(time.Hour))
+	newToken := buildCodexAccessTokenWithJTI(t, "workspace-1", "user-1", "new-token", time.Now().Add(time.Hour))
+	svc := newCodexImportMemoryAdminService([]service.Account{{
+		ID:       14,
+		Name:     "existing",
+		Platform: service.PlatformOpenAI,
+		Type:     service.AccountTypeOAuth,
+		Credentials: map[string]any{
+			"chatgpt_account_id": "workspace-1",
+			"chatgpt_user_id":    "user-1",
+			"access_token":       oldToken,
+			"refresh_token":      "refresh-old",
+		},
+	}})
+	handler := NewAccountHandler(svc, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil)
+	req := CodexSessionImportRequest{SkipDefaultGroupBind: boolPtr(true)}
+	entries := []codexImportEntry{
+		{Index: 1, Value: map[string]any{
+			"access_token":  newToken,
+			"refresh_token": "refresh-new",
+		}},
+		{Index: 2, Value: map[string]any{"access_token": oldToken}},
+	}
+
+	result, err := handler.importCodexSessions(context.Background(), req, entries)
+	if err != nil {
+		t.Fatalf("importCodexSessions error = %v", err)
+	}
+	if result.Updated != 1 || result.Created != 1 || result.Failed != 0 {
+		t.Fatalf("result = %+v, want first item updated and stale access token created separately", result)
+	}
+	if len(svc.updatedAccounts) != 1 || svc.updatedAccounts[0].id != 14 {
+		t.Fatalf("updated accounts = %+v, want account 14 updated once", svc.updatedAccounts)
+	}
+	stored, err := svc.GetAccount(context.Background(), 14)
+	if err != nil {
+		t.Fatalf("GetAccount error = %v", err)
+	}
+	if got := stored.Credentials["access_token"]; got != newToken {
+		t.Fatalf("stored access_token rolled back = %v, want new token", got)
+	}
+	if got := stored.Credentials["refresh_token"]; got != "refresh-new" {
+		t.Fatalf("stored refresh_token = %v, want refresh-new", got)
+	}
+}
+
+func TestImportCodexSessionsWithRefreshTokenKeepsExistingDedup(t *testing.T) {
+	existingToken := buildCodexAccessToken(t, "workspace-1", "user-1", time.Now().Add(time.Hour))
+	svc := newCodexImportMemoryAdminService([]service.Account{{
+		ID:       11,
+		Name:     "existing",
+		Platform: service.PlatformOpenAI,
+		Type:     service.AccountTypeOAuth,
+		Credentials: map[string]any{
+			"chatgpt_account_id": "workspace-1",
+			"chatgpt_user_id":    "user-1",
+			"access_token":       existingToken,
+			"refresh_token":      "refresh-old",
+		},
+	}})
+	handler := NewAccountHandler(svc, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil)
+	req := CodexSessionImportRequest{SkipDefaultGroupBind: boolPtr(true)}
+	entries := []codexImportEntry{
+		{Index: 1, Value: buildCodexRefreshImportValue(t, "workspace-1", "user-1", "refresh-new")},
+	}
+
+	result, err := handler.importCodexSessions(context.Background(), req, entries)
+	if err != nil {
+		t.Fatalf("importCodexSessions error = %v", err)
+	}
+	if result.Created != 0 || result.Updated != 1 || result.Failed != 0 {
+		t.Fatalf("result = %+v, want one updated account", result)
+	}
+	if got := svc.updatedAccounts[0].input.Credentials["refresh_token"]; got != "refresh-new" {
+		t.Fatalf("updated refresh_token = %v, want refresh-new", got)
+	}
+}
+
+type codexImportMemoryAdminService struct {
+	*stubAdminService
+	nextID          int64
+	updatedAccounts []struct {
+		id    int64
+		input *service.UpdateAccountInput
+	}
+}
+
+func newCodexImportMemoryAdminService(accounts []service.Account) *codexImportMemoryAdminService {
+	stub := newStubAdminService()
+	stub.accounts = append([]service.Account(nil), accounts...)
+	return &codexImportMemoryAdminService{
+		stubAdminService: stub,
+		nextID:           100,
+	}
+}
+
+func (s *codexImportMemoryAdminService) CreateAccount(ctx context.Context, input *service.CreateAccountInput) (*service.Account, error) {
+	s.createdAccounts = append(s.createdAccounts, input)
+	if s.createAccountErr != nil {
+		return nil, s.createAccountErr
+	}
+	account := service.Account{
+		ID:          s.nextID,
+		Name:        input.Name,
+		Platform:    input.Platform,
+		Type:        input.Type,
+		Status:      service.StatusActive,
+		Credentials: cloneCodexImportTestMap(input.Credentials),
+		Extra:       cloneCodexImportTestMap(input.Extra),
+	}
+	s.nextID++
+	s.accounts = append(s.accounts, account)
+	return &account, nil
+}
+
+func (s *codexImportMemoryAdminService) UpdateAccount(ctx context.Context, id int64, input *service.UpdateAccountInput) (*service.Account, error) {
+	s.updatedAccounts = append(s.updatedAccounts, struct {
+		id    int64
+		input *service.UpdateAccountInput
+	}{id: id, input: input})
+	if s.updateAccountErr != nil {
+		return nil, s.updateAccountErr
+	}
+	for idx := range s.accounts {
+		if s.accounts[idx].ID == id {
+			s.accounts[idx].Credentials = cloneCodexImportTestMap(input.Credentials)
+			s.accounts[idx].Extra = cloneCodexImportTestMap(input.Extra)
+			return &s.accounts[idx], nil
+		}
+	}
+	account := service.Account{ID: id, Status: service.StatusActive, Credentials: cloneCodexImportTestMap(input.Credentials)}
+	return &account, nil
+}
+
+func (s *codexImportMemoryAdminService) GetAccount(ctx context.Context, id int64) (*service.Account, error) {
+	for idx := range s.accounts {
+		if s.accounts[idx].ID == id {
+			return &s.accounts[idx], nil
+		}
+	}
+	return s.stubAdminService.GetAccount(ctx, id)
+}
+
+func buildCodexAccessOnlyImportValue(t *testing.T, accountID, userID string) map[string]any {
+	t.Helper()
+	return map[string]any{
+		"access_token": buildCodexAccessToken(t, accountID, userID, time.Now().Add(time.Hour)),
+	}
+}
+
+func buildCodexRefreshImportValue(t *testing.T, accountID, userID, refreshToken string) map[string]any {
+	t.Helper()
+	return map[string]any{
+		"access_token":  buildCodexAccessToken(t, accountID, userID, time.Now().Add(time.Hour)),
+		"refresh_token": refreshToken,
+	}
+}
+
+func buildCodexAccessToken(t *testing.T, accountID, userID string, exp time.Time) string {
+	t.Helper()
+	return buildCodexAccessTokenWithJTI(t, accountID, userID, "", exp)
+}
+
+func buildCodexAccessTokenWithJTI(t *testing.T, accountID, userID, jti string, exp time.Time) string {
+	t.Helper()
+	claims := map[string]any{
+		"sub": userID,
+		"https://api.openai.com/auth": map[string]any{
+			"chatgpt_account_id": accountID,
+		},
+	}
+	if jti != "" {
+		claims["jti"] = jti
+	}
+	return buildCodexImportTestJWT(t, exp, claims)
+}
+
+func cloneCodexImportTestMap(input map[string]any) map[string]any {
+	if input == nil {
+		return nil
+	}
+	out := make(map[string]any, len(input))
+	for key, value := range input {
+		out[key] = value
+	}
+	return out
+}
+
+func boolPtr(v bool) *bool {
+	return &v
+}
+
 func buildCodexImportTestJWT(t *testing.T, exp time.Time, extraClaims map[string]any) string {
 	t.Helper()
 	header := map[string]any{

From 4dd3aee5c22d34d0f4ef9c6c680b8d05e0218296 Mon Sep 17 00:00:00 2001
From: wucm667 <stevenwucongmin@gmail.com>
Date: Sat, 4 Jul 2026 10:04:42 +0800
Subject: [PATCH 15/28] fix(openai): use mapped billing model for responses

---
 .../openai_gateway_record_usage_test.go       |  60 ++++++++
 .../service/openai_gateway_service.go         |   4 +
 .../openai_gateway_service_hotpath_test.go    | 138 ++++++++++++++++++
 3 files changed, 202 insertions(+)

diff --git a/backend/internal/service/openai_gateway_record_usage_test.go b/backend/internal/service/openai_gateway_record_usage_test.go
index a7004f5d54..8323035ffe 100644
--- a/backend/internal/service/openai_gateway_record_usage_test.go
+++ b/backend/internal/service/openai_gateway_record_usage_test.go
@@ -1306,6 +1306,66 @@ func TestOpenAIGatewayServiceRecordUsage_ChannelMappedOverridesBillingModelWhenM
 	require.True(t, usageRepo.lastLog.ActualCost > 0, "cost must not be zero")
 }
 
+func TestOpenAIGatewayServiceRecordUsage_ResponsesMappedBillingModelHonorsBillingModelSource(t *testing.T) {
+	usage := OpenAIUsage{InputTokens: 20, OutputTokens: 10}
+	tokens := UsageTokens{InputTokens: 20, OutputTokens: 10}
+
+	tests := []struct {
+		name               string
+		billingModelSource string
+		wantBillingModel   string
+	}{
+		{
+			name:               "upstream uses mapped billing model",
+			billingModelSource: BillingModelSourceUpstream,
+			wantBillingModel:   "gpt-5.5",
+		},
+		{
+			name:               "requested overrides mapped billing model",
+			billingModelSource: BillingModelSourceRequested,
+			wantBillingModel:   "gpt-5.4",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			usageRepo := &openAIRecordUsageLogRepoStub{inserted: true}
+			userRepo := &openAIRecordUsageUserRepoStub{}
+			subRepo := &openAIRecordUsageSubRepoStub{}
+			svc := newOpenAIRecordUsageServiceForTest(usageRepo, userRepo, subRepo, nil)
+
+			expectedCost, err := svc.billingService.CalculateCost(tt.wantBillingModel, tokens, 1.1)
+			require.NoError(t, err)
+
+			err = svc.RecordUsage(context.Background(), &OpenAIRecordUsageInput{
+				Result: &OpenAIForwardResult{
+					RequestID:     "resp_mapped_billing_model_source",
+					Model:         "gpt-5.4",
+					BillingModel:  "gpt-5.5",
+					UpstreamModel: "gpt-5.5",
+					Usage:         usage,
+					Duration:      time.Second,
+				},
+				APIKey:  &APIKey{ID: 10},
+				User:    &User{ID: 20},
+				Account: &Account{ID: 30},
+				ChannelUsageFields: ChannelUsageFields{
+					OriginalModel:      "gpt-5.4",
+					ChannelMappedModel: "gpt-5.4",
+					BillingModelSource: tt.billingModelSource,
+				},
+			})
+
+			require.NoError(t, err)
+			require.NotNil(t, usageRepo.lastLog)
+			require.Equal(t, "gpt-5.4", usageRepo.lastLog.Model)
+			require.InDelta(t, expectedCost.ActualCost, usageRepo.lastLog.ActualCost, 1e-12)
+			require.InDelta(t, expectedCost.ActualCost, userRepo.lastAmount, 1e-12)
+			require.True(t, usageRepo.lastLog.ActualCost > 0, "cost must not be zero")
+		})
+	}
+}
+
 func TestOpenAIGatewayServiceRecordUsage_BillsCompactOpenAIModelAlias(t *testing.T) {
 	usageRepo := &openAIRecordUsageLogRepoStub{inserted: true}
 	userRepo := &openAIRecordUsageUserRepoStub{}
diff --git a/backend/internal/service/openai_gateway_service.go b/backend/internal/service/openai_gateway_service.go
index 6bf04ef20f..dd3d48aed0 100644
--- a/backend/internal/service/openai_gateway_service.go
+++ b/backend/internal/service/openai_gateway_service.go
@@ -3233,6 +3233,9 @@ func (s *OpenAIGatewayService) Forward(ctx context.Context, c *gin.Context, acco
 				wsAttempts,
 			)
 			wsResult.UpstreamModel = upstreamModel
+			if wsResult.BillingModel == "" {
+				wsResult.BillingModel = billingModel
+			}
 			if wsResult.ImageCount > 0 {
 				wsResult.ImageSize = imageSizeTier
 				wsResult.ImageInputSize = imageInputSize
@@ -3380,6 +3383,7 @@ func (s *OpenAIGatewayService) Forward(ctx context.Context, c *gin.Context, acco
 			ResponseID:      responseID,
 			Usage:           *usage,
 			Model:           originalModel,
+			BillingModel:    billingModel,
 			UpstreamModel:   upstreamModel,
 			ServiceTier:     serviceTier,
 			ReasoningEffort: reasoningEffort,
diff --git a/backend/internal/service/openai_gateway_service_hotpath_test.go b/backend/internal/service/openai_gateway_service_hotpath_test.go
index aee69fffa9..1dde60c9f0 100644
--- a/backend/internal/service/openai_gateway_service_hotpath_test.go
+++ b/backend/internal/service/openai_gateway_service_hotpath_test.go
@@ -196,6 +196,144 @@ func TestOpenAIGatewayService_Forward_MappedImageModelUsesImageGate(t *testing.T
 	require.Equal(t, http.StatusForbidden, rec.Code)
 }
 
+func TestOpenAIGatewayService_Forward_TextResponsesSetsBillingModelToMappedModel(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	upstream := &httpUpstreamRecorder{
+		resp: &http.Response{
+			StatusCode: http.StatusOK,
+			Header:     http.Header{"Content-Type": []string{"application/json"}, "x-request-id": []string{"rid_text_mapped_billing"}},
+			Body: io.NopCloser(strings.NewReader(
+				`{"id":"resp_text_mapped","object":"response","model":"gpt-5.5","status":"completed","output":[{"type":"message","role":"assistant","content":[{"type":"output_text","text":"ok"}]}],"usage":{"input_tokens":20,"output_tokens":10,"total_tokens":30}}`,
+			)),
+		},
+	}
+	cfg := &config.Config{}
+	cfg.Security.URLAllowlist.Enabled = false
+	svc := &OpenAIGatewayService{cfg: cfg, httpUpstream: upstream}
+	account := &Account{
+		ID:          4,
+		Name:        "openai-apikey",
+		Platform:    PlatformOpenAI,
+		Type:        AccountTypeAPIKey,
+		Concurrency: 1,
+		Credentials: map[string]any{
+			"api_key":       "sk-test",
+			"base_url":      "https://example.com",
+			"model_mapping": map[string]any{"gpt-5.4": "gpt-5.5"},
+		},
+		Extra: map[string]any{"use_responses_api": true},
+	}
+	rec := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(rec)
+	c.Request = httptest.NewRequest(http.MethodPost, "/openai/v1/responses", nil)
+	SetOpenAIClientTransport(c, OpenAIClientTransportHTTP)
+
+	body := []byte(`{"model":"gpt-5.4","stream":false,"input":"hello"}`)
+	result, err := svc.Forward(context.Background(), c, account, body)
+	require.NoError(t, err)
+	require.NotNil(t, result)
+	require.Equal(t, "gpt-5.4", result.Model)
+	require.Equal(t, "gpt-5.5", result.BillingModel)
+	require.Equal(t, "gpt-5.5", result.UpstreamModel)
+	require.Equal(t, "gpt-5.5", gjson.GetBytes(upstream.lastBody, "model").String())
+	require.Equal(t, 0, result.ImageCount)
+}
+
+func TestOpenAIGatewayService_Forward_TextResponsesWithoutMappingKeepsRequestedBillingModel(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	upstream := &httpUpstreamRecorder{
+		resp: &http.Response{
+			StatusCode: http.StatusOK,
+			Header:     http.Header{"Content-Type": []string{"application/json"}, "x-request-id": []string{"rid_text_unmapped_billing"}},
+			Body:       io.NopCloser(strings.NewReader(`{"id":"resp_text_unmapped","object":"response","model":"gpt-5.4","status":"completed","usage":{"input_tokens":20,"output_tokens":10,"total_tokens":30}}`)),
+		},
+	}
+	cfg := &config.Config{}
+	cfg.Security.URLAllowlist.Enabled = false
+	svc := &OpenAIGatewayService{cfg: cfg, httpUpstream: upstream}
+	account := &Account{
+		ID:          4,
+		Name:        "openai-apikey",
+		Platform:    PlatformOpenAI,
+		Type:        AccountTypeAPIKey,
+		Concurrency: 1,
+		Credentials: map[string]any{
+			"api_key":  "sk-test",
+			"base_url": "https://example.com",
+		},
+		Extra: map[string]any{"use_responses_api": true},
+	}
+	rec := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(rec)
+	c.Request = httptest.NewRequest(http.MethodPost, "/openai/v1/responses", nil)
+	SetOpenAIClientTransport(c, OpenAIClientTransportHTTP)
+
+	result, err := svc.Forward(context.Background(), c, account, []byte(`{"model":"gpt-5.4","stream":false,"input":"hello"}`))
+	require.NoError(t, err)
+	require.NotNil(t, result)
+	require.Equal(t, "gpt-5.4", result.Model)
+	require.Equal(t, "gpt-5.4", result.BillingModel)
+	require.Equal(t, "gpt-5.4", result.UpstreamModel)
+}
+
+func TestOpenAIGatewayService_Forward_TextResponsesBillingModelMatchesChatCompletions(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	cfg := &config.Config{}
+	cfg.Security.URLAllowlist.Enabled = false
+	account := &Account{
+		ID:          5,
+		Name:        "openai-apikey",
+		Platform:    PlatformOpenAI,
+		Type:        AccountTypeAPIKey,
+		Concurrency: 1,
+		Credentials: map[string]any{
+			"api_key":       "sk-test",
+			"base_url":      "https://example.com",
+			"model_mapping": map[string]any{"gpt-5.4": "gpt-5.5"},
+		},
+		Extra: map[string]any{"use_responses_api": true},
+	}
+
+	responsesUpstream := &httpUpstreamRecorder{
+		resp: &http.Response{
+			StatusCode: http.StatusOK,
+			Header:     http.Header{"Content-Type": []string{"application/json"}, "x-request-id": []string{"rid_responses_mapped_billing"}},
+			Body: io.NopCloser(strings.NewReader(
+				`{"id":"resp_native","object":"response","model":"gpt-5.5","status":"completed","output":[{"type":"message","role":"assistant","content":[{"type":"output_text","text":"ok"}]}],"usage":{"input_tokens":20,"output_tokens":10,"total_tokens":30}}`,
+			)),
+		},
+	}
+	responsesSvc := &OpenAIGatewayService{cfg: cfg, httpUpstream: responsesUpstream}
+	responsesRecorder := httptest.NewRecorder()
+	responsesCtx, _ := gin.CreateTestContext(responsesRecorder)
+	responsesCtx.Request = httptest.NewRequest(http.MethodPost, "/openai/v1/responses", nil)
+	SetOpenAIClientTransport(responsesCtx, OpenAIClientTransportHTTP)
+	responsesResult, err := responsesSvc.Forward(context.Background(), responsesCtx, account, []byte(`{"model":"gpt-5.4","stream":false,"input":"hello"}`))
+	require.NoError(t, err)
+	require.NotNil(t, responsesResult)
+
+	chatUpstream := &httpUpstreamRecorder{
+		resp: &http.Response{
+			StatusCode: http.StatusOK,
+			Header:     http.Header{"Content-Type": []string{"text/event-stream"}, "x-request-id": []string{"rid_chat_mapped_billing"}},
+			Body: io.NopCloser(strings.NewReader(
+				`data: {"type":"response.completed","response":{"id":"resp_chat","object":"response","model":"gpt-5.5","status":"completed","output":[{"type":"message","role":"assistant","content":[{"type":"output_text","text":"ok"}]}],"usage":{"input_tokens":20,"output_tokens":10,"total_tokens":30}}}` + "\n\n",
+			)),
+		},
+	}
+	chatSvc := &OpenAIGatewayService{cfg: cfg, httpUpstream: chatUpstream}
+	chatRecorder := httptest.NewRecorder()
+	chatCtx, _ := gin.CreateTestContext(chatRecorder)
+	chatCtx.Request = httptest.NewRequest(http.MethodPost, "/openai/v1/chat/completions", nil)
+	chatResult, err := chatSvc.ForwardAsChatCompletions(context.Background(), chatCtx, account, []byte(`{"model":"gpt-5.4","stream":false,"messages":[{"role":"user","content":"hello"}]}`), "", "")
+	require.NoError(t, err)
+	require.NotNil(t, chatResult)
+
+	require.Equal(t, chatResult.BillingModel, responsesResult.BillingModel)
+	require.Equal(t, "gpt-5.5", responsesResult.BillingModel)
+	require.Equal(t, "gpt-5.5", chatResult.BillingModel)
+}
+
 func TestOpenAIGatewayService_Forward_TextDataImageDoesNotForceMapMarshal(t *testing.T) {
 	gin.SetMode(gin.TestMode)
 	upstream := &httpUpstreamRecorder{

From dec709fb3afe807da87d91478c7373d0b4dd18ff Mon Sep 17 00:00:00 2001
From: shaw <dev@sub2api.org>
Date: Sat, 4 Jul 2026 10:06:11 +0800
Subject: [PATCH 16/28] chore: update sponsors

---
 README.md    | 4 ++--
 README_CN.md | 5 +++--
 README_JA.md | 4 ++--
 3 files changed, 7 insertions(+), 6 deletions(-)

diff --git a/README.md b/README.md
index 21ea27dee3..1ae4db33e6 100644
--- a/README.md
+++ b/README.md
@@ -26,7 +26,7 @@ Please read the following carefully before using this project:
 
 ## ❤️ Sponsors
 
-> [Want to appear here?](mailto:support@pincc.ai)
+> [Want to appear here?](mailto:support@sub2api.org)
 
 <table>
 
@@ -140,7 +140,7 @@ Model authenticity: no content intervention or secondary filtering — experienc
 
 <tr>
 <td width="180"><a href="https://www.proxy4free.com/?keyword=4yjqecpc"><img src="assets/partners/logos/proxy4free.png" alt="proxy4free" width="150"></a></td>
-<td>Proxy4Free is a data proxy service provider for developers and AI applications, offering residential proxies, static residential proxies, ISP proxies, and datacenter proxies for scenarios such as Web Scraping, Browser Automation, and AI Agents. With global IP resources, stable connections, and flexible switching, it helps developers improve data collection success rates and reduce the risk of IP bans. Register via <a href="https://www.proxy4free.com/?keyword=4yjqecpc">this link</a> to get started and easily build more stable and efficient automation workflows.
+<td>Thanks to Proxy4Free for sponsoring this project! Proxy4Free is a data proxy service provider for developers and AI applications, offering residential proxies, static residential proxies, ISP proxies, and datacenter proxies for scenarios such as Web Scraping, Browser Automation, and AI Agents. With global IP resources, stable connections, and flexible switching, it helps developers improve data collection success rates and reduce the risk of IP bans. Register via <a href="https://www.proxy4free.com/?keyword=4yjqecpc">this link</a> to get started and easily build more stable and efficient automation workflows.
 </td>
 </tr>
 
diff --git a/README_CN.md b/README_CN.md
index 2ba06a5299..c24003a8f2 100644
--- a/README_CN.md
+++ b/README_CN.md
@@ -27,7 +27,7 @@
 
 ## ❤️ 赞助商
 
-> [想出现在这里？](mailto:support@pincc.ai)
+> [想出现在这里？](mailto:support@sub2api.org)
 
 <table>
 
@@ -143,10 +143,11 @@
 
 <tr>
 <td width="180"><a href="https://www.proxy4free.com/?keyword=4yjqecpc"><img src="assets/partners/logos/proxy4free.png" alt="proxy4free" width="150"></a></td>
-<td>Proxy4Free 是面向开发者和 AI 应用的数据代理服务商，提供住宅代理、静态住宅代理、ISP 代理及数据中心代理等多种代理解决方案，适用于 Web Scraping、Browser Automation、AI Agent 等场景。支持全球 IP 资源、稳定连接与灵活切换，帮助开发者提升数据采集成功率，降低 IP 封禁风险。通过<a href="https://www.proxy4free.com/?keyword=4yjqecpc">此链接注册</a>即可开始体验，轻松构建更稳定、高效的自动化工作流。
+<td>感谢 Proxy4Free 赞助本项目！Proxy4Free 是面向开发者和 AI 应用的数据代理服务商，提供住宅代理、静态住宅代理、ISP 代理及数据中心代理等多种代理解决方案，适用于 Web Scraping、Browser Automation、AI Agent 等场景。支持全球 IP 资源、稳定连接与灵活切换，帮助开发者提升数据采集成功率，降低 IP 封禁风险。通过<a href="https://www.proxy4free.com/?keyword=4yjqecpc">此链接注册</a>即可开始体验，轻松构建更稳定、高效的自动化工作流。
 </td>
 </tr>
 
+
 </table>
 
 ## 项目概述
diff --git a/README_JA.md b/README_JA.md
index 3772b93957..1607310c9f 100644
--- a/README_JA.md
+++ b/README_JA.md
@@ -26,7 +26,7 @@
 
 ## ❤️ スポンサー
 
-> [こちらに掲載しませんか？](mailto:support@pincc.ai)
+> [こちらに掲載しませんか？](mailto:support@sub2api.org)
 
 <table>
 
@@ -138,7 +138,7 @@
 
 <tr>
 <td width="180"><a href="https://www.proxy4free.com/?keyword=4yjqecpc"><img src="assets/partners/logos/proxy4free.png" alt="proxy4free" width="150"></a></td>
-<td>Proxy4Free は開発者と AI アプリケーション向けのデータプロキシサービスプロバイダーで、住宅プロキシ、静的住宅プロキシ、ISP プロキシ、データセンタープロキシなど多様なプロキシソリューションを提供しており、Web Scraping、Browser Automation、AI Agent などのシナリオに適しています。グローバル IP リソース、安定した接続、柔軟な切り替えをサポートし、開発者のデータ収集成功率の向上と IP ブロックリスクの低減を支援します。<a href="https://www.proxy4free.com/?keyword=4yjqecpc">こちらのリンクから登録</a>して、より安定した効率的な自動化ワークフローを簡単に構築しましょう。
+<td>Proxy4Free のご支援に感謝します！Proxy4Free は開発者と AI アプリケーション向けのデータプロキシサービスプロバイダーで、住宅プロキシ、静的住宅プロキシ、ISP プロキシ、データセンタープロキシなど多様なプロキシソリューションを提供しており、Web Scraping、Browser Automation、AI Agent などのシナリオに適しています。グローバル IP リソース、安定した接続、柔軟な切り替えをサポートし、開発者のデータ収集成功率の向上と IP ブロックリスクの低減を支援します。<a href="https://www.proxy4free.com/?keyword=4yjqecpc">こちらのリンクから登録</a>して、より安定した効率的な自動化ワークフローを簡単に構築しましょう。
 </td>
 </tr>
 

From cbe0c4665fddabcdbf1862ddcf2cda34adffef4a Mon Sep 17 00:00:00 2001
From: shaw <dev@sub2api.org>
Date: Sat, 4 Jul 2026 10:18:41 +0800
Subject: [PATCH 17/28] chore: update readme

---
 README.md    | 17 +----------------
 README_CN.md | 15 +--------------
 README_JA.md | 17 +----------------
 3 files changed, 3 insertions(+), 46 deletions(-)

diff --git a/README.md b/README.md
index 1ae4db33e6..9aafad6264 100644
--- a/README.md
+++ b/README.md
@@ -23,6 +23,7 @@ Please read the following carefully before using this project:
 - **🚨 Terms of Service Risk**: Using this project may violate the terms of service of Anthropic and other upstream providers. Please review the relevant providers' user agreements before use; all risks arising from such use are borne solely by the user.
 - **⚖️ Compliant Use**: Use this project only in compliance with the laws and regulations of your country or region. Any unlawful use is strictly prohibited.
 - **📖 Disclaimer**: This project is provided for technical learning and research purposes only. The authors assume no liability for account bans, service interruptions, data loss, or any other direct or indirect damages resulting from the use of this project.
+- **🚫 No Commercial Authorization**: The developers of this project have never authorized any individual or organization to conduct any form of commercial operation based on this project. Any commercial activity conducted in the name of or based on this project is unrelated to this project and its developers, and all resulting disputes, losses, and legal liabilities shall be borne solely by the party conducting such activity.
 
 ## ❤️ Sponsors
 
@@ -691,12 +692,6 @@ Antigravity accounts support optional **hybrid scheduling**. When enabled, the g
 
 > **⚠️ Warning**: Anthropic Claude and Antigravity Claude **cannot be mixed within the same conversation context**. Use groups to isolate them properly.
 
-### Known Issues
-
-In Claude Code, Plan Mode cannot exit automatically. (Normally when using the native Claude API, after planning is complete, Claude Code will pop up options for users to approve or reject the plan.)
-
-**Workaround**: Press `Shift + Tab` to manually exit Plan Mode, then type your response to approve or reject the plan.
-
 ---
 
 ## Project Structure
@@ -727,16 +722,6 @@ sub2api/
     └── install.sh            # One-click installation script
 ```
 
-## Disclaimer
-
-> **Please read carefully before using this project:**
->
-> :rotating_light: **Terms of Service Risk**: Using this project may violate Anthropic's Terms of Service. Please read Anthropic's user agreement carefully before use. All risks arising from the use of this project are borne solely by the user.
->
-> :book: **Disclaimer**: This project is for technical learning and research purposes only. The author assumes no responsibility for account suspension, service interruption, or any other losses caused by the use of this project.
-
----
-
 ## Star History
 
 <a href="https://star-history.com/#Wei-Shaw/sub2api&Date">
diff --git a/README_CN.md b/README_CN.md
index c24003a8f2..1f8cac8b03 100644
--- a/README_CN.md
+++ b/README_CN.md
@@ -24,6 +24,7 @@
 - **🚨 服务条款风险**：使用本项目可能违反 Anthropic 等上游服务商的服务条款。请在使用前仔细阅读相关服务商的用户协议，由此产生的一切风险由用户自行承担。
 - **⚖️ 合规使用**：请在符合您所在国家或地区法律法规的前提下使用本项目，严禁将其用于任何违法违规用途。
 - **📖 免责声明**：本项目仅供技术学习与研究使用，作者不对因使用本项目导致的账户封禁、服务中断、数据丢失或其他任何直接或间接损失承担责任。
+- **🚫 无商业授权**：本项目从未授权任何个人或组织基于本项目开展任何形式的商业化运营。任何以本项目名义或基于本项目从事的商业行为均与本项目及其开发者无关，由此产生的一切纠纷、损失和法律责任由行为主体自行承担。
 
 ## ❤️ 赞助商
 
@@ -710,10 +711,6 @@ Antigravity 账户支持可选的**混合调度**功能。开启后，通用端
 
 > **⚠️ 注意**：Anthropic Claude 和 Antigravity Claude **不能在同一上下文中混合使用**，请通过分组功能做好隔离。
 
-
-### 已知问题
-在 Claude Code 中，无法自动退出Plan Mode。（正常使用原生Claude Api时，Plan 完成后，Claude Code会弹出弹出选项让用户同意或拒绝Plan。） 
-解决办法：shift + Tab，手动退出Plan mode，然后输入内容 告诉 Claude Code 同意或拒绝 Plan
 ---
 
 ## 项目结构
@@ -744,16 +741,6 @@ sub2api/
     └── install.sh            # 一键安装脚本
 ```
 
-## 免责声明
-
-> **使用本项目前请仔细阅读：**
->
-> :rotating_light: **服务条款风险**: 使用本项目可能违反 Anthropic 的服务条款。请在使用前仔细阅读 Anthropic 的用户协议，使用本项目的一切风险由用户自行承担。
->
-> :book: **免责声明**: 本项目仅供技术学习和研究使用，作者不对因使用本项目导致的账户封禁、服务中断或其他损失承担任何责任。
-
----
-
 ## Star History
 
 <a href="https://star-history.com/#Wei-Shaw/sub2api&Date">
diff --git a/README_JA.md b/README_JA.md
index 1607310c9f..684e685c8f 100644
--- a/README_JA.md
+++ b/README_JA.md
@@ -23,6 +23,7 @@
 - **🚨 利用規約のリスク**：本プロジェクトの使用は、Anthropic をはじめとする上流プロバイダーの利用規約に違反する可能性があります。ご利用前に各プロバイダーのユーザー規約を必ずご確認ください。使用により生じるすべてのリスクはユーザーご自身が負うものとします。
 - **⚖️ 法令遵守**：お住まいの国または地域の法令を遵守した上で本プロジェクトをご利用ください。いかなる違法な目的での使用も固く禁じます。
 - **📖 免責事項**：本プロジェクトは技術的な学習および研究の目的でのみ提供されます。本プロジェクトの使用により生じたアカウントの停止、サービスの中断、データの損失、その他一切の直接的または間接的な損害について、作者は一切の責任を負いません。
+- **🚫 商用利用の非許諾**：本プロジェクトの開発者は、いかなる個人または組織に対しても、本プロジェクトを利用したいかなる形態の商業運営も一切許諾していません。本プロジェクトの名義で、または本プロジェクトに基づいて行われる商業行為はすべて本プロジェクトおよびその開発者とは無関係であり、それにより生じる一切の紛争、損失、法的責任は行為者自身が負うものとします。
 
 ## ❤️ スポンサー
 
@@ -640,12 +641,6 @@ Antigravity アカウントはオプションの**ハイブリッドスケジュ
 
 > **⚠️ 警告**: Anthropic Claude と Antigravity Claude は**同じ会話コンテキスト内で混在させることはできません**。グループを使用して適切に分離してください。
 
-### 既知の問題
-
-Claude Code では、Plan Mode を自動的に終了できません。（通常、ネイティブの Claude API を使用する場合、計画が完了すると Claude Code はユーザーに計画を承認または拒否するオプションをポップアップ表示します。）
-
-**回避策**: `Shift + Tab` を押して手動で Plan Mode を終了し、計画を承認または拒否するためのレスポンスを入力してください。
-
 ---
 
 ## プロジェクト構成
@@ -676,16 +671,6 @@ sub2api/
     └── install.sh            # ワンクリックインストールスクリプト
 ```
 
-## 免責事項
-
-> **本プロジェクトをご利用の前に、以下をよくお読みください:**
->
-> :rotating_light: **利用規約違反のリスク**: 本プロジェクトの使用は Anthropic の利用規約に違反する可能性があります。使用前に Anthropic のユーザー契約をよくお読みください。本プロジェクトの使用に起因するすべてのリスクは、ユーザー自身が負うものとします。
->
-> :book: **免責事項**: 本プロジェクトは技術的な学習および研究目的のみで提供されています。作者は、本プロジェクトの使用によるアカウント停止、サービス中断、その他の損失について一切の責任を負いません。
-
----
-
 ## スター履歴
 
 <a href="https://star-history.com/#Wei-Shaw/sub2api&Date">

From b650bdd68d25bad3e502b2e34efe775555da2eba Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Sat, 4 Jul 2026 03:49:44 +0000
Subject: [PATCH 18/28] chore: sync VERSION to 0.1.144 [skip ci]

---
 backend/cmd/server/VERSION | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/backend/cmd/server/VERSION b/backend/cmd/server/VERSION
index 3170382e8b..3ef481c326 100644
--- a/backend/cmd/server/VERSION
+++ b/backend/cmd/server/VERSION
@@ -1 +1 @@
-0.1.143
+0.1.144

From 20008264febd36d442ff677093e2f42139b8f1bd Mon Sep 17 00:00:00 2001
From: feitianbubu <feitianbubu@qq.com>
Date: Sat, 4 Jul 2026 13:33:53 +0800
Subject: [PATCH 19/28] =?UTF-8?q?feat:=20=E7=82=B9=E5=87=BB=E4=BE=A7?=
 =?UTF-8?q?=E8=BE=B9=E6=A0=8F=20Logo/=E7=AB=99=E7=82=B9=E5=90=8D=E8=BF=94?=
 =?UTF-8?q?=E5=9B=9E=E9=A6=96=E9=A1=B5?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 frontend/src/components/layout/AppSidebar.vue | 18 ++++++++++++++----
 1 file changed, 14 insertions(+), 4 deletions(-)

diff --git a/frontend/src/components/layout/AppSidebar.vue b/frontend/src/components/layout/AppSidebar.vue
index 3d7f1604c7..4e11008832 100644
--- a/frontend/src/components/layout/AppSidebar.vue
+++ b/frontend/src/components/layout/AppSidebar.vue
@@ -9,13 +9,21 @@
     <!-- Logo/Brand -->
     <div class="sidebar-header" :class="{ 'sidebar-header-collapsed': sidebarCollapsed }">
       <!-- Custom Logo or Default Logo -->
-      <div class="sidebar-logo flex h-9 w-9 items-center justify-center overflow-hidden rounded-xl shadow-glow">
+      <router-link
+        :to="homePath"
+        class="sidebar-logo flex h-9 w-9 items-center justify-center overflow-hidden rounded-xl shadow-glow transition-opacity hover:opacity-80"
+        @click="handleMenuItemClick(homePath)"
+      >
         <img v-if="settingsLoaded" :src="siteLogo || '/logo.png'" alt="Logo" class="h-full w-full object-contain" />
-      </div>
+      </router-link>
       <div class="sidebar-brand" :class="{ 'sidebar-brand-collapsed': sidebarCollapsed }" :aria-hidden="sidebarCollapsed ? 'true' : 'false'">
-        <span class="sidebar-brand-title text-lg font-bold text-gray-900 dark:text-white">
+        <router-link
+          :to="homePath"
+          class="sidebar-brand-title text-lg font-bold text-gray-900 transition-colors hover:text-primary-600 dark:text-white dark:hover:text-primary-400"
+          @click="handleMenuItemClick(homePath)"
+        >
           {{ siteName }}
-        </span>
+        </router-link>
         <!-- Version Badge -->
         <VersionBadge :version="siteVersion" />
       </div>
@@ -238,6 +246,8 @@ const mobileOpen = computed(() => appStore.mobileOpen)
 const isAdmin = computed(() => authStore.isAdmin)
 const isDark = ref(document.documentElement.classList.contains('dark'))
 
+const homePath = computed(() => (isAdmin.value ? '/admin/dashboard' : '/dashboard'))
+
 // Track which parent nav groups are expanded
 const expandedGroups = ref<Set<string>>(new Set())
 

From 498f010ec34e95ef0070326cec9331254bfed7a8 Mon Sep 17 00:00:00 2001
From: shaw <dev@sub2api.org>
Date: Sat, 4 Jul 2026 13:51:37 +0800
Subject: [PATCH 20/28] =?UTF-8?q?fix(=E9=83=A8=E7=BD=B2):=20=E7=BB=9F?=
 =?UTF-8?q?=E4=B8=80=20Docker=20=E9=83=A8=E7=BD=B2=20URL=20=E5=AE=89?=
 =?UTF-8?q?=E5=85=A8=E9=BB=98=E8=AE=A4=E5=80=BC=E4=B8=BA=E5=BC=80=E5=8F=91?=
 =?UTF-8?q?=E5=8F=8B=E5=A5=BD=E6=A8=A1=E5=BC=8F?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

docker-compose.yml / docker-compose.local.yml 中
SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP 与
SECURITY_URL_ALLOWLIST_ALLOW_PRIVATE_HOSTS 的兜底值由 false 改为 true，
与代码默认值（0c7a58fc）保持一致，避免未配置 .env 的 Docker 部署
在测试账号连接时因 http base URL 报 "invalid url scheme: http"。

同步更新 README（三语）、.env.example、config.example.yaml 中
过时的"默认拒绝 HTTP"描述，改为默认允许并指导生产环境显式收紧。
---
 README.md                       | 8 ++++----
 README_CN.md                    | 8 ++++----
 README_JA.md                    | 8 ++++----
 deploy/.env.example             | 8 ++++----
 deploy/config.example.yaml      | 4 ++--
 deploy/docker-compose.local.yml | 8 ++++----
 deploy/docker-compose.yml       | 8 ++++----
 7 files changed, 26 insertions(+), 26 deletions(-)

diff --git a/README.md b/README.md
index 9aafad6264..eb24b0d3d6 100644
--- a/README.md
+++ b/README.md
@@ -523,20 +523,20 @@ Additional security-related options are available in `config.yaml`:
 
 **⚠️ Security Warning: HTTP URL Configuration**
 
-When `security.url_allowlist.enabled=false`, the system performs minimal URL validation by default, **rejecting HTTP URLs** and only allowing HTTPS. To allow HTTP URLs (e.g., for development or internal testing), you must explicitly set:
+When `security.url_allowlist.enabled=false`, the system performs minimal URL validation and **allows HTTP URLs by default** (dev-friendly mode; Docker Compose deployments use the same default). For production, explicitly tighten this to HTTPS-only:
 
 ```yaml
 security:
   url_allowlist:
     enabled: false                # Disable allowlist checks
-    allow_insecure_http: true     # Allow HTTP URLs (⚠️ INSECURE)
+    allow_insecure_http: false    # HTTPS only (recommended for production)
 ```
 
 **Or via environment variable:**
 
 ```bash
 SECURITY_URL_ALLOWLIST_ENABLED=false
-SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP=true
+SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP=false
 ```
 
 **Risks of allowing HTTP:**
@@ -550,7 +550,7 @@ SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP=true
 - ✅ Testing account connectivity before obtaining HTTPS
 - ❌ Production environments (use HTTPS only)
 
-**Example error without this setting:**
+**Example error for HTTP URLs when `allow_insecure_http: false` is set:**
 ```
 Invalid base URL: invalid url scheme: http
 ```
diff --git a/README_CN.md b/README_CN.md
index 1f8cac8b03..a93056282b 100644
--- a/README_CN.md
+++ b/README_CN.md
@@ -568,20 +568,20 @@ gateway:
 
 **⚠️ 安全警告：HTTP URL 配置**
 
-当 `security.url_allowlist.enabled=false` 时，系统默认执行最小 URL 校验，**拒绝 HTTP URL**，仅允许 HTTPS。要允许 HTTP URL（例如用于开发或内网测试），必须显式设置：
+当 `security.url_allowlist.enabled=false` 时，系统仅执行最小 URL 校验，且**默认允许 HTTP URL**（开发友好模式，Docker Compose 部署的默认值一致）。生产环境建议显式收紧为仅允许 HTTPS：
 
 ```yaml
 security:
   url_allowlist:
     enabled: false                # 禁用白名单检查
-    allow_insecure_http: true     # 允许 HTTP URL（⚠️ 不安全）
+    allow_insecure_http: false    # 仅允许 HTTPS（生产环境推荐）
 ```
 
 **或通过环境变量：**
 
 ```bash
 SECURITY_URL_ALLOWLIST_ENABLED=false
-SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP=true
+SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP=false
 ```
 
 **允许 HTTP 的风险：**
@@ -595,7 +595,7 @@ SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP=true
 - ✅ 获取 HTTPS 前测试账号连通性
 - ❌ 生产环境（仅使用 HTTPS）
 
-**未设置此项时的错误示例：**
+**设置 `allow_insecure_http: false` 后，HTTP URL 会返回如下错误：**
 ```
 Invalid base URL: invalid url scheme: http
 ```
diff --git a/README_JA.md b/README_JA.md
index 684e685c8f..bd154a9a6b 100644
--- a/README_JA.md
+++ b/README_JA.md
@@ -521,20 +521,20 @@ default:
 
 **⚠️ セキュリティ警告: HTTP URL 設定**
 
-`security.url_allowlist.enabled=false` の場合、システムはデフォルトで最小限の URL バリデーションを行い、**HTTP URL を拒否**して HTTPS のみを許可します。HTTP URL を許可するには（開発環境や内部テスト用など）、以下を明示的に設定する必要があります:
+`security.url_allowlist.enabled=false` の場合、システムは最小限の URL バリデーションのみを行い、**デフォルトで HTTP URL を許可**します（開発フレンドリーモード。Docker Compose デプロイのデフォルトも同じです）。本番環境では、以下のように明示的に HTTPS のみに制限することを推奨します:
 
 ```yaml
 security:
   url_allowlist:
     enabled: false                # 許可リストチェックを無効化
-    allow_insecure_http: true     # HTTP URL を許可（⚠️ セキュリティリスクあり）
+    allow_insecure_http: false    # HTTPS のみ許可（本番環境推奨）
 ```
 
 **または環境変数で設定:**
 
 ```bash
 SECURITY_URL_ALLOWLIST_ENABLED=false
-SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP=true
+SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP=false
 ```
 
 **HTTP を許可するリスク:**
@@ -548,7 +548,7 @@ SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP=true
 - ✅ HTTPS 取得前のアカウント接続テスト
 - ❌ 本番環境（HTTPS のみを使用）
 
-**この設定なしで表示されるエラー例:**
+**`allow_insecure_http: false` 設定時に HTTP URL で表示されるエラー例:**
 ```
 Invalid base URL: invalid url scheme: http
 ```
diff --git a/deploy/.env.example b/deploy/.env.example
index d8892dbecf..5925f0abb4 100644
--- a/deploy/.env.example
+++ b/deploy/.env.example
@@ -359,11 +359,11 @@ DASHBOARD_AGGREGATION_RETENTION_DAILY_DAYS=730
 # 启用 URL 白名单验证（false 则跳过白名单检查，仅做基本格式校验）
 SECURITY_URL_ALLOWLIST_ENABLED=false
 
-# 关闭白名单时，是否允许 http:// URL（默认 false，只允许 https://）
-# ⚠️ 警告：允许 HTTP 存在安全风险（明文传输），仅建议在开发/测试环境或可信内网中使用
-# Allow insecure HTTP URLs when allowlist is disabled (default: false, requires https)
+# 关闭白名单时，是否允许 http:// URL（默认 true，设为 false 则只允许 https://）
+# ⚠️ 警告：允许 HTTP 存在安全风险（明文传输），生产环境建议设为 false
+# Allow insecure HTTP URLs when allowlist is disabled (default: true; set to false to require https)
 # ⚠️ WARNING: Allowing HTTP has security risks (plaintext transmission)
-#             Only recommended for dev/test environments or trusted networks
+#             Recommended to set false in production
 SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP=true
 
 # 是否允许本地/私有 IP 地址用于上游/定价/CRS（仅在可信网络中使用）
diff --git a/deploy/config.example.yaml b/deploy/config.example.yaml
index 5a3afb0315..eff4bfb598 100644
--- a/deploy/config.example.yaml
+++ b/deploy/config.example.yaml
@@ -108,8 +108,8 @@ security:
     # Allow localhost/private IPs for upstream/pricing/CRS (use only in trusted networks)
     # 允许本地/私有 IP 地址用于上游/定价/CRS（仅在可信网络中使用）
     allow_private_hosts: true
-    # Allow http:// URLs when allowlist is disabled (default: false, require https)
-    # 白名单禁用时是否允许 http:// URL（默认: false，要求 https）
+    # Allow http:// URLs when allowlist is disabled (default: true; set to false to require https)
+    # 白名单禁用时是否允许 http:// URL（默认: true，设为 false 则仅允许 https）
     allow_insecure_http: true
   response_headers:
     # Enable configurable response header filtering (default: true)
diff --git a/deploy/docker-compose.local.yml b/deploy/docker-compose.local.yml
index 21f46e3760..042752e857 100644
--- a/deploy/docker-compose.local.yml
+++ b/deploy/docker-compose.local.yml
@@ -139,10 +139,10 @@ services:
       # =======================================================================
       # Enable URL allowlist validation (false to skip allowlist checks)
       - SECURITY_URL_ALLOWLIST_ENABLED=${SECURITY_URL_ALLOWLIST_ENABLED:-false}
-      # Allow insecure HTTP URLs when allowlist is disabled (default: false, requires https)
-      - SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP=${SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP:-false}
-      # Allow private IP addresses for upstream/pricing/CRS (for internal deployments)
-      - SECURITY_URL_ALLOWLIST_ALLOW_PRIVATE_HOSTS=${SECURITY_URL_ALLOWLIST_ALLOW_PRIVATE_HOSTS:-false}
+      # Allow insecure HTTP URLs when allowlist is disabled (default: true; set to false to require https)
+      - SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP=${SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP:-true}
+      # Allow private IP addresses for upstream/pricing/CRS (default: true; set to false to block private hosts)
+      - SECURITY_URL_ALLOWLIST_ALLOW_PRIVATE_HOSTS=${SECURITY_URL_ALLOWLIST_ALLOW_PRIVATE_HOSTS:-true}
       # Upstream hosts whitelist (comma-separated, only used when enabled=true)
       - SECURITY_URL_ALLOWLIST_UPSTREAM_HOSTS=${SECURITY_URL_ALLOWLIST_UPSTREAM_HOSTS:-}
 
diff --git a/deploy/docker-compose.yml b/deploy/docker-compose.yml
index 3da3bcbfb0..22713c59aa 100644
--- a/deploy/docker-compose.yml
+++ b/deploy/docker-compose.yml
@@ -135,10 +135,10 @@ services:
       # =======================================================================
       # Enable URL allowlist validation (false to skip allowlist checks)
       - SECURITY_URL_ALLOWLIST_ENABLED=${SECURITY_URL_ALLOWLIST_ENABLED:-false}
-      # Allow insecure HTTP URLs when allowlist is disabled (default: false, requires https)
-      - SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP=${SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP:-false}
-      # Allow private IP addresses for upstream/pricing/CRS (for internal deployments)
-      - SECURITY_URL_ALLOWLIST_ALLOW_PRIVATE_HOSTS=${SECURITY_URL_ALLOWLIST_ALLOW_PRIVATE_HOSTS:-false}
+      # Allow insecure HTTP URLs when allowlist is disabled (default: true; set to false to require https)
+      - SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP=${SECURITY_URL_ALLOWLIST_ALLOW_INSECURE_HTTP:-true}
+      # Allow private IP addresses for upstream/pricing/CRS (default: true; set to false to block private hosts)
+      - SECURITY_URL_ALLOWLIST_ALLOW_PRIVATE_HOSTS=${SECURITY_URL_ALLOWLIST_ALLOW_PRIVATE_HOSTS:-true}
       # Upstream hosts whitelist (comma-separated, only used when enabled=true)
       - SECURITY_URL_ALLOWLIST_UPSTREAM_HOSTS=${SECURITY_URL_ALLOWLIST_UPSTREAM_HOSTS:-}
 

From e76e0499d75b6359abadda4d0ec787854d976960 Mon Sep 17 00:00:00 2001
From: MoRanHuiShou-pug <131538457+MoRanHuiShou-pug@users.noreply.github.com>
Date: Sun, 5 Jul 2026 03:55:40 +0800
Subject: [PATCH 21/28] fix: sanitize payment response NUL bytes

---
 backend/internal/service/payment_order.go     | 17 +++++++++
 .../service/payment_order_result_test.go      | 35 +++++++++++++++++++
 2 files changed, 52 insertions(+)

diff --git a/backend/internal/service/payment_order.go b/backend/internal/service/payment_order.go
index 29fe40b1b6..154159b932 100644
--- a/backend/internal/service/payment_order.go
+++ b/backend/internal/service/payment_order.go
@@ -453,6 +453,7 @@ func (s *PaymentService) invokeProvider(ctx context.Context, order *dbent.Paymen
 		}
 		return nil, classifyCreatePaymentError(req, sel.ProviderKey, err)
 	}
+	sanitizeCreatePaymentResponseDetails(pr)
 	_, err = s.entClient.PaymentOrder.UpdateOneID(order.ID).
 		SetNillablePaymentTradeNo(psNilIfEmpty(pr.TradeNo)).
 		SetNillablePayURL(psNilIfEmpty(pr.PayURL)).
@@ -480,6 +481,22 @@ func (s *PaymentService) invokeProvider(ctx context.Context, order *dbent.Paymen
 	return resp, nil
 }
 
+func sanitizeCreatePaymentResponseDetails(pr *payment.CreatePaymentResponse) {
+	if pr == nil {
+		return
+	}
+	pr.TradeNo = removePostgresTextNUL(pr.TradeNo)
+	pr.PayURL = removePostgresTextNUL(pr.PayURL)
+	pr.QRCode = removePostgresTextNUL(pr.QRCode)
+}
+
+func removePostgresTextNUL(value string) string {
+	if !strings.ContainsRune(value, 0) {
+		return value
+	}
+	return strings.ReplaceAll(value, "\x00", "")
+}
+
 func buildProviderCreatePaymentRequest(req CreateOrderRequest, sel *payment.InstanceSelection, orderID, amount, subject string) payment.CreatePaymentRequest {
 	return payment.CreatePaymentRequest{
 		OrderID:            orderID,
diff --git a/backend/internal/service/payment_order_result_test.go b/backend/internal/service/payment_order_result_test.go
index b7545ee45d..14192bd6cd 100644
--- a/backend/internal/service/payment_order_result_test.go
+++ b/backend/internal/service/payment_order_result_test.go
@@ -91,6 +91,41 @@ func TestBuildCreateOrderResponseCopiesJSAPIPayload(t *testing.T) {
 	}
 }
 
+func TestSanitizeCreatePaymentResponseDetailsRemovesNULBytes(t *testing.T) {
+	t.Parallel()
+
+	resp := &payment.CreatePaymentResponse{
+		TradeNo:      "trade\x00-no",
+		PayURL:       "https://pay.example.com/\x00checkout",
+		QRCode:       "wxp://payment-token\x00",
+		ClientSecret: "secret\x00unchanged",
+	}
+
+	sanitizeCreatePaymentResponseDetails(resp)
+
+	if strings.ContainsRune(resp.TradeNo, 0) {
+		t.Fatalf("trade_no still contains NUL: %q", resp.TradeNo)
+	}
+	if strings.ContainsRune(resp.PayURL, 0) {
+		t.Fatalf("pay_url still contains NUL: %q", resp.PayURL)
+	}
+	if strings.ContainsRune(resp.QRCode, 0) {
+		t.Fatalf("qr_code still contains NUL: %q", resp.QRCode)
+	}
+	if resp.TradeNo != "trade-no" {
+		t.Fatalf("trade_no = %q, want trade-no", resp.TradeNo)
+	}
+	if resp.PayURL != "https://pay.example.com/checkout" {
+		t.Fatalf("pay_url = %q, want sanitized URL", resp.PayURL)
+	}
+	if resp.QRCode != "wxp://payment-token" {
+		t.Fatalf("qr_code = %q, want sanitized QR code", resp.QRCode)
+	}
+	if resp.ClientSecret != "secret\x00unchanged" {
+		t.Fatalf("client_secret = %q, should not be touched by payment detail sanitization", resp.ClientSecret)
+	}
+}
+
 func TestValidateSelectedCreateOrderAmountCurrencyRejectsFractionalZeroDecimal(t *testing.T) {
 	t.Parallel()
 

From b23475ac0c553089d37adaf53f750c54fa0224e4 Mon Sep 17 00:00:00 2001
From: wucm667 <stevenwucongmin@gmail.com>
Date: Sun, 5 Jul 2026 08:34:34 +0800
Subject: [PATCH 22/28] fix(antigravity): refresh server-invalidated tokens

---
 .../service/antigravity_token_refresher.go    |  30 ++++
 backend/internal/service/ratelimit_service.go |  14 ++
 .../service/ratelimit_service_401_test.go     |  13 ++
 .../internal/service/token_refresh_service.go |  27 ++++
 .../service/token_refresh_service_test.go     | 133 ++++++++++++++++++
 5 files changed, 217 insertions(+)

diff --git a/backend/internal/service/antigravity_token_refresher.go b/backend/internal/service/antigravity_token_refresher.go
index 7ce0ccf0fe..f40df7d95f 100644
--- a/backend/internal/service/antigravity_token_refresher.go
+++ b/backend/internal/service/antigravity_token_refresher.go
@@ -12,6 +12,10 @@ const (
 	// antigravityRefreshWindow Antigravity token 提前刷新窗口：15分钟
 	// Google OAuth token 有效期55分钟，提前15分钟刷新
 	antigravityRefreshWindow = 15 * time.Minute
+
+	antigravityForceTokenRefreshExtraKey       = "antigravity_force_token_refresh"
+	antigravityForceTokenRefreshReasonExtraKey = "antigravity_force_token_refresh_reason"
+	antigravityForceTokenRefreshAtExtraKey     = "antigravity_force_token_refresh_at"
 )
 
 // AntigravityTokenRefresher 实现 TokenRefresher 接口
@@ -41,6 +45,9 @@ func (r *AntigravityTokenRefresher) NeedsRefresh(account *Account, _ time.Durati
 	if !r.CanRefresh(account) {
 		return false
 	}
+	if accountNeedsAntigravityForceTokenRefresh(account) {
+		return true
+	}
 	expiresAt := account.GetCredentialAsTime("expires_at")
 	if expiresAt == nil {
 		return false
@@ -54,6 +61,29 @@ func (r *AntigravityTokenRefresher) NeedsRefresh(account *Account, _ time.Durati
 	return needsRefresh
 }
 
+func accountNeedsAntigravityForceTokenRefresh(account *Account) bool {
+	return account != nil &&
+		account.Platform == PlatformAntigravity &&
+		account.Type == AccountTypeOAuth &&
+		account.getExtraBool(antigravityForceTokenRefreshExtraKey)
+}
+
+func antigravityForceTokenRefreshExtra(reason string) map[string]any {
+	return map[string]any{
+		antigravityForceTokenRefreshExtraKey:       true,
+		antigravityForceTokenRefreshReasonExtraKey: reason,
+		antigravityForceTokenRefreshAtExtraKey:     time.Now().UTC().Format(time.RFC3339),
+	}
+}
+
+func clearAntigravityForceTokenRefreshExtra() map[string]any {
+	return map[string]any{
+		antigravityForceTokenRefreshExtraKey:       false,
+		antigravityForceTokenRefreshReasonExtraKey: "",
+		antigravityForceTokenRefreshAtExtraKey:     "",
+	}
+}
+
 // Refresh 执行 token 刷新
 func (r *AntigravityTokenRefresher) Refresh(ctx context.Context, account *Account) (map[string]any, error) {
 	tokenInfo, err := r.antigravityOAuthService.RefreshAccountToken(ctx, account)
diff --git a/backend/internal/service/ratelimit_service.go b/backend/internal/service/ratelimit_service.go
index 2b9c85a2fa..81a915be67 100644
--- a/backend/internal/service/ratelimit_service.go
+++ b/backend/internal/service/ratelimit_service.go
@@ -292,6 +292,20 @@ func (s *RateLimitService) HandleUpstreamError(ctx context.Context, account *Acc
 			if upstreamMsg != "" {
 				msg = "OAuth 401: " + upstreamMsg
 			}
+			if authAccount.Platform == PlatformAntigravity {
+				extraUpdates := antigravityForceTokenRefreshExtra("401_invalid")
+				if err := s.accountRepo.UpdateExtra(ctx, authAccount.ID, extraUpdates); err != nil {
+					slog.Warn("antigravity_401_force_refresh_mark_failed", "account_id", authAccount.ID, "error", err)
+				} else {
+					if authAccount.Extra == nil {
+						authAccount.Extra = make(map[string]any, len(extraUpdates))
+					}
+					for k, v := range extraUpdates {
+						authAccount.Extra[k] = v
+					}
+					slog.Info("antigravity_401_force_refresh_marked", "account_id", authAccount.ID)
+				}
+			}
 			cooldownMinutes := s.cfg.RateLimit.OAuth401CooldownMinutes
 			if cooldownMinutes <= 0 {
 				cooldownMinutes = 10
diff --git a/backend/internal/service/ratelimit_service_401_test.go b/backend/internal/service/ratelimit_service_401_test.go
index 09afb9314b..48e6a41def 100644
--- a/backend/internal/service/ratelimit_service_401_test.go
+++ b/backend/internal/service/ratelimit_service_401_test.go
@@ -18,7 +18,9 @@ type rateLimitAccountRepoStub struct {
 	setErrorCalls          int
 	tempCalls              int
 	updateCredentialsCalls int
+	updateExtraCalls       int
 	lastCredentials        map[string]any
+	lastExtraUpdates       map[string]any
 	lastErrorMsg           string
 	lastTempReason         string
 	lastErrorID            int64
@@ -45,6 +47,12 @@ func (r *rateLimitAccountRepoStub) UpdateCredentials(ctx context.Context, id int
 	return nil
 }
 
+func (r *rateLimitAccountRepoStub) UpdateExtra(ctx context.Context, id int64, updates map[string]any) error {
+	r.updateExtraCalls++
+	r.lastExtraUpdates = shallowCopyMap(updates)
+	return nil
+}
+
 type tokenCacheInvalidatorRecorder struct {
 	accounts []*Account
 	err      error
@@ -133,6 +141,10 @@ func TestRateLimitService_HandleUpstreamError_OAuth401SetsTempUnschedulable(t *t
 		require.Equal(t, 1, repo.tempCalls)
 		require.Equal(t, int64(100), repo.lastTempID)
 		require.Contains(t, repo.lastTempReason, "invalid or expired credentials")
+		require.Equal(t, 1, repo.updateExtraCalls)
+		require.Equal(t, true, repo.lastExtraUpdates[antigravityForceTokenRefreshExtraKey])
+		require.Equal(t, "401_invalid", repo.lastExtraUpdates[antigravityForceTokenRefreshReasonExtraKey])
+		require.Equal(t, true, account.Extra[antigravityForceTokenRefreshExtraKey])
 		require.Len(t, invalidator.accounts, 1)
 		require.Equal(t, int64(100), invalidator.accounts[0].ID)
 	})
@@ -245,6 +257,7 @@ func TestRateLimitService_HandleUpstreamError_OAuth401DoesNotOverwriteCredential
 
 	require.True(t, shouldDisable)
 	require.Equal(t, 0, repo.updateCredentialsCalls, "401 handler must not write credentials back from the request-start snapshot")
+	require.Equal(t, 0, repo.updateExtraCalls, "OpenAI 401 must not set Antigravity force-refresh marker")
 	require.Equal(t, 1, repo.tempCalls, "401 handler should still set temp-unschedulable cooldown")
 	require.Nil(t, repo.lastCredentials, "no credentials should have been persisted")
 }
diff --git a/backend/internal/service/token_refresh_service.go b/backend/internal/service/token_refresh_service.go
index 9ca369a52a..2d4a026692 100644
--- a/backend/internal/service/token_refresh_service.go
+++ b/backend/internal/service/token_refresh_service.go
@@ -312,6 +312,7 @@ func (s *TokenRefreshService) refreshWithRetry(ctx context.Context, account *Acc
 		if isNonRetryableRefreshError(err) {
 			errorMsg := "Token refresh failed (non-retryable): " + logredact.RedactText(err.Error())
 			s.notifyAccountSchedulingBlocked(account, time.Time{}, "token_refresh_non_retryable")
+			s.clearAntigravityForceTokenRefresh(ctx, account, "non_retryable")
 			if setErr := s.accountRepo.SetError(ctx, account.ID, errorMsg); setErr != nil {
 				slog.Error("token_refresh.set_error_status_failed",
 					"account_id", account.ID,
@@ -369,6 +370,8 @@ func (s *TokenRefreshService) refreshWithRetry(ctx context.Context, account *Acc
 
 // postRefreshActions 刷新成功后的后续动作（清除错误状态、缓存失效、调度器同步等）
 func (s *TokenRefreshService) postRefreshActions(ctx context.Context, account *Account) {
+	s.clearAntigravityForceTokenRefresh(ctx, account, "success")
+
 	// Antigravity 账户：如果之前是因为缺少 project_id 而标记为 error，现在成功获取到了，清除错误状态
 	if account.Platform == PlatformAntigravity &&
 		account.Status == StatusError &&
@@ -432,6 +435,30 @@ func (s *TokenRefreshService) postRefreshActions(ctx context.Context, account *A
 	s.ensureAntigravityPrivacy(ctx, account)
 }
 
+func (s *TokenRefreshService) clearAntigravityForceTokenRefresh(ctx context.Context, account *Account, outcome string) {
+	if s == nil || account == nil || !accountNeedsAntigravityForceTokenRefresh(account) {
+		return
+	}
+	updates := clearAntigravityForceTokenRefreshExtra()
+	if err := s.accountRepo.UpdateExtra(ctx, account.ID, updates); err != nil {
+		slog.Warn("token_refresh.clear_antigravity_force_refresh_failed",
+			"account_id", account.ID,
+			"outcome", outcome,
+			"error", err,
+		)
+		return
+	}
+	if account.Extra != nil {
+		for k, v := range updates {
+			account.Extra[k] = v
+		}
+	}
+	slog.Info("token_refresh.cleared_antigravity_force_refresh",
+		"account_id", account.ID,
+		"outcome", outcome,
+	)
+}
+
 // errRefreshSkipped 表示刷新被跳过（锁竞争或已被其他路径刷新），不计入 failed 或 refreshed
 var errRefreshSkipped = fmt.Errorf("refresh skipped")
 
diff --git a/backend/internal/service/token_refresh_service_test.go b/backend/internal/service/token_refresh_service_test.go
index 6f227352f9..bd6a521576 100644
--- a/backend/internal/service/token_refresh_service_test.go
+++ b/backend/internal/service/token_refresh_service_test.go
@@ -20,8 +20,10 @@ type tokenRefreshAccountRepo struct {
 	setErrorCalls          int
 	clearTempCalls         int
 	setTempUnschedCalls    int
+	updateExtraCalls       int
 	lastErrorMessage       string
 	lastTempUnschedReason  string
+	lastExtraUpdates       map[string]any
 	lastAccount            *Account
 	updateErr              error
 }
@@ -68,6 +70,22 @@ func (r *tokenRefreshAccountRepo) SetTempUnschedulable(ctx context.Context, id i
 	return nil
 }
 
+func (r *tokenRefreshAccountRepo) UpdateExtra(ctx context.Context, id int64, updates map[string]any) error {
+	r.updateExtraCalls++
+	r.lastExtraUpdates = shallowCopyMap(updates)
+	if r.accountsByID != nil {
+		if acc, ok := r.accountsByID[id]; ok && acc != nil {
+			if acc.Extra == nil {
+				acc.Extra = make(map[string]any, len(updates))
+			}
+			for k, v := range updates {
+				acc.Extra[k] = v
+			}
+		}
+	}
+	return nil
+}
+
 type tokenCacheInvalidatorStub struct {
 	calls int
 	err   error
@@ -233,6 +251,121 @@ func TestTokenRefreshService_RefreshWithRetry_Antigravity(t *testing.T) {
 	require.Equal(t, 1, invalidator.calls) // Antigravity 也应触发缓存失效
 }
 
+func TestAntigravityTokenRefresher_NeedsRefresh_ForceRefreshMarker(t *testing.T) {
+	refresher := NewAntigravityTokenRefresher(nil)
+	account := &Account{
+		ID:       3675,
+		Platform: PlatformAntigravity,
+		Type:     AccountTypeOAuth,
+		Credentials: map[string]any{
+			"expires_at": time.Now().Add(time.Hour).Format(time.RFC3339),
+		},
+		Extra: map[string]any{
+			antigravityForceTokenRefreshExtraKey: true,
+		},
+	}
+
+	require.True(t, refresher.NeedsRefresh(account, 0), "server-invalidated token must refresh even before expires_at")
+}
+
+func TestAntigravityTokenRefresher_NeedsRefresh_NormalExpiryRulesUnchanged(t *testing.T) {
+	refresher := NewAntigravityTokenRefresher(nil)
+
+	t.Run("normal_unexpired_without_marker_does_not_refresh", func(t *testing.T) {
+		account := &Account{
+			ID:       3707,
+			Platform: PlatformAntigravity,
+			Type:     AccountTypeOAuth,
+			Credentials: map[string]any{
+				"expires_at": time.Now().Add(time.Hour).Format(time.RFC3339),
+			},
+		}
+
+		require.False(t, refresher.NeedsRefresh(account, 0))
+	})
+
+	t.Run("normal_expiring_refreshes", func(t *testing.T) {
+		account := &Account{
+			ID:       3708,
+			Platform: PlatformAntigravity,
+			Type:     AccountTypeOAuth,
+			Credentials: map[string]any{
+				"expires_at": time.Now().Add(5 * time.Minute).Format(time.RFC3339),
+			},
+		}
+
+		require.True(t, refresher.NeedsRefresh(account, 0))
+	})
+}
+
+func TestTokenRefreshService_RefreshWithRetry_AntigravityClearsForceRefreshOnSuccess(t *testing.T) {
+	repo := &tokenRefreshAccountRepo{}
+	cfg := &config.Config{
+		TokenRefresh: config.TokenRefreshConfig{
+			MaxRetries:          1,
+			RetryBackoffSeconds: 0,
+		},
+	}
+	service := NewTokenRefreshService(repo, nil, nil, nil, nil, nil, nil, cfg, nil)
+	until := time.Now().Add(10 * time.Minute)
+	account := &Account{
+		ID:                     3709,
+		Platform:               PlatformAntigravity,
+		Type:                   AccountTypeOAuth,
+		TempUnschedulableUntil: &until,
+		Extra: map[string]any{
+			antigravityForceTokenRefreshExtraKey:       true,
+			antigravityForceTokenRefreshReasonExtraKey: "401_invalid",
+			"privacy_mode": AntigravityPrivacySet,
+		},
+	}
+	refresher := &tokenRefresherStub{
+		credentials: map[string]any{
+			"access_token": "new-ag-token",
+		},
+	}
+
+	err := service.refreshWithRetry(context.Background(), account, refresher, refresher, time.Hour)
+	require.NoError(t, err)
+	require.Equal(t, 1, repo.updateCredentialsCalls)
+	require.Equal(t, 1, repo.updateExtraCalls)
+	require.Equal(t, false, repo.lastExtraUpdates[antigravityForceTokenRefreshExtraKey])
+	require.Equal(t, "", repo.lastExtraUpdates[antigravityForceTokenRefreshReasonExtraKey])
+	require.Equal(t, false, account.Extra[antigravityForceTokenRefreshExtraKey])
+	require.Equal(t, 1, repo.clearTempCalls, "successful refresh should restore schedulability")
+}
+
+func TestTokenRefreshService_RefreshWithRetry_AntigravityForceRefreshInvalidGrantSetsError(t *testing.T) {
+	repo := &tokenRefreshAccountRepo{}
+	cfg := &config.Config{
+		TokenRefresh: config.TokenRefreshConfig{
+			MaxRetries:          3,
+			RetryBackoffSeconds: 0,
+		},
+	}
+	service := NewTokenRefreshService(repo, nil, nil, nil, nil, nil, nil, cfg, nil)
+	account := &Account{
+		ID:       3710,
+		Platform: PlatformAntigravity,
+		Type:     AccountTypeOAuth,
+		Extra: map[string]any{
+			antigravityForceTokenRefreshExtraKey:       true,
+			antigravityForceTokenRefreshReasonExtraKey: "401_invalid",
+		},
+	}
+	refresher := &tokenRefresherStub{
+		err: errors.New("invalid_grant: token revoked"),
+	}
+
+	err := service.refreshWithRetry(context.Background(), account, refresher, refresher, time.Hour)
+	require.Error(t, err)
+	require.Equal(t, 1, repo.setErrorCalls)
+	require.Equal(t, 0, repo.setTempUnschedCalls)
+	require.Equal(t, 1, repo.updateExtraCalls)
+	require.Equal(t, false, repo.lastExtraUpdates[antigravityForceTokenRefreshExtraKey])
+	require.Contains(t, repo.lastErrorMessage, "non-retryable")
+}
+
 // TestTokenRefreshService_RefreshWithRetry_NonOAuthAccount 测试非 OAuth 账号不触发缓存失效
 func TestTokenRefreshService_RefreshWithRetry_NonOAuthAccount(t *testing.T) {
 	repo := &tokenRefreshAccountRepo{}

From 41cdd438d79db0d5cde749abc4775f7f0ff9427f Mon Sep 17 00:00:00 2001
From: wucm667 <stevenwucongmin@gmail.com>
Date: Sun, 5 Jul 2026 08:35:16 +0800
Subject: [PATCH 23/28] fix(gateway): honor Anthropic custom models list

---
 backend/internal/handler/gateway_handler.go   |  38 ++++-
 .../internal/handler/gateway_models_test.go   | 143 ++++++++++++++++++
 2 files changed, 180 insertions(+), 1 deletion(-)

diff --git a/backend/internal/handler/gateway_handler.go b/backend/internal/handler/gateway_handler.go
index b65dedf9cc..b20d9ef652 100644
--- a/backend/internal/handler/gateway_handler.go
+++ b/backend/internal/handler/gateway_handler.go
@@ -1006,7 +1006,8 @@ func (h *GatewayHandler) Models(c *gin.Context) {
 	// Get available models from account configurations for the selected group platform.
 	availableModels := h.gatewayService.GetAvailableModels(c.Request.Context(), groupID, platform)
 	if apiKey != nil && apiKey.Group != nil && apiKey.Group.CustomModelsListEnabled() {
-		availableModels = filterModelsByCustomList(availableModels, defaultModelIDsForPlatform(platform), apiKey.Group.ModelsListConfig.Models)
+		fallbackModels := defaultModelIDsForPlatform(platform)
+		availableModels = filterModelsByCustomList(customModelsListSource(platform, availableModels, fallbackModels), fallbackModels, apiKey.Group.ModelsListConfig.Models)
 		writeCustomModelsList(c, platform, availableModels)
 		return
 	}
@@ -1090,6 +1091,13 @@ func writeOpenAIModelsList(c *gin.Context, modelIDs []string) {
 	})
 }
 
+func customModelsListSource(platform string, availableModels, fallbackModels []string) []string {
+	if platform == service.PlatformAnthropic && len(availableModels) > 0 {
+		return mergeModelIDs(availableModels, fallbackModels)
+	}
+	return availableModels
+}
+
 func filterModelsByCustomList(availableModels, fallbackModels, selectedModels []string) []string {
 	if len(selectedModels) == 0 {
 		return availableModels
@@ -1158,6 +1166,15 @@ func defaultModelIDsForPlatform(platform string) []string {
 			ids = append(ids, model.ID)
 		}
 		return ids
+	case service.PlatformAnthropic:
+		ids := make([]string, 0, len(claude.DefaultModels)+len(antigravity.DefaultModels()))
+		for _, model := range claude.DefaultModels {
+			ids = append(ids, model.ID)
+		}
+		for _, model := range antigravity.DefaultModels() {
+			ids = append(ids, model.ID)
+		}
+		return mergeModelIDs(ids, nil)
 	case service.PlatformGrok:
 		return xai.DefaultModelIDs()
 	default:
@@ -1169,6 +1186,25 @@ func defaultModelIDsForPlatform(platform string) []string {
 	}
 }
 
+func mergeModelIDs(primary, secondary []string) []string {
+	seen := make(map[string]struct{}, len(primary)+len(secondary))
+	merged := make([]string, 0, len(primary)+len(secondary))
+	for _, models := range [][]string{primary, secondary} {
+		for _, model := range models {
+			model = strings.TrimSpace(model)
+			if model == "" {
+				continue
+			}
+			if _, ok := seen[model]; ok {
+				continue
+			}
+			seen[model] = struct{}{}
+			merged = append(merged, model)
+		}
+	}
+	return merged
+}
+
 // AntigravityModels 返回 Antigravity 支持的全部模型
 // GET /antigravity/models
 func (h *GatewayHandler) AntigravityModels(c *gin.Context) {
diff --git a/backend/internal/handler/gateway_models_test.go b/backend/internal/handler/gateway_models_test.go
index c5238f2a2f..6011e13027 100644
--- a/backend/internal/handler/gateway_models_test.go
+++ b/backend/internal/handler/gateway_models_test.go
@@ -269,6 +269,149 @@ func TestGatewayModels_CustomModelsListKeepsConcreteModelAllowedByWildcardMappin
 	require.Equal(t, []string{"claude-sonnet-4-6"}, modelIDsForTest(got.Data))
 }
 
+func TestGatewayModels_AnthropicCustomModelsListIncludesOAuthClaudeAndMappedDeepSeek(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+
+	groupID := int64(28)
+	h := newGatewayModelsHandlerForTest(
+		&gatewayModelsAccountRepoStub{
+			byGroup: map[int64][]service.Account{
+				groupID: {
+					{
+						ID:       1,
+						Platform: service.PlatformAnthropic,
+						Type:     service.AccountTypeOAuth,
+					},
+					{
+						ID:       2,
+						Platform: service.PlatformAnthropic,
+						Type:     service.AccountTypeAPIKey,
+						Credentials: map[string]any{
+							"model_mapping": map[string]any{
+								"deepseek-v4-pro": "deepseek-v4-pro",
+							},
+						},
+					},
+				},
+			},
+		},
+	)
+
+	rec := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(rec)
+	c.Request = httptest.NewRequest(http.MethodGet, "/v1/models", nil)
+	c.Set(string(middleware2.ContextKeyAPIKey), &service.APIKey{
+		Group: &service.Group{
+			ID:       groupID,
+			Platform: service.PlatformAnthropic,
+			ModelsListConfig: service.GroupModelsListConfig{
+				Enabled: true,
+				Models:  []string{"claude-fable-5", "claude-opus-4-8", "deepseek-v4-pro"},
+			},
+		},
+	})
+
+	h.Models(c)
+
+	require.Equal(t, http.StatusOK, rec.Code)
+
+	var got gatewayModelsResponseForTest
+	require.NoError(t, json.Unmarshal(rec.Body.Bytes(), &got))
+	require.Equal(t, []string{"claude-fable-5", "claude-opus-4-8", "deepseek-v4-pro"}, modelIDsForTest(got.Data))
+}
+
+func TestGatewayModels_AnthropicCustomModelsListDisabledKeepsMappedModelList(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+
+	groupID := int64(29)
+	h := newGatewayModelsHandlerForTest(
+		&gatewayModelsAccountRepoStub{
+			byGroup: map[int64][]service.Account{
+				groupID: {
+					{
+						ID:       1,
+						Platform: service.PlatformAnthropic,
+						Type:     service.AccountTypeOAuth,
+					},
+					{
+						ID:       2,
+						Platform: service.PlatformAnthropic,
+						Type:     service.AccountTypeAPIKey,
+						Credentials: map[string]any{
+							"model_mapping": map[string]any{
+								"deepseek-v4-pro": "deepseek-v4-pro",
+							},
+						},
+					},
+				},
+			},
+		},
+	)
+
+	rec := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(rec)
+	c.Request = httptest.NewRequest(http.MethodGet, "/v1/models", nil)
+	c.Set(string(middleware2.ContextKeyAPIKey), &service.APIKey{
+		Group: &service.Group{
+			ID:       groupID,
+			Platform: service.PlatformAnthropic,
+			ModelsListConfig: service.GroupModelsListConfig{
+				Enabled: false,
+				Models:  []string{"claude-fable-5", "deepseek-v4-pro"},
+			},
+		},
+	})
+
+	h.Models(c)
+
+	require.Equal(t, http.StatusOK, rec.Code)
+
+	var got gatewayModelsResponseForTest
+	require.NoError(t, json.Unmarshal(rec.Body.Bytes(), &got))
+	require.Equal(t, []string{"deepseek-v4-pro"}, modelIDsForTest(got.Data))
+}
+
+func TestGatewayModels_AnthropicCustomModelsListIncludesOAuthClaudeWithoutMappings(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+
+	groupID := int64(30)
+	h := newGatewayModelsHandlerForTest(
+		&gatewayModelsAccountRepoStub{
+			byGroup: map[int64][]service.Account{
+				groupID: {
+					{
+						ID:       1,
+						Platform: service.PlatformAnthropic,
+						Type:     service.AccountTypeOAuth,
+					},
+				},
+			},
+		},
+	)
+
+	rec := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(rec)
+	c.Request = httptest.NewRequest(http.MethodGet, "/v1/models", nil)
+	c.Set(string(middleware2.ContextKeyAPIKey), &service.APIKey{
+		Group: &service.Group{
+			ID:       groupID,
+			Platform: service.PlatformAnthropic,
+			ModelsListConfig: service.GroupModelsListConfig{
+				Enabled: true,
+				Models:  []string{"claude-opus-4-6-thinking", "claude-sonnet-4-5"},
+			},
+		},
+	})
+
+	h.Models(c)
+
+	require.Equal(t, http.StatusOK, rec.Code)
+
+	var got gatewayModelsResponseForTest
+	require.NoError(t, json.Unmarshal(rec.Body.Bytes(), &got))
+	require.Equal(t, []string{"claude-opus-4-6-thinking", "claude-sonnet-4-5"}, modelIDsForTest(got.Data))
+}
+
 func TestGatewayModels_CustomModelsListCanReturnEmptyWhenSelectionsUnavailable(t *testing.T) {
 	gin.SetMode(gin.TestMode)
 

From aee9a7ba98ab4e3da4e6e618ea0175fc048a67a5 Mon Sep 17 00:00:00 2001
From: wucm667 <stevenwucongmin@gmail.com>
Date: Sun, 5 Jul 2026 08:36:15 +0800
Subject: [PATCH 24/28] fix(usage): add UTF-8 BOM to CSV export

---
 frontend/src/views/user/UsageView.vue               | 2 +-
 frontend/src/views/user/__tests__/UsageView.spec.ts | 5 +++++
 2 files changed, 6 insertions(+), 1 deletion(-)

diff --git a/frontend/src/views/user/UsageView.vue b/frontend/src/views/user/UsageView.vue
index bbf58532ab..eb2b124839 100644
--- a/frontend/src/views/user/UsageView.vue
+++ b/frontend/src/views/user/UsageView.vue
@@ -677,7 +677,7 @@ const exportToCSV = async () => {
       headers.map(escapeCSVValue).join(','),
       ...rows.map((row) => row.join(',')),
     ].join('\n')
-    const blob = new Blob([csvContent], { type: 'text/csv;charset=utf-8;' })
+    const blob = new Blob(['\uFEFF' + csvContent], { type: 'text/csv;charset=utf-8;' })
     const url = window.URL.createObjectURL(blob)
     const link = document.createElement('a')
     link.href = url
diff --git a/frontend/src/views/user/__tests__/UsageView.spec.ts b/frontend/src/views/user/__tests__/UsageView.spec.ts
index 555ea4b0f6..23fc3c256a 100644
--- a/frontend/src/views/user/__tests__/UsageView.spec.ts
+++ b/frontend/src/views/user/__tests__/UsageView.spec.ts
@@ -237,6 +237,11 @@ describe('user UsageView', () => {
     }))
     expect(clickSpy).toHaveBeenCalled()
     expect(showSuccess).toHaveBeenCalled()
+    expect(csvContent.startsWith('\uFEFF')).toBe(true)
+    expect(csvContent.slice(1)).toBe([
+      'Time,API Key Name,Model,Reasoning Effort,Inbound Endpoint,IP Address,Type,Billing Mode,Input Tokens,Output Tokens,Cache Read Tokens,Cache Creation Tokens,Rate Multiplier,Billed Cost,Original Cost,First Token (ms),Duration (ms)',
+      '2026-03-08T00:00:00Z,demo-key,gpt-5.4,"\'-",,203.0.113.10,Sync,Token,4057,101,278272,4,1,0.09288300,0.09288300,12,345',
+    ].join('\n'))
     expect(csvContent).toContain('IP Address')
     expect(csvContent).toContain('203.0.113.10')
     expect(csvContent).toContain('Billed Cost')

From f26ca5661e9be80921e15b217af668dd1c232905 Mon Sep 17 00:00:00 2001
From: linshuboy <linshuboy@qq.com>
Date: Fri, 3 Jul 2026 23:52:10 +0800
Subject: [PATCH 25/28] feat: add OpenAI advanced scheduler controls

Related: #1089, #408, #123
---
 backend/internal/config/config.go             |   9 +-
 .../internal/handler/admin/account_handler.go | 209 ++++-
 .../admin/account_handler_list_test.go        | 220 +++++
 .../handler/admin/admin_service_stub_test.go  |  95 +-
 .../internal/handler/admin/setting_handler.go | 843 ++++++++++--------
 ...tting_handler_auth_source_defaults_test.go |  15 +-
 backend/internal/handler/dto/settings.go      |  24 +-
 .../handler/openai_gateway_handler.go         |  11 +-
 backend/internal/repository/account_repo.go   |  15 +-
 .../internal/repository/scheduler_cache.go    |   2 +-
 .../repository/scheduler_cache_test.go        |  37 +
 backend/internal/server/api_contract_test.go  |  64 +-
 backend/internal/service/account.go           |  12 +
 backend/internal/service/admin_service.go     |  23 +
 backend/internal/service/domain_constants.go  |  14 +
 .../service/openai_account_scheduler.go       | 698 +++++++++++++--
 .../openai_account_scheduler_reset_test.go    |  11 +-
 .../service/openai_account_scheduler_test.go  | 520 ++++++++++-
 .../internal/service/openai_ws_forwarder.go   | 194 ++--
 backend/internal/service/ratelimit_service.go |   8 +
 backend/internal/service/setting_service.go   | 175 +++-
 .../service/setting_service_update_test.go    | 104 ++-
 backend/internal/service/settings_view.go     |  24 +-
 frontend/src/api/admin/settings.ts            |  34 +
 frontend/src/i18n/locales/en.ts               |  26 +-
 frontend/src/i18n/locales/zh.ts               |  26 +-
 frontend/src/types/index.ts                   |  17 +
 frontend/src/views/admin/AccountsView.vue     |  49 +-
 frontend/src/views/admin/SettingsView.vue     | 220 +++++
 .../admin/__tests__/SettingsView.spec.ts      |  39 +
 30 files changed, 3158 insertions(+), 580 deletions(-)
 create mode 100644 backend/internal/repository/scheduler_cache_test.go

diff --git a/backend/internal/config/config.go b/backend/internal/config/config.go
index b8b880eab0..99fedb5b1c 100644
--- a/backend/internal/config/config.go
+++ b/backend/internal/config/config.go
@@ -969,6 +969,9 @@ type GatewayOpenAIWSSchedulerScoreWeights struct {
 	Reset float64 `mapstructure:"reset"`
 	// QuotaHeadroom 倾向 7d 剩余额度更健康的账号；默认 0（关闭，不改变原有行为）。
 	QuotaHeadroom float64 `mapstructure:"quota_headroom"`
+	// PreviousResponse/SessionSticky 仅在开启 OpenAI 高级调度的粘性加权时生效。
+	PreviousResponse float64 `mapstructure:"previous_response"`
+	SessionSticky    float64 `mapstructure:"session_sticky"`
 }
 
 // GatewayOpenAISchedulerConfig OpenAI 高级调度器配置。
@@ -1891,6 +1894,8 @@ func setDefaults() {
 	viper.SetDefault("gateway.openai_ws.scheduler_score_weights.ttft", 0.5)
 	viper.SetDefault("gateway.openai_ws.scheduler_score_weights.reset", 0.0)
 	viper.SetDefault("gateway.openai_ws.scheduler_score_weights.quota_headroom", 0.0)
+	viper.SetDefault("gateway.openai_ws.scheduler_score_weights.previous_response", 5.0)
+	viper.SetDefault("gateway.openai_ws.scheduler_score_weights.session_sticky", 3.0)
 	// OpenAI HTTP upstream protocol strategy
 	viper.SetDefault("gateway.openai_http2.enabled", true)
 	viper.SetDefault("gateway.openai_http2.allow_proxy_fallback_to_http1", true)
@@ -2674,7 +2679,9 @@ func (c *Config) Validate() error {
 		c.Gateway.OpenAIWS.SchedulerScoreWeights.Queue < 0 ||
 		c.Gateway.OpenAIWS.SchedulerScoreWeights.ErrorRate < 0 ||
 		c.Gateway.OpenAIWS.SchedulerScoreWeights.TTFT < 0 ||
-		c.Gateway.OpenAIWS.SchedulerScoreWeights.QuotaHeadroom < 0 {
+		c.Gateway.OpenAIWS.SchedulerScoreWeights.QuotaHeadroom < 0 ||
+		c.Gateway.OpenAIWS.SchedulerScoreWeights.PreviousResponse < 0 ||
+		c.Gateway.OpenAIWS.SchedulerScoreWeights.SessionSticky < 0 {
 		return fmt.Errorf("gateway.openai_ws.scheduler_score_weights.* must be non-negative")
 	}
 	weightSum := c.Gateway.OpenAIWS.SchedulerScoreWeights.Priority +
diff --git a/backend/internal/handler/admin/account_handler.go b/backend/internal/handler/admin/account_handler.go
index 8d4a5dfea6..efe95801c3 100644
--- a/backend/internal/handler/admin/account_handler.go
+++ b/backend/internal/handler/admin/account_handler.go
@@ -171,15 +171,39 @@ type CheckMixedChannelRequest struct {
 // AccountWithConcurrency extends Account with real-time concurrency info
 type AccountWithConcurrency struct {
 	*dto.Account
-	CurrentConcurrency int `json:"current_concurrency"`
+	CurrentConcurrency int                          `json:"current_concurrency"`
+	SchedulerScore     *AccountSchedulerScore       `json:"scheduler_score,omitempty"`
+	SchedulerScores    []AccountSchedulerGroupScore `json:"scheduler_scores,omitempty"`
 	// 以下字段仅对 Anthropic OAuth/SetupToken 账号有效，且仅在启用相应功能时返回
 	CurrentWindowCost *float64 `json:"current_window_cost,omitempty"` // 当前窗口费用
 	ActiveSessions    *int     `json:"active_sessions,omitempty"`     // 当前活跃会话数
 	CurrentRPM        *int     `json:"current_rpm,omitempty"`         // 当前分钟 RPM 计数
 }
 
+type AccountSchedulerScore struct {
+	BaseScore             float64 `json:"base_score"`
+	StickyScore           float64 `json:"sticky_score"`
+	StickyScoreInfinity   bool    `json:"sticky_score_infinity"`
+	StickyWeightedEnabled bool    `json:"sticky_weighted_enabled"`
+}
+
+type AccountSchedulerGroupScore struct {
+	GroupID       *int64 `json:"group_id"`
+	GroupName     string `json:"group_name,omitempty"`
+	GroupPriority *int   `json:"group_priority,omitempty"`
+	AccountSchedulerScore
+}
+
 const accountListGroupUngroupedQueryValue = "ungrouped"
 
+type openAIAccountSchedulerScorePoolLister interface {
+	ListOpenAISchedulableAccountsForSchedulerScore(ctx context.Context, groupID *int64) ([]service.Account, error)
+}
+
+type accountSchedulerScoreFilterPoolLister interface {
+	ListAccountsForSchedulerScoreFilter(ctx context.Context, platform, accountType, status, search string, groupID int64, privacyMode string) ([]service.Account, error)
+}
+
 func (h *AccountHandler) buildAccountResponseWithRuntime(ctx context.Context, account *service.Account) AccountWithConcurrency {
 	item := AccountWithConcurrency{
 		Account:            dto.AccountFromService(account),
@@ -226,6 +250,185 @@ func (h *AccountHandler) buildAccountResponseWithRuntime(ctx context.Context, ac
 	return item
 }
 
+func (h *AccountHandler) scoreOpenAIAccountSchedulerPool(ctx context.Context, accounts []service.Account) map[int64]AccountSchedulerScore {
+	if len(accounts) == 0 {
+		return nil
+	}
+
+	openAIAccounts := make([]*service.Account, 0, len(accounts))
+	loadReq := make([]service.AccountWithConcurrency, 0, len(accounts))
+	for i := range accounts {
+		account := &accounts[i]
+		if account.Platform != service.PlatformOpenAI {
+			continue
+		}
+		openAIAccounts = append(openAIAccounts, account)
+		loadReq = append(loadReq, service.AccountWithConcurrency{
+			ID:             account.ID,
+			MaxConcurrency: account.EffectiveLoadFactor(),
+		})
+	}
+	if len(openAIAccounts) == 0 {
+		return nil
+	}
+
+	loadMap := map[int64]*service.AccountLoadInfo{}
+	if h.concurrencyService != nil {
+		if batchLoad, err := h.concurrencyService.GetAccountsLoadBatch(ctx, loadReq); err == nil && batchLoad != nil {
+			loadMap = batchLoad
+		}
+	}
+
+	var scores map[int64]service.OpenAIAccountSchedulerScoreSnapshot
+	if h.rateLimitService != nil {
+		scores = h.rateLimitService.BuildOpenAIAccountSchedulerScoreSnapshot(ctx, openAIAccounts, loadMap)
+	} else {
+		scores = service.BuildOpenAIAccountSchedulerScoreSnapshot(openAIAccounts, loadMap)
+	}
+	result := make(map[int64]AccountSchedulerScore, len(scores))
+	for accountID, score := range scores {
+		result[accountID] = AccountSchedulerScore{
+			BaseScore:             score.BaseScore,
+			StickyScore:           score.StickyScore,
+			StickyScoreInfinity:   score.StickyScoreInfinity,
+			StickyWeightedEnabled: score.StickyWeightedEnabled,
+		}
+	}
+	return result
+}
+
+func (h *AccountHandler) buildOpenAIAccountSchedulerScores(
+	ctx context.Context,
+	accounts []service.Account,
+	filterPool []service.Account,
+) (map[int64]*AccountSchedulerScore, map[int64][]AccountSchedulerGroupScore) {
+	if len(accounts) == 0 {
+		return nil, nil
+	}
+	if len(filterPool) == 0 {
+		filterPool = accounts
+	}
+
+	baseScores := make(map[int64]*AccountSchedulerScore)
+	for accountID, score := range h.scoreOpenAIAccountSchedulerPool(ctx, filterPool) {
+		copiedScore := score
+		baseScores[accountID] = &copiedScore
+	}
+
+	pageOpenAIAccountIDs := make(map[int64]struct{})
+	groupIDs := make(map[int64]struct{})
+	for i := range accounts {
+		account := &accounts[i]
+		if account.Platform != service.PlatformOpenAI {
+			continue
+		}
+		pageOpenAIAccountIDs[account.ID] = struct{}{}
+		if len(account.AccountGroups) == 0 && len(account.GroupIDs) == 0 {
+			continue
+		}
+		for _, accountGroup := range account.AccountGroups {
+			if accountGroup.GroupID > 0 {
+				groupIDs[accountGroup.GroupID] = struct{}{}
+			}
+		}
+		for _, groupID := range account.GroupIDs {
+			if groupID > 0 {
+				groupIDs[groupID] = struct{}{}
+			}
+		}
+	}
+	if len(pageOpenAIAccountIDs) == 0 {
+		return baseScores, nil
+	}
+
+	groupScoresByAccount := make(map[int64][]AccountSchedulerGroupScore)
+	scoreGroupPool := func(groupID *int64, groupNameByID map[int64]string, groupPriorityByAccount map[int64]int, pool []service.Account) {
+		if len(pool) == 0 {
+			return
+		}
+		scores := h.scoreOpenAIAccountSchedulerPool(ctx, pool)
+		for accountID, schedulerScore := range scores {
+			if _, ok := pageOpenAIAccountIDs[accountID]; !ok {
+				continue
+			}
+			groupScore := AccountSchedulerGroupScore{
+				GroupID:               groupID,
+				AccountSchedulerScore: schedulerScore,
+			}
+			if groupID != nil {
+				groupScore.GroupName = groupNameByID[*groupID]
+				if priority, ok := groupPriorityByAccount[accountID]; ok {
+					groupScore.GroupPriority = &priority
+				}
+			}
+			groupScoresByAccount[accountID] = append(groupScoresByAccount[accountID], groupScore)
+		}
+	}
+
+	if lister, ok := h.adminService.(openAIAccountSchedulerScorePoolLister); ok {
+		groupIDList := make([]int64, 0, len(groupIDs))
+		for groupID := range groupIDs {
+			groupIDList = append(groupIDList, groupID)
+		}
+		sort.Slice(groupIDList, func(i, j int) bool { return groupIDList[i] < groupIDList[j] })
+
+		for _, groupID := range groupIDList {
+			gid := groupID
+			pool, err := lister.ListOpenAISchedulableAccountsForSchedulerScore(ctx, &gid)
+			if err != nil {
+				slog.Warn("openai_scheduler_group_score_pool_failed", "group_id", gid, "error", err)
+				continue
+			}
+			groupNameByID := make(map[int64]string)
+			groupPriorityByAccount := make(map[int64]int)
+			for i := range pool {
+				account := &pool[i]
+				for _, accountGroup := range account.AccountGroups {
+					if accountGroup.GroupID != gid {
+						continue
+					}
+					groupPriorityByAccount[account.ID] = accountGroup.Priority
+					if accountGroup.Group != nil {
+						groupNameByID[gid] = accountGroup.Group.Name
+					}
+				}
+			}
+			scoreGroupPool(&gid, groupNameByID, groupPriorityByAccount, pool)
+		}
+
+	}
+
+	for accountID := range groupScoresByAccount {
+		sort.SliceStable(groupScoresByAccount[accountID], func(i, j int) bool {
+			left := groupScoresByAccount[accountID][i]
+			right := groupScoresByAccount[accountID][j]
+			return *left.GroupID < *right.GroupID
+		})
+	}
+	return baseScores, groupScoresByAccount
+}
+
+func (h *AccountHandler) listAccountSchedulerScoreFilterPool(
+	ctx context.Context,
+	platform, accountType, status, search string,
+	groupID int64,
+	privacyMode string,
+) []service.Account {
+	if h.adminService == nil || (platform != "" && platform != service.PlatformOpenAI) {
+		return nil
+	}
+	lister, ok := h.adminService.(accountSchedulerScoreFilterPoolLister)
+	if !ok {
+		return nil
+	}
+	accounts, err := lister.ListAccountsForSchedulerScoreFilter(ctx, platform, accountType, status, search, groupID, privacyMode)
+	if err != nil {
+		slog.Warn("openai_scheduler_filter_score_pool_failed", "error", err)
+		return nil
+	}
+	return accounts
+}
+
 // List handles listing all accounts with pagination
 // GET /api/v1/admin/accounts
 func (h *AccountHandler) List(c *gin.Context) {
@@ -278,6 +481,8 @@ func (h *AccountHandler) List(c *gin.Context) {
 	var windowCosts map[int64]float64
 	var activeSessions map[int64]int
 	var rpmCounts map[int64]int
+	schedulerFilterPool := h.listAccountSchedulerScoreFilterPool(c.Request.Context(), platform, accountType, status, search, groupID, privacyMode)
+	schedulerScores, schedulerGroupScores := h.buildOpenAIAccountSchedulerScores(c.Request.Context(), accounts, schedulerFilterPool)
 
 	// 始终获取并发数（Redis ZCARD，极低开销）
 	if h.concurrencyService != nil {
@@ -358,6 +563,8 @@ func (h *AccountHandler) List(c *gin.Context) {
 		item := AccountWithConcurrency{
 			Account:            dto.AccountFromService(acc),
 			CurrentConcurrency: concurrencyCounts[acc.ID],
+			SchedulerScore:     schedulerScores[acc.ID],
+			SchedulerScores:    schedulerGroupScores[acc.ID],
 		}
 
 		// 添加窗口费用（仅当启用时）
diff --git a/backend/internal/handler/admin/account_handler_list_test.go b/backend/internal/handler/admin/account_handler_list_test.go
index 4d628365df..29e36ad865 100644
--- a/backend/internal/handler/admin/account_handler_list_test.go
+++ b/backend/internal/handler/admin/account_handler_list_test.go
@@ -8,6 +8,7 @@ import (
 	"testing"
 	"time"
 
+	"github.com/Wei-Shaw/sub2api/internal/service"
 	"github.com/gin-gonic/gin"
 	"github.com/stretchr/testify/require"
 )
@@ -50,3 +51,222 @@ func TestAccountHandlerListIncludesCreatedAt(t *testing.T) {
 	_, offset := parsed.Zone()
 	require.Equal(t, 0, offset)
 }
+
+func TestAccountHandlerListReturnsSchedulerScoresPerGroup(t *testing.T) {
+	router, adminSvc := setupAccountListRouter()
+	now := time.Now().UTC()
+	groupID := int64(41)
+	adminSvc.accounts = []service.Account{
+		{
+			ID:          101,
+			Name:        "account-high-priority",
+			Platform:    service.PlatformOpenAI,
+			Type:        service.AccountTypeAPIKey,
+			Status:      service.StatusActive,
+			Schedulable: true,
+			Concurrency: 10,
+			Priority:    1,
+			AccountGroups: []service.AccountGroup{
+				{AccountID: 101, GroupID: groupID, Priority: 100, Group: &service.Group{ID: groupID, Name: "openai"}},
+			},
+			GroupIDs:  []int64{groupID},
+			CreatedAt: now,
+			UpdatedAt: now,
+		},
+		{
+			ID:          102,
+			Name:        "account-low-priority",
+			Platform:    service.PlatformOpenAI,
+			Type:        service.AccountTypeAPIKey,
+			Status:      service.StatusActive,
+			Schedulable: true,
+			Concurrency: 10,
+			Priority:    100000,
+			AccountGroups: []service.AccountGroup{
+				{AccountID: 102, GroupID: groupID, Priority: 1, Group: &service.Group{ID: groupID, Name: "openai"}},
+			},
+			GroupIDs:  []int64{groupID},
+			CreatedAt: now,
+			UpdatedAt: now,
+		},
+	}
+
+	rec := httptest.NewRecorder()
+	req := httptest.NewRequest(http.MethodGet, "/api/v1/admin/accounts?page=1&page_size=20&platform=openai", nil)
+	router.ServeHTTP(rec, req)
+
+	require.Equal(t, http.StatusOK, rec.Code)
+	var payload struct {
+		Data struct {
+			Items []struct {
+				ID             int64 `json:"id"`
+				SchedulerScore struct {
+					BaseScore float64 `json:"base_score"`
+				} `json:"scheduler_score"`
+				SchedulerScores []struct {
+					GroupID       *int64  `json:"group_id"`
+					GroupName     string  `json:"group_name"`
+					GroupPriority *int    `json:"group_priority"`
+					BaseScore     float64 `json:"base_score"`
+				} `json:"scheduler_scores"`
+			} `json:"items"`
+		} `json:"data"`
+	}
+	require.NoError(t, json.Unmarshal(rec.Body.Bytes(), &payload))
+	require.Len(t, payload.Data.Items, 2)
+
+	var high, low *struct {
+		ID             int64 `json:"id"`
+		SchedulerScore struct {
+			BaseScore float64 `json:"base_score"`
+		} `json:"scheduler_score"`
+		SchedulerScores []struct {
+			GroupID       *int64  `json:"group_id"`
+			GroupName     string  `json:"group_name"`
+			GroupPriority *int    `json:"group_priority"`
+			BaseScore     float64 `json:"base_score"`
+		} `json:"scheduler_scores"`
+	}
+	for i := range payload.Data.Items {
+		item := &payload.Data.Items[i]
+		switch item.ID {
+		case 101:
+			high = item
+		case 102:
+			low = item
+		}
+	}
+	require.NotNil(t, high)
+	require.NotNil(t, low)
+	require.Len(t, high.SchedulerScores, 1)
+	require.Len(t, low.SchedulerScores, 1)
+	require.Equal(t, groupID, *high.SchedulerScores[0].GroupID)
+	require.Equal(t, "openai", high.SchedulerScores[0].GroupName)
+	require.Equal(t, 100, *high.SchedulerScores[0].GroupPriority)
+	require.Equal(t, 1, *low.SchedulerScores[0].GroupPriority)
+	require.Greater(t, high.SchedulerScores[0].BaseScore, low.SchedulerScores[0].BaseScore)
+}
+
+func TestAccountHandlerListKeepsSchedulerScoreScopedToFilter(t *testing.T) {
+	router, adminSvc := setupAccountListRouter()
+	now := time.Now().UTC()
+	groupID := int64(42)
+	visibleAccount := service.Account{
+		ID:          201,
+		Name:        "visible-low-priority",
+		Platform:    service.PlatformOpenAI,
+		Type:        service.AccountTypeAPIKey,
+		Status:      service.StatusActive,
+		Schedulable: true,
+		Concurrency: 10,
+		Priority:    100000,
+		AccountGroups: []service.AccountGroup{
+			{AccountID: 201, GroupID: groupID, Priority: 1, Group: &service.Group{ID: groupID, Name: "openai"}},
+		},
+		GroupIDs:  []int64{groupID},
+		CreatedAt: now,
+		UpdatedAt: now,
+	}
+	hiddenGroupPeer := service.Account{
+		ID:          202,
+		Name:        "hidden-high-priority",
+		Platform:    service.PlatformOpenAI,
+		Type:        service.AccountTypeAPIKey,
+		Status:      service.StatusActive,
+		Schedulable: true,
+		Concurrency: 10,
+		Priority:    1,
+		AccountGroups: []service.AccountGroup{
+			{AccountID: 202, GroupID: groupID, Priority: 2, Group: &service.Group{ID: groupID, Name: "openai"}},
+		},
+		GroupIDs:  []int64{groupID},
+		CreatedAt: now,
+		UpdatedAt: now,
+	}
+	adminSvc.accounts = []service.Account{visibleAccount}
+	adminSvc.accountSchedulerScoreFilterAccounts = []service.Account{visibleAccount, hiddenGroupPeer}
+	adminSvc.openAISchedulerScorePoolAccounts = []service.Account{visibleAccount, hiddenGroupPeer}
+
+	rec := httptest.NewRecorder()
+	req := httptest.NewRequest(http.MethodGet, "/api/v1/admin/accounts?page=1&page_size=1&platform=openai", nil)
+	router.ServeHTTP(rec, req)
+
+	require.Equal(t, http.StatusOK, rec.Code)
+	var payload struct {
+		Data struct {
+			Items []struct {
+				ID             int64 `json:"id"`
+				SchedulerScore struct {
+					BaseScore float64 `json:"base_score"`
+				} `json:"scheduler_score"`
+				SchedulerScores []struct {
+					GroupID   *int64  `json:"group_id"`
+					BaseScore float64 `json:"base_score"`
+				} `json:"scheduler_scores"`
+			} `json:"items"`
+		} `json:"data"`
+	}
+	require.NoError(t, json.Unmarshal(rec.Body.Bytes(), &payload))
+	require.Len(t, payload.Data.Items, 1)
+	item := payload.Data.Items[0]
+	require.Equal(t, int64(201), item.ID)
+	require.Len(t, item.SchedulerScores, 1)
+	require.Equal(t, groupID, *item.SchedulerScores[0].GroupID)
+	require.Equal(t, item.SchedulerScores[0].BaseScore, item.SchedulerScore.BaseScore)
+}
+
+func TestAccountHandlerListSchedulerScoreIgnoresPagination(t *testing.T) {
+	router, adminSvc := setupAccountListRouter()
+	now := time.Now().UTC()
+	visibleAccount := service.Account{
+		ID:          301,
+		Name:        "visible-low-priority",
+		Platform:    service.PlatformOpenAI,
+		Type:        service.AccountTypeAPIKey,
+		Status:      service.StatusActive,
+		Schedulable: true,
+		Concurrency: 10,
+		Priority:    100000,
+		CreatedAt:   now,
+		UpdatedAt:   now,
+	}
+	hiddenFilterPeer := service.Account{
+		ID:          302,
+		Name:        "hidden-high-priority",
+		Platform:    service.PlatformOpenAI,
+		Type:        service.AccountTypeAPIKey,
+		Status:      service.StatusActive,
+		Schedulable: true,
+		Concurrency: 10,
+		Priority:    1,
+		CreatedAt:   now,
+		UpdatedAt:   now,
+	}
+	adminSvc.accounts = []service.Account{visibleAccount}
+	adminSvc.accountSchedulerScoreFilterAccounts = []service.Account{visibleAccount, hiddenFilterPeer}
+
+	rec := httptest.NewRecorder()
+	req := httptest.NewRequest(http.MethodGet, "/api/v1/admin/accounts?page=1&page_size=1&platform=openai", nil)
+	router.ServeHTTP(rec, req)
+
+	require.Equal(t, http.StatusOK, rec.Code)
+	var payload struct {
+		Data struct {
+			Items []struct {
+				ID             int64 `json:"id"`
+				SchedulerScore struct {
+					BaseScore float64 `json:"base_score"`
+				} `json:"scheduler_score"`
+				SchedulerScores []struct {
+					GroupID   *int64  `json:"group_id"`
+					BaseScore float64 `json:"base_score"`
+				} `json:"scheduler_scores"`
+			} `json:"items"`
+		} `json:"data"`
+	}
+	require.NoError(t, json.Unmarshal(rec.Body.Bytes(), &payload))
+	require.Len(t, payload.Data.Items, 1)
+	require.Equal(t, int64(301), payload.Data.Items[0].ID)
+	require.Less(t, payload.Data.Items[0].SchedulerScore.BaseScore, 3.75)
+	require.Empty(t, payload.Data.Items[0].SchedulerScores)
+}
diff --git a/backend/internal/handler/admin/admin_service_stub_test.go b/backend/internal/handler/admin/admin_service_stub_test.go
index f7bab13253..187925e33a 100644
--- a/backend/internal/handler/admin/admin_service_stub_test.go
+++ b/backend/internal/handler/admin/admin_service_stub_test.go
@@ -10,27 +10,29 @@ import (
 )
 
 type stubAdminService struct {
-	users                []service.User
-	apiKeys              []service.APIKey
-	groups               []service.Group
-	accounts             []service.Account
-	proxies              []service.Proxy
-	proxyCounts          []service.ProxyWithAccountCount
-	redeems              []service.RedeemCode
-	boundAuthIdentity    *service.AdminBindAuthIdentityInput
-	boundAuthIdentityFor int64
-	createdAccounts      []*service.CreateAccountInput
-	createdProxies       []*service.CreateProxyInput
-	updatedProxyIDs      []int64
-	updatedProxies       []*service.UpdateProxyInput
-	testedProxyIDs       []int64
-	getUserErr           error
-	createAccountErr     error
-	createSparkShadowErr error
-	updateAccountErr     error
-	bulkUpdateAccountErr error
-	checkMixedErr        error
-	lastMixedCheck       struct {
+	users                               []service.User
+	apiKeys                             []service.APIKey
+	groups                              []service.Group
+	accounts                            []service.Account
+	accountSchedulerScoreFilterAccounts []service.Account
+	openAISchedulerScorePoolAccounts    []service.Account
+	proxies                             []service.Proxy
+	proxyCounts                         []service.ProxyWithAccountCount
+	redeems                             []service.RedeemCode
+	boundAuthIdentity                   *service.AdminBindAuthIdentityInput
+	boundAuthIdentityFor                int64
+	createdAccounts                     []*service.CreateAccountInput
+	createdProxies                      []*service.CreateProxyInput
+	updatedProxyIDs                     []int64
+	updatedProxies                      []*service.UpdateProxyInput
+	testedProxyIDs                      []int64
+	getUserErr                          error
+	createAccountErr                    error
+	createSparkShadowErr                error
+	updateAccountErr                    error
+	bulkUpdateAccountErr                error
+	checkMixedErr                       error
+	lastMixedCheck                      struct {
 		accountID int64
 		platform  string
 		groupIDs  []int64
@@ -329,7 +331,56 @@ func (s *stubAdminService) ListAccounts(ctx context.Context, page, pageSize int,
 	s.lastListAccounts.sortBy = sortBy
 	s.lastListAccounts.sortOrder = sortOrder
 	s.lastListAccounts.calls++
-	return s.accounts, int64(len(s.accounts)), nil
+	accounts := s.accounts
+	total := len(accounts)
+	if page < 1 {
+		page = 1
+	}
+	if pageSize < 1 {
+		pageSize = total
+	}
+	start := (page - 1) * pageSize
+	if start >= total {
+		return []service.Account{}, int64(total), nil
+	}
+	end := start + pageSize
+	if end > total {
+		end = total
+	}
+	return accounts[start:end], int64(total), nil
+}
+
+func (s *stubAdminService) ListAccountsForSchedulerScoreFilter(_ context.Context, platform, accountType, status, search string, groupID int64, privacyMode string) ([]service.Account, error) {
+	if s.accountSchedulerScoreFilterAccounts != nil {
+		return s.accountSchedulerScoreFilterAccounts, nil
+	}
+	return s.accounts, nil
+}
+
+func (s *stubAdminService) ListOpenAISchedulableAccountsForSchedulerScore(_ context.Context, groupID *int64) ([]service.Account, error) {
+	accounts := s.openAISchedulerScorePoolAccounts
+	if accounts == nil {
+		accounts = s.accounts
+	}
+	out := make([]service.Account, 0, len(accounts))
+	for _, account := range accounts {
+		if account.Platform != service.PlatformOpenAI || !account.IsSchedulable() {
+			continue
+		}
+		if groupID == nil {
+			if len(account.AccountGroups) == 0 && len(account.GroupIDs) == 0 {
+				out = append(out, account)
+			}
+			continue
+		}
+		for _, accountGroup := range account.AccountGroups {
+			if accountGroup.GroupID == *groupID {
+				out = append(out, account)
+				break
+			}
+		}
+	}
+	return out, nil
 }
 
 func (s *stubAdminService) GetAccount(ctx context.Context, id int64) (*service.Account, error) {
diff --git a/backend/internal/handler/admin/setting_handler.go b/backend/internal/handler/admin/setting_handler.go
index 0f50ab723b..624fddb1e6 100644
--- a/backend/internal/handler/admin/setting_handler.go
+++ b/backend/internal/handler/admin/setting_handler.go
@@ -119,188 +119,210 @@ func (h *SettingHandler) GetSettings(c *gin.Context) {
 	}
 
 	payload := dto.SystemSettings{
-		RegistrationEnabled:                    settings.RegistrationEnabled,
-		EmailVerifyEnabled:                     settings.EmailVerifyEnabled,
-		RegistrationEmailSuffixWhitelist:       settings.RegistrationEmailSuffixWhitelist,
-		PromoCodeEnabled:                       settings.PromoCodeEnabled,
-		PasswordResetEnabled:                   settings.PasswordResetEnabled,
-		FrontendURL:                            settings.FrontendURL,
-		InvitationCodeEnabled:                  settings.InvitationCodeEnabled,
-		TotpEnabled:                            settings.TotpEnabled,
-		TotpEncryptionKeyConfigured:            h.settingService.IsTotpEncryptionKeyConfigured(),
-		LoginAgreementEnabled:                  settings.LoginAgreementEnabled,
-		LoginAgreementMode:                     settings.LoginAgreementMode,
-		LoginAgreementUpdatedAt:                settings.LoginAgreementUpdatedAt,
-		LoginAgreementDocuments:                loginAgreementDocumentsToDTO(settings.LoginAgreementDocuments),
-		SMTPHost:                               settings.SMTPHost,
-		SMTPPort:                               settings.SMTPPort,
-		SMTPUsername:                           settings.SMTPUsername,
-		SMTPPasswordConfigured:                 settings.SMTPPasswordConfigured,
-		SMTPFrom:                               settings.SMTPFrom,
-		SMTPFromName:                           settings.SMTPFromName,
-		SMTPUseTLS:                             settings.SMTPUseTLS,
-		TurnstileEnabled:                       settings.TurnstileEnabled,
-		TurnstileSiteKey:                       settings.TurnstileSiteKey,
-		TurnstileSecretKeyConfigured:           settings.TurnstileSecretKeyConfigured,
-		APIKeyACLTrustForwardedIP:              settings.APIKeyACLTrustForwardedIP,
-		LinuxDoConnectEnabled:                  settings.LinuxDoConnectEnabled,
-		LinuxDoConnectClientID:                 settings.LinuxDoConnectClientID,
-		LinuxDoConnectClientSecretConfigured:   settings.LinuxDoConnectClientSecretConfigured,
-		LinuxDoConnectRedirectURL:              settings.LinuxDoConnectRedirectURL,
-		DingTalkConnectEnabled:                 settings.DingTalkConnectEnabled,
-		DingTalkConnectClientID:                settings.DingTalkConnectClientID,
-		DingTalkConnectClientSecretConfigured:  settings.DingTalkConnectClientSecretConfigured,
-		DingTalkConnectRedirectURL:             settings.DingTalkConnectRedirectURL,
-		DingTalkConnectCorpRestrictionPolicy:   settings.DingTalkConnectCorpRestrictionPolicy,
-		DingTalkConnectInternalCorpID:          settings.DingTalkConnectInternalCorpID,
-		DingTalkConnectBypassRegistration:      settings.DingTalkConnectBypassRegistration,
-		DingTalkConnectSyncCorpEmail:           settings.DingTalkConnectSyncCorpEmail,
-		DingTalkConnectSyncDisplayName:         settings.DingTalkConnectSyncDisplayName,
-		DingTalkConnectSyncDept:                settings.DingTalkConnectSyncDept,
-		DingTalkConnectSyncCorpEmailAttrKey:    settings.DingTalkConnectSyncCorpEmailAttrKey,
-		DingTalkConnectSyncDisplayNameAttrKey:  settings.DingTalkConnectSyncDisplayNameAttrKey,
-		DingTalkConnectSyncDeptAttrKey:         settings.DingTalkConnectSyncDeptAttrKey,
-		DingTalkConnectSyncCorpEmailAttrName:   settings.DingTalkConnectSyncCorpEmailAttrName,
-		DingTalkConnectSyncDisplayNameAttrName: settings.DingTalkConnectSyncDisplayNameAttrName,
-		DingTalkConnectSyncDeptAttrName:        settings.DingTalkConnectSyncDeptAttrName,
-		WeChatConnectEnabled:                   settings.WeChatConnectEnabled,
-		WeChatConnectAppID:                     settings.WeChatConnectAppID,
-		WeChatConnectAppSecretConfigured:       settings.WeChatConnectAppSecretConfigured,
-		WeChatConnectOpenAppID:                 settings.WeChatConnectOpenAppID,
-		WeChatConnectOpenAppSecretConfigured:   settings.WeChatConnectOpenAppSecretConfigured,
-		WeChatConnectMPAppID:                   settings.WeChatConnectMPAppID,
-		WeChatConnectMPAppSecretConfigured:     settings.WeChatConnectMPAppSecretConfigured,
-		WeChatConnectMobileAppID:               settings.WeChatConnectMobileAppID,
-		WeChatConnectMobileAppSecretConfigured: settings.WeChatConnectMobileAppSecretConfigured,
-		WeChatConnectOpenEnabled:               settings.WeChatConnectOpenEnabled,
-		WeChatConnectMPEnabled:                 settings.WeChatConnectMPEnabled,
-		WeChatConnectMobileEnabled:             settings.WeChatConnectMobileEnabled,
-		WeChatConnectMode:                      settings.WeChatConnectMode,
-		WeChatConnectScopes:                    settings.WeChatConnectScopes,
-		WeChatConnectRedirectURL:               settings.WeChatConnectRedirectURL,
-		WeChatConnectFrontendRedirectURL:       settings.WeChatConnectFrontendRedirectURL,
-		OIDCConnectEnabled:                     settings.OIDCConnectEnabled,
-		OIDCConnectProviderName:                settings.OIDCConnectProviderName,
-		OIDCConnectClientID:                    settings.OIDCConnectClientID,
-		OIDCConnectClientSecretConfigured:      settings.OIDCConnectClientSecretConfigured,
-		OIDCConnectIssuerURL:                   settings.OIDCConnectIssuerURL,
-		OIDCConnectDiscoveryURL:                settings.OIDCConnectDiscoveryURL,
-		OIDCConnectAuthorizeURL:                settings.OIDCConnectAuthorizeURL,
-		OIDCConnectTokenURL:                    settings.OIDCConnectTokenURL,
-		OIDCConnectUserInfoURL:                 settings.OIDCConnectUserInfoURL,
-		OIDCConnectJWKSURL:                     settings.OIDCConnectJWKSURL,
-		OIDCConnectScopes:                      settings.OIDCConnectScopes,
-		OIDCConnectRedirectURL:                 settings.OIDCConnectRedirectURL,
-		OIDCConnectFrontendRedirectURL:         settings.OIDCConnectFrontendRedirectURL,
-		OIDCConnectTokenAuthMethod:             settings.OIDCConnectTokenAuthMethod,
-		OIDCConnectUsePKCE:                     settings.OIDCConnectUsePKCE,
-		OIDCConnectValidateIDToken:             settings.OIDCConnectValidateIDToken,
-		OIDCConnectAllowedSigningAlgs:          settings.OIDCConnectAllowedSigningAlgs,
-		OIDCConnectClockSkewSeconds:            settings.OIDCConnectClockSkewSeconds,
-		OIDCConnectRequireEmailVerified:        settings.OIDCConnectRequireEmailVerified,
-		OIDCConnectUserInfoEmailPath:           settings.OIDCConnectUserInfoEmailPath,
-		OIDCConnectUserInfoIDPath:              settings.OIDCConnectUserInfoIDPath,
-		OIDCConnectUserInfoUsernamePath:        settings.OIDCConnectUserInfoUsernamePath,
-		GitHubOAuthEnabled:                     settings.GitHubOAuthEnabled,
-		GitHubOAuthClientID:                    settings.GitHubOAuthClientID,
-		GitHubOAuthClientSecretConfigured:      settings.GitHubOAuthClientSecretConfigured,
-		GitHubOAuthRedirectURL:                 settings.GitHubOAuthRedirectURL,
-		GitHubOAuthFrontendRedirectURL:         settings.GitHubOAuthFrontendRedirectURL,
-		GoogleOAuthEnabled:                     settings.GoogleOAuthEnabled,
-		GoogleOAuthClientID:                    settings.GoogleOAuthClientID,
-		GoogleOAuthClientSecretConfigured:      settings.GoogleOAuthClientSecretConfigured,
-		GoogleOAuthRedirectURL:                 settings.GoogleOAuthRedirectURL,
-		GoogleOAuthFrontendRedirectURL:         settings.GoogleOAuthFrontendRedirectURL,
-		SiteName:                               settings.SiteName,
-		SiteLogo:                               settings.SiteLogo,
-		SiteSubtitle:                           settings.SiteSubtitle,
-		APIBaseURL:                             settings.APIBaseURL,
-		ContactInfo:                            settings.ContactInfo,
-		DocURL:                                 settings.DocURL,
-		HomeContent:                            settings.HomeContent,
-		HideCcsImportButton:                    settings.HideCcsImportButton,
-		PurchaseSubscriptionEnabled:            settings.PurchaseSubscriptionEnabled,
-		PurchaseSubscriptionURL:                settings.PurchaseSubscriptionURL,
-		TableDefaultPageSize:                   settings.TableDefaultPageSize,
-		TablePageSizeOptions:                   settings.TablePageSizeOptions,
-		CustomMenuItems:                        dto.ParseCustomMenuItems(settings.CustomMenuItems),
-		CustomEndpoints:                        dto.ParseCustomEndpoints(settings.CustomEndpoints),
-		DefaultConcurrency:                     settings.DefaultConcurrency,
-		DefaultBalance:                         settings.DefaultBalance,
-		RiskControlEnabled:                     settings.RiskControlEnabled,
-		CyberSessionBlockEnabled:               settings.CyberSessionBlockEnabled,
-		CyberSessionBlockTTLSeconds:            settings.CyberSessionBlockTTLSeconds,
-		AffiliateRebateRate:                    settings.AffiliateRebateRate,
-		AffiliateRebateFreezeHours:             settings.AffiliateRebateFreezeHours,
-		AffiliateRebateDurationDays:            settings.AffiliateRebateDurationDays,
-		AffiliateRebatePerInviteeCap:           settings.AffiliateRebatePerInviteeCap,
-		DefaultUserRPMLimit:                    settings.DefaultUserRPMLimit,
-		DefaultSubscriptions:                   defaultSubscriptions,
-		EnableModelFallback:                    settings.EnableModelFallback,
-		FallbackModelAnthropic:                 settings.FallbackModelAnthropic,
-		FallbackModelOpenAI:                    settings.FallbackModelOpenAI,
-		FallbackModelGemini:                    settings.FallbackModelGemini,
-		FallbackModelAntigravity:               settings.FallbackModelAntigravity,
-		EnableIdentityPatch:                    settings.EnableIdentityPatch,
-		IdentityPatchPrompt:                    settings.IdentityPatchPrompt,
-		OpsMonitoringEnabled:                   opsEnabled && settings.OpsMonitoringEnabled,
-		OpsRealtimeMonitoringEnabled:           settings.OpsRealtimeMonitoringEnabled,
-		OpsQueryModeDefault:                    settings.OpsQueryModeDefault,
-		OpsMetricsIntervalSeconds:              settings.OpsMetricsIntervalSeconds,
-		MinClaudeCodeVersion:                   settings.MinClaudeCodeVersion,
-		MaxClaudeCodeVersion:                   settings.MaxClaudeCodeVersion,
-		AllowUngroupedKeyScheduling:            settings.AllowUngroupedKeyScheduling,
-		BackendModeEnabled:                     settings.BackendModeEnabled,
-		EnableFingerprintUnification:           settings.EnableFingerprintUnification,
-		EnableMetadataPassthrough:              settings.EnableMetadataPassthrough,
-		EnableCCHSigning:                       settings.EnableCCHSigning,
-		EnableClaudeOAuthSystemPromptInjection: settings.EnableClaudeOAuthSystemPromptInjection,
-		ClaudeOAuthSystemPrompt:                settings.ClaudeOAuthSystemPrompt,
-		ClaudeOAuthSystemPromptBlocks:          settings.ClaudeOAuthSystemPromptBlocks,
-		EnableAnthropicCacheTTL1hInjection:     settings.EnableAnthropicCacheTTL1hInjection,
-		RewriteMessageCacheControl:             settings.RewriteMessageCacheControl,
-		EnableClientDatelineNormalization:      settings.EnableClientDatelineNormalization,
-		AntigravityUserAgentVersion:            settings.AntigravityUserAgentVersion,
-		OpenAICodexUserAgent:                   settings.OpenAICodexUserAgent,
-		MinCodexVersion:                        settings.MinCodexVersion,
-		MaxCodexVersion:                        settings.MaxCodexVersion,
-		CodexCLIOnlyBlacklist:                  settings.CodexCLIOnlyBlacklist,
-		CodexCLIOnlyWhitelist:                  settings.CodexCLIOnlyWhitelist,
-		CodexCLIOnlyAllowAppServerClients:      settings.CodexCLIOnlyAllowAppServerClients,
-		CodexCLIOnlyEngineFingerprintSignals:   settings.CodexCLIOnlyEngineFingerprintSignals,
-		WebSearchEmulationEnabled:              settings.WebSearchEmulationEnabled,
-		PaymentVisibleMethodAlipaySource:       settings.PaymentVisibleMethodAlipaySource,
-		PaymentVisibleMethodWxpaySource:        settings.PaymentVisibleMethodWxpaySource,
-		PaymentVisibleMethodAlipayEnabled:      settings.PaymentVisibleMethodAlipayEnabled,
-		PaymentVisibleMethodWxpayEnabled:       settings.PaymentVisibleMethodWxpayEnabled,
-		OpenAIAdvancedSchedulerEnabled:         settings.OpenAIAdvancedSchedulerEnabled,
-		BalanceLowNotifyEnabled:                settings.BalanceLowNotifyEnabled,
-		BalanceLowNotifyThreshold:              settings.BalanceLowNotifyThreshold,
-		BalanceLowNotifyRechargeURL:            settings.BalanceLowNotifyRechargeURL,
-		SubscriptionExpiryNotifyEnabled:        settings.SubscriptionExpiryNotifyEnabled,
-		AccountQuotaNotifyEnabled:              settings.AccountQuotaNotifyEnabled,
-		AccountQuotaNotifyEmails:               dto.NotifyEmailEntriesFromService(settings.AccountQuotaNotifyEmails),
-		PaymentEnabled:                         paymentCfg.Enabled,
-		PaymentMinAmount:                       paymentCfg.MinAmount,
-		PaymentMaxAmount:                       paymentCfg.MaxAmount,
-		PaymentDailyLimit:                      paymentCfg.DailyLimit,
-		PaymentOrderTimeoutMin:                 paymentCfg.OrderTimeoutMin,
-		PaymentMaxPendingOrders:                paymentCfg.MaxPendingOrders,
-		PaymentEnabledTypes:                    paymentCfg.EnabledTypes,
-		PaymentBalanceDisabled:                 paymentCfg.BalanceDisabled,
-		PaymentBalanceRechargeMultiplier:       paymentCfg.BalanceRechargeMultiplier,
-		PaymentRechargeFeeRate:                 paymentCfg.RechargeFeeRate,
-		PaymentLoadBalanceStrat:                paymentCfg.LoadBalanceStrategy,
-		PaymentProductNamePrefix:               paymentCfg.ProductNamePrefix,
-		PaymentProductNameSuffix:               paymentCfg.ProductNameSuffix,
-		PaymentHelpImageURL:                    paymentCfg.HelpImageURL,
-		PaymentHelpText:                        paymentCfg.HelpText,
-		PaymentCancelRateLimitEnabled:          paymentCfg.CancelRateLimitEnabled,
-		PaymentCancelRateLimitMax:              paymentCfg.CancelRateLimitMax,
-		PaymentCancelRateLimitWindow:           paymentCfg.CancelRateLimitWindow,
-		PaymentCancelRateLimitUnit:             paymentCfg.CancelRateLimitUnit,
-		PaymentCancelRateLimitMode:             paymentCfg.CancelRateLimitMode,
-		PaymentAlipayForceQRCode:               paymentCfg.AlipayForceQRCode,
+		RegistrationEnabled:                                    settings.RegistrationEnabled,
+		EmailVerifyEnabled:                                     settings.EmailVerifyEnabled,
+		RegistrationEmailSuffixWhitelist:                       settings.RegistrationEmailSuffixWhitelist,
+		PromoCodeEnabled:                                       settings.PromoCodeEnabled,
+		PasswordResetEnabled:                                   settings.PasswordResetEnabled,
+		FrontendURL:                                            settings.FrontendURL,
+		InvitationCodeEnabled:                                  settings.InvitationCodeEnabled,
+		TotpEnabled:                                            settings.TotpEnabled,
+		TotpEncryptionKeyConfigured:                            h.settingService.IsTotpEncryptionKeyConfigured(),
+		LoginAgreementEnabled:                                  settings.LoginAgreementEnabled,
+		LoginAgreementMode:                                     settings.LoginAgreementMode,
+		LoginAgreementUpdatedAt:                                settings.LoginAgreementUpdatedAt,
+		LoginAgreementDocuments:                                loginAgreementDocumentsToDTO(settings.LoginAgreementDocuments),
+		SMTPHost:                                               settings.SMTPHost,
+		SMTPPort:                                               settings.SMTPPort,
+		SMTPUsername:                                           settings.SMTPUsername,
+		SMTPPasswordConfigured:                                 settings.SMTPPasswordConfigured,
+		SMTPFrom:                                               settings.SMTPFrom,
+		SMTPFromName:                                           settings.SMTPFromName,
+		SMTPUseTLS:                                             settings.SMTPUseTLS,
+		TurnstileEnabled:                                       settings.TurnstileEnabled,
+		TurnstileSiteKey:                                       settings.TurnstileSiteKey,
+		TurnstileSecretKeyConfigured:                           settings.TurnstileSecretKeyConfigured,
+		APIKeyACLTrustForwardedIP:                              settings.APIKeyACLTrustForwardedIP,
+		LinuxDoConnectEnabled:                                  settings.LinuxDoConnectEnabled,
+		LinuxDoConnectClientID:                                 settings.LinuxDoConnectClientID,
+		LinuxDoConnectClientSecretConfigured:                   settings.LinuxDoConnectClientSecretConfigured,
+		LinuxDoConnectRedirectURL:                              settings.LinuxDoConnectRedirectURL,
+		DingTalkConnectEnabled:                                 settings.DingTalkConnectEnabled,
+		DingTalkConnectClientID:                                settings.DingTalkConnectClientID,
+		DingTalkConnectClientSecretConfigured:                  settings.DingTalkConnectClientSecretConfigured,
+		DingTalkConnectRedirectURL:                             settings.DingTalkConnectRedirectURL,
+		DingTalkConnectCorpRestrictionPolicy:                   settings.DingTalkConnectCorpRestrictionPolicy,
+		DingTalkConnectInternalCorpID:                          settings.DingTalkConnectInternalCorpID,
+		DingTalkConnectBypassRegistration:                      settings.DingTalkConnectBypassRegistration,
+		DingTalkConnectSyncCorpEmail:                           settings.DingTalkConnectSyncCorpEmail,
+		DingTalkConnectSyncDisplayName:                         settings.DingTalkConnectSyncDisplayName,
+		DingTalkConnectSyncDept:                                settings.DingTalkConnectSyncDept,
+		DingTalkConnectSyncCorpEmailAttrKey:                    settings.DingTalkConnectSyncCorpEmailAttrKey,
+		DingTalkConnectSyncDisplayNameAttrKey:                  settings.DingTalkConnectSyncDisplayNameAttrKey,
+		DingTalkConnectSyncDeptAttrKey:                         settings.DingTalkConnectSyncDeptAttrKey,
+		DingTalkConnectSyncCorpEmailAttrName:                   settings.DingTalkConnectSyncCorpEmailAttrName,
+		DingTalkConnectSyncDisplayNameAttrName:                 settings.DingTalkConnectSyncDisplayNameAttrName,
+		DingTalkConnectSyncDeptAttrName:                        settings.DingTalkConnectSyncDeptAttrName,
+		WeChatConnectEnabled:                                   settings.WeChatConnectEnabled,
+		WeChatConnectAppID:                                     settings.WeChatConnectAppID,
+		WeChatConnectAppSecretConfigured:                       settings.WeChatConnectAppSecretConfigured,
+		WeChatConnectOpenAppID:                                 settings.WeChatConnectOpenAppID,
+		WeChatConnectOpenAppSecretConfigured:                   settings.WeChatConnectOpenAppSecretConfigured,
+		WeChatConnectMPAppID:                                   settings.WeChatConnectMPAppID,
+		WeChatConnectMPAppSecretConfigured:                     settings.WeChatConnectMPAppSecretConfigured,
+		WeChatConnectMobileAppID:                               settings.WeChatConnectMobileAppID,
+		WeChatConnectMobileAppSecretConfigured:                 settings.WeChatConnectMobileAppSecretConfigured,
+		WeChatConnectOpenEnabled:                               settings.WeChatConnectOpenEnabled,
+		WeChatConnectMPEnabled:                                 settings.WeChatConnectMPEnabled,
+		WeChatConnectMobileEnabled:                             settings.WeChatConnectMobileEnabled,
+		WeChatConnectMode:                                      settings.WeChatConnectMode,
+		WeChatConnectScopes:                                    settings.WeChatConnectScopes,
+		WeChatConnectRedirectURL:                               settings.WeChatConnectRedirectURL,
+		WeChatConnectFrontendRedirectURL:                       settings.WeChatConnectFrontendRedirectURL,
+		OIDCConnectEnabled:                                     settings.OIDCConnectEnabled,
+		OIDCConnectProviderName:                                settings.OIDCConnectProviderName,
+		OIDCConnectClientID:                                    settings.OIDCConnectClientID,
+		OIDCConnectClientSecretConfigured:                      settings.OIDCConnectClientSecretConfigured,
+		OIDCConnectIssuerURL:                                   settings.OIDCConnectIssuerURL,
+		OIDCConnectDiscoveryURL:                                settings.OIDCConnectDiscoveryURL,
+		OIDCConnectAuthorizeURL:                                settings.OIDCConnectAuthorizeURL,
+		OIDCConnectTokenURL:                                    settings.OIDCConnectTokenURL,
+		OIDCConnectUserInfoURL:                                 settings.OIDCConnectUserInfoURL,
+		OIDCConnectJWKSURL:                                     settings.OIDCConnectJWKSURL,
+		OIDCConnectScopes:                                      settings.OIDCConnectScopes,
+		OIDCConnectRedirectURL:                                 settings.OIDCConnectRedirectURL,
+		OIDCConnectFrontendRedirectURL:                         settings.OIDCConnectFrontendRedirectURL,
+		OIDCConnectTokenAuthMethod:                             settings.OIDCConnectTokenAuthMethod,
+		OIDCConnectUsePKCE:                                     settings.OIDCConnectUsePKCE,
+		OIDCConnectValidateIDToken:                             settings.OIDCConnectValidateIDToken,
+		OIDCConnectAllowedSigningAlgs:                          settings.OIDCConnectAllowedSigningAlgs,
+		OIDCConnectClockSkewSeconds:                            settings.OIDCConnectClockSkewSeconds,
+		OIDCConnectRequireEmailVerified:                        settings.OIDCConnectRequireEmailVerified,
+		OIDCConnectUserInfoEmailPath:                           settings.OIDCConnectUserInfoEmailPath,
+		OIDCConnectUserInfoIDPath:                              settings.OIDCConnectUserInfoIDPath,
+		OIDCConnectUserInfoUsernamePath:                        settings.OIDCConnectUserInfoUsernamePath,
+		GitHubOAuthEnabled:                                     settings.GitHubOAuthEnabled,
+		GitHubOAuthClientID:                                    settings.GitHubOAuthClientID,
+		GitHubOAuthClientSecretConfigured:                      settings.GitHubOAuthClientSecretConfigured,
+		GitHubOAuthRedirectURL:                                 settings.GitHubOAuthRedirectURL,
+		GitHubOAuthFrontendRedirectURL:                         settings.GitHubOAuthFrontendRedirectURL,
+		GoogleOAuthEnabled:                                     settings.GoogleOAuthEnabled,
+		GoogleOAuthClientID:                                    settings.GoogleOAuthClientID,
+		GoogleOAuthClientSecretConfigured:                      settings.GoogleOAuthClientSecretConfigured,
+		GoogleOAuthRedirectURL:                                 settings.GoogleOAuthRedirectURL,
+		GoogleOAuthFrontendRedirectURL:                         settings.GoogleOAuthFrontendRedirectURL,
+		SiteName:                                               settings.SiteName,
+		SiteLogo:                                               settings.SiteLogo,
+		SiteSubtitle:                                           settings.SiteSubtitle,
+		APIBaseURL:                                             settings.APIBaseURL,
+		ContactInfo:                                            settings.ContactInfo,
+		DocURL:                                                 settings.DocURL,
+		HomeContent:                                            settings.HomeContent,
+		HideCcsImportButton:                                    settings.HideCcsImportButton,
+		PurchaseSubscriptionEnabled:                            settings.PurchaseSubscriptionEnabled,
+		PurchaseSubscriptionURL:                                settings.PurchaseSubscriptionURL,
+		TableDefaultPageSize:                                   settings.TableDefaultPageSize,
+		TablePageSizeOptions:                                   settings.TablePageSizeOptions,
+		CustomMenuItems:                                        dto.ParseCustomMenuItems(settings.CustomMenuItems),
+		CustomEndpoints:                                        dto.ParseCustomEndpoints(settings.CustomEndpoints),
+		DefaultConcurrency:                                     settings.DefaultConcurrency,
+		DefaultBalance:                                         settings.DefaultBalance,
+		RiskControlEnabled:                                     settings.RiskControlEnabled,
+		CyberSessionBlockEnabled:                               settings.CyberSessionBlockEnabled,
+		CyberSessionBlockTTLSeconds:                            settings.CyberSessionBlockTTLSeconds,
+		AffiliateRebateRate:                                    settings.AffiliateRebateRate,
+		AffiliateRebateFreezeHours:                             settings.AffiliateRebateFreezeHours,
+		AffiliateRebateDurationDays:                            settings.AffiliateRebateDurationDays,
+		AffiliateRebatePerInviteeCap:                           settings.AffiliateRebatePerInviteeCap,
+		DefaultUserRPMLimit:                                    settings.DefaultUserRPMLimit,
+		DefaultSubscriptions:                                   defaultSubscriptions,
+		EnableModelFallback:                                    settings.EnableModelFallback,
+		FallbackModelAnthropic:                                 settings.FallbackModelAnthropic,
+		FallbackModelOpenAI:                                    settings.FallbackModelOpenAI,
+		FallbackModelGemini:                                    settings.FallbackModelGemini,
+		FallbackModelAntigravity:                               settings.FallbackModelAntigravity,
+		EnableIdentityPatch:                                    settings.EnableIdentityPatch,
+		IdentityPatchPrompt:                                    settings.IdentityPatchPrompt,
+		OpsMonitoringEnabled:                                   opsEnabled && settings.OpsMonitoringEnabled,
+		OpsRealtimeMonitoringEnabled:                           settings.OpsRealtimeMonitoringEnabled,
+		OpsQueryModeDefault:                                    settings.OpsQueryModeDefault,
+		OpsMetricsIntervalSeconds:                              settings.OpsMetricsIntervalSeconds,
+		MinClaudeCodeVersion:                                   settings.MinClaudeCodeVersion,
+		MaxClaudeCodeVersion:                                   settings.MaxClaudeCodeVersion,
+		AllowUngroupedKeyScheduling:                            settings.AllowUngroupedKeyScheduling,
+		BackendModeEnabled:                                     settings.BackendModeEnabled,
+		EnableFingerprintUnification:                           settings.EnableFingerprintUnification,
+		EnableMetadataPassthrough:                              settings.EnableMetadataPassthrough,
+		EnableCCHSigning:                                       settings.EnableCCHSigning,
+		EnableClaudeOAuthSystemPromptInjection:                 settings.EnableClaudeOAuthSystemPromptInjection,
+		ClaudeOAuthSystemPrompt:                                settings.ClaudeOAuthSystemPrompt,
+		ClaudeOAuthSystemPromptBlocks:                          settings.ClaudeOAuthSystemPromptBlocks,
+		EnableAnthropicCacheTTL1hInjection:                     settings.EnableAnthropicCacheTTL1hInjection,
+		RewriteMessageCacheControl:                             settings.RewriteMessageCacheControl,
+		EnableClientDatelineNormalization:                      settings.EnableClientDatelineNormalization,
+		AntigravityUserAgentVersion:                            settings.AntigravityUserAgentVersion,
+		OpenAICodexUserAgent:                                   settings.OpenAICodexUserAgent,
+		MinCodexVersion:                                        settings.MinCodexVersion,
+		MaxCodexVersion:                                        settings.MaxCodexVersion,
+		CodexCLIOnlyBlacklist:                                  settings.CodexCLIOnlyBlacklist,
+		CodexCLIOnlyWhitelist:                                  settings.CodexCLIOnlyWhitelist,
+		CodexCLIOnlyAllowAppServerClients:                      settings.CodexCLIOnlyAllowAppServerClients,
+		CodexCLIOnlyEngineFingerprintSignals:                   settings.CodexCLIOnlyEngineFingerprintSignals,
+		WebSearchEmulationEnabled:                              settings.WebSearchEmulationEnabled,
+		PaymentVisibleMethodAlipaySource:                       settings.PaymentVisibleMethodAlipaySource,
+		PaymentVisibleMethodWxpaySource:                        settings.PaymentVisibleMethodWxpaySource,
+		PaymentVisibleMethodAlipayEnabled:                      settings.PaymentVisibleMethodAlipayEnabled,
+		PaymentVisibleMethodWxpayEnabled:                       settings.PaymentVisibleMethodWxpayEnabled,
+		OpenAIAdvancedSchedulerEnabled:                         settings.OpenAIAdvancedSchedulerEnabled,
+		OpenAIAdvancedSchedulerStickyWeightedEnabled:           settings.OpenAIAdvancedSchedulerStickyWeightedEnabled,
+		OpenAIAdvancedSchedulerSubscriptionPriorityEnabled:     settings.OpenAIAdvancedSchedulerSubscriptionPriorityEnabled,
+		OpenAIAdvancedSchedulerLBTopK:                          settings.OpenAIAdvancedSchedulerLBTopK,
+		OpenAIAdvancedSchedulerWeightPriority:                  settings.OpenAIAdvancedSchedulerWeightPriority,
+		OpenAIAdvancedSchedulerWeightLoad:                      settings.OpenAIAdvancedSchedulerWeightLoad,
+		OpenAIAdvancedSchedulerWeightQueue:                     settings.OpenAIAdvancedSchedulerWeightQueue,
+		OpenAIAdvancedSchedulerWeightErrorRate:                 settings.OpenAIAdvancedSchedulerWeightErrorRate,
+		OpenAIAdvancedSchedulerWeightTTFT:                      settings.OpenAIAdvancedSchedulerWeightTTFT,
+		OpenAIAdvancedSchedulerWeightReset:                     settings.OpenAIAdvancedSchedulerWeightReset,
+		OpenAIAdvancedSchedulerWeightQuotaHeadroom:             settings.OpenAIAdvancedSchedulerWeightQuotaHeadroom,
+		OpenAIAdvancedSchedulerWeightPreviousResponse:          settings.OpenAIAdvancedSchedulerWeightPreviousResponse,
+		OpenAIAdvancedSchedulerWeightSessionSticky:             settings.OpenAIAdvancedSchedulerWeightSessionSticky,
+		OpenAIAdvancedSchedulerEffectiveLBTopK:                 settings.OpenAIAdvancedSchedulerEffectiveLBTopK,
+		OpenAIAdvancedSchedulerEffectiveWeightPriority:         settings.OpenAIAdvancedSchedulerEffectiveWeightPriority,
+		OpenAIAdvancedSchedulerEffectiveWeightLoad:             settings.OpenAIAdvancedSchedulerEffectiveWeightLoad,
+		OpenAIAdvancedSchedulerEffectiveWeightQueue:            settings.OpenAIAdvancedSchedulerEffectiveWeightQueue,
+		OpenAIAdvancedSchedulerEffectiveWeightErrorRate:        settings.OpenAIAdvancedSchedulerEffectiveWeightErrorRate,
+		OpenAIAdvancedSchedulerEffectiveWeightTTFT:             settings.OpenAIAdvancedSchedulerEffectiveWeightTTFT,
+		OpenAIAdvancedSchedulerEffectiveWeightReset:            settings.OpenAIAdvancedSchedulerEffectiveWeightReset,
+		OpenAIAdvancedSchedulerEffectiveWeightQuotaHeadroom:    settings.OpenAIAdvancedSchedulerEffectiveWeightQuotaHeadroom,
+		OpenAIAdvancedSchedulerEffectiveWeightPreviousResponse: settings.OpenAIAdvancedSchedulerEffectiveWeightPreviousResponse,
+		OpenAIAdvancedSchedulerEffectiveWeightSessionSticky:    settings.OpenAIAdvancedSchedulerEffectiveWeightSessionSticky,
+		BalanceLowNotifyEnabled:                                settings.BalanceLowNotifyEnabled,
+		BalanceLowNotifyThreshold:                              settings.BalanceLowNotifyThreshold,
+		BalanceLowNotifyRechargeURL:                            settings.BalanceLowNotifyRechargeURL,
+		SubscriptionExpiryNotifyEnabled:                        settings.SubscriptionExpiryNotifyEnabled,
+		AccountQuotaNotifyEnabled:                              settings.AccountQuotaNotifyEnabled,
+		AccountQuotaNotifyEmails:                               dto.NotifyEmailEntriesFromService(settings.AccountQuotaNotifyEmails),
+		PaymentEnabled:                                         paymentCfg.Enabled,
+		PaymentMinAmount:                                       paymentCfg.MinAmount,
+		PaymentMaxAmount:                                       paymentCfg.MaxAmount,
+		PaymentDailyLimit:                                      paymentCfg.DailyLimit,
+		PaymentOrderTimeoutMin:                                 paymentCfg.OrderTimeoutMin,
+		PaymentMaxPendingOrders:                                paymentCfg.MaxPendingOrders,
+		PaymentEnabledTypes:                                    paymentCfg.EnabledTypes,
+		PaymentBalanceDisabled:                                 paymentCfg.BalanceDisabled,
+		PaymentBalanceRechargeMultiplier:                       paymentCfg.BalanceRechargeMultiplier,
+		PaymentRechargeFeeRate:                                 paymentCfg.RechargeFeeRate,
+		PaymentLoadBalanceStrat:                                paymentCfg.LoadBalanceStrategy,
+		PaymentProductNamePrefix:                               paymentCfg.ProductNamePrefix,
+		PaymentProductNameSuffix:                               paymentCfg.ProductNameSuffix,
+		PaymentHelpImageURL:                                    paymentCfg.HelpImageURL,
+		PaymentHelpText:                                        paymentCfg.HelpText,
+		PaymentCancelRateLimitEnabled:                          paymentCfg.CancelRateLimitEnabled,
+		PaymentCancelRateLimitMax:                              paymentCfg.CancelRateLimitMax,
+		PaymentCancelRateLimitWindow:                           paymentCfg.CancelRateLimitWindow,
+		PaymentCancelRateLimitUnit:                             paymentCfg.CancelRateLimitUnit,
+		PaymentCancelRateLimitMode:                             paymentCfg.CancelRateLimitMode,
+		PaymentAlipayForceQRCode:                               paymentCfg.AlipayForceQRCode,
 
 		ChannelMonitorEnabled:                settings.ChannelMonitorEnabled,
 		ChannelMonitorDefaultIntervalSeconds: settings.ChannelMonitorDefaultIntervalSeconds,
@@ -618,7 +640,19 @@ type UpdateSettingsRequest struct {
 	PaymentVisibleMethodWxpayEnabled  *bool   `json:"payment_visible_method_wxpay_enabled"`
 
 	// OpenAI account scheduling
-	OpenAIAdvancedSchedulerEnabled *bool `json:"openai_advanced_scheduler_enabled"`
+	OpenAIAdvancedSchedulerEnabled                     *bool   `json:"openai_advanced_scheduler_enabled"`
+	OpenAIAdvancedSchedulerStickyWeightedEnabled       *bool   `json:"openai_advanced_scheduler_sticky_weighted_enabled"`
+	OpenAIAdvancedSchedulerSubscriptionPriorityEnabled *bool   `json:"openai_advanced_scheduler_subscription_priority_enabled"`
+	OpenAIAdvancedSchedulerLBTopK                      *string `json:"openai_advanced_scheduler_lb_top_k"`
+	OpenAIAdvancedSchedulerWeightPriority              *string `json:"openai_advanced_scheduler_weight_priority"`
+	OpenAIAdvancedSchedulerWeightLoad                  *string `json:"openai_advanced_scheduler_weight_load"`
+	OpenAIAdvancedSchedulerWeightQueue                 *string `json:"openai_advanced_scheduler_weight_queue"`
+	OpenAIAdvancedSchedulerWeightErrorRate             *string `json:"openai_advanced_scheduler_weight_error_rate"`
+	OpenAIAdvancedSchedulerWeightTTFT                  *string `json:"openai_advanced_scheduler_weight_ttft"`
+	OpenAIAdvancedSchedulerWeightReset                 *string `json:"openai_advanced_scheduler_weight_reset"`
+	OpenAIAdvancedSchedulerWeightQuotaHeadroom         *string `json:"openai_advanced_scheduler_weight_quota_headroom"`
+	OpenAIAdvancedSchedulerWeightPreviousResponse      *string `json:"openai_advanced_scheduler_weight_previous_response"`
+	OpenAIAdvancedSchedulerWeightSessionSticky         *string `json:"openai_advanced_scheduler_weight_session_sticky"`
 
 	// 余额不足提醒
 	BalanceLowNotifyEnabled         *bool                   `json:"balance_low_notify_enabled"`
@@ -1792,6 +1826,28 @@ func (h *SettingHandler) UpdateSettings(c *gin.Context) {
 			}
 			return previousSettings.OpenAIAdvancedSchedulerEnabled
 		}(),
+		OpenAIAdvancedSchedulerStickyWeightedEnabled: func() bool {
+			if req.OpenAIAdvancedSchedulerStickyWeightedEnabled != nil {
+				return *req.OpenAIAdvancedSchedulerStickyWeightedEnabled
+			}
+			return previousSettings.OpenAIAdvancedSchedulerStickyWeightedEnabled
+		}(),
+		OpenAIAdvancedSchedulerSubscriptionPriorityEnabled: func() bool {
+			if req.OpenAIAdvancedSchedulerSubscriptionPriorityEnabled != nil {
+				return *req.OpenAIAdvancedSchedulerSubscriptionPriorityEnabled
+			}
+			return previousSettings.OpenAIAdvancedSchedulerSubscriptionPriorityEnabled
+		}(),
+		OpenAIAdvancedSchedulerLBTopK:                 stringSetting(req.OpenAIAdvancedSchedulerLBTopK, previousSettings.OpenAIAdvancedSchedulerLBTopK),
+		OpenAIAdvancedSchedulerWeightPriority:         stringSetting(req.OpenAIAdvancedSchedulerWeightPriority, previousSettings.OpenAIAdvancedSchedulerWeightPriority),
+		OpenAIAdvancedSchedulerWeightLoad:             stringSetting(req.OpenAIAdvancedSchedulerWeightLoad, previousSettings.OpenAIAdvancedSchedulerWeightLoad),
+		OpenAIAdvancedSchedulerWeightQueue:            stringSetting(req.OpenAIAdvancedSchedulerWeightQueue, previousSettings.OpenAIAdvancedSchedulerWeightQueue),
+		OpenAIAdvancedSchedulerWeightErrorRate:        stringSetting(req.OpenAIAdvancedSchedulerWeightErrorRate, previousSettings.OpenAIAdvancedSchedulerWeightErrorRate),
+		OpenAIAdvancedSchedulerWeightTTFT:             stringSetting(req.OpenAIAdvancedSchedulerWeightTTFT, previousSettings.OpenAIAdvancedSchedulerWeightTTFT),
+		OpenAIAdvancedSchedulerWeightReset:            stringSetting(req.OpenAIAdvancedSchedulerWeightReset, previousSettings.OpenAIAdvancedSchedulerWeightReset),
+		OpenAIAdvancedSchedulerWeightQuotaHeadroom:    stringSetting(req.OpenAIAdvancedSchedulerWeightQuotaHeadroom, previousSettings.OpenAIAdvancedSchedulerWeightQuotaHeadroom),
+		OpenAIAdvancedSchedulerWeightPreviousResponse: stringSetting(req.OpenAIAdvancedSchedulerWeightPreviousResponse, previousSettings.OpenAIAdvancedSchedulerWeightPreviousResponse),
+		OpenAIAdvancedSchedulerWeightSessionSticky:    stringSetting(req.OpenAIAdvancedSchedulerWeightSessionSticky, previousSettings.OpenAIAdvancedSchedulerWeightSessionSticky),
 		BalanceLowNotifyEnabled: func() bool {
 			if req.BalanceLowNotifyEnabled != nil {
 				return *req.BalanceLowNotifyEnabled
@@ -2014,184 +2070,206 @@ func (h *SettingHandler) UpdateSettings(c *gin.Context) {
 	}
 
 	payload := dto.SystemSettings{
-		RegistrationEnabled:                    updatedSettings.RegistrationEnabled,
-		EmailVerifyEnabled:                     updatedSettings.EmailVerifyEnabled,
-		RegistrationEmailSuffixWhitelist:       updatedSettings.RegistrationEmailSuffixWhitelist,
-		PromoCodeEnabled:                       updatedSettings.PromoCodeEnabled,
-		PasswordResetEnabled:                   updatedSettings.PasswordResetEnabled,
-		FrontendURL:                            updatedSettings.FrontendURL,
-		InvitationCodeEnabled:                  updatedSettings.InvitationCodeEnabled,
-		TotpEnabled:                            updatedSettings.TotpEnabled,
-		TotpEncryptionKeyConfigured:            h.settingService.IsTotpEncryptionKeyConfigured(),
-		LoginAgreementEnabled:                  updatedSettings.LoginAgreementEnabled,
-		LoginAgreementMode:                     updatedSettings.LoginAgreementMode,
-		LoginAgreementUpdatedAt:                updatedSettings.LoginAgreementUpdatedAt,
-		LoginAgreementDocuments:                loginAgreementDocumentsToDTO(updatedSettings.LoginAgreementDocuments),
-		SMTPHost:                               updatedSettings.SMTPHost,
-		SMTPPort:                               updatedSettings.SMTPPort,
-		SMTPUsername:                           updatedSettings.SMTPUsername,
-		SMTPPasswordConfigured:                 updatedSettings.SMTPPasswordConfigured,
-		SMTPFrom:                               updatedSettings.SMTPFrom,
-		SMTPFromName:                           updatedSettings.SMTPFromName,
-		SMTPUseTLS:                             updatedSettings.SMTPUseTLS,
-		TurnstileEnabled:                       updatedSettings.TurnstileEnabled,
-		TurnstileSiteKey:                       updatedSettings.TurnstileSiteKey,
-		TurnstileSecretKeyConfigured:           updatedSettings.TurnstileSecretKeyConfigured,
-		APIKeyACLTrustForwardedIP:              updatedSettings.APIKeyACLTrustForwardedIP,
-		LinuxDoConnectEnabled:                  updatedSettings.LinuxDoConnectEnabled,
-		LinuxDoConnectClientID:                 updatedSettings.LinuxDoConnectClientID,
-		LinuxDoConnectClientSecretConfigured:   updatedSettings.LinuxDoConnectClientSecretConfigured,
-		LinuxDoConnectRedirectURL:              updatedSettings.LinuxDoConnectRedirectURL,
-		DingTalkConnectEnabled:                 updatedSettings.DingTalkConnectEnabled,
-		DingTalkConnectClientID:                updatedSettings.DingTalkConnectClientID,
-		DingTalkConnectClientSecretConfigured:  updatedSettings.DingTalkConnectClientSecretConfigured,
-		DingTalkConnectRedirectURL:             updatedSettings.DingTalkConnectRedirectURL,
-		DingTalkConnectCorpRestrictionPolicy:   updatedSettings.DingTalkConnectCorpRestrictionPolicy,
-		DingTalkConnectInternalCorpID:          updatedSettings.DingTalkConnectInternalCorpID,
-		DingTalkConnectBypassRegistration:      updatedSettings.DingTalkConnectBypassRegistration,
-		DingTalkConnectSyncCorpEmail:           updatedSettings.DingTalkConnectSyncCorpEmail,
-		DingTalkConnectSyncDisplayName:         updatedSettings.DingTalkConnectSyncDisplayName,
-		DingTalkConnectSyncDept:                updatedSettings.DingTalkConnectSyncDept,
-		DingTalkConnectSyncCorpEmailAttrKey:    updatedSettings.DingTalkConnectSyncCorpEmailAttrKey,
-		DingTalkConnectSyncDisplayNameAttrKey:  updatedSettings.DingTalkConnectSyncDisplayNameAttrKey,
-		DingTalkConnectSyncDeptAttrKey:         updatedSettings.DingTalkConnectSyncDeptAttrKey,
-		DingTalkConnectSyncCorpEmailAttrName:   updatedSettings.DingTalkConnectSyncCorpEmailAttrName,
-		DingTalkConnectSyncDisplayNameAttrName: updatedSettings.DingTalkConnectSyncDisplayNameAttrName,
-		DingTalkConnectSyncDeptAttrName:        updatedSettings.DingTalkConnectSyncDeptAttrName,
-		WeChatConnectEnabled:                   updatedSettings.WeChatConnectEnabled,
-		WeChatConnectAppID:                     updatedSettings.WeChatConnectAppID,
-		WeChatConnectAppSecretConfigured:       updatedSettings.WeChatConnectAppSecretConfigured,
-		WeChatConnectOpenAppID:                 updatedSettings.WeChatConnectOpenAppID,
-		WeChatConnectOpenAppSecretConfigured:   updatedSettings.WeChatConnectOpenAppSecretConfigured,
-		WeChatConnectMPAppID:                   updatedSettings.WeChatConnectMPAppID,
-		WeChatConnectMPAppSecretConfigured:     updatedSettings.WeChatConnectMPAppSecretConfigured,
-		WeChatConnectMobileAppID:               updatedSettings.WeChatConnectMobileAppID,
-		WeChatConnectMobileAppSecretConfigured: updatedSettings.WeChatConnectMobileAppSecretConfigured,
-		WeChatConnectOpenEnabled:               updatedSettings.WeChatConnectOpenEnabled,
-		WeChatConnectMPEnabled:                 updatedSettings.WeChatConnectMPEnabled,
-		WeChatConnectMobileEnabled:             updatedSettings.WeChatConnectMobileEnabled,
-		WeChatConnectMode:                      updatedSettings.WeChatConnectMode,
-		WeChatConnectScopes:                    updatedSettings.WeChatConnectScopes,
-		WeChatConnectRedirectURL:               updatedSettings.WeChatConnectRedirectURL,
-		WeChatConnectFrontendRedirectURL:       updatedSettings.WeChatConnectFrontendRedirectURL,
-		OIDCConnectEnabled:                     updatedSettings.OIDCConnectEnabled,
-		OIDCConnectProviderName:                updatedSettings.OIDCConnectProviderName,
-		OIDCConnectClientID:                    updatedSettings.OIDCConnectClientID,
-		OIDCConnectClientSecretConfigured:      updatedSettings.OIDCConnectClientSecretConfigured,
-		OIDCConnectIssuerURL:                   updatedSettings.OIDCConnectIssuerURL,
-		OIDCConnectDiscoveryURL:                updatedSettings.OIDCConnectDiscoveryURL,
-		OIDCConnectAuthorizeURL:                updatedSettings.OIDCConnectAuthorizeURL,
-		OIDCConnectTokenURL:                    updatedSettings.OIDCConnectTokenURL,
-		OIDCConnectUserInfoURL:                 updatedSettings.OIDCConnectUserInfoURL,
-		OIDCConnectJWKSURL:                     updatedSettings.OIDCConnectJWKSURL,
-		OIDCConnectScopes:                      updatedSettings.OIDCConnectScopes,
-		OIDCConnectRedirectURL:                 updatedSettings.OIDCConnectRedirectURL,
-		OIDCConnectFrontendRedirectURL:         updatedSettings.OIDCConnectFrontendRedirectURL,
-		OIDCConnectTokenAuthMethod:             updatedSettings.OIDCConnectTokenAuthMethod,
-		OIDCConnectUsePKCE:                     updatedSettings.OIDCConnectUsePKCE,
-		OIDCConnectValidateIDToken:             updatedSettings.OIDCConnectValidateIDToken,
-		OIDCConnectAllowedSigningAlgs:          updatedSettings.OIDCConnectAllowedSigningAlgs,
-		OIDCConnectClockSkewSeconds:            updatedSettings.OIDCConnectClockSkewSeconds,
-		OIDCConnectRequireEmailVerified:        updatedSettings.OIDCConnectRequireEmailVerified,
-		OIDCConnectUserInfoEmailPath:           updatedSettings.OIDCConnectUserInfoEmailPath,
-		OIDCConnectUserInfoIDPath:              updatedSettings.OIDCConnectUserInfoIDPath,
-		OIDCConnectUserInfoUsernamePath:        updatedSettings.OIDCConnectUserInfoUsernamePath,
-		GitHubOAuthEnabled:                     updatedSettings.GitHubOAuthEnabled,
-		GitHubOAuthClientID:                    updatedSettings.GitHubOAuthClientID,
-		GitHubOAuthClientSecretConfigured:      updatedSettings.GitHubOAuthClientSecretConfigured,
-		GitHubOAuthRedirectURL:                 updatedSettings.GitHubOAuthRedirectURL,
-		GitHubOAuthFrontendRedirectURL:         updatedSettings.GitHubOAuthFrontendRedirectURL,
-		GoogleOAuthEnabled:                     updatedSettings.GoogleOAuthEnabled,
-		GoogleOAuthClientID:                    updatedSettings.GoogleOAuthClientID,
-		GoogleOAuthClientSecretConfigured:      updatedSettings.GoogleOAuthClientSecretConfigured,
-		GoogleOAuthRedirectURL:                 updatedSettings.GoogleOAuthRedirectURL,
-		GoogleOAuthFrontendRedirectURL:         updatedSettings.GoogleOAuthFrontendRedirectURL,
-		SiteName:                               updatedSettings.SiteName,
-		SiteLogo:                               updatedSettings.SiteLogo,
-		SiteSubtitle:                           updatedSettings.SiteSubtitle,
-		APIBaseURL:                             updatedSettings.APIBaseURL,
-		ContactInfo:                            updatedSettings.ContactInfo,
-		DocURL:                                 updatedSettings.DocURL,
-		HomeContent:                            updatedSettings.HomeContent,
-		HideCcsImportButton:                    updatedSettings.HideCcsImportButton,
-		PurchaseSubscriptionEnabled:            updatedSettings.PurchaseSubscriptionEnabled,
-		PurchaseSubscriptionURL:                updatedSettings.PurchaseSubscriptionURL,
-		TableDefaultPageSize:                   updatedSettings.TableDefaultPageSize,
-		TablePageSizeOptions:                   updatedSettings.TablePageSizeOptions,
-		CustomMenuItems:                        dto.ParseCustomMenuItems(updatedSettings.CustomMenuItems),
-		CustomEndpoints:                        dto.ParseCustomEndpoints(updatedSettings.CustomEndpoints),
-		DefaultConcurrency:                     updatedSettings.DefaultConcurrency,
-		DefaultBalance:                         updatedSettings.DefaultBalance,
-		AffiliateRebateRate:                    updatedSettings.AffiliateRebateRate,
-		AffiliateRebateFreezeHours:             updatedSettings.AffiliateRebateFreezeHours,
-		AffiliateRebateDurationDays:            updatedSettings.AffiliateRebateDurationDays,
-		AffiliateRebatePerInviteeCap:           updatedSettings.AffiliateRebatePerInviteeCap,
-		DefaultUserRPMLimit:                    updatedSettings.DefaultUserRPMLimit,
-		DefaultSubscriptions:                   updatedDefaultSubscriptions,
-		EnableModelFallback:                    updatedSettings.EnableModelFallback,
-		FallbackModelAnthropic:                 updatedSettings.FallbackModelAnthropic,
-		FallbackModelOpenAI:                    updatedSettings.FallbackModelOpenAI,
-		FallbackModelGemini:                    updatedSettings.FallbackModelGemini,
-		FallbackModelAntigravity:               updatedSettings.FallbackModelAntigravity,
-		EnableIdentityPatch:                    updatedSettings.EnableIdentityPatch,
-		IdentityPatchPrompt:                    updatedSettings.IdentityPatchPrompt,
-		OpsMonitoringEnabled:                   updatedSettings.OpsMonitoringEnabled,
-		OpsRealtimeMonitoringEnabled:           updatedSettings.OpsRealtimeMonitoringEnabled,
-		OpsQueryModeDefault:                    updatedSettings.OpsQueryModeDefault,
-		OpsMetricsIntervalSeconds:              updatedSettings.OpsMetricsIntervalSeconds,
-		MinClaudeCodeVersion:                   updatedSettings.MinClaudeCodeVersion,
-		MaxClaudeCodeVersion:                   updatedSettings.MaxClaudeCodeVersion,
-		AllowUngroupedKeyScheduling:            updatedSettings.AllowUngroupedKeyScheduling,
-		BackendModeEnabled:                     updatedSettings.BackendModeEnabled,
-		EnableFingerprintUnification:           updatedSettings.EnableFingerprintUnification,
-		EnableMetadataPassthrough:              updatedSettings.EnableMetadataPassthrough,
-		EnableCCHSigning:                       updatedSettings.EnableCCHSigning,
-		EnableClaudeOAuthSystemPromptInjection: updatedSettings.EnableClaudeOAuthSystemPromptInjection,
-		ClaudeOAuthSystemPrompt:                updatedSettings.ClaudeOAuthSystemPrompt,
-		ClaudeOAuthSystemPromptBlocks:          updatedSettings.ClaudeOAuthSystemPromptBlocks,
-		EnableAnthropicCacheTTL1hInjection:     updatedSettings.EnableAnthropicCacheTTL1hInjection,
-		RewriteMessageCacheControl:             updatedSettings.RewriteMessageCacheControl,
-		EnableClientDatelineNormalization:      updatedSettings.EnableClientDatelineNormalization,
-		AntigravityUserAgentVersion:            updatedSettings.AntigravityUserAgentVersion,
-		OpenAICodexUserAgent:                   updatedSettings.OpenAICodexUserAgent,
-		MinCodexVersion:                        updatedSettings.MinCodexVersion,
-		MaxCodexVersion:                        updatedSettings.MaxCodexVersion,
-		CodexCLIOnlyBlacklist:                  updatedSettings.CodexCLIOnlyBlacklist,
-		CodexCLIOnlyWhitelist:                  updatedSettings.CodexCLIOnlyWhitelist,
-		CodexCLIOnlyAllowAppServerClients:      updatedSettings.CodexCLIOnlyAllowAppServerClients,
-		CodexCLIOnlyEngineFingerprintSignals:   updatedSettings.CodexCLIOnlyEngineFingerprintSignals,
-		PaymentVisibleMethodAlipaySource:       updatedSettings.PaymentVisibleMethodAlipaySource,
-		PaymentVisibleMethodWxpaySource:        updatedSettings.PaymentVisibleMethodWxpaySource,
-		PaymentVisibleMethodAlipayEnabled:      updatedSettings.PaymentVisibleMethodAlipayEnabled,
-		PaymentVisibleMethodWxpayEnabled:       updatedSettings.PaymentVisibleMethodWxpayEnabled,
-		OpenAIAdvancedSchedulerEnabled:         updatedSettings.OpenAIAdvancedSchedulerEnabled,
-		BalanceLowNotifyEnabled:                updatedSettings.BalanceLowNotifyEnabled,
-		BalanceLowNotifyThreshold:              updatedSettings.BalanceLowNotifyThreshold,
-		BalanceLowNotifyRechargeURL:            updatedSettings.BalanceLowNotifyRechargeURL,
-		SubscriptionExpiryNotifyEnabled:        updatedSettings.SubscriptionExpiryNotifyEnabled,
-		AccountQuotaNotifyEnabled:              updatedSettings.AccountQuotaNotifyEnabled,
-		AccountQuotaNotifyEmails:               dto.NotifyEmailEntriesFromService(updatedSettings.AccountQuotaNotifyEmails),
-		PaymentEnabled:                         updatedPaymentCfg.Enabled,
-		PaymentMinAmount:                       updatedPaymentCfg.MinAmount,
-		PaymentMaxAmount:                       updatedPaymentCfg.MaxAmount,
-		PaymentDailyLimit:                      updatedPaymentCfg.DailyLimit,
-		PaymentOrderTimeoutMin:                 updatedPaymentCfg.OrderTimeoutMin,
-		PaymentMaxPendingOrders:                updatedPaymentCfg.MaxPendingOrders,
-		PaymentEnabledTypes:                    updatedPaymentCfg.EnabledTypes,
-		PaymentBalanceDisabled:                 updatedPaymentCfg.BalanceDisabled,
-		PaymentBalanceRechargeMultiplier:       updatedPaymentCfg.BalanceRechargeMultiplier,
-		PaymentRechargeFeeRate:                 updatedPaymentCfg.RechargeFeeRate,
-		PaymentLoadBalanceStrat:                updatedPaymentCfg.LoadBalanceStrategy,
-		PaymentProductNamePrefix:               updatedPaymentCfg.ProductNamePrefix,
-		PaymentProductNameSuffix:               updatedPaymentCfg.ProductNameSuffix,
-		PaymentHelpImageURL:                    updatedPaymentCfg.HelpImageURL,
-		PaymentHelpText:                        updatedPaymentCfg.HelpText,
-		PaymentCancelRateLimitEnabled:          updatedPaymentCfg.CancelRateLimitEnabled,
-		PaymentCancelRateLimitMax:              updatedPaymentCfg.CancelRateLimitMax,
-		PaymentCancelRateLimitWindow:           updatedPaymentCfg.CancelRateLimitWindow,
-		PaymentCancelRateLimitUnit:             updatedPaymentCfg.CancelRateLimitUnit,
-		PaymentCancelRateLimitMode:             updatedPaymentCfg.CancelRateLimitMode,
-		PaymentAlipayForceQRCode:               updatedPaymentCfg.AlipayForceQRCode,
+		RegistrationEnabled:                                    updatedSettings.RegistrationEnabled,
+		EmailVerifyEnabled:                                     updatedSettings.EmailVerifyEnabled,
+		RegistrationEmailSuffixWhitelist:                       updatedSettings.RegistrationEmailSuffixWhitelist,
+		PromoCodeEnabled:                                       updatedSettings.PromoCodeEnabled,
+		PasswordResetEnabled:                                   updatedSettings.PasswordResetEnabled,
+		FrontendURL:                                            updatedSettings.FrontendURL,
+		InvitationCodeEnabled:                                  updatedSettings.InvitationCodeEnabled,
+		TotpEnabled:                                            updatedSettings.TotpEnabled,
+		TotpEncryptionKeyConfigured:                            h.settingService.IsTotpEncryptionKeyConfigured(),
+		LoginAgreementEnabled:                                  updatedSettings.LoginAgreementEnabled,
+		LoginAgreementMode:                                     updatedSettings.LoginAgreementMode,
+		LoginAgreementUpdatedAt:                                updatedSettings.LoginAgreementUpdatedAt,
+		LoginAgreementDocuments:                                loginAgreementDocumentsToDTO(updatedSettings.LoginAgreementDocuments),
+		SMTPHost:                                               updatedSettings.SMTPHost,
+		SMTPPort:                                               updatedSettings.SMTPPort,
+		SMTPUsername:                                           updatedSettings.SMTPUsername,
+		SMTPPasswordConfigured:                                 updatedSettings.SMTPPasswordConfigured,
+		SMTPFrom:                                               updatedSettings.SMTPFrom,
+		SMTPFromName:                                           updatedSettings.SMTPFromName,
+		SMTPUseTLS:                                             updatedSettings.SMTPUseTLS,
+		TurnstileEnabled:                                       updatedSettings.TurnstileEnabled,
+		TurnstileSiteKey:                                       updatedSettings.TurnstileSiteKey,
+		TurnstileSecretKeyConfigured:                           updatedSettings.TurnstileSecretKeyConfigured,
+		APIKeyACLTrustForwardedIP:                              updatedSettings.APIKeyACLTrustForwardedIP,
+		LinuxDoConnectEnabled:                                  updatedSettings.LinuxDoConnectEnabled,
+		LinuxDoConnectClientID:                                 updatedSettings.LinuxDoConnectClientID,
+		LinuxDoConnectClientSecretConfigured:                   updatedSettings.LinuxDoConnectClientSecretConfigured,
+		LinuxDoConnectRedirectURL:                              updatedSettings.LinuxDoConnectRedirectURL,
+		DingTalkConnectEnabled:                                 updatedSettings.DingTalkConnectEnabled,
+		DingTalkConnectClientID:                                updatedSettings.DingTalkConnectClientID,
+		DingTalkConnectClientSecretConfigured:                  updatedSettings.DingTalkConnectClientSecretConfigured,
+		DingTalkConnectRedirectURL:                             updatedSettings.DingTalkConnectRedirectURL,
+		DingTalkConnectCorpRestrictionPolicy:                   updatedSettings.DingTalkConnectCorpRestrictionPolicy,
+		DingTalkConnectInternalCorpID:                          updatedSettings.DingTalkConnectInternalCorpID,
+		DingTalkConnectBypassRegistration:                      updatedSettings.DingTalkConnectBypassRegistration,
+		DingTalkConnectSyncCorpEmail:                           updatedSettings.DingTalkConnectSyncCorpEmail,
+		DingTalkConnectSyncDisplayName:                         updatedSettings.DingTalkConnectSyncDisplayName,
+		DingTalkConnectSyncDept:                                updatedSettings.DingTalkConnectSyncDept,
+		DingTalkConnectSyncCorpEmailAttrKey:                    updatedSettings.DingTalkConnectSyncCorpEmailAttrKey,
+		DingTalkConnectSyncDisplayNameAttrKey:                  updatedSettings.DingTalkConnectSyncDisplayNameAttrKey,
+		DingTalkConnectSyncDeptAttrKey:                         updatedSettings.DingTalkConnectSyncDeptAttrKey,
+		DingTalkConnectSyncCorpEmailAttrName:                   updatedSettings.DingTalkConnectSyncCorpEmailAttrName,
+		DingTalkConnectSyncDisplayNameAttrName:                 updatedSettings.DingTalkConnectSyncDisplayNameAttrName,
+		DingTalkConnectSyncDeptAttrName:                        updatedSettings.DingTalkConnectSyncDeptAttrName,
+		WeChatConnectEnabled:                                   updatedSettings.WeChatConnectEnabled,
+		WeChatConnectAppID:                                     updatedSettings.WeChatConnectAppID,
+		WeChatConnectAppSecretConfigured:                       updatedSettings.WeChatConnectAppSecretConfigured,
+		WeChatConnectOpenAppID:                                 updatedSettings.WeChatConnectOpenAppID,
+		WeChatConnectOpenAppSecretConfigured:                   updatedSettings.WeChatConnectOpenAppSecretConfigured,
+		WeChatConnectMPAppID:                                   updatedSettings.WeChatConnectMPAppID,
+		WeChatConnectMPAppSecretConfigured:                     updatedSettings.WeChatConnectMPAppSecretConfigured,
+		WeChatConnectMobileAppID:                               updatedSettings.WeChatConnectMobileAppID,
+		WeChatConnectMobileAppSecretConfigured:                 updatedSettings.WeChatConnectMobileAppSecretConfigured,
+		WeChatConnectOpenEnabled:                               updatedSettings.WeChatConnectOpenEnabled,
+		WeChatConnectMPEnabled:                                 updatedSettings.WeChatConnectMPEnabled,
+		WeChatConnectMobileEnabled:                             updatedSettings.WeChatConnectMobileEnabled,
+		WeChatConnectMode:                                      updatedSettings.WeChatConnectMode,
+		WeChatConnectScopes:                                    updatedSettings.WeChatConnectScopes,
+		WeChatConnectRedirectURL:                               updatedSettings.WeChatConnectRedirectURL,
+		WeChatConnectFrontendRedirectURL:                       updatedSettings.WeChatConnectFrontendRedirectURL,
+		OIDCConnectEnabled:                                     updatedSettings.OIDCConnectEnabled,
+		OIDCConnectProviderName:                                updatedSettings.OIDCConnectProviderName,
+		OIDCConnectClientID:                                    updatedSettings.OIDCConnectClientID,
+		OIDCConnectClientSecretConfigured:                      updatedSettings.OIDCConnectClientSecretConfigured,
+		OIDCConnectIssuerURL:                                   updatedSettings.OIDCConnectIssuerURL,
+		OIDCConnectDiscoveryURL:                                updatedSettings.OIDCConnectDiscoveryURL,
+		OIDCConnectAuthorizeURL:                                updatedSettings.OIDCConnectAuthorizeURL,
+		OIDCConnectTokenURL:                                    updatedSettings.OIDCConnectTokenURL,
+		OIDCConnectUserInfoURL:                                 updatedSettings.OIDCConnectUserInfoURL,
+		OIDCConnectJWKSURL:                                     updatedSettings.OIDCConnectJWKSURL,
+		OIDCConnectScopes:                                      updatedSettings.OIDCConnectScopes,
+		OIDCConnectRedirectURL:                                 updatedSettings.OIDCConnectRedirectURL,
+		OIDCConnectFrontendRedirectURL:                         updatedSettings.OIDCConnectFrontendRedirectURL,
+		OIDCConnectTokenAuthMethod:                             updatedSettings.OIDCConnectTokenAuthMethod,
+		OIDCConnectUsePKCE:                                     updatedSettings.OIDCConnectUsePKCE,
+		OIDCConnectValidateIDToken:                             updatedSettings.OIDCConnectValidateIDToken,
+		OIDCConnectAllowedSigningAlgs:                          updatedSettings.OIDCConnectAllowedSigningAlgs,
+		OIDCConnectClockSkewSeconds:                            updatedSettings.OIDCConnectClockSkewSeconds,
+		OIDCConnectRequireEmailVerified:                        updatedSettings.OIDCConnectRequireEmailVerified,
+		OIDCConnectUserInfoEmailPath:                           updatedSettings.OIDCConnectUserInfoEmailPath,
+		OIDCConnectUserInfoIDPath:                              updatedSettings.OIDCConnectUserInfoIDPath,
+		OIDCConnectUserInfoUsernamePath:                        updatedSettings.OIDCConnectUserInfoUsernamePath,
+		GitHubOAuthEnabled:                                     updatedSettings.GitHubOAuthEnabled,
+		GitHubOAuthClientID:                                    updatedSettings.GitHubOAuthClientID,
+		GitHubOAuthClientSecretConfigured:                      updatedSettings.GitHubOAuthClientSecretConfigured,
+		GitHubOAuthRedirectURL:                                 updatedSettings.GitHubOAuthRedirectURL,
+		GitHubOAuthFrontendRedirectURL:                         updatedSettings.GitHubOAuthFrontendRedirectURL,
+		GoogleOAuthEnabled:                                     updatedSettings.GoogleOAuthEnabled,
+		GoogleOAuthClientID:                                    updatedSettings.GoogleOAuthClientID,
+		GoogleOAuthClientSecretConfigured:                      updatedSettings.GoogleOAuthClientSecretConfigured,
+		GoogleOAuthRedirectURL:                                 updatedSettings.GoogleOAuthRedirectURL,
+		GoogleOAuthFrontendRedirectURL:                         updatedSettings.GoogleOAuthFrontendRedirectURL,
+		SiteName:                                               updatedSettings.SiteName,
+		SiteLogo:                                               updatedSettings.SiteLogo,
+		SiteSubtitle:                                           updatedSettings.SiteSubtitle,
+		APIBaseURL:                                             updatedSettings.APIBaseURL,
+		ContactInfo:                                            updatedSettings.ContactInfo,
+		DocURL:                                                 updatedSettings.DocURL,
+		HomeContent:                                            updatedSettings.HomeContent,
+		HideCcsImportButton:                                    updatedSettings.HideCcsImportButton,
+		PurchaseSubscriptionEnabled:                            updatedSettings.PurchaseSubscriptionEnabled,
+		PurchaseSubscriptionURL:                                updatedSettings.PurchaseSubscriptionURL,
+		TableDefaultPageSize:                                   updatedSettings.TableDefaultPageSize,
+		TablePageSizeOptions:                                   updatedSettings.TablePageSizeOptions,
+		CustomMenuItems:                                        dto.ParseCustomMenuItems(updatedSettings.CustomMenuItems),
+		CustomEndpoints:                                        dto.ParseCustomEndpoints(updatedSettings.CustomEndpoints),
+		DefaultConcurrency:                                     updatedSettings.DefaultConcurrency,
+		DefaultBalance:                                         updatedSettings.DefaultBalance,
+		AffiliateRebateRate:                                    updatedSettings.AffiliateRebateRate,
+		AffiliateRebateFreezeHours:                             updatedSettings.AffiliateRebateFreezeHours,
+		AffiliateRebateDurationDays:                            updatedSettings.AffiliateRebateDurationDays,
+		AffiliateRebatePerInviteeCap:                           updatedSettings.AffiliateRebatePerInviteeCap,
+		DefaultUserRPMLimit:                                    updatedSettings.DefaultUserRPMLimit,
+		DefaultSubscriptions:                                   updatedDefaultSubscriptions,
+		EnableModelFallback:                                    updatedSettings.EnableModelFallback,
+		FallbackModelAnthropic:                                 updatedSettings.FallbackModelAnthropic,
+		FallbackModelOpenAI:                                    updatedSettings.FallbackModelOpenAI,
+		FallbackModelGemini:                                    updatedSettings.FallbackModelGemini,
+		FallbackModelAntigravity:                               updatedSettings.FallbackModelAntigravity,
+		EnableIdentityPatch:                                    updatedSettings.EnableIdentityPatch,
+		IdentityPatchPrompt:                                    updatedSettings.IdentityPatchPrompt,
+		OpsMonitoringEnabled:                                   updatedSettings.OpsMonitoringEnabled,
+		OpsRealtimeMonitoringEnabled:                           updatedSettings.OpsRealtimeMonitoringEnabled,
+		OpsQueryModeDefault:                                    updatedSettings.OpsQueryModeDefault,
+		OpsMetricsIntervalSeconds:                              updatedSettings.OpsMetricsIntervalSeconds,
+		MinClaudeCodeVersion:                                   updatedSettings.MinClaudeCodeVersion,
+		MaxClaudeCodeVersion:                                   updatedSettings.MaxClaudeCodeVersion,
+		AllowUngroupedKeyScheduling:                            updatedSettings.AllowUngroupedKeyScheduling,
+		BackendModeEnabled:                                     updatedSettings.BackendModeEnabled,
+		EnableFingerprintUnification:                           updatedSettings.EnableFingerprintUnification,
+		EnableMetadataPassthrough:                              updatedSettings.EnableMetadataPassthrough,
+		EnableCCHSigning:                                       updatedSettings.EnableCCHSigning,
+		EnableClaudeOAuthSystemPromptInjection:                 updatedSettings.EnableClaudeOAuthSystemPromptInjection,
+		ClaudeOAuthSystemPrompt:                                updatedSettings.ClaudeOAuthSystemPrompt,
+		ClaudeOAuthSystemPromptBlocks:                          updatedSettings.ClaudeOAuthSystemPromptBlocks,
+		EnableAnthropicCacheTTL1hInjection:                     updatedSettings.EnableAnthropicCacheTTL1hInjection,
+		RewriteMessageCacheControl:                             updatedSettings.RewriteMessageCacheControl,
+		EnableClientDatelineNormalization:                      updatedSettings.EnableClientDatelineNormalization,
+		AntigravityUserAgentVersion:                            updatedSettings.AntigravityUserAgentVersion,
+		OpenAICodexUserAgent:                                   updatedSettings.OpenAICodexUserAgent,
+		MinCodexVersion:                                        updatedSettings.MinCodexVersion,
+		MaxCodexVersion:                                        updatedSettings.MaxCodexVersion,
+		CodexCLIOnlyBlacklist:                                  updatedSettings.CodexCLIOnlyBlacklist,
+		CodexCLIOnlyWhitelist:                                  updatedSettings.CodexCLIOnlyWhitelist,
+		CodexCLIOnlyAllowAppServerClients:                      updatedSettings.CodexCLIOnlyAllowAppServerClients,
+		CodexCLIOnlyEngineFingerprintSignals:                   updatedSettings.CodexCLIOnlyEngineFingerprintSignals,
+		PaymentVisibleMethodAlipaySource:                       updatedSettings.PaymentVisibleMethodAlipaySource,
+		PaymentVisibleMethodWxpaySource:                        updatedSettings.PaymentVisibleMethodWxpaySource,
+		PaymentVisibleMethodAlipayEnabled:                      updatedSettings.PaymentVisibleMethodAlipayEnabled,
+		PaymentVisibleMethodWxpayEnabled:                       updatedSettings.PaymentVisibleMethodWxpayEnabled,
+		OpenAIAdvancedSchedulerEnabled:                         updatedSettings.OpenAIAdvancedSchedulerEnabled,
+		OpenAIAdvancedSchedulerStickyWeightedEnabled:           updatedSettings.OpenAIAdvancedSchedulerStickyWeightedEnabled,
+		OpenAIAdvancedSchedulerSubscriptionPriorityEnabled:     updatedSettings.OpenAIAdvancedSchedulerSubscriptionPriorityEnabled,
+		OpenAIAdvancedSchedulerLBTopK:                          updatedSettings.OpenAIAdvancedSchedulerLBTopK,
+		OpenAIAdvancedSchedulerWeightPriority:                  updatedSettings.OpenAIAdvancedSchedulerWeightPriority,
+		OpenAIAdvancedSchedulerWeightLoad:                      updatedSettings.OpenAIAdvancedSchedulerWeightLoad,
+		OpenAIAdvancedSchedulerWeightQueue:                     updatedSettings.OpenAIAdvancedSchedulerWeightQueue,
+		OpenAIAdvancedSchedulerWeightErrorRate:                 updatedSettings.OpenAIAdvancedSchedulerWeightErrorRate,
+		OpenAIAdvancedSchedulerWeightTTFT:                      updatedSettings.OpenAIAdvancedSchedulerWeightTTFT,
+		OpenAIAdvancedSchedulerWeightReset:                     updatedSettings.OpenAIAdvancedSchedulerWeightReset,
+		OpenAIAdvancedSchedulerWeightQuotaHeadroom:             updatedSettings.OpenAIAdvancedSchedulerWeightQuotaHeadroom,
+		OpenAIAdvancedSchedulerWeightPreviousResponse:          updatedSettings.OpenAIAdvancedSchedulerWeightPreviousResponse,
+		OpenAIAdvancedSchedulerWeightSessionSticky:             updatedSettings.OpenAIAdvancedSchedulerWeightSessionSticky,
+		OpenAIAdvancedSchedulerEffectiveLBTopK:                 updatedSettings.OpenAIAdvancedSchedulerEffectiveLBTopK,
+		OpenAIAdvancedSchedulerEffectiveWeightPriority:         updatedSettings.OpenAIAdvancedSchedulerEffectiveWeightPriority,
+		OpenAIAdvancedSchedulerEffectiveWeightLoad:             updatedSettings.OpenAIAdvancedSchedulerEffectiveWeightLoad,
+		OpenAIAdvancedSchedulerEffectiveWeightQueue:            updatedSettings.OpenAIAdvancedSchedulerEffectiveWeightQueue,
+		OpenAIAdvancedSchedulerEffectiveWeightErrorRate:        updatedSettings.OpenAIAdvancedSchedulerEffectiveWeightErrorRate,
+		OpenAIAdvancedSchedulerEffectiveWeightTTFT:             updatedSettings.OpenAIAdvancedSchedulerEffectiveWeightTTFT,
+		OpenAIAdvancedSchedulerEffectiveWeightReset:            updatedSettings.OpenAIAdvancedSchedulerEffectiveWeightReset,
+		OpenAIAdvancedSchedulerEffectiveWeightQuotaHeadroom:    updatedSettings.OpenAIAdvancedSchedulerEffectiveWeightQuotaHeadroom,
+		OpenAIAdvancedSchedulerEffectiveWeightPreviousResponse: updatedSettings.OpenAIAdvancedSchedulerEffectiveWeightPreviousResponse,
+		OpenAIAdvancedSchedulerEffectiveWeightSessionSticky:    updatedSettings.OpenAIAdvancedSchedulerEffectiveWeightSessionSticky,
+		BalanceLowNotifyEnabled:                                updatedSettings.BalanceLowNotifyEnabled,
+		BalanceLowNotifyThreshold:                              updatedSettings.BalanceLowNotifyThreshold,
+		BalanceLowNotifyRechargeURL:                            updatedSettings.BalanceLowNotifyRechargeURL,
+		SubscriptionExpiryNotifyEnabled:                        updatedSettings.SubscriptionExpiryNotifyEnabled,
+		AccountQuotaNotifyEnabled:                              updatedSettings.AccountQuotaNotifyEnabled,
+		AccountQuotaNotifyEmails:                               dto.NotifyEmailEntriesFromService(updatedSettings.AccountQuotaNotifyEmails),
+		PaymentEnabled:                                         updatedPaymentCfg.Enabled,
+		PaymentMinAmount:                                       updatedPaymentCfg.MinAmount,
+		PaymentMaxAmount:                                       updatedPaymentCfg.MaxAmount,
+		PaymentDailyLimit:                                      updatedPaymentCfg.DailyLimit,
+		PaymentOrderTimeoutMin:                                 updatedPaymentCfg.OrderTimeoutMin,
+		PaymentMaxPendingOrders:                                updatedPaymentCfg.MaxPendingOrders,
+		PaymentEnabledTypes:                                    updatedPaymentCfg.EnabledTypes,
+		PaymentBalanceDisabled:                                 updatedPaymentCfg.BalanceDisabled,
+		PaymentBalanceRechargeMultiplier:                       updatedPaymentCfg.BalanceRechargeMultiplier,
+		PaymentRechargeFeeRate:                                 updatedPaymentCfg.RechargeFeeRate,
+		PaymentLoadBalanceStrat:                                updatedPaymentCfg.LoadBalanceStrategy,
+		PaymentProductNamePrefix:                               updatedPaymentCfg.ProductNamePrefix,
+		PaymentProductNameSuffix:                               updatedPaymentCfg.ProductNameSuffix,
+		PaymentHelpImageURL:                                    updatedPaymentCfg.HelpImageURL,
+		PaymentHelpText:                                        updatedPaymentCfg.HelpText,
+		PaymentCancelRateLimitEnabled:                          updatedPaymentCfg.CancelRateLimitEnabled,
+		PaymentCancelRateLimitMax:                              updatedPaymentCfg.CancelRateLimitMax,
+		PaymentCancelRateLimitWindow:                           updatedPaymentCfg.CancelRateLimitWindow,
+		PaymentCancelRateLimitUnit:                             updatedPaymentCfg.CancelRateLimitUnit,
+		PaymentCancelRateLimitMode:                             updatedPaymentCfg.CancelRateLimitMode,
+		PaymentAlipayForceQRCode:                               updatedPaymentCfg.AlipayForceQRCode,
 
 		ChannelMonitorEnabled:                updatedSettings.ChannelMonitorEnabled,
 		ChannelMonitorDefaultIntervalSeconds: updatedSettings.ChannelMonitorDefaultIntervalSeconds,
@@ -2677,6 +2755,42 @@ func diffSettings(before *service.SystemSettings, after *service.SystemSettings,
 	if before.OpenAIAdvancedSchedulerEnabled != after.OpenAIAdvancedSchedulerEnabled {
 		changed = append(changed, "openai_advanced_scheduler_enabled")
 	}
+	if before.OpenAIAdvancedSchedulerStickyWeightedEnabled != after.OpenAIAdvancedSchedulerStickyWeightedEnabled {
+		changed = append(changed, "openai_advanced_scheduler_sticky_weighted_enabled")
+	}
+	if before.OpenAIAdvancedSchedulerSubscriptionPriorityEnabled != after.OpenAIAdvancedSchedulerSubscriptionPriorityEnabled {
+		changed = append(changed, "openai_advanced_scheduler_subscription_priority_enabled")
+	}
+	if before.OpenAIAdvancedSchedulerLBTopK != after.OpenAIAdvancedSchedulerLBTopK {
+		changed = append(changed, "openai_advanced_scheduler_lb_top_k")
+	}
+	if before.OpenAIAdvancedSchedulerWeightPriority != after.OpenAIAdvancedSchedulerWeightPriority {
+		changed = append(changed, "openai_advanced_scheduler_weight_priority")
+	}
+	if before.OpenAIAdvancedSchedulerWeightLoad != after.OpenAIAdvancedSchedulerWeightLoad {
+		changed = append(changed, "openai_advanced_scheduler_weight_load")
+	}
+	if before.OpenAIAdvancedSchedulerWeightQueue != after.OpenAIAdvancedSchedulerWeightQueue {
+		changed = append(changed, "openai_advanced_scheduler_weight_queue")
+	}
+	if before.OpenAIAdvancedSchedulerWeightErrorRate != after.OpenAIAdvancedSchedulerWeightErrorRate {
+		changed = append(changed, "openai_advanced_scheduler_weight_error_rate")
+	}
+	if before.OpenAIAdvancedSchedulerWeightTTFT != after.OpenAIAdvancedSchedulerWeightTTFT {
+		changed = append(changed, "openai_advanced_scheduler_weight_ttft")
+	}
+	if before.OpenAIAdvancedSchedulerWeightReset != after.OpenAIAdvancedSchedulerWeightReset {
+		changed = append(changed, "openai_advanced_scheduler_weight_reset")
+	}
+	if before.OpenAIAdvancedSchedulerWeightQuotaHeadroom != after.OpenAIAdvancedSchedulerWeightQuotaHeadroom {
+		changed = append(changed, "openai_advanced_scheduler_weight_quota_headroom")
+	}
+	if before.OpenAIAdvancedSchedulerWeightPreviousResponse != after.OpenAIAdvancedSchedulerWeightPreviousResponse {
+		changed = append(changed, "openai_advanced_scheduler_weight_previous_response")
+	}
+	if before.OpenAIAdvancedSchedulerWeightSessionSticky != after.OpenAIAdvancedSchedulerWeightSessionSticky {
+		changed = append(changed, "openai_advanced_scheduler_weight_session_sticky")
+	}
 	// 余额、订阅到期与账号限额通知
 	if before.BalanceLowNotifyEnabled != after.BalanceLowNotifyEnabled {
 		changed = append(changed, "balance_low_notify_enabled")
@@ -3829,3 +3943,10 @@ func equalPlatformQuotaSettings(before, after map[string]*service.DefaultPlatfor
 	}
 	return true
 }
+
+func stringSetting(value *string, fallback string) string {
+	if value == nil {
+		return fallback
+	}
+	return *value
+}
diff --git a/backend/internal/handler/admin/setting_handler_auth_source_defaults_test.go b/backend/internal/handler/admin/setting_handler_auth_source_defaults_test.go
index f953f76760..1626007f19 100644
--- a/backend/internal/handler/admin/setting_handler_auth_source_defaults_test.go
+++ b/backend/internal/handler/admin/setting_handler_auth_source_defaults_test.go
@@ -217,12 +217,13 @@ func TestSettingHandler_UpdateSettings_PersistsPaymentVisibleMethodsAndAdvancedS
 	handler := NewSettingHandler(svc, nil, nil, nil, nil, nil, nil)
 
 	body := map[string]any{
-		"promo_code_enabled":                    true,
-		"payment_visible_method_alipay_source":  "easypay",
-		"payment_visible_method_wxpay_source":   "wxpay",
-		"payment_visible_method_alipay_enabled": true,
-		"payment_visible_method_wxpay_enabled":  false,
-		"openai_advanced_scheduler_enabled":     true,
+		"promo_code_enabled":                                      true,
+		"payment_visible_method_alipay_source":                    "easypay",
+		"payment_visible_method_wxpay_source":                     "wxpay",
+		"payment_visible_method_alipay_enabled":                   true,
+		"payment_visible_method_wxpay_enabled":                    false,
+		"openai_advanced_scheduler_enabled":                       true,
+		"openai_advanced_scheduler_subscription_priority_enabled": true,
 	}
 	rawBody, err := json.Marshal(body)
 	require.NoError(t, err)
@@ -240,6 +241,7 @@ func TestSettingHandler_UpdateSettings_PersistsPaymentVisibleMethodsAndAdvancedS
 	require.Equal(t, "true", repo.values[service.SettingPaymentVisibleMethodAlipayEnabled])
 	require.Equal(t, "false", repo.values[service.SettingPaymentVisibleMethodWxpayEnabled])
 	require.Equal(t, "true", repo.values["openai_advanced_scheduler_enabled"])
+	require.Equal(t, "true", repo.values[service.SettingKeyOpenAIAdvancedSchedulerSubscriptionPriorityEnabled])
 
 	var resp response.Response
 	require.NoError(t, json.Unmarshal(rec.Body.Bytes(), &resp))
@@ -250,6 +252,7 @@ func TestSettingHandler_UpdateSettings_PersistsPaymentVisibleMethodsAndAdvancedS
 	require.Equal(t, true, data["payment_visible_method_alipay_enabled"])
 	require.Equal(t, false, data["payment_visible_method_wxpay_enabled"])
 	require.Equal(t, true, data["openai_advanced_scheduler_enabled"])
+	require.Equal(t, true, data["openai_advanced_scheduler_subscription_priority_enabled"])
 }
 
 func TestSettingHandler_UpdateSettings_PreservesLegacyBlankPaymentVisibleMethodSource(t *testing.T) {
diff --git a/backend/internal/handler/dto/settings.go b/backend/internal/handler/dto/settings.go
index 1d6f73bb99..9a20191a0c 100644
--- a/backend/internal/handler/dto/settings.go
+++ b/backend/internal/handler/dto/settings.go
@@ -208,7 +208,29 @@ type SystemSettings struct {
 	PaymentVisibleMethodWxpayEnabled  bool   `json:"payment_visible_method_wxpay_enabled"`
 
 	// OpenAI account scheduling
-	OpenAIAdvancedSchedulerEnabled bool `json:"openai_advanced_scheduler_enabled"`
+	OpenAIAdvancedSchedulerEnabled                         bool   `json:"openai_advanced_scheduler_enabled"`
+	OpenAIAdvancedSchedulerStickyWeightedEnabled           bool   `json:"openai_advanced_scheduler_sticky_weighted_enabled"`
+	OpenAIAdvancedSchedulerSubscriptionPriorityEnabled     bool   `json:"openai_advanced_scheduler_subscription_priority_enabled"`
+	OpenAIAdvancedSchedulerLBTopK                          string `json:"openai_advanced_scheduler_lb_top_k"`
+	OpenAIAdvancedSchedulerWeightPriority                  string `json:"openai_advanced_scheduler_weight_priority"`
+	OpenAIAdvancedSchedulerWeightLoad                      string `json:"openai_advanced_scheduler_weight_load"`
+	OpenAIAdvancedSchedulerWeightQueue                     string `json:"openai_advanced_scheduler_weight_queue"`
+	OpenAIAdvancedSchedulerWeightErrorRate                 string `json:"openai_advanced_scheduler_weight_error_rate"`
+	OpenAIAdvancedSchedulerWeightTTFT                      string `json:"openai_advanced_scheduler_weight_ttft"`
+	OpenAIAdvancedSchedulerWeightReset                     string `json:"openai_advanced_scheduler_weight_reset"`
+	OpenAIAdvancedSchedulerWeightQuotaHeadroom             string `json:"openai_advanced_scheduler_weight_quota_headroom"`
+	OpenAIAdvancedSchedulerWeightPreviousResponse          string `json:"openai_advanced_scheduler_weight_previous_response"`
+	OpenAIAdvancedSchedulerWeightSessionSticky             string `json:"openai_advanced_scheduler_weight_session_sticky"`
+	OpenAIAdvancedSchedulerEffectiveLBTopK                 string `json:"openai_advanced_scheduler_effective_lb_top_k"`
+	OpenAIAdvancedSchedulerEffectiveWeightPriority         string `json:"openai_advanced_scheduler_effective_weight_priority"`
+	OpenAIAdvancedSchedulerEffectiveWeightLoad             string `json:"openai_advanced_scheduler_effective_weight_load"`
+	OpenAIAdvancedSchedulerEffectiveWeightQueue            string `json:"openai_advanced_scheduler_effective_weight_queue"`
+	OpenAIAdvancedSchedulerEffectiveWeightErrorRate        string `json:"openai_advanced_scheduler_effective_weight_error_rate"`
+	OpenAIAdvancedSchedulerEffectiveWeightTTFT             string `json:"openai_advanced_scheduler_effective_weight_ttft"`
+	OpenAIAdvancedSchedulerEffectiveWeightReset            string `json:"openai_advanced_scheduler_effective_weight_reset"`
+	OpenAIAdvancedSchedulerEffectiveWeightQuotaHeadroom    string `json:"openai_advanced_scheduler_effective_weight_quota_headroom"`
+	OpenAIAdvancedSchedulerEffectiveWeightPreviousResponse string `json:"openai_advanced_scheduler_effective_weight_previous_response"`
+	OpenAIAdvancedSchedulerEffectiveWeightSessionSticky    string `json:"openai_advanced_scheduler_effective_weight_session_sticky"`
 
 	// Payment configuration
 	PaymentEnabled                   bool     `json:"payment_enabled"`
diff --git a/backend/internal/handler/openai_gateway_handler.go b/backend/internal/handler/openai_gateway_handler.go
index 8eec667fa6..ccafde7b02 100644
--- a/backend/internal/handler/openai_gateway_handler.go
+++ b/backend/internal/handler/openai_gateway_handler.go
@@ -1266,6 +1266,8 @@ func (h *OpenAIGatewayHandler) ResponsesWebSocket(c *gin.Context) {
 		closeOpenAIClientWS(wsConn, coderws.StatusPolicyViolation, "previous_response_id must be a response.id (resp_*), not a message id")
 		return
 	}
+	firstMessageToolContext := service.ValidateFunctionCallOutputContextBytes(firstMessage)
+	previousResponseCanMove := !firstMessageToolContext.HasFunctionCallOutput || firstMessageToolContext.HasToolCallContext
 	reqLog = reqLog.With(
 		zap.Bool("ws_ingress", true),
 		zap.String("model", reqModel),
@@ -1382,6 +1384,12 @@ func (h *OpenAIGatewayHandler) ResponsesWebSocket(c *gin.Context) {
 			service.OpenAIEndpointCapabilityChatCompletions,
 			false,
 			requestPlatform,
+			func() string {
+				if previousResponseCanMove {
+					return "previous_response_can_move"
+				}
+				return ""
+			}(),
 		)
 		if err != nil {
 			reqLog.Warn("openai.websocket_account_select_failed",
@@ -1581,8 +1589,7 @@ func (h *OpenAIGatewayHandler) ResponsesWebSocket(c *gin.Context) {
 		// 说明该会话链不属于本次调度到的账号，原样转发会触发上游会话链鉴权失败（“鉴权失败，请检查 API Key”）。
 		// 故剥离首包里的 previous_response_id，改用首包内 input 重建上下文；带 function_call_output 的
 		// 工具续链无法重建，保持原样。仅作用于首轮首包，后续 turn 的续链由 WS 转发层既有逻辑处理。
-		if previousResponseID != "" && !scheduleDecision.StickyPreviousHit &&
-			!service.ValidateFunctionCallOutputContextBytes(wsFirstMessage).HasFunctionCallOutput {
+		if previousResponseID != "" && !scheduleDecision.StickyPreviousHit && previousResponseCanMove {
 			wsFirstMessage = service.RemovePreviousResponseIDFromBody(wsFirstMessage)
 			reqLog.Debug("openai.websocket_previous_response_id_stripped_cross_group",
 				zap.Int64("account_id", account.ID),
diff --git a/backend/internal/repository/account_repo.go b/backend/internal/repository/account_repo.go
index 1a21b5893c..a8015ddba0 100644
--- a/backend/internal/repository/account_repo.go
+++ b/backend/internal/repository/account_repo.go
@@ -482,7 +482,7 @@ func (r *accountRepository) List(ctx context.Context, params pagination.Paginati
 	return r.ListWithFilters(ctx, params, "", "", "", "", 0, "")
 }
 
-func (r *accountRepository) ListWithFilters(ctx context.Context, params pagination.PaginationParams, platform, accountType, status, search string, groupID int64, privacyMode string) ([]service.Account, *pagination.PaginationResult, error) {
+func (r *accountRepository) accountListFilteredQuery(platform, accountType, status, search string, groupID int64, privacyMode string) *dbent.AccountQuery {
 	q := r.client.Account.Query()
 
 	if platform != "" {
@@ -575,6 +575,11 @@ func (r *accountRepository) ListWithFilters(ctx context.Context, params paginati
 		}))
 	}
 
+	return q
+}
+
+func (r *accountRepository) ListWithFilters(ctx context.Context, params pagination.PaginationParams, platform, accountType, status, search string, groupID int64, privacyMode string) ([]service.Account, *pagination.PaginationResult, error) {
+	q := r.accountListFilteredQuery(platform, accountType, status, search, groupID, privacyMode)
 	// Clone before Count so interceptor-appended predicates (SoftDeleteMixin's
 	// deleted_at IS NULL) don't accumulate on the shared builder and pollute the
 	// subsequent list query. Same pattern used in group_repo/promo_code_repo/user_repo
@@ -603,6 +608,14 @@ func (r *accountRepository) ListWithFilters(ctx context.Context, params paginati
 	return outAccounts, paginationResultFromTotal(int64(total), params), nil
 }
 
+func (r *accountRepository) ListAllWithFilters(ctx context.Context, platform, accountType, status, search string, groupID int64, privacyMode string) ([]service.Account, error) {
+	accounts, err := r.accountListFilteredQuery(platform, accountType, status, search, groupID, privacyMode).All(ctx)
+	if err != nil {
+		return nil, err
+	}
+	return r.accountsToService(ctx, accounts)
+}
+
 func (r *accountRepository) ListOpsAccountsForStats(ctx context.Context, platformFilter string, groupIDFilter *int64) ([]service.Account, error) {
 	if r == nil || r.client == nil {
 		return []service.Account{}, nil
diff --git a/backend/internal/repository/scheduler_cache.go b/backend/internal/repository/scheduler_cache.go
index c508f09c71..c8e1fe14e0 100644
--- a/backend/internal/repository/scheduler_cache.go
+++ b/backend/internal/repository/scheduler_cache.go
@@ -518,7 +518,7 @@ func filterSchedulerCredentials(credentials map[string]any) map[string]any {
 	if len(credentials) == 0 {
 		return nil
 	}
-	keys := []string{"model_mapping", "compact_model_mapping", "api_key", "project_id", "oauth_type"}
+	keys := []string{"model_mapping", "compact_model_mapping", "api_key", "project_id", "oauth_type", "plan_type"}
 	filtered := make(map[string]any)
 	for _, key := range keys {
 		if value, ok := credentials[key]; ok && value != nil {
diff --git a/backend/internal/repository/scheduler_cache_test.go b/backend/internal/repository/scheduler_cache_test.go
new file mode 100644
index 0000000000..f438b2f456
--- /dev/null
+++ b/backend/internal/repository/scheduler_cache_test.go
@@ -0,0 +1,37 @@
+package repository
+
+import (
+	"testing"
+
+	"github.com/Wei-Shaw/sub2api/internal/service"
+	"github.com/stretchr/testify/require"
+)
+
+func TestFilterSchedulerCredentialsKeepsSubscriptionPlanType(t *testing.T) {
+	filtered := filterSchedulerCredentials(map[string]any{
+		"plan_type":     "plus",
+		"access_token":  "secret-access-token",
+		"refresh_token": "secret-refresh-token",
+	})
+
+	require.Equal(t, "plus", filtered["plan_type"])
+	require.NotContains(t, filtered, "access_token")
+	require.NotContains(t, filtered, "refresh_token")
+}
+
+func TestSchedulerMetadataAccountKeepsOpenAISubscriptionIdentity(t *testing.T) {
+	account := service.Account{
+		ID:       24,
+		Platform: service.PlatformOpenAI,
+		Type:     service.AccountTypeOAuth,
+		Credentials: map[string]any{
+			"plan_type":    "plus",
+			"access_token": "secret-access-token",
+		},
+	}
+
+	metadata := buildSchedulerMetadataAccount(account)
+
+	require.True(t, metadata.IsOpenAIChatGPTSubscription())
+	require.Empty(t, metadata.GetCredential("access_token"))
+}
diff --git a/backend/internal/server/api_contract_test.go b/backend/internal/server/api_contract_test.go
index 26f976c87b..dfa480dd18 100644
--- a/backend/internal/server/api_contract_test.go
+++ b/backend/internal/server/api_contract_test.go
@@ -661,15 +661,17 @@ func TestAPIContracts(t *testing.T) {
 					service.SettingKeyTableDefaultPageSize: "20",
 					service.SettingKeyTablePageSizeOptions: "[10,20,50,100]",
 
-					service.SettingKeyOpsMonitoringEnabled:           "false",
-					service.SettingKeyOpsRealtimeMonitoringEnabled:   "true",
-					service.SettingKeyOpsQueryModeDefault:            "auto",
-					service.SettingKeyOpsMetricsIntervalSeconds:      "60",
-					service.SettingPaymentVisibleMethodAlipaySource:  service.VisibleMethodSourceEasyPayAlipay,
-					service.SettingPaymentVisibleMethodWxpaySource:   service.VisibleMethodSourceOfficialWechat,
-					service.SettingPaymentVisibleMethodAlipayEnabled: "true",
-					service.SettingPaymentVisibleMethodWxpayEnabled:  "false",
-					"openai_advanced_scheduler_enabled":              "true",
+					service.SettingKeyOpsMonitoringEnabled:                               "false",
+					service.SettingKeyOpsRealtimeMonitoringEnabled:                       "true",
+					service.SettingKeyOpsQueryModeDefault:                                "auto",
+					service.SettingKeyOpsMetricsIntervalSeconds:                          "60",
+					service.SettingPaymentVisibleMethodAlipaySource:                      service.VisibleMethodSourceEasyPayAlipay,
+					service.SettingPaymentVisibleMethodWxpaySource:                       service.VisibleMethodSourceOfficialWechat,
+					service.SettingPaymentVisibleMethodAlipayEnabled:                     "true",
+					service.SettingPaymentVisibleMethodWxpayEnabled:                      "false",
+					"openai_advanced_scheduler_enabled":                                  "true",
+					service.SettingKeyOpenAIAdvancedSchedulerStickyWeightedEnabled:       "false",
+					service.SettingKeyOpenAIAdvancedSchedulerSubscriptionPriorityEnabled: "false",
 				})
 			},
 			method:     http.MethodGet,
@@ -861,6 +863,28 @@ func TestAPIContracts(t *testing.T) {
 					"payment_visible_method_alipay_enabled": true,
 					"payment_visible_method_wxpay_enabled": false,
 					"openai_advanced_scheduler_enabled": true,
+					"openai_advanced_scheduler_sticky_weighted_enabled": false,
+					"openai_advanced_scheduler_subscription_priority_enabled": false,
+					"openai_advanced_scheduler_lb_top_k": "",
+					"openai_advanced_scheduler_weight_priority": "",
+					"openai_advanced_scheduler_weight_load": "",
+					"openai_advanced_scheduler_weight_queue": "",
+					"openai_advanced_scheduler_weight_error_rate": "",
+					"openai_advanced_scheduler_weight_ttft": "",
+					"openai_advanced_scheduler_weight_reset": "",
+					"openai_advanced_scheduler_weight_quota_headroom": "",
+					"openai_advanced_scheduler_weight_previous_response": "",
+					"openai_advanced_scheduler_weight_session_sticky": "",
+					"openai_advanced_scheduler_effective_lb_top_k": "7",
+					"openai_advanced_scheduler_effective_weight_priority": "1",
+					"openai_advanced_scheduler_effective_weight_load": "1",
+					"openai_advanced_scheduler_effective_weight_queue": "0.7",
+					"openai_advanced_scheduler_effective_weight_error_rate": "0.8",
+					"openai_advanced_scheduler_effective_weight_ttft": "0.5",
+					"openai_advanced_scheduler_effective_weight_reset": "0",
+					"openai_advanced_scheduler_effective_weight_quota_headroom": "0",
+					"openai_advanced_scheduler_effective_weight_previous_response": "5",
+					"openai_advanced_scheduler_effective_weight_session_sticky": "3",
 					"openai_codex_user_agent":           "",
 					"openai_fast_policy_settings": {
 						"rules": []
@@ -1110,6 +1134,28 @@ func TestAPIContracts(t *testing.T) {
 					"payment_visible_method_alipay_enabled": false,
 					"payment_visible_method_wxpay_enabled": false,
 					"openai_advanced_scheduler_enabled": false,
+					"openai_advanced_scheduler_sticky_weighted_enabled": false,
+					"openai_advanced_scheduler_subscription_priority_enabled": false,
+					"openai_advanced_scheduler_lb_top_k": "",
+					"openai_advanced_scheduler_weight_priority": "",
+					"openai_advanced_scheduler_weight_load": "",
+					"openai_advanced_scheduler_weight_queue": "",
+					"openai_advanced_scheduler_weight_error_rate": "",
+					"openai_advanced_scheduler_weight_ttft": "",
+					"openai_advanced_scheduler_weight_reset": "",
+					"openai_advanced_scheduler_weight_quota_headroom": "",
+					"openai_advanced_scheduler_weight_previous_response": "",
+					"openai_advanced_scheduler_weight_session_sticky": "",
+					"openai_advanced_scheduler_effective_lb_top_k": "7",
+					"openai_advanced_scheduler_effective_weight_priority": "1",
+					"openai_advanced_scheduler_effective_weight_load": "1",
+					"openai_advanced_scheduler_effective_weight_queue": "0.7",
+					"openai_advanced_scheduler_effective_weight_error_rate": "0.8",
+					"openai_advanced_scheduler_effective_weight_ttft": "0.5",
+					"openai_advanced_scheduler_effective_weight_reset": "0",
+					"openai_advanced_scheduler_effective_weight_quota_headroom": "0",
+					"openai_advanced_scheduler_effective_weight_previous_response": "5",
+					"openai_advanced_scheduler_effective_weight_session_sticky": "3",
 					"openai_codex_user_agent":           "",
 					"openai_fast_policy_settings": {
 						"rules": []
diff --git a/backend/internal/service/account.go b/backend/internal/service/account.go
index 26af0a7281..8db5805e9d 100644
--- a/backend/internal/service/account.go
+++ b/backend/internal/service/account.go
@@ -1182,6 +1182,18 @@ func (a *Account) IsOpenAIOAuth() bool {
 	return a.IsOpenAI() && a.Type == AccountTypeOAuth
 }
 
+func (a *Account) IsOpenAIChatGPTSubscription() bool {
+	if !a.IsOpenAIOAuth() {
+		return false
+	}
+	switch strings.ToLower(strings.TrimSpace(a.GetCredential("plan_type"))) {
+	case "", "free", "abnormal":
+		return false
+	default:
+		return true
+	}
+}
+
 func (a *Account) IsOpenAIPersonalAccessToken() bool {
 	if !a.IsOpenAIOAuth() {
 		return false
diff --git a/backend/internal/service/admin_service.go b/backend/internal/service/admin_service.go
index 7685174f3b..ce59c34475 100644
--- a/backend/internal/service/admin_service.go
+++ b/backend/internal/service/admin_service.go
@@ -2618,6 +2618,29 @@ func (s *adminServiceImpl) ListAccounts(ctx context.Context, page, pageSize int,
 	return accounts, result.Total, nil
 }
 
+func (s *adminServiceImpl) ListAccountsForSchedulerScoreFilter(ctx context.Context, platform, accountType, status, search string, groupID int64, privacyMode string) ([]Account, error) {
+	if s == nil || s.accountRepo == nil {
+		return nil, nil
+	}
+	lister, ok := s.accountRepo.(interface {
+		ListAllWithFilters(ctx context.Context, platform, accountType, status, search string, groupID int64, privacyMode string) ([]Account, error)
+	})
+	if !ok {
+		return nil, nil
+	}
+	return lister.ListAllWithFilters(ctx, platform, accountType, status, search, groupID, privacyMode)
+}
+
+func (s *adminServiceImpl) ListOpenAISchedulableAccountsForSchedulerScore(ctx context.Context, groupID *int64) ([]Account, error) {
+	if s == nil || s.accountRepo == nil {
+		return nil, nil
+	}
+	if groupID != nil {
+		return s.accountRepo.ListSchedulableByGroupIDAndPlatform(ctx, *groupID, PlatformOpenAI)
+	}
+	return s.accountRepo.ListSchedulableUngroupedByPlatform(ctx, PlatformOpenAI)
+}
+
 func (s *adminServiceImpl) GetAccount(ctx context.Context, id int64) (*Account, error) {
 	return s.accountRepo.GetByID(ctx, id)
 }
diff --git a/backend/internal/service/domain_constants.go b/backend/internal/service/domain_constants.go
index ab853c280a..bc1db19105 100644
--- a/backend/internal/service/domain_constants.go
+++ b/backend/internal/service/domain_constants.go
@@ -431,6 +431,20 @@ const (
 
 	// SettingKeyAllowUngroupedKeyScheduling 允许未分组 API Key 调度（默认 false：未分组 Key 返回 403）
 	SettingKeyAllowUngroupedKeyScheduling = "allow_ungrouped_key_scheduling"
+	// SettingKeyOpenAIAdvancedSchedulerStickyWeightedEnabled OpenAI 高级调度下是否启用粘性加权。
+	SettingKeyOpenAIAdvancedSchedulerStickyWeightedEnabled = "openai_advanced_scheduler_sticky_weighted_enabled"
+	// SettingKeyOpenAIAdvancedSchedulerSubscriptionPriorityEnabled OpenAI 高级调度下是否优先使用订阅账号池。
+	SettingKeyOpenAIAdvancedSchedulerSubscriptionPriorityEnabled = "openai_advanced_scheduler_subscription_priority_enabled"
+	SettingKeyOpenAIAdvancedSchedulerLBTopK                      = "openai_advanced_scheduler_lb_top_k"
+	SettingKeyOpenAIAdvancedSchedulerWeightPriority              = "openai_advanced_scheduler_weight_priority"
+	SettingKeyOpenAIAdvancedSchedulerWeightLoad                  = "openai_advanced_scheduler_weight_load"
+	SettingKeyOpenAIAdvancedSchedulerWeightQueue                 = "openai_advanced_scheduler_weight_queue"
+	SettingKeyOpenAIAdvancedSchedulerWeightErrorRate             = "openai_advanced_scheduler_weight_error_rate"
+	SettingKeyOpenAIAdvancedSchedulerWeightTTFT                  = "openai_advanced_scheduler_weight_ttft"
+	SettingKeyOpenAIAdvancedSchedulerWeightReset                 = "openai_advanced_scheduler_weight_reset"
+	SettingKeyOpenAIAdvancedSchedulerWeightQuotaHeadroom         = "openai_advanced_scheduler_weight_quota_headroom"
+	SettingKeyOpenAIAdvancedSchedulerWeightPreviousResponse      = "openai_advanced_scheduler_weight_previous_response"
+	SettingKeyOpenAIAdvancedSchedulerWeightSessionSticky         = "openai_advanced_scheduler_weight_session_sticky"
 
 	// SettingKeyBackendModeEnabled Backend 模式：禁用用户注册和自助服务，仅管理员可登录
 	SettingKeyBackendModeEnabled = "backend_mode_enabled"
diff --git a/backend/internal/service/openai_account_scheduler.go b/backend/internal/service/openai_account_scheduler.go
index ab298d4521..dd65163abc 100644
--- a/backend/internal/service/openai_account_scheduler.go
+++ b/backend/internal/service/openai_account_scheduler.go
@@ -36,8 +36,20 @@ const (
 )
 
 type cachedOpenAIAdvancedSchedulerSetting struct {
-	enabled   bool
-	expiresAt int64
+	enabled                     bool
+	stickyWeightedEnabled       bool
+	subscriptionPriorityEnabled bool
+	lbTopKOverride              int
+	weightOverrides             map[string]float64
+	expiresAt                   int64
+}
+
+type openAIAdvancedSchedulerRuntimeSettings struct {
+	enabled                     bool
+	stickyWeightedEnabled       bool
+	subscriptionPriorityEnabled bool
+	lbTopKOverride              int
+	weightOverrides             map[string]float64
 }
 
 var openAIAdvancedSchedulerSettingCache atomic.Value // *cachedOpenAIAdvancedSchedulerSetting
@@ -48,8 +60,12 @@ type OpenAIAccountScheduleRequest struct {
 	Platform                string
 	SessionHash             string
 	StickyAccountID         int64
+	StickyPreviousAccountID int64
+	StickyWeighted          bool
+	SubscriptionPriority    bool
 	PreserveStickyBinding   bool
 	PreviousResponseID      string
+	PreviousResponseCanMove bool
 	RequestedModel          string
 	RequiredTransport       OpenAIUpstreamTransport
 	RequiredCapability      OpenAIEndpointCapability
@@ -111,6 +127,17 @@ type openAIAccountLoadPlan struct {
 	loadSkew                  float64
 }
 
+type openAIAccountLoadSelectionAttempt struct {
+	result              *AccountSelectionResult
+	selectionOrder      []openAIAccountCandidateScore
+	candidateCount      int
+	topK                int
+	loadSkew            float64
+	compactBlocked      bool
+	noCompactCandidates bool
+	err                 error
+}
+
 func (m *openAIAccountSchedulerMetrics) recordSelect(decision OpenAIAccountScheduleDecision) {
 	if m == nil {
 		return
@@ -277,7 +304,8 @@ func (s *defaultOpenAIAccountScheduler) Select(
 	}()
 
 	previousResponseID := strings.TrimSpace(req.PreviousResponseID)
-	if previousResponseID != "" && normalizeOpenAICompatiblePlatform(req.Platform) == PlatformOpenAI {
+	if previousResponseID != "" && normalizeOpenAICompatiblePlatform(req.Platform) == PlatformOpenAI &&
+		(!req.StickyWeighted || !req.PreviousResponseCanMove) {
 		selection, err := s.service.selectAccountByPreviousResponseIDForCapability(
 			ctx,
 			req.GroupID,
@@ -310,19 +338,21 @@ func (s *defaultOpenAIAccountScheduler) Select(
 		}
 	}
 
-	selection, escapedSticky, err := s.selectBySessionHash(ctx, req)
-	if err != nil {
-		return nil, decision, err
-	}
-	if selection != nil && selection.Account != nil {
-		decision.Layer = openAIAccountScheduleLayerSessionSticky
-		decision.StickySessionHit = true
-		decision.SelectedAccountID = selection.Account.ID
-		decision.SelectedAccountType = selection.Account.Type
-		return selection, decision, nil
-	}
-	if escapedSticky {
-		req.PreserveStickyBinding = true
+	if !req.StickyWeighted {
+		selection, escapedSticky, err := s.selectBySessionHash(ctx, req)
+		if err != nil {
+			return nil, decision, err
+		}
+		if selection != nil && selection.Account != nil {
+			decision.Layer = openAIAccountScheduleLayerSessionSticky
+			decision.StickySessionHit = true
+			decision.SelectedAccountID = selection.Account.ID
+			decision.SelectedAccountType = selection.Account.Type
+			return selection, decision, nil
+		}
+		if escapedSticky {
+			req.PreserveStickyBinding = true
+		}
 	}
 
 	selection, candidateCount, topK, loadSkew, err := s.selectByLoadBalance(ctx, req)
@@ -336,6 +366,14 @@ func (s *defaultOpenAIAccountScheduler) Select(
 	if selection != nil && selection.Account != nil {
 		decision.SelectedAccountID = selection.Account.ID
 		decision.SelectedAccountType = selection.Account.Type
+		if req.StickyWeighted {
+			if req.StickyPreviousAccountID > 0 && selection.Account.ID == req.StickyPreviousAccountID {
+				decision.StickyPreviousHit = true
+			}
+			if req.StickyAccountID > 0 && selection.Account.ID == req.StickyAccountID {
+				decision.StickySessionHit = true
+			}
+		}
 	}
 	return selection, decision, nil
 }
@@ -453,6 +491,13 @@ func openAIStickyAccountMatchesGroup(account *Account, groupID *int64) bool {
 	return false
 }
 
+func openAIAccountSchedulingPriority(account *Account) int {
+	if account == nil {
+		return 0
+	}
+	return account.Priority
+}
+
 func (s *defaultOpenAIAccountScheduler) shouldEscapeStickyAccount(accountID int64, cfg openAIStickyEscapeConfig) (reason string, errorRate float64, ttft float64, shouldEscape bool) {
 	if !cfg.enabled || s == nil || s.stats == nil || accountID <= 0 {
 		return "", 0, 0, false
@@ -471,6 +516,7 @@ type openAIAccountCandidateScore struct {
 	account   *Account
 	loadInfo  *AccountLoadInfo
 	score     float64
+	priority  int
 	errorRate float64
 	ttft      float64
 	hasTTFT   bool
@@ -669,6 +715,7 @@ func buildOpenAIWeightedSelectionOrder(
 }
 
 func (s *defaultOpenAIAccountScheduler) buildOpenAIAccountLoadPlan(
+	ctx context.Context,
 	req OpenAIAccountScheduleRequest,
 	filtered []*Account,
 	loadMap map[int64]*AccountLoadInfo,
@@ -716,18 +763,20 @@ func (s *defaultOpenAIAccountScheduler) buildOpenAIAccountLoadPlan(
 		return plan
 	}
 
-	minPriority, maxPriority := candidates[0].account.Priority, candidates[0].account.Priority
+	minPriority, maxPriority := openAIAccountSchedulingPriority(candidates[0].account), openAIAccountSchedulingPriority(candidates[0].account)
 	maxWaiting := 1
 	loadRateSum := 0.0
 	loadRateSumSquares := 0.0
 	minTTFT, maxTTFT := 0.0, 0.0
 	hasTTFTSample := false
-	for _, candidate := range candidates {
-		if candidate.account.Priority < minPriority {
-			minPriority = candidate.account.Priority
+	for i := range candidates {
+		candidate := &candidates[i]
+		candidate.priority = openAIAccountSchedulingPriority(candidate.account)
+		if candidate.priority < minPriority {
+			minPriority = candidate.priority
 		}
-		if candidate.account.Priority > maxPriority {
-			maxPriority = candidate.account.Priority
+		if candidate.priority > maxPriority {
+			maxPriority = candidate.priority
 		}
 		if candidate.loadInfo.WaitingCount > maxWaiting {
 			maxWaiting = candidate.loadInfo.WaitingCount
@@ -751,7 +800,7 @@ func (s *defaultOpenAIAccountScheduler) buildOpenAIAccountLoadPlan(
 	}
 	plan.loadSkew = calcLoadSkewByMoments(loadRateSum, loadRateSumSquares, len(candidates))
 
-	weights := s.service.openAIWSSchedulerWeights()
+	weights := s.service.openAIWSSchedulerWeightsForRequest(ctx)
 
 	// Reset 因子（use-it-or-lose-it）：在拥有「未来会话窗口结束时间」的账号中，
 	// 剩余时间越短 → 因子越接近 1（越早重置越优先用尽）。无活跃窗口的账号因子为 0。
@@ -785,7 +834,7 @@ func (s *defaultOpenAIAccountScheduler) buildOpenAIAccountLoadPlan(
 		item := &candidates[i]
 		priorityFactor := 1.0
 		if maxPriority > minPriority {
-			priorityFactor = 1 - float64(item.account.Priority-minPriority)/float64(maxPriority-minPriority)
+			priorityFactor = 1 - float64(item.priority-minPriority)/float64(maxPriority-minPriority)
 		}
 		loadFactor := 1 - clamp01(float64(item.loadInfo.LoadRate)/100.0)
 		queueFactor := 1 - clamp01(float64(item.loadInfo.WaitingCount)/float64(maxWaiting))
@@ -817,10 +866,18 @@ func (s *defaultOpenAIAccountScheduler) buildOpenAIAccountLoadPlan(
 			weights.TTFT*ttftFactor +
 			weights.Reset*resetFactor +
 			weights.QuotaHeadroom*quotaHeadroomFactor
+		if req.StickyWeighted {
+			if req.PreviousResponseCanMove && req.StickyPreviousAccountID > 0 && item.account.ID == req.StickyPreviousAccountID {
+				item.score += weights.Previous
+			}
+			if req.StickyAccountID > 0 && item.account.ID == req.StickyAccountID {
+				item.score += weights.SessionSticky
+			}
+		}
 	}
 	plan.candidates = candidates
 
-	plan.topK = s.service.openAIWSLBTopK()
+	plan.topK = s.service.openAIWSLBTopKForRequest(ctx)
 	if plan.topK > len(candidates) {
 		plan.topK = len(candidates)
 	}
@@ -845,6 +902,20 @@ func (s *defaultOpenAIAccountScheduler) buildOpenAISelectionOrder(
 			groupTopK = len(pool)
 		}
 		ranked := selectTopKOpenAICandidates(pool, groupTopK)
+		if req.StickyWeighted {
+			for _, stickyID := range []int64{req.StickyPreviousAccountID, req.StickyAccountID} {
+				if stickyID <= 0 {
+					continue
+				}
+				for i, candidate := range ranked {
+					if candidate.account != nil && candidate.account.ID == stickyID {
+						ordered := append([]openAIAccountCandidateScore{candidate}, ranked[:i]...)
+						ordered = append(ordered, ranked[i+1:]...)
+						return ordered
+					}
+				}
+			}
+		}
 		return buildOpenAIWeightedSelectionOrder(ranked, req)
 	}
 
@@ -939,6 +1010,66 @@ func (s *defaultOpenAIAccountScheduler) tryAcquireOpenAISelectionOrder(
 	return nil, compactBlocked, nil
 }
 
+func (s *defaultOpenAIAccountScheduler) tryFallbackToWeightedSticky(
+	ctx context.Context,
+	req OpenAIAccountScheduleRequest,
+) (*AccountSelectionResult, error) {
+	if !req.StickyWeighted {
+		return nil, nil
+	}
+	for _, accountID := range []int64{req.StickyPreviousAccountID, req.StickyAccountID} {
+		if accountID <= 0 {
+			continue
+		}
+		if req.ExcludedIDs != nil {
+			if _, excluded := req.ExcludedIDs[accountID]; excluded {
+				continue
+			}
+		}
+		account, err := s.service.getSchedulableAccount(ctx, accountID)
+		if err != nil || account == nil {
+			continue
+		}
+		if !s.isAccountRequestCompatible(ctx, account, req) || !s.isAccountTransportCompatible(account, req.RequiredTransport) {
+			continue
+		}
+		account = s.service.recheckSelectedOpenAIAccountFromDB(ctx, account, req.Platform, req.RequestedModel, req.RequireCompact, req.RequiredCapability)
+		if account == nil || !s.isAccountRequestCompatible(ctx, account, req) || !s.isAccountTransportCompatible(account, req.RequiredTransport) {
+			continue
+		}
+		if req.RequireCompact && openAICompactSupportTier(account) == 0 {
+			continue
+		}
+		result, acquireErr := s.service.tryAcquireAccountSlot(ctx, account.ID, account.Concurrency)
+		if acquireErr != nil {
+			return nil, acquireErr
+		}
+		if result != nil && result.Acquired {
+			if req.SessionHash != "" && !req.PreserveStickyBinding {
+				_ = s.service.BindStickySession(ctx, req.GroupID, req.SessionHash, account.ID)
+			}
+			return &AccountSelectionResult{
+				Account:     account,
+				Acquired:    true,
+				ReleaseFunc: result.ReleaseFunc,
+			}, nil
+		}
+		if s.service.concurrencyService != nil {
+			cfg := s.service.schedulingConfig()
+			return &AccountSelectionResult{
+				Account: account,
+				WaitPlan: &AccountWaitPlan{
+					AccountID:      account.ID,
+					MaxConcurrency: account.Concurrency,
+					Timeout:        cfg.StickySessionWaitTimeout,
+					MaxWaiting:     cfg.StickySessionMaxWaiting,
+				},
+			}, nil
+		}
+	}
+	return nil, nil
+}
+
 func (s *defaultOpenAIAccountScheduler) selectByLoadBalance(
 	ctx context.Context,
 	req OpenAIAccountScheduleRequest,
@@ -1002,52 +1133,159 @@ func (s *defaultOpenAIAccountScheduler) selectByLoadBalance(
 		}
 	}
 
-	plan := s.buildOpenAIAccountLoadPlan(req, filtered, loadMap)
-	candidateCount := plan.candidateCount
-	topK := plan.topK
-	loadSkew := plan.loadSkew
-	selectionOrder := plan.selectionOrder
-	if req.RequireCompact && len(plan.candidates) == 0 && len(plan.staleSnapshotCompactRetry) == 0 {
-		return nil, 0, 0, 0, ErrNoAvailableCompactAccounts
-	}
-	if req.RequireCompact && len(selectionOrder) == 0 && s.service.schedulerSnapshot == nil {
-		return nil, candidateCount, topK, loadSkew, ErrNoAvailableCompactAccounts
-	}
-	if len(selectionOrder) == 0 {
-		return nil, candidateCount, topK, loadSkew, noAvailableOpenAISelectionError(req.RequestedModel, req.RequireCompact && len(plan.allCandidates) > 0)
+	if req.SubscriptionPriority {
+		subscriptionAccounts, regularAccounts := partitionOpenAIChatGPTSubscriptionAccounts(filtered)
+		if len(subscriptionAccounts) > 0 {
+			attempt := s.trySelectByLoadBalancePool(ctx, req, subscriptionAccounts, loadMap)
+			if attempt.err != nil && (!attempt.noCompactCandidates || len(regularAccounts) <= 0) {
+				return nil, attempt.candidateCount, attempt.topK, attempt.loadSkew, attempt.err
+			}
+			if attempt.result != nil {
+				return attempt.result, attempt.candidateCount, attempt.topK, attempt.loadSkew, nil
+			}
+			if len(regularAccounts) > 0 {
+				regularAttempt := s.trySelectByLoadBalancePool(ctx, req, regularAccounts, loadMap)
+				if regularAttempt.err != nil {
+					return nil, regularAttempt.candidateCount, regularAttempt.topK, regularAttempt.loadSkew, regularAttempt.err
+				}
+				if regularAttempt.result != nil {
+					return regularAttempt.result, regularAttempt.candidateCount, regularAttempt.topK, regularAttempt.loadSkew, nil
+				}
+				return s.finishLoadBalanceSelectionFallback(ctx, req, regularAttempt)
+			}
+			return s.finishLoadBalanceSelectionFallback(ctx, req, attempt)
+		}
 	}
 
-	result, compactBlocked, acquireErr := s.tryAcquireOpenAISelectionOrder(ctx, req, selectionOrder)
+	attempt := s.trySelectByLoadBalancePool(ctx, req, filtered, loadMap)
+	if attempt.err != nil {
+		return nil, attempt.candidateCount, attempt.topK, attempt.loadSkew, attempt.err
+	}
+	if attempt.result != nil {
+		return attempt.result, attempt.candidateCount, attempt.topK, attempt.loadSkew, nil
+	}
+	return s.finishLoadBalanceSelectionFallback(ctx, req, attempt)
+}
+
+func partitionOpenAIChatGPTSubscriptionAccounts(accounts []*Account) ([]*Account, []*Account) {
+	subscriptionAccounts := make([]*Account, 0, len(accounts))
+	regularAccounts := make([]*Account, 0, len(accounts))
+	for _, account := range accounts {
+		if account != nil && account.IsOpenAIChatGPTSubscription() {
+			subscriptionAccounts = append(subscriptionAccounts, account)
+			continue
+		}
+		regularAccounts = append(regularAccounts, account)
+	}
+	return subscriptionAccounts, regularAccounts
+}
+
+func (s *defaultOpenAIAccountScheduler) trySelectByLoadBalancePool(
+	ctx context.Context,
+	req OpenAIAccountScheduleRequest,
+	filtered []*Account,
+	loadMap map[int64]*AccountLoadInfo,
+) openAIAccountLoadSelectionAttempt {
+	plan := s.buildOpenAIAccountLoadPlan(ctx, req, filtered, loadMap)
+	attempt := openAIAccountLoadSelectionAttempt{
+		selectionOrder: plan.selectionOrder,
+		candidateCount: plan.candidateCount,
+		topK:           plan.topK,
+		loadSkew:       plan.loadSkew,
+	}
+	if req.RequireCompact && len(plan.candidates) == 0 && len(plan.staleSnapshotCompactRetry) == 0 {
+		attempt.noCompactCandidates = true
+		attempt.err = ErrNoAvailableCompactAccounts
+		return attempt
+	}
+	if req.RequireCompact && len(attempt.selectionOrder) == 0 && s.service.schedulerSnapshot == nil {
+		attempt.noCompactCandidates = true
+		attempt.err = ErrNoAvailableCompactAccounts
+		return attempt
+	}
+	if len(attempt.selectionOrder) == 0 {
+		attempt.compactBlocked = req.RequireCompact && len(plan.allCandidates) > 0
+		return attempt
+	}
+
+	result, compactBlocked, acquireErr := s.tryAcquireOpenAISelectionOrder(ctx, req, attempt.selectionOrder)
+	attempt.compactBlocked = compactBlocked
 	if acquireErr != nil {
-		return nil, candidateCount, topK, loadSkew, acquireErr
+		attempt.err = acquireErr
+		return attempt
 	}
 	if result != nil {
-		return result, candidateCount, topK, loadSkew, nil
+		attempt.result = result
+		return attempt
 	}
 
 	if s.service.concurrencyService != nil {
+		loadReq := buildOpenAIAccountLoadRequest(filtered)
 		if freshLoadMap, loadErr := s.service.concurrencyService.GetAccountsLoadBatchFresh(ctx, loadReq); loadErr == nil {
-			freshPlan := s.buildOpenAIAccountLoadPlan(req, filtered, freshLoadMap)
+			freshPlan := s.buildOpenAIAccountLoadPlan(ctx, req, filtered, freshLoadMap)
 			if len(freshPlan.selectionOrder) > 0 {
 				freshResult, freshCompactBlocked, freshAcquireErr := s.tryAcquireOpenAISelectionOrder(ctx, req, freshPlan.selectionOrder)
 				if freshAcquireErr != nil {
-					return nil, candidateCount, topK, loadSkew, freshAcquireErr
+					attempt.err = freshAcquireErr
+					return attempt
 				}
 				if freshResult != nil {
-					return freshResult, freshPlan.candidateCount, freshPlan.topK, freshPlan.loadSkew, nil
+					attempt.result = freshResult
+					attempt.selectionOrder = freshPlan.selectionOrder
+					attempt.candidateCount = freshPlan.candidateCount
+					attempt.topK = freshPlan.topK
+					attempt.loadSkew = freshPlan.loadSkew
+					return attempt
 				}
-				compactBlocked = compactBlocked || freshCompactBlocked
-				selectionOrder = freshPlan.selectionOrder
-				candidateCount = freshPlan.candidateCount
-				topK = freshPlan.topK
-				loadSkew = freshPlan.loadSkew
+				attempt.compactBlocked = attempt.compactBlocked || freshCompactBlocked
+				attempt.selectionOrder = freshPlan.selectionOrder
+				attempt.candidateCount = freshPlan.candidateCount
+				attempt.topK = freshPlan.topK
+				attempt.loadSkew = freshPlan.loadSkew
 			}
 		}
 	}
 
+	return attempt
+}
+
+func buildOpenAIAccountLoadRequest(accounts []*Account) []AccountWithConcurrency {
+	loadReq := make([]AccountWithConcurrency, 0, len(accounts))
+	for _, account := range accounts {
+		if account == nil {
+			continue
+		}
+		loadReq = append(loadReq, AccountWithConcurrency{
+			ID:             account.ID,
+			MaxConcurrency: account.EffectiveLoadFactor(),
+		})
+	}
+	return loadReq
+}
+
+func (s *defaultOpenAIAccountScheduler) finishLoadBalanceSelectionFallback(
+	ctx context.Context,
+	req OpenAIAccountScheduleRequest,
+	attempt openAIAccountLoadSelectionAttempt,
+) (*AccountSelectionResult, int, int, float64, error) {
+	candidateCount := attempt.candidateCount
+	topK := attempt.topK
+	loadSkew := attempt.loadSkew
+
+	if len(attempt.selectionOrder) == 0 {
+		return nil, candidateCount, topK, loadSkew, noAvailableOpenAISelectionError(req.RequestedModel, attempt.compactBlocked)
+	}
+
+	if stickyFallback, stickyErr := s.tryFallbackToWeightedSticky(ctx, req); stickyErr != nil {
+		return nil, candidateCount, topK, loadSkew, stickyErr
+	} else if stickyFallback != nil {
+		return stickyFallback, candidateCount, topK, loadSkew, nil
+	}
+
 	cfg := s.service.schedulingConfig()
+	compactBlocked := attempt.compactBlocked
 	// WaitPlan.MaxConcurrency 使用 Concurrency（非 EffectiveLoadFactor），因为 WaitPlan 控制的是 Redis 实际并发槽位等待。
-	for _, candidate := range selectionOrder {
+	for _, candidate := range attempt.selectionOrder {
 		fresh := s.service.resolveFreshSchedulableOpenAIAccount(ctx, candidate.account, req.Platform, req.RequestedModel, false, req.RequiredCapability)
 		if fresh == nil || !s.isAccountTransportCompatible(fresh, req.RequiredTransport) || !s.isAccountRequestCompatible(ctx, fresh, req) {
 			continue
@@ -1184,40 +1422,164 @@ func (s *OpenAIGatewayService) openAIAdvancedSchedulerSettingRepo() SettingRepos
 	return s.rateLimitService.settingService.settingRepo
 }
 
-func (s *OpenAIGatewayService) isOpenAIAdvancedSchedulerEnabled(ctx context.Context) bool {
+func (s *OpenAIGatewayService) openAIAdvancedSchedulerRuntimeSettings(ctx context.Context) openAIAdvancedSchedulerRuntimeSettings {
 	if cached, ok := openAIAdvancedSchedulerSettingCache.Load().(*cachedOpenAIAdvancedSchedulerSetting); ok && cached != nil {
 		if time.Now().UnixNano() < cached.expiresAt {
-			return cached.enabled
+			return openAIAdvancedSchedulerRuntimeSettings{
+				enabled:                     cached.enabled,
+				stickyWeightedEnabled:       cached.stickyWeightedEnabled,
+				subscriptionPriorityEnabled: cached.subscriptionPriorityEnabled,
+				lbTopKOverride:              cached.lbTopKOverride,
+				weightOverrides:             cloneOpenAIAdvancedSchedulerWeightOverrides(cached.weightOverrides),
+			}
 		}
 	}
 
 	result, _, _ := openAIAdvancedSchedulerSettingSF.Do(openAIAdvancedSchedulerSettingKey, func() (any, error) {
 		if cached, ok := openAIAdvancedSchedulerSettingCache.Load().(*cachedOpenAIAdvancedSchedulerSetting); ok && cached != nil {
 			if time.Now().UnixNano() < cached.expiresAt {
-				return cached.enabled, nil
+				return openAIAdvancedSchedulerRuntimeSettings{
+					enabled:                     cached.enabled,
+					stickyWeightedEnabled:       cached.stickyWeightedEnabled,
+					subscriptionPriorityEnabled: cached.subscriptionPriorityEnabled,
+					lbTopKOverride:              cached.lbTopKOverride,
+					weightOverrides:             cloneOpenAIAdvancedSchedulerWeightOverrides(cached.weightOverrides),
+				}, nil
 			}
 		}
 
 		enabled := false
+		stickyWeightedEnabled := false
+		subscriptionPriorityEnabled := false
+		lbTopKOverride := 0
+		weightOverrides := map[string]float64{}
 		if repo := s.openAIAdvancedSchedulerSettingRepo(); repo != nil {
 			dbCtx, cancel := context.WithTimeout(context.WithoutCancel(ctx), openAIAdvancedSchedulerSettingDBTimeout)
 			defer cancel()
 
-			value, err := repo.GetValue(dbCtx, openAIAdvancedSchedulerSettingKey)
-			if err == nil {
-				enabled = strings.EqualFold(strings.TrimSpace(value), "true")
+			if values, err := repo.GetMultiple(dbCtx, openAIAdvancedSchedulerRuntimeSettingKeys()); err == nil {
+				enabled = strings.EqualFold(strings.TrimSpace(values[openAIAdvancedSchedulerSettingKey]), "true")
+				stickyWeightedEnabled = strings.EqualFold(strings.TrimSpace(values[SettingKeyOpenAIAdvancedSchedulerStickyWeightedEnabled]), "true")
+				subscriptionPriorityEnabled = strings.EqualFold(strings.TrimSpace(values[SettingKeyOpenAIAdvancedSchedulerSubscriptionPriorityEnabled]), "true")
+				lbTopKOverride = parsePositiveIntOverride(values[SettingKeyOpenAIAdvancedSchedulerLBTopK])
+				weightOverrides = parseOpenAIAdvancedSchedulerWeightOverrides(values)
+			} else {
+				if value, err := repo.GetValue(dbCtx, openAIAdvancedSchedulerSettingKey); err == nil {
+					enabled = strings.EqualFold(strings.TrimSpace(value), "true")
+				}
+				if value, err := repo.GetValue(dbCtx, SettingKeyOpenAIAdvancedSchedulerStickyWeightedEnabled); err == nil {
+					stickyWeightedEnabled = strings.EqualFold(strings.TrimSpace(value), "true")
+				}
+				if value, err := repo.GetValue(dbCtx, SettingKeyOpenAIAdvancedSchedulerSubscriptionPriorityEnabled); err == nil {
+					subscriptionPriorityEnabled = strings.EqualFold(strings.TrimSpace(value), "true")
+				}
 			}
 		}
 
 		openAIAdvancedSchedulerSettingCache.Store(&cachedOpenAIAdvancedSchedulerSetting{
-			enabled:   enabled,
-			expiresAt: time.Now().Add(openAIAdvancedSchedulerSettingCacheTTL).UnixNano(),
+			enabled:                     enabled,
+			stickyWeightedEnabled:       stickyWeightedEnabled,
+			subscriptionPriorityEnabled: subscriptionPriorityEnabled,
+			lbTopKOverride:              lbTopKOverride,
+			weightOverrides:             cloneOpenAIAdvancedSchedulerWeightOverrides(weightOverrides),
+			expiresAt:                   time.Now().Add(openAIAdvancedSchedulerSettingCacheTTL).UnixNano(),
 		})
-		return enabled, nil
+		return openAIAdvancedSchedulerRuntimeSettings{
+			enabled:                     enabled,
+			stickyWeightedEnabled:       stickyWeightedEnabled,
+			subscriptionPriorityEnabled: subscriptionPriorityEnabled,
+			lbTopKOverride:              lbTopKOverride,
+			weightOverrides:             weightOverrides,
+		}, nil
 	})
 
-	enabled, _ := result.(bool)
-	return enabled
+	settings, _ := result.(openAIAdvancedSchedulerRuntimeSettings)
+	return settings
+}
+
+func (s *OpenAIGatewayService) isOpenAIAdvancedSchedulerEnabled(ctx context.Context) bool {
+	return s.openAIAdvancedSchedulerRuntimeSettings(ctx).enabled
+}
+
+func (s *OpenAIGatewayService) isOpenAIAdvancedSchedulerStickyWeightedEnabled(ctx context.Context) bool {
+	settings := s.openAIAdvancedSchedulerRuntimeSettings(ctx)
+	return settings.enabled && settings.stickyWeightedEnabled
+}
+
+func (s *OpenAIGatewayService) isOpenAIAdvancedSchedulerSubscriptionPriorityEnabled(ctx context.Context) bool {
+	settings := s.openAIAdvancedSchedulerRuntimeSettings(ctx)
+	return settings.enabled && settings.subscriptionPriorityEnabled
+}
+
+func openAIAdvancedSchedulerRuntimeSettingKeys() []string {
+	keys := []string{
+		openAIAdvancedSchedulerSettingKey,
+		SettingKeyOpenAIAdvancedSchedulerStickyWeightedEnabled,
+		SettingKeyOpenAIAdvancedSchedulerSubscriptionPriorityEnabled,
+		SettingKeyOpenAIAdvancedSchedulerLBTopK,
+	}
+	for _, spec := range openAIAdvancedSchedulerWeightOverrideSpecs() {
+		keys = append(keys, spec.key)
+	}
+	return keys
+}
+
+type openAIAdvancedSchedulerWeightOverrideSpec struct {
+	key  string
+	name string
+}
+
+func openAIAdvancedSchedulerWeightOverrideSpecs() []openAIAdvancedSchedulerWeightOverrideSpec {
+	return []openAIAdvancedSchedulerWeightOverrideSpec{
+		{key: SettingKeyOpenAIAdvancedSchedulerWeightPriority, name: "priority"},
+		{key: SettingKeyOpenAIAdvancedSchedulerWeightLoad, name: "load"},
+		{key: SettingKeyOpenAIAdvancedSchedulerWeightQueue, name: "queue"},
+		{key: SettingKeyOpenAIAdvancedSchedulerWeightErrorRate, name: "error_rate"},
+		{key: SettingKeyOpenAIAdvancedSchedulerWeightTTFT, name: "ttft"},
+		{key: SettingKeyOpenAIAdvancedSchedulerWeightReset, name: "reset"},
+		{key: SettingKeyOpenAIAdvancedSchedulerWeightQuotaHeadroom, name: "quota_headroom"},
+		{key: SettingKeyOpenAIAdvancedSchedulerWeightPreviousResponse, name: "previous_response"},
+		{key: SettingKeyOpenAIAdvancedSchedulerWeightSessionSticky, name: "session_sticky"},
+	}
+}
+
+func parsePositiveIntOverride(raw string) int {
+	raw = strings.TrimSpace(raw)
+	if raw == "" {
+		return 0
+	}
+	value, err := strconv.Atoi(raw)
+	if err != nil || value <= 0 {
+		return 0
+	}
+	return value
+}
+
+func parseOpenAIAdvancedSchedulerWeightOverrides(values map[string]string) map[string]float64 {
+	overrides := map[string]float64{}
+	for _, spec := range openAIAdvancedSchedulerWeightOverrideSpecs() {
+		raw := strings.TrimSpace(values[spec.key])
+		if raw == "" {
+			continue
+		}
+		value, err := strconv.ParseFloat(raw, 64)
+		if err != nil || value < 0 || math.IsNaN(value) || math.IsInf(value, 0) {
+			continue
+		}
+		overrides[spec.name] = value
+	}
+	return overrides
+}
+
+func cloneOpenAIAdvancedSchedulerWeightOverrides(in map[string]float64) map[string]float64 {
+	if len(in) == 0 {
+		return nil
+	}
+	out := make(map[string]float64, len(in))
+	for key, value := range in {
+		out[key] = value
+	}
+	return out
 }
 
 func (s *OpenAIGatewayService) getOpenAIAccountScheduler(ctx context.Context) OpenAIAccountScheduler {
@@ -1253,7 +1615,7 @@ func (s *OpenAIGatewayService) SelectAccountWithScheduler(
 	requiredTransport OpenAIUpstreamTransport,
 	requireCompact bool,
 ) (*AccountSelectionResult, OpenAIAccountScheduleDecision, error) {
-	return s.selectAccountWithScheduler(ctx, groupID, previousResponseID, sessionHash, requestedModel, excludedIDs, requiredTransport, "", "", requireCompact, PlatformOpenAI)
+	return s.selectAccountWithScheduler(ctx, groupID, previousResponseID, sessionHash, requestedModel, excludedIDs, requiredTransport, "", "", requireCompact, PlatformOpenAI, false)
 }
 
 func (s *OpenAIGatewayService) SelectAccountWithSchedulerForCapability(
@@ -1269,10 +1631,14 @@ func (s *OpenAIGatewayService) SelectAccountWithSchedulerForCapability(
 	platformOverride ...string,
 ) (*AccountSelectionResult, OpenAIAccountScheduleDecision, error) {
 	platform := PlatformOpenAI
+	previousResponseCanMove := false
 	if len(platformOverride) > 0 {
 		platform = platformOverride[0]
 	}
-	return s.selectAccountWithScheduler(ctx, groupID, previousResponseID, sessionHash, requestedModel, excludedIDs, requiredTransport, requiredCapability, "", requireCompact, platform)
+	if len(platformOverride) > 1 {
+		previousResponseCanMove = strings.EqualFold(platformOverride[1], "previous_response_can_move")
+	}
+	return s.selectAccountWithScheduler(ctx, groupID, previousResponseID, sessionHash, requestedModel, excludedIDs, requiredTransport, requiredCapability, "", requireCompact, platform, previousResponseCanMove)
 }
 
 func (s *OpenAIGatewayService) SelectAccountWithSchedulerForImages(
@@ -1283,13 +1649,13 @@ func (s *OpenAIGatewayService) SelectAccountWithSchedulerForImages(
 	excludedIDs map[int64]struct{},
 	requiredCapability OpenAIImagesCapability,
 ) (*AccountSelectionResult, OpenAIAccountScheduleDecision, error) {
-	selection, decision, err := s.selectAccountWithScheduler(ctx, groupID, "", sessionHash, requestedModel, excludedIDs, OpenAIUpstreamTransportHTTPSSE, "", requiredCapability, false, PlatformOpenAI)
+	selection, decision, err := s.selectAccountWithScheduler(ctx, groupID, "", sessionHash, requestedModel, excludedIDs, OpenAIUpstreamTransportHTTPSSE, "", requiredCapability, false, PlatformOpenAI, false)
 	if err == nil && selection != nil && selection.Account != nil {
 		return selection, decision, nil
 	}
 	// 如果要求 native 能力（如指定了模型）但没有可用的 APIKey 账号，回退到 basic（OAuth 账号）
 	if requiredCapability == OpenAIImagesCapabilityNative {
-		return s.selectAccountWithScheduler(ctx, groupID, "", sessionHash, requestedModel, excludedIDs, OpenAIUpstreamTransportHTTPSSE, "", OpenAIImagesCapabilityBasic, false, PlatformOpenAI)
+		return s.selectAccountWithScheduler(ctx, groupID, "", sessionHash, requestedModel, excludedIDs, OpenAIUpstreamTransportHTTPSSE, "", OpenAIImagesCapabilityBasic, false, PlatformOpenAI, false)
 	}
 	return selection, decision, err
 }
@@ -1306,6 +1672,7 @@ func (s *OpenAIGatewayService) selectAccountWithScheduler(
 	requiredImageCapability OpenAIImagesCapability,
 	requireCompact bool,
 	platform string,
+	previousResponseCanMove bool,
 ) (*AccountSelectionResult, OpenAIAccountScheduleDecision, error) {
 	ctx = s.withOpenAIQuotaAutoPauseContext(ctx)
 	platform = normalizeOpenAICompatiblePlatform(platform)
@@ -1378,13 +1745,23 @@ func (s *OpenAIGatewayService) selectAccountWithScheduler(
 			stickyAccountID = accountID
 		}
 	}
+	stickyWeighted := s.isOpenAIAdvancedSchedulerStickyWeightedEnabled(ctx)
+	subscriptionPriority := s.isOpenAIAdvancedSchedulerSubscriptionPriorityEnabled(ctx)
+	stickyPreviousAccountID := int64(0)
+	if stickyWeighted && previousResponseCanMove && strings.TrimSpace(previousResponseID) != "" && platform == PlatformOpenAI {
+		stickyPreviousAccountID = s.ResolveAccountIDByPreviousResponseIDForScheduler(ctx, groupID, previousResponseID, requestedModel, excludedIDs, requiredCapability, requireCompact)
+	}
 
 	return scheduler.Select(ctx, OpenAIAccountScheduleRequest{
 		GroupID:                 groupID,
 		Platform:                platform,
 		SessionHash:             sessionHash,
 		StickyAccountID:         stickyAccountID,
+		StickyPreviousAccountID: stickyPreviousAccountID,
+		StickyWeighted:          stickyWeighted,
+		SubscriptionPriority:    subscriptionPriority,
 		PreviousResponseID:      previousResponseID,
+		PreviousResponseCanMove: previousResponseCanMove,
 		RequestedModel:          requestedModel,
 		RequiredTransport:       requiredTransport,
 		RequiredCapability:      requiredCapability,
@@ -1473,6 +1850,15 @@ func (s *OpenAIGatewayService) openAIWSLBTopK() int {
 	return 7
 }
 
+func (s *OpenAIGatewayService) openAIWSLBTopKForRequest(ctx context.Context) int {
+	base := s.openAIWSLBTopK()
+	settings := s.openAIAdvancedSchedulerRuntimeSettings(ctx)
+	if settings.lbTopKOverride > 0 {
+		return settings.lbTopKOverride
+	}
+	return base
+}
+
 func (s *OpenAIGatewayService) openAIStickyEscapeConfig() openAIStickyEscapeConfig {
 	if s != nil && s.cfg != nil {
 		cfg := s.cfg.Gateway.OpenAIScheduler
@@ -1514,6 +1900,8 @@ func (s *OpenAIGatewayService) openAIWSSchedulerWeights() GatewayOpenAIWSSchedul
 			TTFT:          s.cfg.Gateway.OpenAIWS.SchedulerScoreWeights.TTFT,
 			Reset:         s.cfg.Gateway.OpenAIWS.SchedulerScoreWeights.Reset,
 			QuotaHeadroom: s.cfg.Gateway.OpenAIWS.SchedulerScoreWeights.QuotaHeadroom,
+			Previous:      s.cfg.Gateway.OpenAIWS.SchedulerScoreWeights.PreviousResponse,
+			SessionSticky: s.cfg.Gateway.OpenAIWS.SchedulerScoreWeights.SessionSticky,
 		}
 	}
 	return GatewayOpenAIWSSchedulerScoreWeightsView{
@@ -1524,9 +1912,46 @@ func (s *OpenAIGatewayService) openAIWSSchedulerWeights() GatewayOpenAIWSSchedul
 		TTFT:          0.5,
 		Reset:         0.0,
 		QuotaHeadroom: 0.0,
+		Previous:      5.0,
+		SessionSticky: 3.0,
 	}
 }
 
+func (s *OpenAIGatewayService) openAIWSSchedulerWeightsForRequest(ctx context.Context) GatewayOpenAIWSSchedulerScoreWeightsView {
+	weights := s.openAIWSSchedulerWeights()
+	settings := s.openAIAdvancedSchedulerRuntimeSettings(ctx)
+	return applyOpenAIAdvancedSchedulerWeightOverrides(weights, settings.weightOverrides)
+}
+
+func applyOpenAIAdvancedSchedulerWeightOverrides(
+	weights GatewayOpenAIWSSchedulerScoreWeightsView,
+	overrides map[string]float64,
+) GatewayOpenAIWSSchedulerScoreWeightsView {
+	for key, value := range overrides {
+		switch key {
+		case "priority":
+			weights.Priority = value
+		case "load":
+			weights.Load = value
+		case "queue":
+			weights.Queue = value
+		case "error_rate":
+			weights.ErrorRate = value
+		case "ttft":
+			weights.TTFT = value
+		case "reset":
+			weights.Reset = value
+		case "quota_headroom":
+			weights.QuotaHeadroom = value
+		case "previous_response":
+			weights.Previous = value
+		case "session_sticky":
+			weights.SessionSticky = value
+		}
+	}
+	return weights
+}
+
 type GatewayOpenAIWSSchedulerScoreWeightsView struct {
 	Priority  float64
 	Load      float64
@@ -1536,6 +1961,149 @@ type GatewayOpenAIWSSchedulerScoreWeightsView struct {
 	// Reset 倾向「会话窗口最早重置」的账号；0 表示关闭（默认）。
 	Reset         float64
 	QuotaHeadroom float64
+	Previous      float64
+	SessionSticky float64
+}
+
+type OpenAIAccountSchedulerScoreSnapshot struct {
+	BaseScore             float64
+	StickyScore           float64
+	StickyScoreInfinity   bool
+	StickyWeightedEnabled bool
+}
+
+func (s *RateLimitService) BuildOpenAIAccountSchedulerScoreSnapshot(
+	ctx context.Context,
+	accounts []*Account,
+	loadMap map[int64]*AccountLoadInfo,
+) map[int64]OpenAIAccountSchedulerScoreSnapshot {
+	gateway := &OpenAIGatewayService{cfg: nil, rateLimitService: s}
+	if s != nil {
+		gateway.cfg = s.cfg
+	}
+	return buildOpenAIAccountSchedulerScoreSnapshot(accounts, loadMap, gateway.openAIWSSchedulerWeightsForRequest(ctx), gateway.isOpenAIAdvancedSchedulerStickyWeightedEnabled(ctx))
+}
+
+func BuildOpenAIAccountSchedulerScoreSnapshot(
+	accounts []*Account,
+	loadMap map[int64]*AccountLoadInfo,
+) map[int64]OpenAIAccountSchedulerScoreSnapshot {
+	gateway := &OpenAIGatewayService{}
+	return buildOpenAIAccountSchedulerScoreSnapshot(accounts, loadMap, gateway.openAIWSSchedulerWeights(), false)
+}
+
+func buildOpenAIAccountSchedulerScoreSnapshot(
+	accounts []*Account,
+	loadMap map[int64]*AccountLoadInfo,
+	weights GatewayOpenAIWSSchedulerScoreWeightsView,
+	stickyWeightedEnabled bool,
+) map[int64]OpenAIAccountSchedulerScoreSnapshot {
+	if len(accounts) == 0 {
+		return nil
+	}
+	candidates := make([]openAIAccountCandidateScore, 0, len(accounts))
+	for _, account := range accounts {
+		if account == nil {
+			continue
+		}
+		loadInfo := loadMap[account.ID]
+		if loadInfo == nil {
+			loadInfo = &AccountLoadInfo{AccountID: account.ID}
+		}
+		candidates = append(candidates, openAIAccountCandidateScore{
+			account:   account,
+			loadInfo:  loadInfo,
+			errorRate: 0,
+			ttft:      0,
+			hasTTFT:   false,
+		})
+	}
+	if len(candidates) == 0 {
+		return nil
+	}
+
+	minPriority, maxPriority := openAIAccountSchedulingPriority(candidates[0].account), openAIAccountSchedulingPriority(candidates[0].account)
+	maxWaiting := 1
+	for i := range candidates {
+		candidate := &candidates[i]
+		candidate.priority = openAIAccountSchedulingPriority(candidate.account)
+		if candidate.priority < minPriority {
+			minPriority = candidate.priority
+		}
+		if candidate.priority > maxPriority {
+			maxPriority = candidate.priority
+		}
+		if candidate.loadInfo.WaitingCount > maxWaiting {
+			maxWaiting = candidate.loadInfo.WaitingCount
+		}
+	}
+
+	minResetRemaining, maxResetRemaining := 0.0, 0.0
+	hasResetSample := false
+	now := time.Now()
+	if weights.Reset > 0 {
+		for _, candidate := range candidates {
+			end := candidate.account.SessionWindowEnd
+			if end == nil || !now.Before(*end) {
+				continue
+			}
+			remaining := end.Sub(now).Seconds()
+			if !hasResetSample {
+				minResetRemaining, maxResetRemaining = remaining, remaining
+				hasResetSample = true
+				continue
+			}
+			if remaining < minResetRemaining {
+				minResetRemaining = remaining
+			}
+			if remaining > maxResetRemaining {
+				maxResetRemaining = remaining
+			}
+		}
+	}
+
+	result := make(map[int64]OpenAIAccountSchedulerScoreSnapshot, len(candidates))
+	for _, candidate := range candidates {
+		priorityFactor := 1.0
+		if maxPriority > minPriority {
+			priorityFactor = 1 - float64(candidate.priority-minPriority)/float64(maxPriority-minPriority)
+		}
+		loadFactor := 1 - clamp01(float64(candidate.loadInfo.LoadRate)/100.0)
+		queueFactor := 1 - clamp01(float64(candidate.loadInfo.WaitingCount)/float64(maxWaiting))
+		errorFactor := 1.0
+		ttftFactor := 0.5
+		resetFactor := 0.0
+		if weights.Reset > 0 && hasResetSample {
+			if end := candidate.account.SessionWindowEnd; end != nil && now.Before(*end) {
+				if maxResetRemaining > minResetRemaining {
+					resetFactor = 1 - clamp01((end.Sub(now).Seconds()-minResetRemaining)/(maxResetRemaining-minResetRemaining))
+				} else {
+					resetFactor = 1
+				}
+			}
+		}
+		quotaHeadroomFactor := 0.0
+		if weights.QuotaHeadroom > 0 {
+			quotaHeadroomFactor = openAIQuotaHeadroomFactor(candidate.account, now)
+		}
+		baseScore := weights.Priority*priorityFactor +
+			weights.Load*loadFactor +
+			weights.Queue*queueFactor +
+			weights.ErrorRate*errorFactor +
+			weights.TTFT*ttftFactor +
+			weights.Reset*resetFactor +
+			weights.QuotaHeadroom*quotaHeadroomFactor
+		score := OpenAIAccountSchedulerScoreSnapshot{
+			BaseScore:             baseScore,
+			StickyWeightedEnabled: stickyWeightedEnabled,
+			StickyScoreInfinity:   !stickyWeightedEnabled,
+		}
+		if stickyWeightedEnabled {
+			score.StickyScore = baseScore + weights.Previous + weights.SessionSticky
+		}
+		result[candidate.account.ID] = score
+	}
+	return result
 }
 
 func openAIQuotaHeadroomFactor(account *Account, now time.Time) float64 {
diff --git a/backend/internal/service/openai_account_scheduler_reset_test.go b/backend/internal/service/openai_account_scheduler_reset_test.go
index 50155f20d6..f1270c8caa 100644
--- a/backend/internal/service/openai_account_scheduler_reset_test.go
+++ b/backend/internal/service/openai_account_scheduler_reset_test.go
@@ -1,6 +1,7 @@
 package service
 
 import (
+	"context"
 	"testing"
 	"time"
 
@@ -49,7 +50,7 @@ func TestBuildOpenAIAccountLoadPlan_ResetWeightPrefersSoonestReset(t *testing.T)
 	}
 	sched := openAIResetTestScheduler(5.0)
 
-	plan := sched.buildOpenAIAccountLoadPlan(OpenAIAccountScheduleRequest{}, filtered, map[int64]*AccountLoadInfo{})
+	plan := sched.buildOpenAIAccountLoadPlan(context.Background(), OpenAIAccountScheduleRequest{}, filtered, map[int64]*AccountLoadInfo{})
 	scores := openAIPlanScores(plan)
 	require.Greater(t, scores[2], scores[1], "重置时间最早的账号（ID=2）得分更高")
 }
@@ -65,7 +66,7 @@ func TestBuildOpenAIAccountLoadPlan_ResetWeightZeroNoEffect(t *testing.T) {
 	}
 	sched := openAIResetTestScheduler(0.0)
 
-	plan := sched.buildOpenAIAccountLoadPlan(OpenAIAccountScheduleRequest{}, filtered, map[int64]*AccountLoadInfo{})
+	plan := sched.buildOpenAIAccountLoadPlan(context.Background(), OpenAIAccountScheduleRequest{}, filtered, map[int64]*AccountLoadInfo{})
 	scores := openAIPlanScores(plan)
 	require.Equal(t, scores[1], scores[2], "Reset 权重为 0 时两账号得分相同")
 }
@@ -80,7 +81,7 @@ func TestBuildOpenAIAccountLoadPlan_ResetWeightIgnoresNilWindow(t *testing.T) {
 	}
 	sched := openAIResetTestScheduler(5.0)
 
-	plan := sched.buildOpenAIAccountLoadPlan(OpenAIAccountScheduleRequest{}, filtered, map[int64]*AccountLoadInfo{})
+	plan := sched.buildOpenAIAccountLoadPlan(context.Background(), OpenAIAccountScheduleRequest{}, filtered, map[int64]*AccountLoadInfo{})
 	scores := openAIPlanScores(plan)
 	require.Greater(t, scores[2], scores[1], "拥有活跃窗口的账号得分高于无窗口账号")
 }
@@ -161,7 +162,7 @@ func TestBuildOpenAIAccountLoadPlan_QuotaHeadroomPrefersHigher7dRemaining(t *tes
 	}
 	sched := openAIQuotaHeadroomTestScheduler(1.0)
 
-	plan := sched.buildOpenAIAccountLoadPlan(OpenAIAccountScheduleRequest{}, filtered, map[int64]*AccountLoadInfo{})
+	plan := sched.buildOpenAIAccountLoadPlan(context.Background(), OpenAIAccountScheduleRequest{}, filtered, map[int64]*AccountLoadInfo{})
 	scores := openAIPlanScores(plan)
 	require.Greater(t, scores[2], scores[1], "7d 剩余额度更高的账号得分应更高")
 }
@@ -190,7 +191,7 @@ func TestBuildOpenAIAccountLoadPlan_QuotaHeadroomZeroNoEffect(t *testing.T) {
 	}
 	sched := openAIResetTestScheduler(0)
 
-	plan := sched.buildOpenAIAccountLoadPlan(OpenAIAccountScheduleRequest{}, filtered, map[int64]*AccountLoadInfo{})
+	plan := sched.buildOpenAIAccountLoadPlan(context.Background(), OpenAIAccountScheduleRequest{}, filtered, map[int64]*AccountLoadInfo{})
 	scores := openAIPlanScores(plan)
 	require.Equal(t, scores[1], scores[2], "quota_headroom 权重为 0 时不应影响打分")
 }
diff --git a/backend/internal/service/openai_account_scheduler_test.go b/backend/internal/service/openai_account_scheduler_test.go
index 0255dbbde3..a61a923053 100644
--- a/backend/internal/service/openai_account_scheduler_test.go
+++ b/backend/internal/service/openai_account_scheduler_test.go
@@ -183,6 +183,17 @@ func newSchedulerTestOpenAIWSV2Config() *config.Config {
 	return cfg
 }
 
+func newSchedulerTestSubscriptionPriorityConfig() *config.Config {
+	cfg := &config.Config{}
+	cfg.Gateway.OpenAIWS.LBTopK = 1
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.Priority = 1
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.Load = 1
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.Queue = 1
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.ErrorRate = 0
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.TTFT = 0
+	return cfg
+}
+
 type openAIAdvancedSchedulerSettingRepoStub struct {
 	values map[string]string
 }
@@ -210,8 +221,14 @@ func (s *openAIAdvancedSchedulerSettingRepoStub) Set(context.Context, string, st
 	panic("unexpected call to Set")
 }
 
-func (s *openAIAdvancedSchedulerSettingRepoStub) GetMultiple(context.Context, []string) (map[string]string, error) {
-	panic("unexpected call to GetMultiple")
+func (s *openAIAdvancedSchedulerSettingRepoStub) GetMultiple(_ context.Context, keys []string) (map[string]string, error) {
+	result := make(map[string]string, len(keys))
+	for _, key := range keys {
+		if value, err := s.GetValue(context.Background(), key); err == nil {
+			result[key] = value
+		}
+	}
+	return result, nil
 }
 
 func (s *openAIAdvancedSchedulerSettingRepoStub) SetMultiple(context.Context, map[string]string) error {
@@ -226,7 +243,7 @@ func (s *openAIAdvancedSchedulerSettingRepoStub) Delete(context.Context, string)
 	panic("unexpected call to Delete")
 }
 
-func newOpenAIAdvancedSchedulerRateLimitService(enabled string) *RateLimitService {
+func newOpenAIAdvancedSchedulerRateLimitService(enabled string, values ...string) *RateLimitService {
 	resetOpenAIAdvancedSchedulerSettingCacheForTest()
 	repo := &openAIAdvancedSchedulerSettingRepoStub{
 		values: map[string]string{},
@@ -234,6 +251,12 @@ func newOpenAIAdvancedSchedulerRateLimitService(enabled string) *RateLimitServic
 	if enabled != "" {
 		repo.values[openAIAdvancedSchedulerSettingKey] = enabled
 	}
+	if len(values) > 0 && values[0] != "" {
+		repo.values[SettingKeyOpenAIAdvancedSchedulerStickyWeightedEnabled] = values[0]
+	}
+	if len(values) > 1 && values[1] != "" {
+		repo.values[SettingKeyOpenAIAdvancedSchedulerSubscriptionPriorityEnabled] = values[1]
+	}
 	return &RateLimitService{
 		settingService: NewSettingService(repo, &config.Config{}),
 	}
@@ -266,6 +289,45 @@ func (s *openAISnapshotCacheStub) GetAccount(ctx context.Context, accountID int6
 	return &cloned, nil
 }
 
+func TestOpenAIGatewayService_OpenAIAdvancedSchedulerRuntimeSettings_DBOverridesConfig(t *testing.T) {
+	resetOpenAIAdvancedSchedulerSettingCacheForTest()
+	defer resetOpenAIAdvancedSchedulerSettingCacheForTest()
+
+	cfg := &config.Config{}
+	cfg.Gateway.OpenAIWS.LBTopK = 11
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights = config.GatewayOpenAIWSSchedulerScoreWeights{
+		Priority:         1,
+		Load:             2,
+		Queue:            3,
+		ErrorRate:        4,
+		TTFT:             5,
+		Reset:            6,
+		QuotaHeadroom:    7,
+		PreviousResponse: 8,
+		SessionSticky:    9,
+	}
+	repo := &openAIAdvancedSchedulerSettingRepoStub{
+		values: map[string]string{
+			openAIAdvancedSchedulerSettingKey:                       "true",
+			SettingKeyOpenAIAdvancedSchedulerLBTopK:                 "3",
+			SettingKeyOpenAIAdvancedSchedulerWeightPriority:         "2.5",
+			SettingKeyOpenAIAdvancedSchedulerWeightPreviousResponse: "12",
+		},
+	}
+	svc := &OpenAIGatewayService{
+		cfg:              cfg,
+		rateLimitService: &RateLimitService{settingService: NewSettingService(repo, cfg)},
+	}
+
+	ctx := context.Background()
+	require.Equal(t, 3, svc.openAIWSLBTopKForRequest(ctx))
+	weights := svc.openAIWSSchedulerWeightsForRequest(ctx)
+	require.Equal(t, 2.5, weights.Priority)
+	require.Equal(t, 2.0, weights.Load)
+	require.Equal(t, 12.0, weights.Previous)
+	require.Equal(t, 9.0, weights.SessionSticky)
+}
+
 func TestOpenAIGatewayService_SelectAccountWithScheduler_DefaultDisabledUsesLegacyLoadAwareness(t *testing.T) {
 	resetOpenAIAdvancedSchedulerSettingCacheForTest()
 
@@ -584,6 +646,247 @@ func TestOpenAIGatewayService_SelectAccountWithScheduler_EnabledUsesAdvancedPrev
 	require.True(t, decision.StickyPreviousHit)
 }
 
+func TestOpenAIGatewayService_SelectAccountWithScheduler_StickyWeightedSessionInTopKUsesStickyFirst(t *testing.T) {
+	resetOpenAIAdvancedSchedulerSettingCacheForTest()
+
+	ctx := context.Background()
+	groupID := int64(101071)
+	accounts := []Account{
+		{
+			ID:          37101,
+			Platform:    PlatformOpenAI,
+			Type:        AccountTypeAPIKey,
+			Status:      StatusActive,
+			Schedulable: true,
+			Concurrency: 1,
+			Priority:    100,
+			GroupIDs:    []int64{groupID},
+		},
+		{
+			ID:          37102,
+			Platform:    PlatformOpenAI,
+			Type:        AccountTypeAPIKey,
+			Status:      StatusActive,
+			Schedulable: true,
+			Concurrency: 1,
+			Priority:    0,
+			GroupIDs:    []int64{groupID},
+		},
+	}
+	cfg := &config.Config{}
+	cfg.Gateway.OpenAIWS.LBTopK = 2
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.Priority = 1
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.Load = 1
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.Queue = 0.7
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.ErrorRate = 0.8
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.TTFT = 0.5
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.SessionSticky = 3
+	cache := &schedulerTestGatewayCache{sessionBindings: map[string]int64{
+		"openai:session_hash_weighted_topk": 37101,
+	}}
+	svc := &OpenAIGatewayService{
+		accountRepo:        schedulerTestOpenAIAccountRepo{accounts: accounts},
+		cache:              cache,
+		cfg:                cfg,
+		rateLimitService:   newOpenAIAdvancedSchedulerRateLimitService("true", "true"),
+		concurrencyService: NewConcurrencyService(schedulerTestConcurrencyCache{}),
+	}
+
+	selection, decision, err := svc.SelectAccountWithScheduler(
+		ctx,
+		&groupID,
+		"",
+		"session_hash_weighted_topk",
+		"gpt-5.1",
+		nil,
+		OpenAIUpstreamTransportAny,
+		false,
+	)
+	require.NoError(t, err)
+	require.NotNil(t, selection)
+	require.NotNil(t, selection.Account)
+	require.Equal(t, int64(37101), selection.Account.ID)
+	require.Equal(t, openAIAccountScheduleLayerLoadBalance, decision.Layer)
+	require.True(t, decision.StickySessionHit)
+	require.Equal(t, 2, decision.TopK)
+	if selection.ReleaseFunc != nil {
+		selection.ReleaseFunc()
+	}
+}
+
+func TestOpenAIGatewayService_SelectAccountWithScheduler_StickyWeightedPreviousRequiresMovableContext(t *testing.T) {
+	resetOpenAIAdvancedSchedulerSettingCacheForTest()
+
+	ctx := context.Background()
+	groupID := int64(101072)
+	accounts := []Account{
+		{
+			ID:          37111,
+			Platform:    PlatformOpenAI,
+			Type:        AccountTypeAPIKey,
+			Status:      StatusActive,
+			Schedulable: true,
+			Concurrency: 1,
+			Priority:    100,
+			GroupIDs:    []int64{groupID},
+			Extra: map[string]any{
+				"openai_apikey_responses_websockets_v2_enabled": true,
+			},
+		},
+		{
+			ID:          37112,
+			Platform:    PlatformOpenAI,
+			Type:        AccountTypeAPIKey,
+			Status:      StatusActive,
+			Schedulable: true,
+			Concurrency: 1,
+			Priority:    0,
+			GroupIDs:    []int64{groupID},
+			Extra: map[string]any{
+				"openai_apikey_responses_websockets_v2_enabled": true,
+			},
+		},
+	}
+	cfg := newSchedulerTestOpenAIWSV2Config()
+	cfg.Gateway.OpenAIWS.LBTopK = 1
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.Priority = 1
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.Load = 1
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.Queue = 0.7
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.ErrorRate = 0.8
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.TTFT = 0.5
+	svc := &OpenAIGatewayService{
+		accountRepo:        schedulerTestOpenAIAccountRepo{accounts: accounts},
+		cache:              &schedulerTestGatewayCache{},
+		cfg:                cfg,
+		rateLimitService:   newOpenAIAdvancedSchedulerRateLimitService("true", "true"),
+		concurrencyService: NewConcurrencyService(schedulerTestConcurrencyCache{}),
+	}
+	store := svc.getOpenAIWSStateStore()
+	require.NoError(t, store.BindResponseAccount(ctx, groupID, "resp_weighted_unmovable", 37111, time.Hour))
+
+	selection, decision, err := svc.SelectAccountWithSchedulerForCapability(
+		ctx,
+		&groupID,
+		"resp_weighted_unmovable",
+		"",
+		"gpt-5.1",
+		nil,
+		OpenAIUpstreamTransportAny,
+		OpenAIEndpointCapabilityChatCompletions,
+		false,
+		PlatformOpenAI,
+	)
+	require.NoError(t, err)
+	require.NotNil(t, selection)
+	require.NotNil(t, selection.Account)
+	require.Equal(t, int64(37111), selection.Account.ID)
+	require.Equal(t, openAIAccountScheduleLayerPreviousResponse, decision.Layer)
+	require.True(t, decision.StickyPreviousHit)
+	if selection.ReleaseFunc != nil {
+		selection.ReleaseFunc()
+	}
+
+	selection, decision, err = svc.SelectAccountWithSchedulerForCapability(
+		ctx,
+		&groupID,
+		"resp_weighted_unmovable",
+		"",
+		"gpt-5.1",
+		nil,
+		OpenAIUpstreamTransportAny,
+		OpenAIEndpointCapabilityChatCompletions,
+		false,
+		PlatformOpenAI,
+		"previous_response_can_move",
+	)
+	require.NoError(t, err)
+	require.NotNil(t, selection)
+	require.NotNil(t, selection.Account)
+	require.Equal(t, int64(37112), selection.Account.ID)
+	require.Equal(t, openAIAccountScheduleLayerLoadBalance, decision.Layer)
+	require.False(t, decision.StickyPreviousHit)
+	if selection.ReleaseFunc != nil {
+		selection.ReleaseFunc()
+	}
+}
+
+func TestOpenAIGatewayService_SelectAccountWithScheduler_PreviousResponseCompactUnsupportedDeletesBinding(t *testing.T) {
+	resetOpenAIAdvancedSchedulerSettingCacheForTest()
+
+	ctx := context.Background()
+	groupID := int64(101073)
+	accounts := []Account{
+		{
+			ID:          37121,
+			Platform:    PlatformOpenAI,
+			Type:        AccountTypeAPIKey,
+			Status:      StatusActive,
+			Schedulable: true,
+			Concurrency: 1,
+			Priority:    0,
+			GroupIDs:    []int64{groupID},
+			Extra: map[string]any{
+				"openai_apikey_responses_websockets_v2_enabled": true,
+				"openai_compact_mode":                           OpenAICompactModeForceOff,
+			},
+		},
+		{
+			ID:          37122,
+			Platform:    PlatformOpenAI,
+			Type:        AccountTypeAPIKey,
+			Status:      StatusActive,
+			Schedulable: true,
+			Concurrency: 1,
+			Priority:    10,
+			GroupIDs:    []int64{groupID},
+			Extra: map[string]any{
+				"openai_apikey_responses_websockets_v2_enabled": true,
+				"openai_compact_mode":                           OpenAICompactModeForceOn,
+			},
+		},
+	}
+	cfg := newSchedulerTestOpenAIWSV2Config()
+	cfg.Gateway.OpenAIWS.LBTopK = 2
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.Priority = 1
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.Load = 1
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.Queue = 0.7
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.ErrorRate = 0.8
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.TTFT = 0.5
+	svc := &OpenAIGatewayService{
+		accountRepo:        schedulerTestOpenAIAccountRepo{accounts: accounts},
+		cache:              &schedulerTestGatewayCache{},
+		cfg:                cfg,
+		rateLimitService:   newOpenAIAdvancedSchedulerRateLimitService("true"),
+		concurrencyService: NewConcurrencyService(schedulerTestConcurrencyCache{}),
+	}
+	store := svc.getOpenAIWSStateStore()
+	require.NoError(t, store.BindResponseAccount(ctx, groupID, "resp_compact_unsupported", 37121, time.Hour))
+
+	selection, decision, err := svc.SelectAccountWithScheduler(
+		ctx,
+		&groupID,
+		"resp_compact_unsupported",
+		"",
+		"gpt-5.1",
+		nil,
+		OpenAIUpstreamTransportAny,
+		true,
+	)
+	require.NoError(t, err)
+	require.NotNil(t, selection)
+	require.NotNil(t, selection.Account)
+	require.Equal(t, int64(37122), selection.Account.ID)
+	require.Equal(t, openAIAccountScheduleLayerLoadBalance, decision.Layer)
+	require.False(t, decision.StickyPreviousHit)
+	if selection.ReleaseFunc != nil {
+		selection.ReleaseFunc()
+	}
+
+	accountID, err := store.GetResponseAccount(ctx, groupID, "resp_compact_unsupported")
+	require.NoError(t, err)
+	require.Zero(t, accountID)
+}
+
 func TestOpenAIGatewayService_SelectAccountWithScheduler_Enabled_EmbeddingsSkipsChatOnlyAccount(t *testing.T) {
 	resetOpenAIAdvancedSchedulerSettingCacheForTest()
 
@@ -1560,6 +1863,217 @@ func TestOpenAIGatewayService_SelectAccountWithScheduler_SessionStickyEscapeDisa
 	require.True(t, decision.StickySessionHit)
 }
 
+func TestOpenAIGatewayService_SelectAccountWithScheduler_SubscriptionPriorityChoosesSubscriptionPoolFirst(t *testing.T) {
+	ctx := context.Background()
+	groupID := int64(10120)
+	accounts := []Account{
+		{
+			ID:          21601,
+			Platform:    PlatformOpenAI,
+			Type:        AccountTypeOAuth,
+			Status:      StatusActive,
+			Schedulable: true,
+			Concurrency: 1,
+			Priority:    10,
+			GroupIDs:    []int64{groupID},
+			Credentials: map[string]any{"plan_type": "plus"},
+		},
+		{
+			ID:          21602,
+			Platform:    PlatformOpenAI,
+			Type:        AccountTypeAPIKey,
+			Status:      StatusActive,
+			Schedulable: true,
+			Concurrency: 1,
+			Priority:    0,
+			GroupIDs:    []int64{groupID},
+		},
+	}
+	concurrencyCache := schedulerTestConcurrencyCache{
+		acquireResults: map[int64]bool{21601: true, 21602: true},
+		loadMap: map[int64]*AccountLoadInfo{
+			21601: {AccountID: 21601, LoadRate: 90, WaitingCount: 1},
+			21602: {AccountID: 21602, LoadRate: 0, WaitingCount: 0},
+		},
+	}
+	svc := &OpenAIGatewayService{
+		accountRepo:        schedulerTestOpenAIAccountRepo{accounts: accounts},
+		cache:              &schedulerTestGatewayCache{},
+		cfg:                newSchedulerTestSubscriptionPriorityConfig(),
+		rateLimitService:   newOpenAIAdvancedSchedulerRateLimitService("true", "", "true"),
+		concurrencyService: NewConcurrencyService(concurrencyCache),
+	}
+
+	selection, decision, err := svc.SelectAccountWithScheduler(ctx, &groupID, "", "session_subscription_first", "gpt-5.1", nil, OpenAIUpstreamTransportAny, false)
+	require.NoError(t, err)
+	require.NotNil(t, selection)
+	require.NotNil(t, selection.Account)
+	require.Equal(t, int64(21601), selection.Account.ID)
+	require.Equal(t, openAIAccountScheduleLayerLoadBalance, decision.Layer)
+	require.Equal(t, 1, decision.TopK)
+	if selection.ReleaseFunc != nil {
+		selection.ReleaseFunc()
+	}
+}
+
+func TestOpenAIGatewayService_SelectAccountWithScheduler_SubscriptionPriorityFallsBackWhenSubscriptionFull(t *testing.T) {
+	ctx := context.Background()
+	groupID := int64(10121)
+	accounts := []Account{
+		{
+			ID:          21611,
+			Platform:    PlatformOpenAI,
+			Type:        AccountTypeOAuth,
+			Status:      StatusActive,
+			Schedulable: true,
+			Concurrency: 1,
+			Priority:    0,
+			GroupIDs:    []int64{groupID},
+			Credentials: map[string]any{"plan_type": "team"},
+		},
+		{
+			ID:          21612,
+			Platform:    PlatformOpenAI,
+			Type:        AccountTypeAPIKey,
+			Status:      StatusActive,
+			Schedulable: true,
+			Concurrency: 1,
+			Priority:    9,
+			GroupIDs:    []int64{groupID},
+		},
+	}
+	concurrencyCache := schedulerTestConcurrencyCache{
+		acquireResults: map[int64]bool{21611: false, 21612: true},
+		loadMap: map[int64]*AccountLoadInfo{
+			21611: {AccountID: 21611, LoadRate: 0, WaitingCount: 0},
+			21612: {AccountID: 21612, LoadRate: 90, WaitingCount: 1},
+		},
+	}
+	svc := &OpenAIGatewayService{
+		accountRepo:        schedulerTestOpenAIAccountRepo{accounts: accounts},
+		cache:              &schedulerTestGatewayCache{},
+		cfg:                newSchedulerTestSubscriptionPriorityConfig(),
+		rateLimitService:   newOpenAIAdvancedSchedulerRateLimitService("true", "", "true"),
+		concurrencyService: NewConcurrencyService(concurrencyCache),
+	}
+
+	selection, decision, err := svc.SelectAccountWithScheduler(ctx, &groupID, "", "session_subscription_fallback", "gpt-5.1", nil, OpenAIUpstreamTransportAny, false)
+	require.NoError(t, err)
+	require.NotNil(t, selection)
+	require.NotNil(t, selection.Account)
+	require.Equal(t, int64(21612), selection.Account.ID)
+	require.Equal(t, openAIAccountScheduleLayerLoadBalance, decision.Layer)
+	require.True(t, selection.Acquired)
+	if selection.ReleaseFunc != nil {
+		selection.ReleaseFunc()
+	}
+}
+
+func TestOpenAIGatewayService_SelectAccountWithScheduler_SubscriptionPriorityDisabledUsesScore(t *testing.T) {
+	ctx := context.Background()
+	groupID := int64(10122)
+	accounts := []Account{
+		{
+			ID:          21621,
+			Platform:    PlatformOpenAI,
+			Type:        AccountTypeOAuth,
+			Status:      StatusActive,
+			Schedulable: true,
+			Concurrency: 1,
+			Priority:    10,
+			GroupIDs:    []int64{groupID},
+			Credentials: map[string]any{"plan_type": "pro"},
+		},
+		{
+			ID:          21622,
+			Platform:    PlatformOpenAI,
+			Type:        AccountTypeAPIKey,
+			Status:      StatusActive,
+			Schedulable: true,
+			Concurrency: 1,
+			Priority:    0,
+			GroupIDs:    []int64{groupID},
+		},
+	}
+	concurrencyCache := schedulerTestConcurrencyCache{
+		acquireResults: map[int64]bool{21621: true, 21622: true},
+		loadMap: map[int64]*AccountLoadInfo{
+			21621: {AccountID: 21621, LoadRate: 90, WaitingCount: 1},
+			21622: {AccountID: 21622, LoadRate: 0, WaitingCount: 0},
+		},
+	}
+	svc := &OpenAIGatewayService{
+		accountRepo:        schedulerTestOpenAIAccountRepo{accounts: accounts},
+		cache:              &schedulerTestGatewayCache{},
+		cfg:                newSchedulerTestSubscriptionPriorityConfig(),
+		rateLimitService:   newOpenAIAdvancedSchedulerRateLimitService("true", "", "false"),
+		concurrencyService: NewConcurrencyService(concurrencyCache),
+	}
+
+	selection, decision, err := svc.SelectAccountWithScheduler(ctx, &groupID, "", "session_subscription_disabled", "gpt-5.1", nil, OpenAIUpstreamTransportAny, false)
+	require.NoError(t, err)
+	require.NotNil(t, selection)
+	require.NotNil(t, selection.Account)
+	require.Equal(t, int64(21622), selection.Account.ID)
+	require.Equal(t, openAIAccountScheduleLayerLoadBalance, decision.Layer)
+	if selection.ReleaseFunc != nil {
+		selection.ReleaseFunc()
+	}
+}
+
+func TestOpenAIGatewayService_SelectAccountWithScheduler_UsesAccountPriorityWithinGroupPool(t *testing.T) {
+	ctx := context.Background()
+	groupID := int64(10123)
+	accounts := []Account{
+		{
+			ID:          21631,
+			Platform:    PlatformOpenAI,
+			Type:        AccountTypeAPIKey,
+			Status:      StatusActive,
+			Schedulable: true,
+			Concurrency: 1,
+			Priority:    1,
+			AccountGroups: []AccountGroup{
+				{AccountID: 21631, GroupID: groupID, Priority: 100},
+			},
+			GroupIDs: []int64{groupID},
+		},
+		{
+			ID:          21632,
+			Platform:    PlatformOpenAI,
+			Type:        AccountTypeAPIKey,
+			Status:      StatusActive,
+			Schedulable: true,
+			Concurrency: 1,
+			Priority:    100000,
+			AccountGroups: []AccountGroup{
+				{AccountID: 21632, GroupID: groupID, Priority: 1},
+			},
+			GroupIDs: []int64{groupID},
+		},
+	}
+	cfg := newSchedulerTestSubscriptionPriorityConfig()
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.Load = 0
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights.Queue = 0
+	svc := &OpenAIGatewayService{
+		accountRepo:        schedulerGroupAwareOpenAIAccountRepo{schedulerTestOpenAIAccountRepo{accounts: accounts}},
+		cache:              &schedulerTestGatewayCache{},
+		cfg:                cfg,
+		rateLimitService:   newOpenAIAdvancedSchedulerRateLimitService("true"),
+		concurrencyService: NewConcurrencyService(schedulerTestConcurrencyCache{}),
+	}
+
+	selection, decision, err := svc.SelectAccountWithScheduler(ctx, &groupID, "", "session_group_priority", "gpt-5.1", nil, OpenAIUpstreamTransportAny, false)
+	require.NoError(t, err)
+	require.NotNil(t, selection)
+	require.NotNil(t, selection.Account)
+	require.Equal(t, int64(21631), selection.Account.ID)
+	require.Equal(t, openAIAccountScheduleLayerLoadBalance, decision.Layer)
+	if selection.ReleaseFunc != nil {
+		selection.ReleaseFunc()
+	}
+}
+
 func TestDefaultOpenAIAccountScheduler_ShouldEscapeStickyAccount_ThresholdBoundary(t *testing.T) {
 	stats := newOpenAIAccountRuntimeStats()
 	accountID := int64(21501)
diff --git a/backend/internal/service/openai_ws_forwarder.go b/backend/internal/service/openai_ws_forwarder.go
index 9261db5544..fcc4b98064 100644
--- a/backend/internal/service/openai_ws_forwarder.go
+++ b/backend/internal/service/openai_ws_forwarder.go
@@ -4325,87 +4325,8 @@ func (s *OpenAIGatewayService) selectAccountByPreviousResponseIDForCapability(
 	if s == nil {
 		return nil, nil
 	}
-	responseID := strings.TrimSpace(previousResponseID)
-	if responseID == "" {
-		return nil, nil
-	}
-	store := s.getOpenAIWSStateStore()
-	if store == nil {
-		return nil, nil
-	}
-
-	accountID, err := store.GetResponseAccount(ctx, derefGroupID(groupID), responseID)
-	if err != nil || accountID <= 0 {
-		return nil, nil
-	}
-	if excludedIDs != nil {
-		if _, excluded := excludedIDs[accountID]; excluded {
-			return nil, nil
-		}
-	}
-
-	account, err := s.getSchedulableAccount(ctx, accountID)
-	if err != nil || account == nil {
-		_ = store.DeleteResponseAccount(ctx, derefGroupID(groupID), responseID)
-		return nil, nil
-	}
-	// 非 WSv2 场景（如 force_http/全局关闭）不应使用 previous_response_id 粘连，
-	// 以保持“回滚到 HTTP”后的历史行为一致性。
-	if s.getOpenAIWSProtocolResolver().Resolve(account).Transport != OpenAIUpstreamTransportResponsesWebsocketV2 {
-		return nil, nil
-	}
-	if shouldClearStickySession(account, requestedModel) || !account.IsOpenAI() || !account.IsSchedulable() {
-		_ = store.DeleteResponseAccount(ctx, derefGroupID(groupID), responseID)
-		return nil, nil
-	}
-	if !parentHealthyForShadow(account, s.parentAccountLookup(ctx)) {
-		_ = store.DeleteResponseAccount(ctx, derefGroupID(groupID), responseID)
-		return nil, nil
-	}
-	if requestedModel != "" && !account.IsModelSupported(requestedModel) {
-		return nil, nil
-	}
-	if !account.SupportsOpenAIEndpointCapability(requiredCapability) {
-		return nil, nil
-	}
-	// Quota auto-pause must also gate the previous_response_id sticky path; otherwise an
-	// account over its 5h/7d threshold keeps serving the same response chain even though
-	// normal scheduling skips it. Pause is transient, so fall through to normal scheduling
-	// without deleting the binding (the window may reset before the next turn).
-	if paused, _ := shouldAutoPauseOpenAIAccountByQuota(ctx, account); paused {
-		return nil, nil
-	}
-	if s.schedulerSnapshot != nil && s.accountRepo != nil {
-		latest, latestErr := s.accountRepo.GetByID(ctx, account.ID)
-		if latestErr != nil || latest == nil {
-			_ = store.DeleteResponseAccount(ctx, derefGroupID(groupID), responseID)
-			return nil, nil
-		}
-		if shouldClearStickySession(latest, requestedModel) || !latest.IsOpenAI() || !latest.IsSchedulable() {
-			_ = store.DeleteResponseAccount(ctx, derefGroupID(groupID), responseID)
-			return nil, nil
-		}
-		if !parentHealthyForShadow(latest, s.parentAccountLookup(ctx)) {
-			_ = store.DeleteResponseAccount(ctx, derefGroupID(groupID), responseID)
-			return nil, nil
-		}
-		if requestedModel != "" && !latest.IsModelSupported(requestedModel) {
-			return nil, nil
-		}
-		if !latest.SupportsOpenAIEndpointCapability(requiredCapability) {
-			return nil, nil
-		}
-		if paused, _ := shouldAutoPauseOpenAIAccountByQuota(ctx, latest); paused {
-			return nil, nil
-		}
-		if s.isOpenAIAccountRuntimeBlocked(latest) {
-			_ = store.DeleteResponseAccount(ctx, derefGroupID(groupID), responseID)
-			return nil, nil
-		}
-		account = latest
-	}
-	if requireCompact && openAICompactSupportTier(account) == 0 {
-		_ = store.DeleteResponseAccount(ctx, derefGroupID(groupID), responseID)
+	accountID, account, responseID, store := s.resolveAccountByPreviousResponseIDForCapability(ctx, groupID, previousResponseID, requestedModel, excludedIDs, requiredCapability, requireCompact)
+	if accountID <= 0 || account == nil || store == nil {
 		return nil, nil
 	}
 
@@ -4439,6 +4360,117 @@ func (s *OpenAIGatewayService) selectAccountByPreviousResponseIDForCapability(
 	return nil, nil
 }
 
+func (s *OpenAIGatewayService) ResolveAccountIDByPreviousResponseIDForScheduler(
+	ctx context.Context,
+	groupID *int64,
+	previousResponseID string,
+	requestedModel string,
+	excludedIDs map[int64]struct{},
+	requiredCapability OpenAIEndpointCapability,
+	requireCompact bool,
+) int64 {
+	accountID, _, _, _ := s.resolveAccountByPreviousResponseIDForCapability(ctx, groupID, previousResponseID, requestedModel, excludedIDs, requiredCapability, requireCompact)
+	return accountID
+}
+
+func (s *OpenAIGatewayService) resolveAccountByPreviousResponseIDForCapability(
+	ctx context.Context,
+	groupID *int64,
+	previousResponseID string,
+	requestedModel string,
+	excludedIDs map[int64]struct{},
+	requiredCapability OpenAIEndpointCapability,
+	requireCompact bool,
+) (int64, *Account, string, OpenAIWSStateStore) {
+	if s == nil {
+		return 0, nil, "", nil
+	}
+	responseID := strings.TrimSpace(previousResponseID)
+	if responseID == "" {
+		return 0, nil, "", nil
+	}
+	store := s.getOpenAIWSStateStore()
+	if store == nil {
+		return 0, nil, "", nil
+	}
+
+	accountID, err := store.GetResponseAccount(ctx, derefGroupID(groupID), responseID)
+	if err != nil || accountID <= 0 {
+		return 0, nil, "", nil
+	}
+	if excludedIDs != nil {
+		if _, excluded := excludedIDs[accountID]; excluded {
+			return 0, nil, "", nil
+		}
+	}
+
+	account, err := s.getSchedulableAccount(ctx, accountID)
+	if err != nil || account == nil {
+		_ = store.DeleteResponseAccount(ctx, derefGroupID(groupID), responseID)
+		return 0, nil, "", nil
+	}
+	// 非 WSv2 场景（如 force_http/全局关闭）不应使用 previous_response_id 粘连，
+	// 以保持“回滚到 HTTP”后的历史行为一致性。
+	if s.getOpenAIWSProtocolResolver().Resolve(account).Transport != OpenAIUpstreamTransportResponsesWebsocketV2 {
+		return 0, nil, "", nil
+	}
+	if shouldClearStickySession(account, requestedModel) || !account.IsOpenAI() || !account.IsSchedulable() {
+		_ = store.DeleteResponseAccount(ctx, derefGroupID(groupID), responseID)
+		return 0, nil, "", nil
+	}
+	if !parentHealthyForShadow(account, s.parentAccountLookup(ctx)) {
+		_ = store.DeleteResponseAccount(ctx, derefGroupID(groupID), responseID)
+		return 0, nil, "", nil
+	}
+	if requestedModel != "" && !account.IsModelSupported(requestedModel) {
+		return 0, nil, "", nil
+	}
+	if !account.SupportsOpenAIEndpointCapability(requiredCapability) {
+		return 0, nil, "", nil
+	}
+	// Quota auto-pause must also gate the previous_response_id sticky path; otherwise an
+	// account over its 5h/7d threshold keeps serving the same response chain even though
+	// normal scheduling skips it. Pause is transient, so fall through to normal scheduling
+	// without deleting the binding (the window may reset before the next turn).
+	if paused, _ := shouldAutoPauseOpenAIAccountByQuota(ctx, account); paused {
+		return 0, nil, "", nil
+	}
+	if s.schedulerSnapshot != nil && s.accountRepo != nil {
+		latest, latestErr := s.accountRepo.GetByID(ctx, account.ID)
+		if latestErr != nil || latest == nil {
+			_ = store.DeleteResponseAccount(ctx, derefGroupID(groupID), responseID)
+			return 0, nil, "", nil
+		}
+		if shouldClearStickySession(latest, requestedModel) || !latest.IsOpenAI() || !latest.IsSchedulable() {
+			_ = store.DeleteResponseAccount(ctx, derefGroupID(groupID), responseID)
+			return 0, nil, "", nil
+		}
+		if !parentHealthyForShadow(latest, s.parentAccountLookup(ctx)) {
+			_ = store.DeleteResponseAccount(ctx, derefGroupID(groupID), responseID)
+			return 0, nil, "", nil
+		}
+		if requestedModel != "" && !latest.IsModelSupported(requestedModel) {
+			return 0, nil, "", nil
+		}
+		if !latest.SupportsOpenAIEndpointCapability(requiredCapability) {
+			return 0, nil, "", nil
+		}
+		if paused, _ := shouldAutoPauseOpenAIAccountByQuota(ctx, latest); paused {
+			return 0, nil, "", nil
+		}
+		if s.isOpenAIAccountRuntimeBlocked(latest) {
+			_ = store.DeleteResponseAccount(ctx, derefGroupID(groupID), responseID)
+			return 0, nil, "", nil
+		}
+		account = latest
+	}
+	if requireCompact && openAICompactSupportTier(account) == 0 {
+		_ = store.DeleteResponseAccount(ctx, derefGroupID(groupID), responseID)
+		return 0, nil, "", nil
+	}
+	return accountID, account, responseID, store
+}
+
 func classifyOpenAIWSAcquireError(err error) string {
 	if err == nil {
 		return "acquire_conn"
diff --git a/backend/internal/service/ratelimit_service.go b/backend/internal/service/ratelimit_service.go
index 2b9c85a2fa..c84192368f 100644
--- a/backend/internal/service/ratelimit_service.go
+++ b/backend/internal/service/ratelimit_service.go
@@ -116,6 +116,14 @@ func (s *RateLimitService) SetAccountRuntimeBlocker(blocker AccountRuntimeBlocke
 	s.runtimeBlocker = blocker
 }
 
+func (s *RateLimitService) IsOpenAIAdvancedSchedulerStickyWeightedEnabled(ctx context.Context) bool {
+	if s == nil || s.settingService == nil {
+		return false
+	}
+	gateway := &OpenAIGatewayService{rateLimitService: s}
+	return gateway.isOpenAIAdvancedSchedulerStickyWeightedEnabled(ctx)
+}
+
 func (s *RateLimitService) notifyAccountSchedulingBlocked(account *Account, until time.Time, reason string) {
 	if s == nil || s.runtimeBlocker == nil || account == nil {
 		return
diff --git a/backend/internal/service/setting_service.go b/backend/internal/service/setting_service.go
index 1a5e677a12..1024243bea 100644
--- a/backend/internal/service/setting_service.go
+++ b/backend/internal/service/setting_service.go
@@ -1921,6 +1921,9 @@ func (s *SettingService) buildSystemSettingsUpdates(ctx context.Context, setting
 	if err != nil {
 		return nil, err
 	}
+	if err := normalizeOpenAIAdvancedSchedulerOverrides(settings); err != nil {
+		return nil, err
+	}
 	settings.PaymentVisibleMethodAlipaySource = alipaySource
 	settings.PaymentVisibleMethodWxpaySource = wxpaySource
 	settings.WeChatConnectAppID = strings.TrimSpace(settings.WeChatConnectAppID)
@@ -2229,6 +2232,18 @@ func (s *SettingService) buildSystemSettingsUpdates(ctx context.Context, setting
 	updates[SettingPaymentVisibleMethodAlipayEnabled] = strconv.FormatBool(settings.PaymentVisibleMethodAlipayEnabled)
 	updates[SettingPaymentVisibleMethodWxpayEnabled] = strconv.FormatBool(settings.PaymentVisibleMethodWxpayEnabled)
 	updates[openAIAdvancedSchedulerSettingKey] = strconv.FormatBool(settings.OpenAIAdvancedSchedulerEnabled)
+	updates[SettingKeyOpenAIAdvancedSchedulerStickyWeightedEnabled] = strconv.FormatBool(settings.OpenAIAdvancedSchedulerStickyWeightedEnabled)
+	updates[SettingKeyOpenAIAdvancedSchedulerSubscriptionPriorityEnabled] = strconv.FormatBool(settings.OpenAIAdvancedSchedulerSubscriptionPriorityEnabled)
+	updates[SettingKeyOpenAIAdvancedSchedulerLBTopK] = settings.OpenAIAdvancedSchedulerLBTopK
+	updates[SettingKeyOpenAIAdvancedSchedulerWeightPriority] = settings.OpenAIAdvancedSchedulerWeightPriority
+	updates[SettingKeyOpenAIAdvancedSchedulerWeightLoad] = settings.OpenAIAdvancedSchedulerWeightLoad
+	updates[SettingKeyOpenAIAdvancedSchedulerWeightQueue] = settings.OpenAIAdvancedSchedulerWeightQueue
+	updates[SettingKeyOpenAIAdvancedSchedulerWeightErrorRate] = settings.OpenAIAdvancedSchedulerWeightErrorRate
+	updates[SettingKeyOpenAIAdvancedSchedulerWeightTTFT] = settings.OpenAIAdvancedSchedulerWeightTTFT
+	updates[SettingKeyOpenAIAdvancedSchedulerWeightReset] = settings.OpenAIAdvancedSchedulerWeightReset
+	updates[SettingKeyOpenAIAdvancedSchedulerWeightQuotaHeadroom] = settings.OpenAIAdvancedSchedulerWeightQuotaHeadroom
+	updates[SettingKeyOpenAIAdvancedSchedulerWeightPreviousResponse] = settings.OpenAIAdvancedSchedulerWeightPreviousResponse
+	updates[SettingKeyOpenAIAdvancedSchedulerWeightSessionSticky] = settings.OpenAIAdvancedSchedulerWeightSessionSticky
 
 	// 余额、订阅到期与账号限额通知
 	updates[SettingKeyBalanceLowNotifyEnabled] = strconv.FormatBool(settings.BalanceLowNotifyEnabled)
@@ -2376,7 +2391,21 @@ func (s *SettingService) refreshCachedSettings(settings *SystemSettings) {
 	})
 	openAIAdvancedSchedulerSettingSF.Forget(openAIAdvancedSchedulerSettingKey)
 	openAIAdvancedSchedulerSettingCache.Store(&cachedOpenAIAdvancedSchedulerSetting{
-		enabled:   settings.OpenAIAdvancedSchedulerEnabled,
+		enabled:                     settings.OpenAIAdvancedSchedulerEnabled,
+		stickyWeightedEnabled:       settings.OpenAIAdvancedSchedulerStickyWeightedEnabled,
+		subscriptionPriorityEnabled: settings.OpenAIAdvancedSchedulerSubscriptionPriorityEnabled,
+		lbTopKOverride:              parsePositiveIntOverride(settings.OpenAIAdvancedSchedulerLBTopK),
+		weightOverrides: parseOpenAIAdvancedSchedulerWeightOverrides(map[string]string{
+			SettingKeyOpenAIAdvancedSchedulerWeightPriority:         settings.OpenAIAdvancedSchedulerWeightPriority,
+			SettingKeyOpenAIAdvancedSchedulerWeightLoad:             settings.OpenAIAdvancedSchedulerWeightLoad,
+			SettingKeyOpenAIAdvancedSchedulerWeightQueue:            settings.OpenAIAdvancedSchedulerWeightQueue,
+			SettingKeyOpenAIAdvancedSchedulerWeightErrorRate:        settings.OpenAIAdvancedSchedulerWeightErrorRate,
+			SettingKeyOpenAIAdvancedSchedulerWeightTTFT:             settings.OpenAIAdvancedSchedulerWeightTTFT,
+			SettingKeyOpenAIAdvancedSchedulerWeightReset:            settings.OpenAIAdvancedSchedulerWeightReset,
+			SettingKeyOpenAIAdvancedSchedulerWeightQuotaHeadroom:    settings.OpenAIAdvancedSchedulerWeightQuotaHeadroom,
+			SettingKeyOpenAIAdvancedSchedulerWeightPreviousResponse: settings.OpenAIAdvancedSchedulerWeightPreviousResponse,
+			SettingKeyOpenAIAdvancedSchedulerWeightSessionSticky:    settings.OpenAIAdvancedSchedulerWeightSessionSticky,
+		}),
 		expiresAt: time.Now().Add(openAIAdvancedSchedulerSettingCacheTTL).UnixNano(),
 	})
 	// Invalidate the quota auto-pause cache and let the next read trigger a fresh load.
@@ -3196,17 +3225,29 @@ func (s *SettingService) InitializeDefaultSettings(ctx context.Context) error {
 		SettingKeyCodexCLIOnlyEngineFingerprintSignals: openai.DefaultEngineFingerprintSignalsJSON(),
 
 		// 分组隔离（默认不允许未分组 Key 调度）
-		SettingKeyAllowUngroupedKeyScheduling:        "false",
-		SettingKeyEnableAnthropicCacheTTL1hInjection: "false",
-		SettingKeyRewriteMessageCacheControl:         strconv.FormatBool(s.defaultRewriteMessageCacheControl()),
-		SettingKeyEnableClientDatelineNormalization:  "true",
-		SettingKeyAntigravityUserAgentVersion:        "",
-		SettingKeyOpenAICodexUserAgent:               "",
-		SettingPaymentVisibleMethodAlipaySource:      "",
-		SettingPaymentVisibleMethodWxpaySource:       "",
-		SettingPaymentVisibleMethodAlipayEnabled:     "false",
-		SettingPaymentVisibleMethodWxpayEnabled:      "false",
-		openAIAdvancedSchedulerSettingKey:            "false",
+		SettingKeyAllowUngroupedKeyScheduling:                        "false",
+		SettingKeyEnableAnthropicCacheTTL1hInjection:                 "false",
+		SettingKeyRewriteMessageCacheControl:                         strconv.FormatBool(s.defaultRewriteMessageCacheControl()),
+		SettingKeyEnableClientDatelineNormalization:                  "true",
+		SettingKeyAntigravityUserAgentVersion:                        "",
+		SettingKeyOpenAICodexUserAgent:                               "",
+		SettingPaymentVisibleMethodAlipaySource:                      "",
+		SettingPaymentVisibleMethodWxpaySource:                       "",
+		SettingPaymentVisibleMethodAlipayEnabled:                     "false",
+		SettingPaymentVisibleMethodWxpayEnabled:                      "false",
+		openAIAdvancedSchedulerSettingKey:                            "false",
+		SettingKeyOpenAIAdvancedSchedulerStickyWeightedEnabled:       "false",
+		SettingKeyOpenAIAdvancedSchedulerSubscriptionPriorityEnabled: "false",
+		SettingKeyOpenAIAdvancedSchedulerLBTopK:                      "",
+		SettingKeyOpenAIAdvancedSchedulerWeightPriority:              "",
+		SettingKeyOpenAIAdvancedSchedulerWeightLoad:                  "",
+		SettingKeyOpenAIAdvancedSchedulerWeightQueue:                 "",
+		SettingKeyOpenAIAdvancedSchedulerWeightErrorRate:             "",
+		SettingKeyOpenAIAdvancedSchedulerWeightTTFT:                  "",
+		SettingKeyOpenAIAdvancedSchedulerWeightReset:                 "",
+		SettingKeyOpenAIAdvancedSchedulerWeightQuotaHeadroom:         "",
+		SettingKeyOpenAIAdvancedSchedulerWeightPreviousResponse:      "",
+		SettingKeyOpenAIAdvancedSchedulerWeightSessionSticky:         "",
 
 		SettingKeyAllowUserViewErrorRequests: "false",
 	}
@@ -3769,6 +3810,29 @@ func (s *SettingService) parseSettings(settings map[string]string) *SystemSettin
 	result.PaymentVisibleMethodAlipayEnabled = settings[SettingPaymentVisibleMethodAlipayEnabled] == "true"
 	result.PaymentVisibleMethodWxpayEnabled = settings[SettingPaymentVisibleMethodWxpayEnabled] == "true"
 	result.OpenAIAdvancedSchedulerEnabled = settings[openAIAdvancedSchedulerSettingKey] == "true"
+	result.OpenAIAdvancedSchedulerStickyWeightedEnabled = settings[SettingKeyOpenAIAdvancedSchedulerStickyWeightedEnabled] == "true"
+	result.OpenAIAdvancedSchedulerSubscriptionPriorityEnabled = settings[SettingKeyOpenAIAdvancedSchedulerSubscriptionPriorityEnabled] == "true"
+	result.OpenAIAdvancedSchedulerLBTopK = strings.TrimSpace(settings[SettingKeyOpenAIAdvancedSchedulerLBTopK])
+	result.OpenAIAdvancedSchedulerWeightPriority = strings.TrimSpace(settings[SettingKeyOpenAIAdvancedSchedulerWeightPriority])
+	result.OpenAIAdvancedSchedulerWeightLoad = strings.TrimSpace(settings[SettingKeyOpenAIAdvancedSchedulerWeightLoad])
+	result.OpenAIAdvancedSchedulerWeightQueue = strings.TrimSpace(settings[SettingKeyOpenAIAdvancedSchedulerWeightQueue])
+	result.OpenAIAdvancedSchedulerWeightErrorRate = strings.TrimSpace(settings[SettingKeyOpenAIAdvancedSchedulerWeightErrorRate])
+	result.OpenAIAdvancedSchedulerWeightTTFT = strings.TrimSpace(settings[SettingKeyOpenAIAdvancedSchedulerWeightTTFT])
+	result.OpenAIAdvancedSchedulerWeightReset = strings.TrimSpace(settings[SettingKeyOpenAIAdvancedSchedulerWeightReset])
+	result.OpenAIAdvancedSchedulerWeightQuotaHeadroom = strings.TrimSpace(settings[SettingKeyOpenAIAdvancedSchedulerWeightQuotaHeadroom])
+	result.OpenAIAdvancedSchedulerWeightPreviousResponse = strings.TrimSpace(settings[SettingKeyOpenAIAdvancedSchedulerWeightPreviousResponse])
+	result.OpenAIAdvancedSchedulerWeightSessionSticky = strings.TrimSpace(settings[SettingKeyOpenAIAdvancedSchedulerWeightSessionSticky])
+	result.OpenAIAdvancedSchedulerEffectiveLBTopK = s.openAIAdvancedSchedulerEffectiveLBTopK()
+	effectiveWeights := s.openAIAdvancedSchedulerEffectiveWeights()
+	result.OpenAIAdvancedSchedulerEffectiveWeightPriority = formatOpenAIAdvancedSchedulerFloat(effectiveWeights.Priority)
+	result.OpenAIAdvancedSchedulerEffectiveWeightLoad = formatOpenAIAdvancedSchedulerFloat(effectiveWeights.Load)
+	result.OpenAIAdvancedSchedulerEffectiveWeightQueue = formatOpenAIAdvancedSchedulerFloat(effectiveWeights.Queue)
+	result.OpenAIAdvancedSchedulerEffectiveWeightErrorRate = formatOpenAIAdvancedSchedulerFloat(effectiveWeights.ErrorRate)
+	result.OpenAIAdvancedSchedulerEffectiveWeightTTFT = formatOpenAIAdvancedSchedulerFloat(effectiveWeights.TTFT)
+	result.OpenAIAdvancedSchedulerEffectiveWeightReset = formatOpenAIAdvancedSchedulerFloat(effectiveWeights.Reset)
+	result.OpenAIAdvancedSchedulerEffectiveWeightQuotaHeadroom = formatOpenAIAdvancedSchedulerFloat(effectiveWeights.QuotaHeadroom)
+	result.OpenAIAdvancedSchedulerEffectiveWeightPreviousResponse = formatOpenAIAdvancedSchedulerFloat(effectiveWeights.PreviousResponse)
+	result.OpenAIAdvancedSchedulerEffectiveWeightSessionSticky = formatOpenAIAdvancedSchedulerFloat(effectiveWeights.SessionSticky)
 
 	// 余额、订阅到期与账号限额通知
 	result.BalanceLowNotifyEnabled = settings[SettingKeyBalanceLowNotifyEnabled] == "true"
@@ -3841,6 +3905,93 @@ func normalizeVisibleMethodSettingSource(method, source string, enabled bool) (s
 	return normalized, nil
 }
 
+func (s *SettingService) openAIAdvancedSchedulerEffectiveLBTopK() string {
+	if s != nil && s.cfg != nil && s.cfg.Gateway.OpenAIWS.LBTopK > 0 {
+		return strconv.Itoa(s.cfg.Gateway.OpenAIWS.LBTopK)
+	}
+	return "7"
+}
+
+func (s *SettingService) openAIAdvancedSchedulerEffectiveWeights() config.GatewayOpenAIWSSchedulerScoreWeights {
+	defaults := config.GatewayOpenAIWSSchedulerScoreWeights{
+		Priority:         1.0,
+		Load:             1.0,
+		Queue:            0.7,
+		ErrorRate:        0.8,
+		TTFT:             0.5,
+		Reset:            0.0,
+		QuotaHeadroom:    0.0,
+		PreviousResponse: 5.0,
+		SessionSticky:    3.0,
+	}
+	if s == nil || s.cfg == nil {
+		return defaults
+	}
+
+	weights := s.cfg.Gateway.OpenAIWS.SchedulerScoreWeights
+	baseSum := weights.Priority + weights.Load + weights.Queue + weights.ErrorRate + weights.TTFT + weights.QuotaHeadroom
+	if baseSum <= 0 {
+		return defaults
+	}
+	return weights
+}
+
+func formatOpenAIAdvancedSchedulerFloat(value float64) string {
+	return strconv.FormatFloat(value, 'f', -1, 64)
+}
+
+func normalizeOpenAIAdvancedSchedulerOverrides(settings *SystemSettings) error {
+	lbTopK, err := normalizeOptionalPositiveIntString(settings.OpenAIAdvancedSchedulerLBTopK)
+	if err != nil {
+		return infraerrors.BadRequest("INVALID_OPENAI_ADVANCED_SCHEDULER_LB_TOP_K", "openai advanced scheduler TopK must be a positive integer or empty")
+	}
+	settings.OpenAIAdvancedSchedulerLBTopK = lbTopK
+
+	weights := []*string{
+		&settings.OpenAIAdvancedSchedulerWeightPriority,
+		&settings.OpenAIAdvancedSchedulerWeightLoad,
+		&settings.OpenAIAdvancedSchedulerWeightQueue,
+		&settings.OpenAIAdvancedSchedulerWeightErrorRate,
+		&settings.OpenAIAdvancedSchedulerWeightTTFT,
+		&settings.OpenAIAdvancedSchedulerWeightReset,
+		&settings.OpenAIAdvancedSchedulerWeightQuotaHeadroom,
+		&settings.OpenAIAdvancedSchedulerWeightPreviousResponse,
+		&settings.OpenAIAdvancedSchedulerWeightSessionSticky,
+	}
+	for _, target := range weights {
+		normalized, err := normalizeOptionalNonNegativeFloatString(*target)
+		if err != nil {
+			return infraerrors.BadRequest("INVALID_OPENAI_ADVANCED_SCHEDULER_WEIGHT", "openai advanced scheduler weights must be non-negative numbers or empty")
+		}
+		*target = normalized
+	}
+	return nil
+}
+
+func normalizeOptionalPositiveIntString(raw string) (string, error) {
+	raw = strings.TrimSpace(raw)
+	if raw == "" {
+		return "", nil
+	}
+	value, err := strconv.Atoi(raw)
+	if err != nil || value <= 0 {
+		return "", fmt.Errorf("invalid positive integer")
+	}
+	return strconv.Itoa(value), nil
+}
+
+func normalizeOptionalNonNegativeFloatString(raw string) (string, error) {
+	raw = strings.TrimSpace(raw)
+	if raw == "" {
+		return "", nil
+	}
+	value, err := strconv.ParseFloat(raw, 64)
+	if err != nil || value < 0 || math.IsNaN(value) || math.IsInf(value, 0) {
+		return "", fmt.Errorf("invalid non-negative float")
+	}
+	return strconv.FormatFloat(value, 'f', -1, 64), nil
+}
+
 func parseDefaultSubscriptions(raw string) []DefaultSubscriptionSetting {
 	raw = strings.TrimSpace(raw)
 	if raw == "" {
diff --git a/backend/internal/service/setting_service_update_test.go b/backend/internal/service/setting_service_update_test.go
index 379bf9bc04..a47e1379f7 100644
--- a/backend/internal/service/setting_service_update_test.go
+++ b/backend/internal/service/setting_service_update_test.go
@@ -49,6 +49,42 @@ func (s *settingUpdateRepoStub) Delete(ctx context.Context, key string) error {
 	panic("unexpected Delete call")
 }
 
+type settingGetAllRepoStub struct {
+	values map[string]string
+}
+
+func (s *settingGetAllRepoStub) Get(ctx context.Context, key string) (*Setting, error) {
+	panic("unexpected Get call")
+}
+
+func (s *settingGetAllRepoStub) GetValue(ctx context.Context, key string) (string, error) {
+	panic("unexpected GetValue call")
+}
+
+func (s *settingGetAllRepoStub) Set(ctx context.Context, key, value string) error {
+	panic("unexpected Set call")
+}
+
+func (s *settingGetAllRepoStub) GetMultiple(ctx context.Context, keys []string) (map[string]string, error) {
+	panic("unexpected GetMultiple call")
+}
+
+func (s *settingGetAllRepoStub) SetMultiple(ctx context.Context, settings map[string]string) error {
+	panic("unexpected SetMultiple call")
+}
+
+func (s *settingGetAllRepoStub) GetAll(ctx context.Context) (map[string]string, error) {
+	out := make(map[string]string, len(s.values))
+	for key, value := range s.values {
+		out[key] = value
+	}
+	return out, nil
+}
+
+func (s *settingGetAllRepoStub) Delete(ctx context.Context, key string) error {
+	panic("unexpected Delete call")
+}
+
 type settingAntigravityUARepoStub struct {
 	values map[string]string
 }
@@ -261,15 +297,30 @@ func TestSettingService_UpdateSettings_TablePreferences(t *testing.T) {
 }
 
 func TestSettingService_UpdateSettings_PaymentVisibleMethodsAndAdvancedScheduler(t *testing.T) {
+	resetOpenAIAdvancedSchedulerSettingCacheForTest()
+	defer resetOpenAIAdvancedSchedulerSettingCacheForTest()
+
 	repo := &settingUpdateRepoStub{}
 	svc := NewSettingService(repo, &config.Config{})
 
 	err := svc.UpdateSettings(context.Background(), &SystemSettings{
-		PaymentVisibleMethodAlipaySource:  "alipay",
-		PaymentVisibleMethodWxpaySource:   "easypay",
-		PaymentVisibleMethodAlipayEnabled: true,
-		PaymentVisibleMethodWxpayEnabled:  false,
-		OpenAIAdvancedSchedulerEnabled:    true,
+		PaymentVisibleMethodAlipaySource:                   "alipay",
+		PaymentVisibleMethodWxpaySource:                    "easypay",
+		PaymentVisibleMethodAlipayEnabled:                  true,
+		PaymentVisibleMethodWxpayEnabled:                   false,
+		OpenAIAdvancedSchedulerEnabled:                     true,
+		OpenAIAdvancedSchedulerStickyWeightedEnabled:       true,
+		OpenAIAdvancedSchedulerSubscriptionPriorityEnabled: true,
+		OpenAIAdvancedSchedulerLBTopK:                      " 3 ",
+		OpenAIAdvancedSchedulerWeightPriority:              "2.50",
+		OpenAIAdvancedSchedulerWeightLoad:                  "0",
+		OpenAIAdvancedSchedulerWeightQueue:                 "0.75",
+		OpenAIAdvancedSchedulerWeightErrorRate:             "1.25",
+		OpenAIAdvancedSchedulerWeightTTFT:                  "0.5",
+		OpenAIAdvancedSchedulerWeightReset:                 "",
+		OpenAIAdvancedSchedulerWeightQuotaHeadroom:         "0.2",
+		OpenAIAdvancedSchedulerWeightPreviousResponse:      "8",
+		OpenAIAdvancedSchedulerWeightSessionSticky:         "4",
 	})
 	require.NoError(t, err)
 	require.Equal(t, VisibleMethodSourceOfficialAlipay, repo.updates[SettingPaymentVisibleMethodAlipaySource])
@@ -277,6 +328,49 @@ func TestSettingService_UpdateSettings_PaymentVisibleMethodsAndAdvancedScheduler
 	require.Equal(t, "true", repo.updates[SettingPaymentVisibleMethodAlipayEnabled])
 	require.Equal(t, "false", repo.updates[SettingPaymentVisibleMethodWxpayEnabled])
 	require.Equal(t, "true", repo.updates[openAIAdvancedSchedulerSettingKey])
+	require.Equal(t, "true", repo.updates[SettingKeyOpenAIAdvancedSchedulerStickyWeightedEnabled])
+	require.Equal(t, "true", repo.updates[SettingKeyOpenAIAdvancedSchedulerSubscriptionPriorityEnabled])
+	require.Equal(t, "3", repo.updates[SettingKeyOpenAIAdvancedSchedulerLBTopK])
+	require.Equal(t, "2.5", repo.updates[SettingKeyOpenAIAdvancedSchedulerWeightPriority])
+	require.Equal(t, "0", repo.updates[SettingKeyOpenAIAdvancedSchedulerWeightLoad])
+	require.Equal(t, "0.75", repo.updates[SettingKeyOpenAIAdvancedSchedulerWeightQueue])
+	require.Equal(t, "1.25", repo.updates[SettingKeyOpenAIAdvancedSchedulerWeightErrorRate])
+	require.Equal(t, "0.5", repo.updates[SettingKeyOpenAIAdvancedSchedulerWeightTTFT])
+	require.Equal(t, "", repo.updates[SettingKeyOpenAIAdvancedSchedulerWeightReset])
+	require.Equal(t, "0.2", repo.updates[SettingKeyOpenAIAdvancedSchedulerWeightQuotaHeadroom])
+	require.Equal(t, "8", repo.updates[SettingKeyOpenAIAdvancedSchedulerWeightPreviousResponse])
+	require.Equal(t, "4", repo.updates[SettingKeyOpenAIAdvancedSchedulerWeightSessionSticky])
+}
+
+func TestSettingService_GetAllSettings_OpenAIAdvancedSchedulerEffectiveValuesUseConfig(t *testing.T) {
+	cfg := &config.Config{}
+	cfg.Gateway.OpenAIWS.LBTopK = 13
+	cfg.Gateway.OpenAIWS.SchedulerScoreWeights = config.GatewayOpenAIWSSchedulerScoreWeights{
+		Priority:         2,
+		Load:             3,
+		Queue:            4,
+		ErrorRate:        5,
+		TTFT:             6,
+		Reset:            7,
+		QuotaHeadroom:    8,
+		PreviousResponse: 9,
+		SessionSticky:    10,
+	}
+	svc := NewSettingService(&settingGetAllRepoStub{values: map[string]string{
+		SettingKeyOpenAIAdvancedSchedulerLBTopK:              "3",
+		SettingKeyOpenAIAdvancedSchedulerWeightPriority:      "99",
+		SettingKeyOpenAIAdvancedSchedulerWeightSessionSticky: "88",
+	}}, cfg)
+
+	settings, err := svc.GetAllSettings(context.Background())
+	require.NoError(t, err)
+	require.Equal(t, "3", settings.OpenAIAdvancedSchedulerLBTopK)
+	require.Equal(t, "99", settings.OpenAIAdvancedSchedulerWeightPriority)
+	require.Equal(t, "88", settings.OpenAIAdvancedSchedulerWeightSessionSticky)
+	require.Equal(t, "13", settings.OpenAIAdvancedSchedulerEffectiveLBTopK)
+	require.Equal(t, "2", settings.OpenAIAdvancedSchedulerEffectiveWeightPriority)
+	require.Equal(t, "3", settings.OpenAIAdvancedSchedulerEffectiveWeightLoad)
+	require.Equal(t, "10", settings.OpenAIAdvancedSchedulerEffectiveWeightSessionSticky)
 }
 
 func TestSettingService_UpdateSettings_AntigravityUserAgentVersion(t *testing.T) {
diff --git a/backend/internal/service/settings_view.go b/backend/internal/service/settings_view.go
index ac225e1b14..465760a445 100644
--- a/backend/internal/service/settings_view.go
+++ b/backend/internal/service/settings_view.go
@@ -218,7 +218,29 @@ type SystemSettings struct {
 	PaymentVisibleMethodWxpayEnabled  bool
 
 	// OpenAI 账号调度
-	OpenAIAdvancedSchedulerEnabled bool
+	OpenAIAdvancedSchedulerEnabled                         bool
+	OpenAIAdvancedSchedulerStickyWeightedEnabled           bool
+	OpenAIAdvancedSchedulerSubscriptionPriorityEnabled     bool
+	OpenAIAdvancedSchedulerLBTopK                          string
+	OpenAIAdvancedSchedulerWeightPriority                  string
+	OpenAIAdvancedSchedulerWeightLoad                      string
+	OpenAIAdvancedSchedulerWeightQueue                     string
+	OpenAIAdvancedSchedulerWeightErrorRate                 string
+	OpenAIAdvancedSchedulerWeightTTFT                      string
+	OpenAIAdvancedSchedulerWeightReset                     string
+	OpenAIAdvancedSchedulerWeightQuotaHeadroom             string
+	OpenAIAdvancedSchedulerWeightPreviousResponse          string
+	OpenAIAdvancedSchedulerWeightSessionSticky             string
+	OpenAIAdvancedSchedulerEffectiveLBTopK                 string
+	OpenAIAdvancedSchedulerEffectiveWeightPriority         string
+	OpenAIAdvancedSchedulerEffectiveWeightLoad             string
+	OpenAIAdvancedSchedulerEffectiveWeightQueue            string
+	OpenAIAdvancedSchedulerEffectiveWeightErrorRate        string
+	OpenAIAdvancedSchedulerEffectiveWeightTTFT             string
+	OpenAIAdvancedSchedulerEffectiveWeightReset            string
+	OpenAIAdvancedSchedulerEffectiveWeightQuotaHeadroom    string
+	OpenAIAdvancedSchedulerEffectiveWeightPreviousResponse string
+	OpenAIAdvancedSchedulerEffectiveWeightSessionSticky    string
 
 	// 余额不足提醒
 	BalanceLowNotifyEnabled     bool
diff --git a/frontend/src/api/admin/settings.ts b/frontend/src/api/admin/settings.ts
index 44fbe29187..457775870d 100644
--- a/frontend/src/api/admin/settings.ts
+++ b/frontend/src/api/admin/settings.ts
@@ -606,6 +606,28 @@ export interface SystemSettings {
   payment_visible_method_alipay_enabled?: boolean;
   payment_visible_method_wxpay_enabled?: boolean;
   openai_advanced_scheduler_enabled?: boolean;
+  openai_advanced_scheduler_sticky_weighted_enabled?: boolean;
+  openai_advanced_scheduler_subscription_priority_enabled?: boolean;
+  openai_advanced_scheduler_lb_top_k?: string;
+  openai_advanced_scheduler_weight_priority?: string;
+  openai_advanced_scheduler_weight_load?: string;
+  openai_advanced_scheduler_weight_queue?: string;
+  openai_advanced_scheduler_weight_error_rate?: string;
+  openai_advanced_scheduler_weight_ttft?: string;
+  openai_advanced_scheduler_weight_reset?: string;
+  openai_advanced_scheduler_weight_quota_headroom?: string;
+  openai_advanced_scheduler_weight_previous_response?: string;
+  openai_advanced_scheduler_weight_session_sticky?: string;
+  openai_advanced_scheduler_effective_lb_top_k?: string;
+  openai_advanced_scheduler_effective_weight_priority?: string;
+  openai_advanced_scheduler_effective_weight_load?: string;
+  openai_advanced_scheduler_effective_weight_queue?: string;
+  openai_advanced_scheduler_effective_weight_error_rate?: string;
+  openai_advanced_scheduler_effective_weight_ttft?: string;
+  openai_advanced_scheduler_effective_weight_reset?: string;
+  openai_advanced_scheduler_effective_weight_quota_headroom?: string;
+  openai_advanced_scheduler_effective_weight_previous_response?: string;
+  openai_advanced_scheduler_effective_weight_session_sticky?: string;
 
   // 余额、订阅到期与账号限额通知
   balance_low_notify_enabled: boolean;
@@ -855,6 +877,18 @@ export interface UpdateSettingsRequest {
   payment_visible_method_alipay_enabled?: boolean;
   payment_visible_method_wxpay_enabled?: boolean;
   openai_advanced_scheduler_enabled?: boolean;
+  openai_advanced_scheduler_sticky_weighted_enabled?: boolean;
+  openai_advanced_scheduler_subscription_priority_enabled?: boolean;
+  openai_advanced_scheduler_lb_top_k?: string;
+  openai_advanced_scheduler_weight_priority?: string;
+  openai_advanced_scheduler_weight_load?: string;
+  openai_advanced_scheduler_weight_queue?: string;
+  openai_advanced_scheduler_weight_error_rate?: string;
+  openai_advanced_scheduler_weight_ttft?: string;
+  openai_advanced_scheduler_weight_reset?: string;
+  openai_advanced_scheduler_weight_quota_headroom?: string;
+  openai_advanced_scheduler_weight_previous_response?: string;
+  openai_advanced_scheduler_weight_session_sticky?: string;
   // 余额、订阅到期与账号限额通知
   balance_low_notify_enabled?: boolean;
   balance_low_notify_threshold?: number;
diff --git a/frontend/src/i18n/locales/en.ts b/frontend/src/i18n/locales/en.ts
index ac831278bd..b42acc9c91 100644
--- a/frontend/src/i18n/locales/en.ts
+++ b/frontend/src/i18n/locales/en.ts
@@ -3265,6 +3265,7 @@ export default {
         priority: 'Priority',
         billingRateMultiplier: 'Billing Rate',
         weight: 'Weight',
+        schedulerScore: 'Scheduler Score',
         status: 'Status',
         schedulable: 'Schedulable',
         todayStats: 'Today Stats',
@@ -3276,6 +3277,12 @@ export default {
         expiresAt: 'Expires At',
         actions: 'Actions'
       },
+      schedulerScore: {
+        baseShort: 'Base',
+        stickyShort: 'Sticky',
+        ungrouped: 'Ungrouped',
+        hint: 'Displayed as "group / base score / sticky bonus". The base score is computed within the current filtered candidate set and includes priority, load, queue depth, error rate, first-token latency, reset window, quota headroom, and related factors. The sticky bonus applies only when sticky weighting is enabled for previous_response_id or session_hash. Higher scores are preferred.'
+      },
       usageWindowsHint: '"5h / 7d" are the upstream account\'s official rolling usage windows (e.g. OpenAI ChatGPT, Claude). They are imposed by the upstream provider on the account itself — not configured by sub2api, and unrelated to the models you map. Usage resets automatically once each window rolls over, and the limit cannot be lifted from within sub2api.',
       allPrivacyModes: 'All Privacy States',
       privacyUnset: 'Unset',
@@ -6724,7 +6731,24 @@ export default {
       },
       openaiExperimentalScheduler: {
         title: 'OpenAI experimental scheduler policy',
-        description: "Disabled by default. When enabled, this only changes the gateway's experimental account-selection policy for OpenAI traffic; it does not indicate an upstream OpenAI capability."
+        description: "Disabled by default. When enabled, this only changes the gateway's experimental account-selection policy for OpenAI traffic; it does not indicate an upstream OpenAI capability.",
+        stickyWeightedTitle: 'Sticky weighting',
+        stickyWeightedDescription: 'When enabled, previous_response_id and session_hash affinity are scored by the advanced scheduler. When disabled, sticky accounts keep the legacy hard-hit behavior.',
+        subscriptionPriorityTitle: 'Subscription priority',
+        subscriptionPriorityDescription: 'When enabled, the scheduler scores ChatGPT subscription accounts first and falls back to non-subscription accounts only if no subscription slot can be acquired.',
+        weightsTitle: 'Scheduler weight overrides',
+        weightsDescription: 'Blank values use config/environment values; when config is not set, built-in defaults apply. Non-blank page settings take priority.',
+        defaultPlaceholder: 'config/default: {value}',
+        topKLabel: 'TopK',
+        priorityWeight: 'Priority',
+        loadWeight: 'Load',
+        queueWeight: 'Queue',
+        errorRateWeight: 'Error rate',
+        ttftWeight: 'TTFT',
+        resetWeight: 'Reset window',
+        quotaHeadroomWeight: 'Quota headroom',
+        previousResponseWeight: 'previous_response sticky',
+        sessionStickyWeight: 'session_hash sticky'
       },
       usageRecords: {
         title: 'Usage Records',
diff --git a/frontend/src/i18n/locales/zh.ts b/frontend/src/i18n/locales/zh.ts
index dc090a458f..44194168d9 100644
--- a/frontend/src/i18n/locales/zh.ts
+++ b/frontend/src/i18n/locales/zh.ts
@@ -3296,6 +3296,7 @@ export default {
         priority: '优先级',
         billingRateMultiplier: '账号倍率',
         weight: '权重',
+        schedulerScore: '调度权值',
         status: '状态',
         schedulable: '调度',
         todayStats: '今日统计',
@@ -3307,6 +3308,12 @@ export default {
         expiresAt: '过期时间',
         actions: '操作'
       },
+      schedulerScore: {
+        baseShort: '普通',
+        stickyShort: '粘性',
+        ungrouped: '未分组',
+        hint: '显示格式为“分组名 / 基础分 / 粘性加分”。基础分按当前筛选条件限定的候选账号计算，包含优先级、负载、排队、错误率、首包延迟、重置窗口、额度余量等因子；粘性加分只在开启粘性加权时用于 previous_response_id 或 session_hash。分数越大越优先。'
+      },
       usageWindowsHint: '“5h / 7d”是上游账号（如 OpenAI ChatGPT、Claude）官方的滚动用量窗口限制，由上游对账号设定，并非 sub2api 配置，也与你映射的模型无关。窗口滚动到期后用量会自动重置，无法在 sub2api 端解除该限制。',
       allPrivacyModes: '全部Privacy状态',
       privacyUnset: '未设置',
@@ -6878,7 +6885,24 @@ export default {
       },
       openaiExperimentalScheduler: {
         title: 'OpenAI 实验调度策略',
-        description: '默认关闭。开启后仅影响本网关在 OpenAI 账号间的实验性调度选择逻辑，不代表上游 OpenAI 官方能力。'
+        description: '默认关闭。开启后仅影响本网关在 OpenAI 账号间的实验性调度选择逻辑，不代表上游 OpenAI 官方能力。',
+        stickyWeightedTitle: '粘性加权',
+        stickyWeightedDescription: '开启后 previous_response_id 和 session_hash 粘性进入高级调度打分；关闭时仍按旧逻辑硬命中粘性账号。',
+        subscriptionPriorityTitle: '订阅优先',
+        subscriptionPriorityDescription: '开启后先在 ChatGPT 订阅账号池中按权值选取；订阅池拿不到席位时再回退到非订阅账号池。',
+        weightsTitle: '调度权值覆盖',
+        weightsDescription: '留空时使用配置/环境变量值；配置未设置时使用内置默认值。页面非空设置优先。',
+        defaultPlaceholder: '配置/默认：{value}',
+        topKLabel: 'TopK',
+        priorityWeight: '优先级',
+        loadWeight: '负载',
+        queueWeight: '排队',
+        errorRateWeight: '错误率',
+        ttftWeight: '首包延迟',
+        resetWeight: '重置窗口',
+        quotaHeadroomWeight: '额度余量',
+        previousResponseWeight: 'previous_response 粘性',
+        sessionStickyWeight: 'session_hash 粘性'
       },
       usageRecords: {
         title: '使用记录',
diff --git a/frontend/src/types/index.ts b/frontend/src/types/index.ts
index 315db76224..cff836d041 100644
--- a/frontend/src/types/index.ts
+++ b/frontend/src/types/index.ts
@@ -857,6 +857,13 @@ export interface Account {
   concurrency: number
   load_factor?: number | null
   current_concurrency?: number // Real-time concurrency count from Redis
+  scheduler_score?: {
+    base_score: number
+    sticky_score?: number
+    sticky_score_infinity?: boolean
+    sticky_weighted_enabled: boolean
+  } | null
+  scheduler_scores?: AccountSchedulerGroupScore[] | null
   priority: number
   rate_multiplier?: number // Account billing multiplier (>=0, 0 means free)
   status: 'active' | 'inactive' | 'error'
@@ -947,6 +954,16 @@ export interface Account {
   parent_chatgpt_account_id?: string
 }
 
+export interface AccountSchedulerGroupScore {
+  group_id?: number | null
+  group_name?: string
+  group_priority?: number | null
+  base_score: number
+  sticky_score?: number
+  sticky_score_infinity?: boolean
+  sticky_weighted_enabled: boolean
+}
+
 // Account Usage types
 export interface WindowStats {
   requests: number
diff --git a/frontend/src/views/admin/AccountsView.vue b/frontend/src/views/admin/AccountsView.vue
index e8cd05cb2b..bd8fd6067a 100644
--- a/frontend/src/views/admin/AccountsView.vue
+++ b/frontend/src/views/admin/AccountsView.vue
@@ -322,6 +322,29 @@
           <template #cell-priority="{ value }">
             <span class="text-sm text-gray-700 dark:text-gray-300">{{ value }}</span>
           </template>
+          <template #header-scheduler_score="{ column }">
+            <div class="flex items-center">
+              <span>{{ column.label }}</span>
+              <HelpTooltip :content="t('admin.accounts.schedulerScore.hint')" width-class="w-80" />
+            </div>
+          </template>
+          <template #cell-scheduler_score="{ row }">
+            <div v-if="getSchedulerScoreRows(row).length" class="flex min-w-[7rem] flex-col gap-0.5 font-mono text-[11px] leading-4">
+              <div
+                v-for="score in getSchedulerScoreRows(row)"
+                :key="String(score.group_id)"
+                class="flex items-center gap-1 whitespace-nowrap text-gray-700 dark:text-gray-300"
+                :title="`${formatSchedulerScoreGroup(score)} / ${formatSchedulerScore(score.base_score)} / ${formatStickySchedulerScore(score)}`"
+              >
+                <span class="max-w-[4.75rem] truncate text-gray-500 dark:text-dark-400">{{ formatSchedulerScoreGroup(score) }}</span>
+                <span class="text-gray-300 dark:text-gray-600">/</span>
+                <span>{{ formatSchedulerScore(score.base_score) }}</span>
+                <span class="text-gray-300 dark:text-gray-600">/</span>
+                <span class="text-primary-700 dark:text-primary-300">{{ formatStickySchedulerScore(score) }}</span>
+              </div>
+            </div>
+            <span v-else class="text-sm text-gray-400 dark:text-dark-500">-</span>
+          </template>
           <template #cell-last_used_at="{ value }">
             <span class="text-sm text-gray-500 dark:text-dark-400">{{ formatRelativeTime(value) }}</span>
           </template>
@@ -441,7 +464,7 @@ import TLSFingerprintProfilesModal from '@/components/admin/TLSFingerprintProfil
 import { buildOpenAIUsageRefreshKey } from '@/utils/accountUsageRefresh'
 import { formatDateTime, formatRelativeTime } from '@/utils/format'
 import { proxyExpiryBadgeClass, proxyExpiryLabelKey } from '@/utils/proxyExpiry'
-import type { Account, AccountPlatform, AccountType, Proxy as AccountProxy, AdminGroup, WindowStats, ClaudeModel } from '@/types'
+import type { Account, AccountPlatform, AccountSchedulerGroupScore, AccountType, Proxy as AccountProxy, AdminGroup, WindowStats, ClaudeModel } from '@/types'
 
 const { t } = useI18n()
 const appStore = useAppStore()
@@ -643,6 +666,29 @@ const autoRefreshIntervalLabel = (sec: number) => {
   return `${sec}s`
 }
 
+const formatSchedulerScore = (value: unknown): string => {
+  const num = Number(value)
+  if (!Number.isFinite(num)) return '-'
+  return num.toFixed(6).replace(/\.?0+$/, '')
+}
+
+const formatStickySchedulerScore = (score: AccountSchedulerGroupScore): string => {
+  if (!score) return '-'
+  if (score.sticky_score_infinity) return '+∞'
+  return formatSchedulerScore(score.sticky_score)
+}
+
+const getSchedulerScoreRows = (account: Account): AccountSchedulerGroupScore[] => {
+  if (!Array.isArray(account.scheduler_scores)) return []
+  return account.scheduler_scores.filter(score => score.group_id != null)
+}
+
+const formatSchedulerScoreGroup = (score: AccountSchedulerGroupScore): string => {
+  if ('group_name' in score && score.group_name) return score.group_name
+  if ('group_id' in score && score.group_id != null) return `#${score.group_id}`
+  return '-'
+}
+
 const loadSavedColumns = () => {
   try {
     const saved = localStorage.getItem(HIDDEN_COLUMNS_KEY)
@@ -1166,6 +1212,7 @@ const allColumns = computed(() => {
     { key: 'usage', label: t('admin.accounts.columns.usageWindows'), sortable: false },
     { key: 'proxy', label: t('admin.accounts.columns.proxy'), sortable: false },
     { key: 'priority', label: t('admin.accounts.columns.priority'), sortable: true },
+    { key: 'scheduler_score', label: t('admin.accounts.columns.schedulerScore'), sortable: false },
     { key: 'rate_multiplier', label: t('admin.accounts.columns.billingRateMultiplier'), sortable: true },
     { key: 'last_used_at', label: t('admin.accounts.columns.lastUsed'), sortable: true },
     { key: 'created_at', label: t('admin.accounts.columns.createdAt'), sortable: true },
diff --git a/frontend/src/views/admin/SettingsView.vue b/frontend/src/views/admin/SettingsView.vue
index 650c169c92..78f642d761 100644
--- a/frontend/src/views/admin/SettingsView.vue
+++ b/frontend/src/views/admin/SettingsView.vue
@@ -4014,6 +4014,81 @@
                 </div>
                 <Toggle v-model="form.openai_advanced_scheduler_enabled" />
               </div>
+
+              <div
+                v-if="form.openai_advanced_scheduler_enabled"
+                class="flex items-center justify-between border-t border-gray-100 pt-5 dark:border-dark-700"
+              >
+                <div>
+                  <label
+                    class="text-sm font-medium text-gray-700 dark:text-gray-300"
+                  >
+                    {{ t("admin.settings.openaiExperimentalScheduler.stickyWeightedTitle") }}
+                  </label>
+                  <p class="mt-0.5 text-xs text-gray-500 dark:text-gray-400">
+                    {{
+                      t("admin.settings.openaiExperimentalScheduler.stickyWeightedDescription")
+                    }}
+                  </p>
+                </div>
+                <Toggle v-model="form.openai_advanced_scheduler_sticky_weighted_enabled" />
+              </div>
+
+              <div
+                v-if="form.openai_advanced_scheduler_enabled"
+                class="flex items-center justify-between border-t border-gray-100 pt-5 dark:border-dark-700"
+              >
+                <div>
+                  <label
+                    class="text-sm font-medium text-gray-700 dark:text-gray-300"
+                  >
+                    {{ t("admin.settings.openaiExperimentalScheduler.subscriptionPriorityTitle") }}
+                  </label>
+                  <p class="mt-0.5 text-xs text-gray-500 dark:text-gray-400">
+                    {{
+                      t("admin.settings.openaiExperimentalScheduler.subscriptionPriorityDescription")
+                    }}
+                  </p>
+                </div>
+                <Toggle v-model="form.openai_advanced_scheduler_subscription_priority_enabled" />
+              </div>
+
+              <div
+                v-if="form.openai_advanced_scheduler_enabled"
+                class="border-t border-gray-100 pt-5 dark:border-dark-700"
+              >
+                <div>
+                  <label
+                    class="text-sm font-medium text-gray-700 dark:text-gray-300"
+                  >
+                    {{ t("admin.settings.openaiExperimentalScheduler.weightsTitle") }}
+                  </label>
+                  <p class="mt-0.5 text-xs text-gray-500 dark:text-gray-400">
+                    {{
+                      t("admin.settings.openaiExperimentalScheduler.weightsDescription")
+                    }}
+                  </p>
+                </div>
+
+                <div class="mt-4 grid grid-cols-1 gap-4 md:grid-cols-2 xl:grid-cols-5">
+                  <label
+                    v-for="field in openAIAdvancedSchedulerWeightFields"
+                    :key="field.key"
+                    class="block"
+                  >
+                    <span class="text-xs font-medium text-gray-600 dark:text-gray-400">
+                      {{ field.label }}
+                    </span>
+                    <input
+                      v-model="form[field.key]"
+                      class="input mt-1"
+                      inputmode="decimal"
+                      :placeholder="field.placeholder"
+                      type="text"
+                    />
+                  </label>
+                </div>
+              </div>
             </div>
           </div>
 
@@ -7903,6 +7978,18 @@ type SettingsForm = Omit<
   google_oauth_client_secret: string;
   force_email_on_third_party_signup: boolean;
   openai_advanced_scheduler_enabled: boolean;
+  openai_advanced_scheduler_sticky_weighted_enabled: boolean;
+  openai_advanced_scheduler_subscription_priority_enabled: boolean;
+  openai_advanced_scheduler_lb_top_k: string;
+  openai_advanced_scheduler_weight_priority: string;
+  openai_advanced_scheduler_weight_load: string;
+  openai_advanced_scheduler_weight_queue: string;
+  openai_advanced_scheduler_weight_error_rate: string;
+  openai_advanced_scheduler_weight_ttft: string;
+  openai_advanced_scheduler_weight_reset: string;
+  openai_advanced_scheduler_weight_quota_headroom: string;
+  openai_advanced_scheduler_weight_previous_response: string;
+  openai_advanced_scheduler_weight_session_sticky: string;
   // 系统全局平台限额 map；form 内始终归一化为全 4 平台对象（模板非空绑定依赖此不变量）
   default_platform_quotas: DefaultPlatformQuotasMap;
 };
@@ -8094,6 +8181,18 @@ const form = reactive<SettingsForm>({
   // 分组隔离
   allow_ungrouped_key_scheduling: false,
   openai_advanced_scheduler_enabled: false,
+  openai_advanced_scheduler_sticky_weighted_enabled: false,
+  openai_advanced_scheduler_subscription_priority_enabled: false,
+  openai_advanced_scheduler_lb_top_k: "",
+  openai_advanced_scheduler_weight_priority: "",
+  openai_advanced_scheduler_weight_load: "",
+  openai_advanced_scheduler_weight_queue: "",
+  openai_advanced_scheduler_weight_error_rate: "",
+  openai_advanced_scheduler_weight_ttft: "",
+  openai_advanced_scheduler_weight_reset: "",
+  openai_advanced_scheduler_weight_quota_headroom: "",
+  openai_advanced_scheduler_weight_previous_response: "",
+  openai_advanced_scheduler_weight_session_sticky: "",
   // Gateway forwarding behavior
   enable_fingerprint_unification: true,
   enable_metadata_passthrough: false,
@@ -8131,6 +8230,103 @@ const form = reactive<SettingsForm>({
   allow_user_view_error_requests: false,
 });
 
+type OpenAIAdvancedSchedulerOverrideKey =
+  | "openai_advanced_scheduler_lb_top_k"
+  | "openai_advanced_scheduler_weight_priority"
+  | "openai_advanced_scheduler_weight_load"
+  | "openai_advanced_scheduler_weight_queue"
+  | "openai_advanced_scheduler_weight_error_rate"
+  | "openai_advanced_scheduler_weight_ttft"
+  | "openai_advanced_scheduler_weight_reset"
+  | "openai_advanced_scheduler_weight_quota_headroom"
+  | "openai_advanced_scheduler_weight_previous_response"
+  | "openai_advanced_scheduler_weight_session_sticky";
+
+type OpenAIAdvancedSchedulerEffectiveKey =
+  | "openai_advanced_scheduler_effective_lb_top_k"
+  | "openai_advanced_scheduler_effective_weight_priority"
+  | "openai_advanced_scheduler_effective_weight_load"
+  | "openai_advanced_scheduler_effective_weight_queue"
+  | "openai_advanced_scheduler_effective_weight_error_rate"
+  | "openai_advanced_scheduler_effective_weight_ttft"
+  | "openai_advanced_scheduler_effective_weight_reset"
+  | "openai_advanced_scheduler_effective_weight_quota_headroom"
+  | "openai_advanced_scheduler_effective_weight_previous_response"
+  | "openai_advanced_scheduler_effective_weight_session_sticky";
+
+const openAIAdvancedSchedulerWeightFields = computed<
+  Array<{
+    key: OpenAIAdvancedSchedulerOverrideKey;
+    label: string;
+    placeholder: string;
+  }>
+>(() => {
+  const placeholder = (
+    effectiveKey: OpenAIAdvancedSchedulerEffectiveKey,
+    fallbackValue: string,
+  ) => {
+    const effectiveValue = String(
+      (form as Record<string, unknown>)[effectiveKey] ?? "",
+    ).trim();
+    return t("admin.settings.openaiExperimentalScheduler.defaultPlaceholder", {
+      value: effectiveValue || fallbackValue,
+    });
+  };
+
+  return [
+    {
+      key: "openai_advanced_scheduler_lb_top_k",
+      label: t("admin.settings.openaiExperimentalScheduler.topKLabel"),
+      placeholder: placeholder("openai_advanced_scheduler_effective_lb_top_k", "7"),
+    },
+    {
+      key: "openai_advanced_scheduler_weight_priority",
+      label: t("admin.settings.openaiExperimentalScheduler.priorityWeight"),
+      placeholder: placeholder("openai_advanced_scheduler_effective_weight_priority", "1"),
+    },
+    {
+      key: "openai_advanced_scheduler_weight_load",
+      label: t("admin.settings.openaiExperimentalScheduler.loadWeight"),
+      placeholder: placeholder("openai_advanced_scheduler_effective_weight_load", "1"),
+    },
+    {
+      key: "openai_advanced_scheduler_weight_queue",
+      label: t("admin.settings.openaiExperimentalScheduler.queueWeight"),
+      placeholder: placeholder("openai_advanced_scheduler_effective_weight_queue", "0.7"),
+    },
+    {
+      key: "openai_advanced_scheduler_weight_error_rate",
+      label: t("admin.settings.openaiExperimentalScheduler.errorRateWeight"),
+      placeholder: placeholder("openai_advanced_scheduler_effective_weight_error_rate", "0.8"),
+    },
+    {
+      key: "openai_advanced_scheduler_weight_ttft",
+      label: t("admin.settings.openaiExperimentalScheduler.ttftWeight"),
+      placeholder: placeholder("openai_advanced_scheduler_effective_weight_ttft", "0.5"),
+    },
+    {
+      key: "openai_advanced_scheduler_weight_reset",
+      label: t("admin.settings.openaiExperimentalScheduler.resetWeight"),
+      placeholder: placeholder("openai_advanced_scheduler_effective_weight_reset", "0"),
+    },
+    {
+      key: "openai_advanced_scheduler_weight_quota_headroom",
+      label: t("admin.settings.openaiExperimentalScheduler.quotaHeadroomWeight"),
+      placeholder: placeholder("openai_advanced_scheduler_effective_weight_quota_headroom", "0"),
+    },
+    {
+      key: "openai_advanced_scheduler_weight_previous_response",
+      label: t("admin.settings.openaiExperimentalScheduler.previousResponseWeight"),
+      placeholder: placeholder("openai_advanced_scheduler_effective_weight_previous_response", "5"),
+    },
+    {
+      key: "openai_advanced_scheduler_weight_session_sticky",
+      label: t("admin.settings.openaiExperimentalScheduler.sessionStickyWeight"),
+      placeholder: placeholder("openai_advanced_scheduler_effective_weight_session_sticky", "3"),
+    },
+  ];
+});
+
 const authSourceDefaults = reactive<AuthSourceDefaultsState>(
   buildAuthSourceDefaultsState({}),
 );
@@ -9359,6 +9555,30 @@ async function saveSettings() {
         form.payment_cancel_rate_limit_window_mode,
       payment_alipay_force_qrcode: form.payment_alipay_force_qrcode,
       openai_advanced_scheduler_enabled: form.openai_advanced_scheduler_enabled,
+      openai_advanced_scheduler_sticky_weighted_enabled:
+        form.openai_advanced_scheduler_sticky_weighted_enabled,
+      openai_advanced_scheduler_subscription_priority_enabled:
+        form.openai_advanced_scheduler_subscription_priority_enabled,
+      openai_advanced_scheduler_lb_top_k:
+        form.openai_advanced_scheduler_lb_top_k.trim(),
+      openai_advanced_scheduler_weight_priority:
+        form.openai_advanced_scheduler_weight_priority.trim(),
+      openai_advanced_scheduler_weight_load:
+        form.openai_advanced_scheduler_weight_load.trim(),
+      openai_advanced_scheduler_weight_queue:
+        form.openai_advanced_scheduler_weight_queue.trim(),
+      openai_advanced_scheduler_weight_error_rate:
+        form.openai_advanced_scheduler_weight_error_rate.trim(),
+      openai_advanced_scheduler_weight_ttft:
+        form.openai_advanced_scheduler_weight_ttft.trim(),
+      openai_advanced_scheduler_weight_reset:
+        form.openai_advanced_scheduler_weight_reset.trim(),
+      openai_advanced_scheduler_weight_quota_headroom:
+        form.openai_advanced_scheduler_weight_quota_headroom.trim(),
+      openai_advanced_scheduler_weight_previous_response:
+        form.openai_advanced_scheduler_weight_previous_response.trim(),
+      openai_advanced_scheduler_weight_session_sticky:
+        form.openai_advanced_scheduler_weight_session_sticky.trim(),
       // 余额、订阅到期与账号限额通知
       balance_low_notify_enabled: form.balance_low_notify_enabled,
       balance_low_notify_threshold:
diff --git a/frontend/src/views/admin/__tests__/SettingsView.spec.ts b/frontend/src/views/admin/__tests__/SettingsView.spec.ts
index e458728b66..e28e186e05 100644
--- a/frontend/src/views/admin/__tests__/SettingsView.spec.ts
+++ b/frontend/src/views/admin/__tests__/SettingsView.spec.ts
@@ -161,6 +161,23 @@ vi.mock("vue-i18n", async () => {
     "admin.settings.payment.findProvider": "查看支持的支付方式",
     "admin.settings.openaiExperimentalScheduler.title": "OpenAI 实验调度策略",
     "admin.settings.openaiExperimentalScheduler.description": "默认关闭。开启后仅影响本网关在 OpenAI 账号间的实验性调度选择逻辑，不代表上游 OpenAI 官方能力。",
+    "admin.settings.openaiExperimentalScheduler.stickyWeightedTitle": "粘性加权",
+    "admin.settings.openaiExperimentalScheduler.stickyWeightedDescription": "开启后 previous_response_id 和 session_hash 粘性进入高级调度打分；关闭时仍按旧逻辑硬命中粘性账号。",
+    "admin.settings.openaiExperimentalScheduler.subscriptionPriorityTitle": "订阅优先",
+    "admin.settings.openaiExperimentalScheduler.subscriptionPriorityDescription": "开启后先在 ChatGPT 订阅账号池中按权值选取；订阅池拿不到席位时再回退到非订阅账号池。",
+    "admin.settings.openaiExperimentalScheduler.weightsTitle": "调度权值覆盖",
+    "admin.settings.openaiExperimentalScheduler.weightsDescription": "留空时使用配置/环境变量值；配置未设置时使用内置默认值。页面非空设置优先。",
+    "admin.settings.openaiExperimentalScheduler.defaultPlaceholder": "配置/默认：{value}",
+    "admin.settings.openaiExperimentalScheduler.topKLabel": "TopK",
+    "admin.settings.openaiExperimentalScheduler.priorityWeight": "优先级",
+    "admin.settings.openaiExperimentalScheduler.loadWeight": "负载",
+    "admin.settings.openaiExperimentalScheduler.queueWeight": "排队",
+    "admin.settings.openaiExperimentalScheduler.errorRateWeight": "错误率",
+    "admin.settings.openaiExperimentalScheduler.ttftWeight": "首包延迟",
+    "admin.settings.openaiExperimentalScheduler.resetWeight": "重置窗口",
+    "admin.settings.openaiExperimentalScheduler.quotaHeadroomWeight": "额度余量",
+    "admin.settings.openaiExperimentalScheduler.previousResponseWeight": "previous_response 粘性",
+    "admin.settings.openaiExperimentalScheduler.sessionStickyWeight": "session_hash 粘性",
     "admin.settings.site.uploadImage": "上传图片",
     "admin.settings.site.remove": "移除",
     "admin.settings.platformQuota.platform": "平台",
@@ -411,6 +428,28 @@ const baseSettingsResponse = {
   payment_visible_method_alipay_enabled: true,
   payment_visible_method_wxpay_enabled: true,
   openai_advanced_scheduler_enabled: false,
+  openai_advanced_scheduler_sticky_weighted_enabled: false,
+  openai_advanced_scheduler_subscription_priority_enabled: false,
+  openai_advanced_scheduler_lb_top_k: "",
+  openai_advanced_scheduler_weight_priority: "",
+  openai_advanced_scheduler_weight_load: "",
+  openai_advanced_scheduler_weight_queue: "",
+  openai_advanced_scheduler_weight_error_rate: "",
+  openai_advanced_scheduler_weight_ttft: "",
+  openai_advanced_scheduler_weight_reset: "",
+  openai_advanced_scheduler_weight_quota_headroom: "",
+  openai_advanced_scheduler_weight_previous_response: "",
+  openai_advanced_scheduler_weight_session_sticky: "",
+  openai_advanced_scheduler_effective_lb_top_k: "7",
+  openai_advanced_scheduler_effective_weight_priority: "1",
+  openai_advanced_scheduler_effective_weight_load: "1",
+  openai_advanced_scheduler_effective_weight_queue: "0.7",
+  openai_advanced_scheduler_effective_weight_error_rate: "0.8",
+  openai_advanced_scheduler_effective_weight_ttft: "0.5",
+  openai_advanced_scheduler_effective_weight_reset: "0",
+  openai_advanced_scheduler_effective_weight_quota_headroom: "0",
+  openai_advanced_scheduler_effective_weight_previous_response: "5",
+  openai_advanced_scheduler_effective_weight_session_sticky: "3",
   balance_low_notify_enabled: false,
   balance_low_notify_threshold: 0,
   balance_low_notify_recharge_url: "",

From 2d2b6f032585cdc1733bf6727d4fadcbb3844ef1 Mon Sep 17 00:00:00 2001
From: Yeyin Hu <yinr@yinr.cc>
Date: Mon, 6 Jul 2026 08:12:09 +0800
Subject: [PATCH 26/28] Update revokeDesc to allow subscription recovery

---
 frontend/src/i18n/locales/zh.ts | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/frontend/src/i18n/locales/zh.ts b/frontend/src/i18n/locales/zh.ts
index dc090a458f..8ee89e153e 100644
--- a/frontend/src/i18n/locales/zh.ts
+++ b/frontend/src/i18n/locales/zh.ts
@@ -3182,7 +3182,7 @@ export default {
           resetQuota: '重置配额',
           resetQuotaDesc: '将日/周/月用量归零，重新开始计算',
           revoke: '撤销',
-          revokeDesc: '立即终止该用户的订阅，不可恢复'
+          revokeDesc: '立即终止该用户的订阅，可在已撤销列表中恢复'
         },
         tip: '提示：订阅分组下拉列表中只会显示计费类型为「订阅」且状态为「正常」的分组。如果没有可选项，请先到分组管理中创建。'
       }

From ee90246aa5705ea59be7108a221d7a70b1b91e6b Mon Sep 17 00:00:00 2001
From: Yeyin Hu <yinr@yinr.cc>
Date: Mon, 6 Jul 2026 08:23:43 +0800
Subject: [PATCH 27/28] Modify revokeDesc for subscription termination

Updated revokeDesc to indicate restoration option.
---
 frontend/src/i18n/locales/en.ts | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/frontend/src/i18n/locales/en.ts b/frontend/src/i18n/locales/en.ts
index ac831278bd..46ace87051 100644
--- a/frontend/src/i18n/locales/en.ts
+++ b/frontend/src/i18n/locales/en.ts
@@ -3107,7 +3107,7 @@ export default {
           resetQuota: 'Reset Quota',
           resetQuotaDesc: 'Reset daily/weekly/monthly usage to zero',
           revoke: 'Revoke',
-          revokeDesc: 'Immediately terminate the subscription (irreversible)'
+          revokeDesc: 'Immediately terminate the subscription (restorable from the revoked list)'
         },
         tip: 'Tip: Only groups with billing type "Subscription" and status "Active" appear in the group dropdown. If no options are available, create one in Group Management first.'
       }

From 6752cdc826db7d7b9ca6ef398c4b58f6f293f8a5 Mon Sep 17 00:00:00 2001
From: shaw <dev@sub2api.org>
Date: Mon, 6 Jul 2026 09:36:21 +0800
Subject: [PATCH 28/28] =?UTF-8?q?fix(i18n):=20=E7=94=A8=E9=87=8F=E9=A1=B5?=
 =?UTF-8?q?=E8=B4=B9=E7=94=A8=E6=98=8E=E7=BB=86=E6=8F=90=E7=A4=BA=E5=B0=86?=
 =?UTF-8?q?=E9=9D=A2=E5=90=91=E7=94=A8=E6=88=B7=E7=9A=84"=E6=88=90?=
 =?UTF-8?q?=E6=9C=AC"=E6=8E=AA=E8=BE=9E=E7=BB=9F=E4=B8=80=E4=B8=BA"?=
 =?UTF-8?q?=E8=B4=B9=E7=94=A8"?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

用户 /usage 页费用列悬浮提示中的成本明细/输入成本/输出成本/
缓存创建成本/缓存读取成本/图片输出成本 改为对应的"费用"表述;
成本是运营方视角的词,面向用户应称费用。管理端真正的账号成本
词条(accountCost 等)保持不变。
---
 frontend/src/i18n/locales/zh.ts | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/frontend/src/i18n/locales/zh.ts b/frontend/src/i18n/locales/zh.ts
index dc090a458f..4ce4070711 100644
--- a/frontend/src/i18n/locales/zh.ts
+++ b/frontend/src/i18n/locales/zh.ts
@@ -896,7 +896,7 @@ export default {
   usage: {
     title: '使用记录',
     description: '查看和分析您的 API 使用历史',
-    costDetails: '成本明细',
+    costDetails: '费用明细',
     tokenDetails: 'Token 明细',
     cacheTtlOverriddenHint: '缓存 TTL Override 已启用',
     cacheTtlOverriddenLabel: 'TTL 替换',
@@ -972,7 +972,7 @@ export default {
     imageOutputSize: '输出尺寸',
     imageOutputTokens: '图片输出 Token',
     imageOutputTokenPrice: '图片输出单价',
-    imageOutputCost: '图片输出成本',
+    imageOutputCost: '图片输出费用',
     imageSizeSource: '尺寸来源',
     imageSizeBreakdown: '尺寸明细',
     imageSizeSourceOutput: '上游输出',
@@ -4936,10 +4936,10 @@ export default {
       allAccounts: '全部账户',
       allGroups: '全部分组',
       allTypes: '全部类型',
-      inputCost: '输入成本',
-      outputCost: '输出成本',
-      cacheCreationCost: '缓存创建成本',
-      cacheReadCost: '缓存读取成本',
+      inputCost: '输入费用',
+      outputCost: '输出费用',
+      cacheCreationCost: '缓存创建费用',
+      cacheReadCost: '缓存读取费用',
       inputTokens: '输入 Token',
       outputTokens: '输出 Token',
       cacheCreationTokens: '缓存创建 Token',