test: 更新管理员设置契约字段

Merge pull request #2118 from DaydreamCoding/fix/restore-pagination-localStorage
fix: 恢复表格分页大小 localStorage 持久化
2026-05-04 21:20:51 +08:00 · 2026-04-30 13:48:27 +08:00 · 2026-04-30 13:42:18 +08:00 · 2026-04-30 13:38:22 +08:00 · 2026-04-30 12:02:08 +08:00 · 2026-04-30 11:38:11 +08:00
25 changed files with 726 additions and 79 deletions
--- a/.gitignore
+++ b/.gitignore
@@ -122,7 +122,7 @@ scripts
 .code-review-state
 #openspec/
 code-reviews/
-#AGENTS.md
+AGENTS.md
 backend/cmd/server/server
 deploy/docker-compose.override.yml
 .gocache/
--- a/backend/cmd/server/VERSION
+++ b/backend/cmd/server/VERSION
@@ -1 +1 @@
-0.1.119
+0.1.120
--- a/backend/internal/handler/admin/setting_handler.go
+++ b/backend/internal/handler/admin/setting_handler.go
@@ -209,6 +209,7 @@ func (h *SettingHandler) GetSettings(c *gin.Context) {
 		EnableFingerprintUnification:           settings.EnableFingerprintUnification,
 		EnableMetadataPassthrough:              settings.EnableMetadataPassthrough,
 		EnableCCHSigning:                       settings.EnableCCHSigning,
+		EnableAnthropicCacheTTL1hInjection:     settings.EnableAnthropicCacheTTL1hInjection,
 		WebSearchEmulationEnabled:              settings.WebSearchEmulationEnabled,
 		PaymentVisibleMethodAlipaySource:       settings.PaymentVisibleMethodAlipaySource,
 		PaymentVisibleMethodWxpaySource:        settings.PaymentVisibleMethodWxpaySource,
@@ -441,9 +442,10 @@ type UpdateSettingsRequest struct {
 	BackendModeEnabled bool `json:"backend_mode_enabled"`

 	// Gateway forwarding behavior
-	EnableFingerprintUnification *bool `json:"enable_fingerprint_unification"`
-	EnableMetadataPassthrough    *bool `json:"enable_metadata_passthrough"`
-	EnableCCHSigning             *bool `json:"enable_cch_signing"`
+	EnableFingerprintUnification       *bool `json:"enable_fingerprint_unification"`
+	EnableMetadataPassthrough          *bool `json:"enable_metadata_passthrough"`
+	EnableCCHSigning                   *bool `json:"enable_cch_signing"`
+	EnableAnthropicCacheTTL1hInjection *bool `json:"enable_anthropic_cache_ttl_1h_injection"`

 	// Payment visible method routing
 	PaymentVisibleMethodAlipaySource  *string `json:"payment_visible_method_alipay_source"`
@@ -1273,6 +1275,12 @@ func (h *SettingHandler) UpdateSettings(c *gin.Context) {
 			}
 			return previousSettings.EnableCCHSigning
 		}(),
+		EnableAnthropicCacheTTL1hInjection: func() bool {
+			if req.EnableAnthropicCacheTTL1hInjection != nil {
+				return *req.EnableAnthropicCacheTTL1hInjection
+			}
+			return previousSettings.EnableAnthropicCacheTTL1hInjection
+		}(),
 		PaymentVisibleMethodAlipaySource: func() string {
 			if req.PaymentVisibleMethodAlipaySource != nil {
 				return strings.TrimSpace(*req.PaymentVisibleMethodAlipaySource)
@@ -1570,6 +1578,7 @@ func (h *SettingHandler) UpdateSettings(c *gin.Context) {
 		EnableFingerprintUnification:           updatedSettings.EnableFingerprintUnification,
 		EnableMetadataPassthrough:              updatedSettings.EnableMetadataPassthrough,
 		EnableCCHSigning:                       updatedSettings.EnableCCHSigning,
+		EnableAnthropicCacheTTL1hInjection:     updatedSettings.EnableAnthropicCacheTTL1hInjection,
 		PaymentVisibleMethodAlipaySource:       updatedSettings.PaymentVisibleMethodAlipaySource,
 		PaymentVisibleMethodWxpaySource:        updatedSettings.PaymentVisibleMethodWxpaySource,
 		PaymentVisibleMethodAlipayEnabled:      updatedSettings.PaymentVisibleMethodAlipayEnabled,
@@ -1949,6 +1958,9 @@ func diffSettings(before *service.SystemSettings, after *service.SystemSettings,
 	if before.EnableCCHSigning != after.EnableCCHSigning {
 		changed = append(changed, "enable_cch_signing")
 	}
+	if before.EnableAnthropicCacheTTL1hInjection != after.EnableAnthropicCacheTTL1hInjection {
+		changed = append(changed, "enable_anthropic_cache_ttl_1h_injection")
+	}
 	if before.PaymentVisibleMethodAlipaySource != after.PaymentVisibleMethodAlipaySource {
 		changed = append(changed, "payment_visible_method_alipay_source")
 	}
--- a/backend/internal/handler/dto/settings.go
+++ b/backend/internal/handler/dto/settings.go
@@ -142,9 +142,10 @@ type SystemSettings struct {
 	BackendModeEnabled bool `json:"backend_mode_enabled"`

 	// Gateway forwarding behavior
-	EnableFingerprintUnification bool `json:"enable_fingerprint_unification"`
-	EnableMetadataPassthrough    bool `json:"enable_metadata_passthrough"`
-	EnableCCHSigning             bool `json:"enable_cch_signing"`
+	EnableFingerprintUnification       bool `json:"enable_fingerprint_unification"`
+	EnableMetadataPassthrough          bool `json:"enable_metadata_passthrough"`
+	EnableCCHSigning                   bool `json:"enable_cch_signing"`
+	EnableAnthropicCacheTTL1hInjection bool `json:"enable_anthropic_cache_ttl_1h_injection"`

 	// Web Search Emulation
 	WebSearchEmulationEnabled bool `json:"web_search_emulation_enabled"`
--- a/backend/internal/handler/gateway_handler.go
+++ b/backend/internal/handler/gateway_handler.go
@@ -262,6 +262,12 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 	}
 	sessionHash := h.gatewayService.GenerateSessionHash(parsedReq)

+	// [DEBUG-STICKY] 打印会话 hash 生成结果
+	reqLog.Info("sticky.session_hash_generated",
+		zap.String("session_hash", sessionHash),
+		zap.String("metadata_user_id_raw", parsedReq.MetadataUserID),
+	)
+
 	// 获取平台：优先使用强制平台（/antigravity 路由，中间件已设置 request.Context），否则使用分组平台
 	platform := ""
 	if forcePlatform, ok := middleware2.GetForcePlatformFromContext(c); ok {
@@ -278,6 +284,11 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 	var sessionBoundAccountID int64
 	if sessionKey != "" {
 		sessionBoundAccountID, _ = h.gatewayService.GetCachedSessionAccountID(c.Request.Context(), apiKey.GroupID, sessionKey)
+		// [DEBUG-STICKY] 打印粘性会话查询结果
+		reqLog.Info("sticky.cache_lookup",
+			zap.String("session_key", sessionKey),
+			zap.Int64("bound_account_id", sessionBoundAccountID),
+		)
 		if sessionBoundAccountID > 0 {
 			prefetchedGroupID := int64(0)
 			if apiKey.GroupID != nil {
@@ -286,6 +297,8 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 			ctx := service.WithPrefetchedStickySession(c.Request.Context(), sessionBoundAccountID, prefetchedGroupID, h.metadataBridgeEnabled())
 			c.Request = c.Request.WithContext(ctx)
 		}
+	} else {
+		reqLog.Info("sticky.no_session_key", zap.String("session_hash", sessionHash))
 	}
 	// 判断是否真的绑定了粘性会话：有 sessionKey 且已经绑定到某个账号
 	hasBoundSession := sessionKey != "" && sessionBoundAccountID > 0
@@ -536,6 +549,12 @@ func (h *GatewayHandler) Messages(c *gin.Context) {

 		for {
 			// 选择支持该模型的账号
+			reqLog.Info("sticky.selecting_account",
+				zap.String("session_key", sessionKey),
+				zap.Int64("sticky_bound_account_id", sessionBoundAccountID),
+				zap.Bool("has_bound_session", hasBoundSession),
+				zap.Int("failed_account_count", len(fs.FailedAccountIDs)),
+			)
 			selection, err := h.gatewayService.SelectAccountWithLoadAwareness(c.Request.Context(), currentAPIKey.GroupID, sessionKey, reqModel, fs.FailedAccountIDs, parsedReq.MetadataUserID, subject.UserID)
 			if err != nil {
 				if len(fs.FailedAccountIDs) == 0 {
@@ -569,6 +588,16 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 			account := selection.Account
 			setOpsSelectedAccount(c, account.ID, account.Platform)

+			// [DEBUG-STICKY] 打印账号选择结果
+			reqLog.Info("sticky.account_selected",
+				zap.Int64("selected_account_id", account.ID),
+				zap.String("account_name", account.Name),
+				zap.Bool("slot_acquired", selection.Acquired),
+				zap.Bool("has_wait_plan", selection.WaitPlan != nil),
+				zap.Int64("sticky_bound_account_id", sessionBoundAccountID),
+				zap.Bool("sticky_honored", sessionBoundAccountID > 0 && sessionBoundAccountID == account.ID),
+			)
+
 			// 检查请求拦截（预热请求、SUGGESTION MODE等）
 			if account.IsInterceptWarmupEnabled() {
 				interceptType := detectInterceptType(body, reqModel, parsedReq.MaxTokens, reqStream, isClaudeCodeClient)
@@ -635,6 +664,10 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 				}
 				// Slot acquired: no longer waiting in queue.
 				releaseWait()
+				reqLog.Info("sticky.bind_after_wait",
+					zap.String("session_key", sessionKey),
+					zap.Int64("account_id", account.ID),
+				)
 				if err := h.gatewayService.BindStickySession(c.Request.Context(), currentAPIKey.GroupID, sessionKey, account.ID); err != nil {
 					reqLog.Warn("gateway.bind_sticky_session_failed", zap.Int64("account_id", account.ID), zap.Error(err))
 				}
@@ -829,6 +862,17 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 				}
 			}

+			// 绑定粘性会话（成功转发后绑定/刷新）
+			// - 无现有绑定（首次请求）：创建绑定
+			// - 选中账号与粘性账号一致：刷新 TTL
+			// - 粘性账号因负载/RPM 被跳过、选中了其他账号：不覆盖原绑定，
+			//   下次请求粘性账号恢复后仍可命中
+			if sessionKey != "" && (sessionBoundAccountID == 0 || sessionBoundAccountID == account.ID) {
+				if err := h.gatewayService.BindStickySession(c.Request.Context(), currentAPIKey.GroupID, sessionKey, account.ID); err != nil {
+					reqLog.Warn("gateway.bind_sticky_session_failed", zap.Int64("account_id", account.ID), zap.Error(err))
+				}
+			}
+
 			// 捕获请求信息（用于异步记录，避免在 goroutine 中访问 gin.Context）
 			userAgent := c.GetHeader("User-Agent")
 			clientIP := ip.GetClientIP(c)
--- a/backend/internal/repository/scheduler_cache.go
+++ b/backend/internal/repository/scheduler_cache.go
@@ -449,11 +449,69 @@ func buildSchedulerMetadataAccount(account service.Account) service.Account {
 		SessionWindowStart:      account.SessionWindowStart,
 		SessionWindowEnd:        account.SessionWindowEnd,
 		SessionWindowStatus:     account.SessionWindowStatus,
+		AccountGroups:           filterSchedulerAccountGroups(account.AccountGroups),
+		GroupIDs:                filterSchedulerGroupIDs(account.GroupIDs, account.AccountGroups),
 		Credentials:             filterSchedulerCredentials(account.Credentials),
 		Extra:                   filterSchedulerExtra(account.Extra),
 	}
 }

+func filterSchedulerAccountGroups(accountGroups []service.AccountGroup) []service.AccountGroup {
+	if len(accountGroups) == 0 {
+		return nil
+	}
+
+	filtered := make([]service.AccountGroup, 0, len(accountGroups))
+	for _, ag := range accountGroups {
+		if ag.GroupID <= 0 {
+			continue
+		}
+		filtered = append(filtered, service.AccountGroup{
+			AccountID: ag.AccountID,
+			GroupID:   ag.GroupID,
+			Priority:  ag.Priority,
+			CreatedAt: ag.CreatedAt,
+		})
+	}
+	if len(filtered) == 0 {
+		return nil
+	}
+	return filtered
+}
+
+func filterSchedulerGroupIDs(groupIDs []int64, accountGroups []service.AccountGroup) []int64 {
+	if len(groupIDs) == 0 && len(accountGroups) == 0 {
+		return nil
+	}
+
+	seen := make(map[int64]struct{}, len(groupIDs)+len(accountGroups))
+	filtered := make([]int64, 0, len(groupIDs)+len(accountGroups))
+	for _, id := range groupIDs {
+		if id <= 0 {
+			continue
+		}
+		if _, ok := seen[id]; ok {
+			continue
+		}
+		seen[id] = struct{}{}
+		filtered = append(filtered, id)
+	}
+	for _, ag := range accountGroups {
+		if ag.GroupID <= 0 {
+			continue
+		}
+		if _, ok := seen[ag.GroupID]; ok {
+			continue
+		}
+		seen[ag.GroupID] = struct{}{}
+		filtered = append(filtered, ag.GroupID)
+	}
+	if len(filtered) == 0 {
+		return nil
+	}
+	return filtered
+}
+
 func filterSchedulerCredentials(credentials map[string]any) map[string]any {
 	if len(credentials) == 0 {
 		return nil
--- a/backend/internal/repository/scheduler_cache_integration_test.go
+++ b/backend/internal/repository/scheduler_cache_integration_test.go
@@ -56,6 +56,15 @@ func TestSchedulerCacheSnapshotUsesSlimMetadataButKeepsFullAccount(t *testing.T)
 		SessionWindowStart:     &now,
 		SessionWindowEnd:       &windowEnd,
 		SessionWindowStatus:    "active",
+		GroupIDs:               []int64{bucket.GroupID},
+		AccountGroups: []service.AccountGroup{
+			{
+				AccountID: 101,
+				GroupID:   bucket.GroupID,
+				Priority:  5,
+				Group:     &service.Group{ID: bucket.GroupID, Name: "gemini-group"},
+			},
+		},
 	}

 	require.NoError(t, cache.SetSnapshot(ctx, bucket, []service.Account{account}))
@@ -79,10 +88,17 @@ func TestSchedulerCacheSnapshotUsesSlimMetadataButKeepsFullAccount(t *testing.T)
 	require.Equal(t, 4, got.GetMaxSessions())
 	require.Equal(t, 11, got.GetSessionIdleTimeoutMinutes())
 	require.Nil(t, got.Extra["unused_large_field"])
+	require.Equal(t, []int64{bucket.GroupID}, got.GroupIDs)
+	require.Len(t, got.AccountGroups, 1)
+	require.Equal(t, account.ID, got.AccountGroups[0].AccountID)
+	require.Equal(t, bucket.GroupID, got.AccountGroups[0].GroupID)
+	require.Nil(t, got.AccountGroups[0].Group)

 	full, err := cache.GetAccount(ctx, account.ID)
 	require.NoError(t, err)
 	require.NotNil(t, full)
 	require.Equal(t, "secret-access-token", full.GetCredential("access_token"))
 	require.Equal(t, strings.Repeat("x", 4096), full.GetCredential("huge_blob"))
+	require.Len(t, full.AccountGroups, 1)
+	require.NotNil(t, full.AccountGroups[0].Group)
 }
--- a/backend/internal/repository/scheduler_cache_unit_test.go
+++ b/backend/internal/repository/scheduler_cache_unit_test.go
@@ -31,3 +31,43 @@ func TestBuildSchedulerMetadataAccount_KeepsOpenAIWSFlags(t *testing.T) {
 	require.Equal(t, true, got.Extra["mixed_scheduling"])
 	require.Nil(t, got.Extra["unused_large_field"])
 }
+
+func TestBuildSchedulerMetadataAccount_KeepsSlimGroupMembership(t *testing.T) {
+	account := service.Account{
+		ID:       42,
+		Platform: service.PlatformAnthropic,
+		GroupIDs: []int64{7, 9, 7, 0},
+		AccountGroups: []service.AccountGroup{
+			{
+				AccountID: 42,
+				GroupID:   7,
+				Priority:  2,
+				Account:   &service.Account{ID: 42, Name: "drop-from-metadata"},
+				Group:     &service.Group{ID: 7, Name: "drop-from-metadata"},
+			},
+			{
+				AccountID: 42,
+				GroupID:   11,
+				Priority:  3,
+				Group:     &service.Group{ID: 11, Name: "drop-from-metadata"},
+			},
+			{
+				AccountID: 42,
+				GroupID:   0,
+				Priority:  4,
+			},
+		},
+	}
+
+	got := buildSchedulerMetadataAccount(account)
+
+	require.Equal(t, []int64{7, 9, 11}, got.GroupIDs)
+	require.Len(t, got.AccountGroups, 2)
+	require.Equal(t, int64(42), got.AccountGroups[0].AccountID)
+	require.Equal(t, int64(7), got.AccountGroups[0].GroupID)
+	require.Equal(t, 2, got.AccountGroups[0].Priority)
+	require.Nil(t, got.AccountGroups[0].Account)
+	require.Nil(t, got.AccountGroups[0].Group)
+	require.Equal(t, int64(11), got.AccountGroups[1].GroupID)
+	require.Nil(t, got.Groups)
+}
--- a/backend/internal/server/api_contract_test.go
+++ b/backend/internal/server/api_contract_test.go
@@ -740,6 +740,7 @@ func TestAPIContracts(t *testing.T) {
 					"allow_ungrouped_key_scheduling": false,
 					"backend_mode_enabled": false,
 					"enable_cch_signing": false,
+					"enable_anthropic_cache_ttl_1h_injection": false,
 					"enable_fingerprint_unification": true,
 					"enable_metadata_passthrough": false,
 					"web_search_emulation_enabled": false,
@@ -934,6 +935,7 @@ func TestAPIContracts(t *testing.T) {
 					"enable_fingerprint_unification": true,
 					"enable_metadata_passthrough": false,
 					"enable_cch_signing": false,
+					"enable_anthropic_cache_ttl_1h_injection": false,
 					"web_search_emulation_enabled": false,
 					"payment_visible_method_alipay_source": "",
 					"payment_visible_method_wxpay_source": "",
--- a/backend/internal/service/domain_constants.go
+++ b/backend/internal/service/domain_constants.go
@@ -336,6 +336,8 @@ const (
 	SettingKeyEnableMetadataPassthrough = "enable_metadata_passthrough"
 	// SettingKeyEnableCCHSigning 是否对 billing header 中的 cch 进行 xxHash64 签名（默认 false）
 	SettingKeyEnableCCHSigning = "enable_cch_signing"
+	// SettingKeyEnableAnthropicCacheTTL1hInjection 是否对 Anthropic OAuth/SetupToken 请求体注入 1h cache_control ttl（默认 false）
+	SettingKeyEnableAnthropicCacheTTL1hInjection = "enable_anthropic_cache_ttl_1h_injection"

 	// Balance Low Notification
 	SettingKeyBalanceLowNotifyEnabled     = "balance_low_notify_enabled"      // 全局开关
--- a/backend/internal/service/gateway_body_order_test.go
+++ b/backend/internal/service/gateway_body_order_test.go
@@ -1,13 +1,91 @@
 package service

 import (
+	"context"
+	"errors"
 	"strings"
 	"testing"

+	"github.com/Wei-Shaw/sub2api/internal/config"
 	"github.com/Wei-Shaw/sub2api/internal/pkg/claude"
 	"github.com/stretchr/testify/require"
+	"github.com/tidwall/gjson"
 )

+type gatewayTTLSettingRepo struct {
+	data map[string]string
+}
+
+func (r *gatewayTTLSettingRepo) Get(context.Context, string) (*Setting, error) {
+	return nil, ErrSettingNotFound
+}
+
+func (r *gatewayTTLSettingRepo) GetValue(_ context.Context, key string) (string, error) {
+	if r == nil {
+		return "", ErrSettingNotFound
+	}
+	v, ok := r.data[key]
+	if !ok {
+		return "", ErrSettingNotFound
+	}
+	return v, nil
+}
+
+func (r *gatewayTTLSettingRepo) Set(_ context.Context, key, value string) error {
+	if r == nil {
+		return errors.New("setting repo is nil")
+	}
+	if r.data == nil {
+		r.data = map[string]string{}
+	}
+	r.data[key] = value
+	return nil
+}
+
+func (r *gatewayTTLSettingRepo) GetMultiple(_ context.Context, keys []string) (map[string]string, error) {
+	result := make(map[string]string)
+	if r == nil {
+		return result, nil
+	}
+	for _, key := range keys {
+		if v, ok := r.data[key]; ok {
+			result[key] = v
+		}
+	}
+	return result, nil
+}
+
+func (r *gatewayTTLSettingRepo) SetMultiple(_ context.Context, settings map[string]string) error {
+	if r == nil {
+		return errors.New("setting repo is nil")
+	}
+	if r.data == nil {
+		r.data = map[string]string{}
+	}
+	for key, value := range settings {
+		r.data[key] = value
+	}
+	return nil
+}
+
+func (r *gatewayTTLSettingRepo) GetAll(context.Context) (map[string]string, error) {
+	result := make(map[string]string)
+	if r == nil {
+		return result, nil
+	}
+	for key, value := range r.data {
+		result[key] = value
+	}
+	return result, nil
+}
+
+func (r *gatewayTTLSettingRepo) Delete(_ context.Context, key string) error {
+	if r != nil {
+		delete(r.data, key)
+	}
+	return nil
+}
+
 func assertJSONTokenOrder(t *testing.T, body string, tokens ...string) {
 	t.Helper()

@@ -71,3 +149,60 @@ func TestEnforceCacheControlLimit_PreservesTopLevelFieldOrder(t *testing.T) {
 	assertJSONTokenOrder(t, resultStr, `"alpha"`, `"system"`, `"messages"`, `"omega"`)
 	require.Equal(t, 4, strings.Count(resultStr, `"cache_control"`))
 }
+
+func TestInjectAnthropicCacheControlTTL1h_OnlyUpdatesExistingEphemeralCacheControl(t *testing.T) {
+	body := []byte(`{"alpha":1,"cache_control":{"type":"ephemeral"},"system":[{"type":"text","text":"sys","cache_control":{"type":"ephemeral","ttl":"5m"}},{"type":"text","text":"plain"}],"messages":[{"role":"user","content":[{"type":"text","text":"hi","cache_control":{"type":"ephemeral"}},{"type":"text","text":"non","cache_control":{"type":"persistent","ttl":"5m"}}]}],"tools":[{"name":"a","input_schema":{},"cache_control":{"type":"ephemeral"}}],"omega":2}`)
+
+	result := injectAnthropicCacheControlTTL1h(body)
+	resultStr := string(result)
+
+	assertJSONTokenOrder(t, resultStr, `"alpha"`, `"cache_control"`, `"system"`, `"messages"`, `"tools"`, `"omega"`)
+	require.Equal(t, "1h", gjson.GetBytes(result, "cache_control.ttl").String())
+	require.Equal(t, "1h", gjson.GetBytes(result, "system.0.cache_control.ttl").String())
+	require.False(t, gjson.GetBytes(result, "system.1.cache_control").Exists())
+	require.Equal(t, "1h", gjson.GetBytes(result, "messages.0.content.0.cache_control.ttl").String())
+	require.Equal(t, "5m", gjson.GetBytes(result, "messages.0.content.1.cache_control.ttl").String())
+	require.Equal(t, "1h", gjson.GetBytes(result, "tools.0.cache_control.ttl").String())
+}
+
+func TestGatewayCacheTTLGlobalSetting_TargetResolution(t *testing.T) {
+	repo := &gatewayTTLSettingRepo{data: map[string]string{
+		SettingKeyEnableAnthropicCacheTTL1hInjection: "true",
+	}}
+	gatewayForwardingCache.Store(&cachedGatewayForwardingSettings{})
+	svc := &GatewayService{
+		settingService: NewSettingService(repo, &config.Config{}),
+	}
+	account := &Account{Platform: PlatformAnthropic, Type: AccountTypeOAuth}
+
+	target, ok := svc.resolveCacheTTLUsageOverrideTarget(context.Background(), account)
+	require.True(t, ok)
+	require.Equal(t, cacheTTLTarget5m, target)
+
+	account.Extra = map[string]any{
+		"cache_ttl_override_enabled": true,
+		"cache_ttl_override_target":  "1h",
+	}
+	target, ok = svc.resolveCacheTTLUsageOverrideTarget(context.Background(), account)
+	require.True(t, ok)
+	require.Equal(t, cacheTTLTarget1h, target)
+}
+
+func TestGatewayCacheTTLGlobalSetting_RequestInjectionScope(t *testing.T) {
+	repo := &gatewayTTLSettingRepo{data: map[string]string{
+		SettingKeyEnableAnthropicCacheTTL1hInjection: "true",
+	}}
+	gatewayForwardingCache.Store(&cachedGatewayForwardingSettings{})
+	svc := &GatewayService{
+		settingService: NewSettingService(repo, &config.Config{}),
+	}
+
+	require.True(t, svc.shouldInjectAnthropicCacheTTL1h(context.Background(), &Account{Platform: PlatformAnthropic, Type: AccountTypeOAuth}))
+	require.True(t, svc.shouldInjectAnthropicCacheTTL1h(context.Background(), &Account{Platform: PlatformAnthropic, Type: AccountTypeSetupToken}))
+	require.False(t, svc.shouldInjectAnthropicCacheTTL1h(context.Background(), &Account{Platform: PlatformAnthropic, Type: AccountTypeAPIKey}))
+	require.False(t, svc.shouldInjectAnthropicCacheTTL1h(context.Background(), &Account{Platform: PlatformOpenAI, Type: AccountTypeOAuth}))
+
+	repo.data[SettingKeyEnableAnthropicCacheTTL1hInjection] = "false"
+	gatewayForwardingCache.Store(&cachedGatewayForwardingSettings{})
+	require.False(t, svc.shouldInjectAnthropicCacheTTL1h(context.Background(), &Account{Platform: PlatformAnthropic, Type: AccountTypeOAuth}))
+}
--- a/backend/internal/service/gateway_service.go
+++ b/backend/internal/service/gateway_service.go
@@ -62,6 +62,11 @@ const (
 	claudeMimicDebugInfoKey = "claude_mimic_debug_info"
 )

+const (
+	cacheTTLTarget5m = "5m"
+	cacheTTLTarget1h = "1h"
+)
+
 // ForceCacheBillingContextKey 强制缓存计费上下文键
 // 用于粘性会话切换时，将 input_tokens 转为 cache_read_input_tokens 计费
 type forceCacheBillingKeyType struct{}
@@ -654,15 +659,31 @@ func (s *GatewayService) GenerateSessionHash(parsed *ParsedRequest) string {

 	// 1. 最高优先级：从 metadata.user_id 提取 session_xxx
 	if parsed.MetadataUserID != "" {
-		if uid := ParseMetadataUserID(parsed.MetadataUserID); uid != nil && uid.SessionID != "" {
+		uid := ParseMetadataUserID(parsed.MetadataUserID)
+		if uid != nil && uid.SessionID != "" {
+			slog.Info("sticky.hash_source",
+				"source", "metadata_user_id",
+				"session_id", uid.SessionID,
+				"device_id", uid.DeviceID,
+				"is_new_format", uid.IsNewFormat,
+			)
 			return uid.SessionID
 		}
+		slog.Info("sticky.hash_metadata_parse_failed",
+			"metadata_user_id", parsed.MetadataUserID,
+			"parsed_nil", uid == nil,
+		)
 	}

 	// 2. 提取带 cache_control: {type: "ephemeral"} 的内容
 	cacheableContent := s.extractCacheableContent(parsed)
 	if cacheableContent != "" {
-		return s.hashContent(cacheableContent)
+		hash := s.hashContent(cacheableContent)
+		slog.Info("sticky.hash_source",
+			"source", "cacheable_content",
+			"hash", hash,
+		)
+		return hash
 	}

 	// 3. 最后 fallback: 使用 session上下文 + system + 所有消息的完整摘要串
@@ -702,7 +723,13 @@ func (s *GatewayService) GenerateSessionHash(parsed *ParsedRequest) string {
 		}
 	}
 	if combined.Len() > 0 {
-		return s.hashContent(combined.String())
+		hash := s.hashContent(combined.String())
+		slog.Info("sticky.hash_source",
+			"source", "message_content_fallback",
+			"hash", hash,
+			"content_len", combined.Len(),
+		)
+		return hash
 	}

 	return ""
@@ -1406,14 +1433,29 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 	}

 	var stickyAccountID int64
+	var stickySource string
 	if prefetch := prefetchedStickyAccountIDFromContext(ctx, groupID); prefetch > 0 {
 		stickyAccountID = prefetch
+		stickySource = "prefetch"
 	} else if sessionHash != "" && s.cache != nil {
 		if accountID, err := s.cache.GetSessionAccountID(ctx, derefGroupID(groupID), sessionHash); err == nil {
 			stickyAccountID = accountID
+			stickySource = "cache"
 		}
 	}

+	// [DEBUG-STICKY] 调度器入口日志
+	slog.Info("sticky.scheduler_entry",
+		"group_id", derefGroupID(groupID),
+		"session_hash", shortSessionHash(sessionHash),
+		"sticky_account_id", stickyAccountID,
+		"sticky_source", stickySource,
+		"model", requestedModel,
+		"load_batch", cfg.LoadBatchEnabled,
+		"has_concurrency_svc", s.concurrencyService != nil,
+		"excluded_count", len(excludedIDs),
+	)
+
 	if s.debugModelRoutingEnabled() && requestedModel != "" {
 		groupPlatform := ""
 		if group != nil {
@@ -1589,6 +1631,13 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 		if len(routingCandidates) > 0 {
 			// 1.5. 在路由账号范围内检查粘性会话
 			if sessionHash != "" && stickyAccountID > 0 {
+				slog.Debug("sticky.layer1_5_checking",
+					"sticky_account_id", stickyAccountID,
+					"in_routing_list", containsInt64(routingAccountIDs, stickyAccountID),
+					"is_excluded", isExcluded(stickyAccountID),
+					"in_account_map", func() bool { _, ok := accountByID[stickyAccountID]; return ok }(),
+					"session", shortSessionHash(sessionHash),
+				)
 				if containsInt64(routingAccountIDs, stickyAccountID) && !isExcluded(stickyAccountID) {
 					// 粘性账号在路由列表中，优先使用
 					if stickyAccount, ok := accountByID[stickyAccountID]; ok {
@@ -1612,6 +1661,11 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 									stickyCacheMissReason = "session_limit"
 									// 继续到负载感知选择
 								} else {
+									slog.Debug("sticky.layer1_5_hit",
+										"account_id", stickyAccountID,
+										"session", shortSessionHash(sessionHash),
+										"result", "slot_acquired",
+									)
 									if s.debugModelRoutingEnabled() {
 										logger.LegacyPrintf("service.gateway", "[ModelRoutingDebug] routed sticky hit: group_id=%v model=%s session=%s account=%d", derefGroupID(groupID), requestedModel, shortSessionHash(sessionHash), stickyAccountID)
 									}
@@ -1762,27 +1816,65 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 				// 检查账户是否需要清理粘性会话绑定
 				clearSticky := shouldClearStickySession(account, requestedModel)
 				if clearSticky {
+					slog.Debug("sticky.layer1_5_no_routing_clear",
+						"account_id", accountID,
+						"reason", "should_clear_sticky_session",
+						"session", shortSessionHash(sessionHash),
+					)
 					_ = s.cache.DeleteSessionAccountID(ctx, derefGroupID(groupID), sessionHash)
 				}
-				if !clearSticky && s.isAccountInGroup(account, groupID) &&
-					s.isAccountAllowedForPlatform(account, platform, useMixed) &&
-					(requestedModel == "" || s.isModelSupportedByAccountWithContext(ctx, account, requestedModel)) &&
-					s.isAccountSchedulableForModelSelection(ctx, account, requestedModel) &&
-					s.isAccountSchedulableForQuota(account) &&
-					s.isAccountSchedulableForWindowCost(ctx, account, true) &&

-					s.isAccountSchedulableForRPM(ctx, account, true) { // 粘性会话窗口费用+RPM 检查
+				// 注意：不再检查 isAccountInGroup，因为 accountByID 已经从按分组过滤的
+				// accounts 列表构建，账号一定在分组内。而 scheduler snapshot 缓存
+				// 反序列化后 AccountGroups 字段为空，导致 isAccountInGroup 永远返回 false。
+				platformOK := s.isAccountAllowedForPlatform(account, platform, useMixed)
+				modelSupported := requestedModel == "" || s.isModelSupportedByAccountWithContext(ctx, account, requestedModel)
+				modelSchedulable := s.isAccountSchedulableForModelSelection(ctx, account, requestedModel)
+				quotaOK := s.isAccountSchedulableForQuota(account)
+				windowCostOK := s.isAccountSchedulableForWindowCost(ctx, account, true)
+				rpmOK := s.isAccountSchedulableForRPM(ctx, account, true)
+				schedulable := s.isAccountSchedulableForSelection(account)
+
+				slog.Debug("sticky.layer1_5_no_routing_checks",
+					"account_id", accountID,
+					"session", shortSessionHash(sessionHash),
+					"clear_sticky", clearSticky,
+					"schedulable", schedulable,
+					"platform_ok", platformOK,
+					"model_supported", modelSupported,
+					"model_schedulable", modelSchedulable,
+					"quota_ok", quotaOK,
+					"window_cost_ok", windowCostOK,
+					"rpm_ok", rpmOK,
+				)
+
+				if !clearSticky && platformOK && modelSupported && modelSchedulable && quotaOK && windowCostOK && rpmOK && schedulable {
 					result, err := s.tryAcquireAccountSlot(ctx, accountID, account.Concurrency)
 					if err == nil && result.Acquired {
 						// 会话数量限制检查
 						if !s.checkAndRegisterSession(ctx, account, sessionHash) {
 							result.ReleaseFunc() // 释放槽位，继续到 Layer 2
+							slog.Debug("sticky.layer1_5_no_routing_miss",
+								"account_id", accountID,
+								"reason", "session_limit",
+								"session", shortSessionHash(sessionHash),
+							)
 						} else {
+							slog.Debug("sticky.layer1_5_no_routing_hit",
+								"account_id", accountID,
+								"session", shortSessionHash(sessionHash),
+								"result", "slot_acquired",
+							)
 							if s.cache != nil {
 								_ = s.cache.RefreshSessionTTL(ctx, derefGroupID(groupID), sessionHash, stickySessionTTL)
 							}
 							return s.newSelectionResult(ctx, account, true, result.ReleaseFunc, nil)
 						}
+					} else {
+						slog.Debug("sticky.layer1_5_no_routing_slot_busy",
+							"account_id", accountID,
+							"session", shortSessionHash(sessionHash),
+						)
 					}

 					waitingCount, _ := s.concurrencyService.GetAccountWaitingCount(ctx, accountID)
@@ -1791,6 +1883,11 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 						if !s.checkAndRegisterSession(ctx, account, sessionHash) {
 							// 会话限制已满，继续到 Layer 2
 						} else {
+							slog.Debug("sticky.layer1_5_no_routing_hit",
+								"account_id", accountID,
+								"session", shortSessionHash(sessionHash),
+								"result", "wait_plan",
+							)
 							return s.newSelectionResult(ctx, account, false, nil, &AccountWaitPlan{
 								AccountID:      accountID,
 								MaxConcurrency: account.Concurrency,
@@ -1799,12 +1896,42 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
 							})
 						}
 					}
+				} else if !clearSticky {
+					slog.Debug("sticky.layer1_5_no_routing_miss",
+						"account_id", accountID,
+						"reason", "gate_check_failed",
+						"session", shortSessionHash(sessionHash),
+					)
 				}
+			} else {
+				slog.Debug("sticky.layer1_5_no_routing_miss",
+					"account_id", accountID,
+					"reason", "account_not_in_map",
+					"session", shortSessionHash(sessionHash),
+				)
 			}
 		}
+	} else if len(routingAccountIDs) == 0 && sessionHash != "" {
+		slog.Debug("sticky.layer1_5_no_routing_skip",
+			"sticky_account_id", stickyAccountID,
+			"is_excluded", func() bool { return stickyAccountID > 0 && isExcluded(stickyAccountID) }(),
+			"session", shortSessionHash(sessionHash),
+			"reason", func() string {
+				if stickyAccountID == 0 {
+					return "no_sticky_binding"
+				}
+				return "sticky_account_excluded"
+			}(),
+		)
 	}

 	// ============ Layer 2: 负载感知选择 ============
+	slog.Debug("sticky.layer2_fallback",
+		"session", shortSessionHash(sessionHash),
+		"sticky_account_id", stickyAccountID,
+		"reason", "sticky_not_used_falling_back_to_load_balance",
+		"total_accounts", len(accounts),
+	)
 	candidates := make([]*Account, 0, len(accounts))
 	for i := range accounts {
 		acc := &accounts[i]
@@ -4104,6 +4231,87 @@ func enforceCacheControlLimit(body []byte) []byte {
 	return body
 }

+// injectAnthropicCacheControlTTL1h 将已有 ephemeral cache_control 块的 ttl 强制写为 1h。
+// 仅修改已经存在的 cache_control，不新增缓存断点。
+func injectAnthropicCacheControlTTL1h(body []byte) []byte {
+	return forceEphemeralCacheControlTTL(body, cacheTTLTarget1h)
+}
+
+func forceEphemeralCacheControlTTL(body []byte, ttl string) []byte {
+	if len(body) == 0 || ttl == "" {
+		return body
+	}
+	out := body
+	var paths []string
+	addPath := func(path string, value gjson.Result) {
+		cc := value.Get("cache_control")
+		if !cc.Exists() || cc.Get("type").String() != "ephemeral" {
+			return
+		}
+		if cc.Get("ttl").String() == ttl {
+			return
+		}
+		paths = append(paths, path+".cache_control.ttl")
+	}
+
+	if topCC := gjson.GetBytes(body, "cache_control"); topCC.Exists() && topCC.Get("type").String() == "ephemeral" && topCC.Get("ttl").String() != ttl {
+		paths = append(paths, "cache_control.ttl")
+	}
+
+	system := gjson.GetBytes(body, "system")
+	if system.IsArray() {
+		idx := -1
+		system.ForEach(func(_, block gjson.Result) bool {
+			idx++
+			addPath(fmt.Sprintf("system.%d", idx), block)
+			return true
+		})
+	}
+
+	messages := gjson.GetBytes(body, "messages")
+	if messages.IsArray() {
+		msgIdx := -1
+		messages.ForEach(func(_, msg gjson.Result) bool {
+			msgIdx++
+			content := msg.Get("content")
+			if !content.IsArray() {
+				return true
+			}
+			contentIdx := -1
+			content.ForEach(func(_, block gjson.Result) bool {
+				contentIdx++
+				addPath(fmt.Sprintf("messages.%d.content.%d", msgIdx, contentIdx), block)
+				return true
+			})
+			return true
+		})
+	}
+
+	tools := gjson.GetBytes(body, "tools")
+	if tools.IsArray() {
+		idx := -1
+		tools.ForEach(func(_, tool gjson.Result) bool {
+			idx++
+			addPath(fmt.Sprintf("tools.%d", idx), tool)
+			return true
+		})
+	}
+
+	for _, path := range paths {
+		if next, err := sjson.SetBytes(out, path, ttl); err == nil {
+			out = next
+		}
+	}
+	return out
+}
+
+func (s *GatewayService) shouldInjectAnthropicCacheTTL1h(ctx context.Context, account *Account) bool {
+	if account == nil || !account.IsAnthropicOAuthOrSetupToken() || s == nil || s.settingService == nil {
+		return false
+	}
+	return s.settingService.IsAnthropicCacheTTL1hInjectionEnabled(ctx)
+}
+
 // Forward 转发请求到Claude API
 func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *Account, parsed *ParsedRequest) (*ForwardResult, error) {
 	startTime := time.Now()
@@ -4263,6 +4471,10 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A
 		logger.LegacyPrintf("service.gateway", "Model mapping applied: %s -> %s (account: %s, source=%s)", originalModel, mappedModel, account.Name, mappingSource)
 	}

+	if s.shouldInjectAnthropicCacheTTL1h(ctx, account) {
+		body = injectAnthropicCacheControlTTL1h(body)
+	}
+
 	// 获取凭证
 	token, tokenType, err := s.GetAccessToken(ctx, account)
 	if err != nil {
@@ -7103,9 +7315,9 @@ func (s *GatewayService) handleStreamingResponse(ctx context.Context, resp *http
 			}
 		}

-		// Cache TTL Override: 重写 SSE 事件中的 cache_creation 分类
-		if account.IsCacheTTLOverrideEnabled() {
-			overrideTarget := account.GetCacheTTLOverrideTarget()
+		// Cache TTL Override: 重写 SSE 事件中的 cache_creation 分类。
+		// 账号级设置优先；全局 1h 请求注入开启时，默认把 usage 计费归回 5m。
+		if overrideTarget, ok := s.resolveCacheTTLUsageOverrideTarget(ctx, account); ok {
 			if eventType == "message_start" {
 				if msg, ok := event["message"].(map[string]any); ok {
 					if u, ok := msg["usage"].(map[string]any); ok {
@@ -7512,6 +7724,19 @@ func rewriteCacheCreationJSON(usageObj map[string]any, target string) bool {
 	return true
 }

+func (s *GatewayService) resolveCacheTTLUsageOverrideTarget(ctx context.Context, account *Account) (string, bool) {
+	if account == nil {
+		return "", false
+	}
+	if account.IsCacheTTLOverrideEnabled() {
+		return account.GetCacheTTLOverrideTarget(), true
+	}
+	if account.IsAnthropicOAuthOrSetupToken() && s != nil && s.settingService != nil && s.settingService.IsAnthropicCacheTTL1hInjectionEnabled(ctx) {
+		return cacheTTLTarget5m, true
+	}
+	return "", false
+}
+
 func (s *GatewayService) handleNonStreamingResponse(ctx context.Context, resp *http.Response, c *gin.Context, account *Account, originalModel, mappedModel string) (*ClaudeUsage, error) {
 	// 更新5h窗口状态
 	s.rateLimitService.UpdateSessionWindow(ctx, account, resp.Header)
@@ -7548,9 +7773,9 @@ func (s *GatewayService) handleNonStreamingResponse(ctx context.Context, resp *h
 		}
 	}

-	// Cache TTL Override: 重写 non-streaming 响应中的 cache_creation 分类
-	if account.IsCacheTTLOverrideEnabled() {
-		overrideTarget := account.GetCacheTTLOverrideTarget()
+	// Cache TTL Override: 重写 non-streaming 响应中的 cache_creation 分类。
+	// 账号级设置优先；全局 1h 请求注入开启时，默认把 usage 计费归回 5m。
+	if overrideTarget, ok := s.resolveCacheTTLUsageOverrideTarget(ctx, account); ok {
 		if applyCacheTTLOverride(&response.Usage, overrideTarget) {
 			// 同步更新 body JSON 中的嵌套 cache_creation 对象
 			if newBody, err := sjson.SetBytes(body, "usage.cache_creation.ephemeral_5m_input_tokens", response.Usage.CacheCreation5mTokens); err == nil {
@@ -8118,10 +8343,11 @@ func (s *GatewayService) recordUsageCore(ctx context.Context, input *recordUsage
 		result.Usage.InputTokens = 0
 	}

-	// Cache TTL Override: 确保计费时 token 分类与账号设置一致
+	// Cache TTL Override: 确保计费时 token 分类与账号设置一致。
+	// 账号级设置优先；全局 1h 请求注入开启时，默认把 usage 计费归回 5m。
 	cacheTTLOverridden := false
-	if account.IsCacheTTLOverrideEnabled() {
-		applyCacheTTLOverride(&result.Usage, account.GetCacheTTLOverrideTarget())
+	if overrideTarget, ok := s.resolveCacheTTLUsageOverrideTarget(ctx, account); ok {
+		applyCacheTTLOverride(&result.Usage, overrideTarget)
 		cacheTTLOverridden = (result.Usage.CacheCreation5mTokens + result.Usage.CacheCreation1hTokens) > 0
 	}

--- a/backend/internal/service/openai_ws_forwarder.go
+++ b/backend/internal/service/openai_ws_forwarder.go
@@ -1379,10 +1379,12 @@ func shouldInferIngressFunctionCallOutputPreviousResponseID(
 	if signals.HasFunctionCallOutputMissingCallID {
 		return false
 	}
-	// If the client already sent tool-call context or item_reference anchors,
-	// treat this as a full replay / self-contained continuation payload rather
-	// than downgrading it into an inferred delta continuation.
-	if signals.HasToolCallContext || signals.HasItemReferenceForAllCallIDs {
+	// If the client already sent the actual tool-call context, treat this as
+	// a full replay / self-contained continuation payload rather than
+	// downgrading it into an inferred delta continuation. item_reference alone
+	// is not enough on the store=false WS path: it still needs a valid prior
+	// response anchor so upstream can resolve the referenced function_call.
+	if signals.HasToolCallContext {
 		return false
 	}
 	return strings.TrimSpace(expectedPreviousResponseID) != ""
--- a/backend/internal/service/openai_ws_forwarder_ingress_session_test.go
+++ b/backend/internal/service/openai_ws_forwarder_ingress_session_test.go
@@ -1488,7 +1488,7 @@ func TestOpenAIGatewayService_ProxyResponsesWebSocketFromClient_StoreDisabledFun
 	require.False(t, gjson.Get(requestToJSONString(captureConn.writes[1]), "previous_response_id").Exists(), "请求已包含 function_call 上下文时不应自动补齐 previous_response_id")
 }

-func TestOpenAIGatewayService_ProxyResponsesWebSocketFromClient_StoreDisabledFunctionCallOutputSkipsAutoAttachWhenItemReferencesPresent(t *testing.T) {
+func TestOpenAIGatewayService_ProxyResponsesWebSocketFromClient_StoreDisabledFunctionCallOutputAutoAttachWhenOnlyItemReferencesPresent(t *testing.T) {
 	gin.SetMode(gin.TestMode)

 	cfg := &config.Config{}
@@ -1619,7 +1619,7 @@ func TestOpenAIGatewayService_ProxyResponsesWebSocketFromClient_StoreDisabledFun

 	require.Equal(t, 1, captureDialer.DialCount())
 	require.Len(t, captureConn.writes, 2)
-	require.False(t, gjson.Get(requestToJSONString(captureConn.writes[1]), "previous_response_id").Exists(), "请求已包含 item_reference 锚点时不应自动补齐 previous_response_id")
+	require.Equal(t, "resp_auto_prev_ref_1", gjson.Get(requestToJSONString(captureConn.writes[1]), "previous_response_id").String(), "仅有 item_reference 不足以自包含 function_call_output，应回填上一轮响应 ID")
 }

 func TestOpenAIGatewayService_ProxyResponsesWebSocketFromClient_PreflightPingFailReconnectsBeforeTurn(t *testing.T) {
--- a/backend/internal/service/openai_ws_forwarder_ingress_test.go
+++ b/backend/internal/service/openai_ws_forwarder_ingress_test.go
@@ -303,12 +303,12 @@ func TestShouldInferIngressFunctionCallOutputPreviousResponseID(t *testing.T) {
 			want:             false,
 		},
 		{
-			name:             "skip_when_item_reference_already_covers_all_call_ids",
+			name:             "infer_when_only_item_reference_covers_call_ids",
 			storeDisabled:    true,
 			turn:             2,
 			signals:          ToolContinuationSignals{HasFunctionCallOutput: true, HasItemReferenceForAllCallIDs: true},
 			expectedPrevious: "resp_2",
-			want:             false,
+			want:             true,
 		},
 		{
 			name:             "skip_when_function_call_output_missing_call_id",
--- a/backend/internal/service/setting_service.go
+++ b/backend/internal/service/setting_service.go
@@ -82,10 +82,11 @@ const backendModeDBTimeout = 5 * time.Second

 // cachedGatewayForwardingSettings 缓存网关转发行为设置（进程内缓存，60s TTL）
 type cachedGatewayForwardingSettings struct {
-	fingerprintUnification bool
-	metadataPassthrough    bool
-	cchSigning             bool
-	expiresAt              int64 // unix nano
+	fingerprintUnification       bool
+	metadataPassthrough          bool
+	cchSigning                   bool
+	anthropicCacheTTL1hInjection bool
+	expiresAt                    int64 // unix nano
 }

 var gatewayForwardingCache atomic.Value // *cachedGatewayForwardingSettings
@@ -1245,6 +1246,7 @@ func (s *SettingService) buildSystemSettingsUpdates(ctx context.Context, setting
 	updates[SettingKeyEnableFingerprintUnification] = strconv.FormatBool(settings.EnableFingerprintUnification)
 	updates[SettingKeyEnableMetadataPassthrough] = strconv.FormatBool(settings.EnableMetadataPassthrough)
 	updates[SettingKeyEnableCCHSigning] = strconv.FormatBool(settings.EnableCCHSigning)
+	updates[SettingKeyEnableAnthropicCacheTTL1hInjection] = strconv.FormatBool(settings.EnableAnthropicCacheTTL1hInjection)
 	updates[SettingPaymentVisibleMethodAlipaySource] = settings.PaymentVisibleMethodAlipaySource
 	updates[SettingPaymentVisibleMethodWxpaySource] = settings.PaymentVisibleMethodWxpaySource
 	updates[SettingPaymentVisibleMethodAlipayEnabled] = strconv.FormatBool(settings.PaymentVisibleMethodAlipayEnabled)
@@ -1305,10 +1307,11 @@ func (s *SettingService) refreshCachedSettings(settings *SystemSettings) {
 	})
 	gatewayForwardingSF.Forget("gateway_forwarding")
 	gatewayForwardingCache.Store(&cachedGatewayForwardingSettings{
-		fingerprintUnification: settings.EnableFingerprintUnification,
-		metadataPassthrough:    settings.EnableMetadataPassthrough,
-		cchSigning:             settings.EnableCCHSigning,
-		expiresAt:              time.Now().Add(gatewayForwardingCacheTTL).UnixNano(),
+		fingerprintUnification:       settings.EnableFingerprintUnification,
+		metadataPassthrough:          settings.EnableMetadataPassthrough,
+		cchSigning:                   settings.EnableCCHSigning,
+		anthropicCacheTTL1hInjection: settings.EnableAnthropicCacheTTL1hInjection,
+		expiresAt:                    time.Now().Add(gatewayForwardingCacheTTL).UnixNano(),
 	})
 	openAIAdvancedSchedulerSettingSF.Forget(openAIAdvancedSchedulerSettingKey)
 	openAIAdvancedSchedulerSettingCache.Store(&cachedOpenAIAdvancedSchedulerSetting{
@@ -1415,22 +1418,30 @@ func (s *SettingService) IsBackendModeEnabled(ctx context.Context) bool {
 	return false
 }

-// GetGatewayForwardingSettings returns cached gateway forwarding settings.
-// Uses in-process atomic.Value cache with 60s TTL, zero-lock hot path.
-// Returns (fingerprintUnification, metadataPassthrough, cchSigning).
-func (s *SettingService) GetGatewayForwardingSettings(ctx context.Context) (fingerprintUnification, metadataPassthrough, cchSigning bool) {
+type gatewayForwardingSettingsResult struct {
+	fp, mp, cch, cacheTTL1h bool
+}
+
+func (s *SettingService) getGatewayForwardingSettingsCached(ctx context.Context) gatewayForwardingSettingsResult {
 	if cached, ok := gatewayForwardingCache.Load().(*cachedGatewayForwardingSettings); ok && cached != nil {
 		if time.Now().UnixNano() < cached.expiresAt {
-			return cached.fingerprintUnification, cached.metadataPassthrough, cached.cchSigning
+			return gatewayForwardingSettingsResult{
+				fp:         cached.fingerprintUnification,
+				mp:         cached.metadataPassthrough,
+				cch:        cached.cchSigning,
+				cacheTTL1h: cached.anthropicCacheTTL1hInjection,
+			}
 		}
 	}
-	type gwfResult struct {
-		fp, mp, cch bool
-	}
 	val, _, _ := gatewayForwardingSF.Do("gateway_forwarding", func() (any, error) {
 		if cached, ok := gatewayForwardingCache.Load().(*cachedGatewayForwardingSettings); ok && cached != nil {
 			if time.Now().UnixNano() < cached.expiresAt {
-				return gwfResult{cached.fingerprintUnification, cached.metadataPassthrough, cached.cchSigning}, nil
+				return gatewayForwardingSettingsResult{
+					fp:         cached.fingerprintUnification,
+					mp:         cached.metadataPassthrough,
+					cch:        cached.cchSigning,
+					cacheTTL1h: cached.anthropicCacheTTL1hInjection,
+				}, nil
 			}
 		}
 		dbCtx, cancel := context.WithTimeout(context.WithoutCancel(ctx), gatewayForwardingDBTimeout)
@@ -1439,16 +1450,18 @@ func (s *SettingService) GetGatewayForwardingSettings(ctx context.Context) (fing
 			SettingKeyEnableFingerprintUnification,
 			SettingKeyEnableMetadataPassthrough,
 			SettingKeyEnableCCHSigning,
+			SettingKeyEnableAnthropicCacheTTL1hInjection,
 		})
 		if err != nil {
 			slog.Warn("failed to get gateway forwarding settings", "error", err)
 			gatewayForwardingCache.Store(&cachedGatewayForwardingSettings{
-				fingerprintUnification: true,
-				metadataPassthrough:    false,
-				cchSigning:             false,
-				expiresAt:              time.Now().Add(gatewayForwardingErrorTTL).UnixNano(),
+				fingerprintUnification:       true,
+				metadataPassthrough:          false,
+				cchSigning:                   false,
+				anthropicCacheTTL1hInjection: false,
+				expiresAt:                    time.Now().Add(gatewayForwardingErrorTTL).UnixNano(),
 			})
-			return gwfResult{true, false, false}, nil
+			return gatewayForwardingSettingsResult{fp: true}, nil
 		}
 		fp := true
 		if v, ok := values[SettingKeyEnableFingerprintUnification]; ok && v != "" {
@@ -1456,18 +1469,33 @@ func (s *SettingService) GetGatewayForwardingSettings(ctx context.Context) (fing
 		}
 		mp := values[SettingKeyEnableMetadataPassthrough] == "true"
 		cch := values[SettingKeyEnableCCHSigning] == "true"
+		cacheTTL1h := values[SettingKeyEnableAnthropicCacheTTL1hInjection] == "true"
 		gatewayForwardingCache.Store(&cachedGatewayForwardingSettings{
-			fingerprintUnification: fp,
-			metadataPassthrough:    mp,
-			cchSigning:             cch,
-			expiresAt:              time.Now().Add(gatewayForwardingCacheTTL).UnixNano(),
+			fingerprintUnification:       fp,
+			metadataPassthrough:          mp,
+			cchSigning:                   cch,
+			anthropicCacheTTL1hInjection: cacheTTL1h,
+			expiresAt:                    time.Now().Add(gatewayForwardingCacheTTL).UnixNano(),
 		})
-		return gwfResult{fp, mp, cch}, nil
+		return gatewayForwardingSettingsResult{fp: fp, mp: mp, cch: cch, cacheTTL1h: cacheTTL1h}, nil
 	})
-	if r, ok := val.(gwfResult); ok {
-		return r.fp, r.mp, r.cch
+	if r, ok := val.(gatewayForwardingSettingsResult); ok {
+		return r
 	}
-	return true, false, false // fail-open defaults
+	return gatewayForwardingSettingsResult{fp: true}
+}
+
+// GetGatewayForwardingSettings returns cached gateway forwarding settings.
+// Uses in-process atomic.Value cache with 60s TTL, zero-lock hot path.
+// Returns (fingerprintUnification, metadataPassthrough, cchSigning).
+func (s *SettingService) GetGatewayForwardingSettings(ctx context.Context) (fingerprintUnification, metadataPassthrough, cchSigning bool) {
+	result := s.getGatewayForwardingSettingsCached(ctx)
+	return result.fp, result.mp, result.cch
+}
+
+// IsAnthropicCacheTTL1hInjectionEnabled 检查是否对 Anthropic OAuth/SetupToken 请求体注入 1h cache_control ttl。
+func (s *SettingService) IsAnthropicCacheTTL1hInjectionEnabled(ctx context.Context) bool {
+	return s.getGatewayForwardingSettingsCached(ctx).cacheTTL1h
 }

 // IsEmailVerifyEnabled 检查是否开启邮件验证
@@ -1880,12 +1908,13 @@ func (s *SettingService) InitializeDefaultSettings(ctx context.Context) error {
 		SettingKeyMaxClaudeCodeVersion: "",

 		// 分组隔离（默认不允许未分组 Key 调度）
-		SettingKeyAllowUngroupedKeyScheduling:    "false",
-		SettingPaymentVisibleMethodAlipaySource:  "",
-		SettingPaymentVisibleMethodWxpaySource:   "",
-		SettingPaymentVisibleMethodAlipayEnabled: "false",
-		SettingPaymentVisibleMethodWxpayEnabled:  "false",
-		openAIAdvancedSchedulerSettingKey:        "false",
+		SettingKeyAllowUngroupedKeyScheduling:        "false",
+		SettingKeyEnableAnthropicCacheTTL1hInjection: "false",
+		SettingPaymentVisibleMethodAlipaySource:      "",
+		SettingPaymentVisibleMethodWxpaySource:       "",
+		SettingPaymentVisibleMethodAlipayEnabled:     "false",
+		SettingPaymentVisibleMethodWxpayEnabled:      "false",
+		openAIAdvancedSchedulerSettingKey:            "false",
 	}

 	return s.settingRepo.SetMultiple(ctx, defaults)
@@ -2228,6 +2257,7 @@ func (s *SettingService) parseSettings(settings map[string]string) *SystemSettin
 	}
 	result.EnableMetadataPassthrough = settings[SettingKeyEnableMetadataPassthrough] == "true"
 	result.EnableCCHSigning = settings[SettingKeyEnableCCHSigning] == "true"
+	result.EnableAnthropicCacheTTL1hInjection = settings[SettingKeyEnableAnthropicCacheTTL1hInjection] == "true"

 	// Web search emulation: quick enabled check from the JSON config
 	if raw := settings[SettingKeyWebSearchEmulationConfig]; raw != "" {
--- a/backend/internal/service/settings_view.go
+++ b/backend/internal/service/settings_view.go
@@ -149,9 +149,10 @@ type SystemSettings struct {
 	BackendModeEnabled bool

 	// Gateway forwarding behavior
-	EnableFingerprintUnification bool // 是否统一 OAuth 账号的指纹头（默认 true）
-	EnableMetadataPassthrough    bool // 是否透传客户端原始 metadata（默认 false）
-	EnableCCHSigning             bool // 是否对 billing header cch 进行签名（默认 false）
+	EnableFingerprintUnification       bool // 是否统一 OAuth 账号的指纹头（默认 true）
+	EnableMetadataPassthrough          bool // 是否透传客户端原始 metadata（默认 false）
+	EnableCCHSigning                   bool // 是否对 billing header cch 进行签名（默认 false）
+	EnableAnthropicCacheTTL1hInjection bool // 是否对 Anthropic OAuth/SetupToken 请求体注入 1h cache_control ttl（默认 false）

 	// Web Search Emulation
 	WebSearchEmulationEnabled bool // 是否启用 web search 模拟
--- a/frontend/src/api/admin/settings.ts
+++ b/frontend/src/api/admin/settings.ts
@@ -439,6 +439,7 @@ export interface SystemSettings {
  enable_fingerprint_unification: boolean;
  enable_metadata_passthrough: boolean;
  enable_cch_signing: boolean;
+  enable_anthropic_cache_ttl_1h_injection: boolean;
  web_search_emulation_enabled?: boolean;

  // Payment configuration
@@ -609,6 +610,7 @@ export interface UpdateSettingsRequest {
  enable_fingerprint_unification?: boolean;
  enable_metadata_passthrough?: boolean;
  enable_cch_signing?: boolean;
+  enable_anthropic_cache_ttl_1h_injection?: boolean;
  // Payment configuration
  payment_enabled?: boolean;
  payment_min_amount?: number;
--- a/frontend/src/components/common/Pagination.vue
+++ b/frontend/src/components/common/Pagination.vue
@@ -123,6 +123,7 @@ import { useI18n } from 'vue-i18n'
 import Icon from '@/components/icons/Icon.vue'
 import Select from './Select.vue'
 import { getConfiguredTablePageSizeOptions, normalizeTablePageSize } from '@/utils/tablePreferences'
+import { setPersistedPageSize } from '@/composables/usePersistedPageSize'

 const { t } = useI18n()

@@ -224,6 +225,7 @@ const goToPage = (newPage: number) => {
 const handlePageSizeChange = (value: string | number | boolean | null) => {
  if (value === null || typeof value === 'boolean') return
  const newPageSize = normalizeTablePageSize(typeof value === 'string' ? parseInt(value, 10) : value)
+  setPersistedPageSize(newPageSize)
  emit('update:pageSize', newPageSize)
 }

--- a/frontend/src/composables/usePersistedPageSize.ts
+++ b/frontend/src/composables/usePersistedPageSize.ts
@@ -1,9 +1,29 @@
 import { getConfiguredTableDefaultPageSize, normalizeTablePageSize } from '@/utils/tablePreferences'

-/**
- * 读取当前系统配置的表格默认每页条数。
- * 不再使用本地持久化缓存，所有页面统一以通用表格设置为准。
- */
+const STORAGE_KEY = 'table-page-size'
+
 export function getPersistedPageSize(fallback = getConfiguredTableDefaultPageSize()): number {
+  if (typeof window !== 'undefined') {
+    try {
+      const stored = window.localStorage.getItem(STORAGE_KEY)
+      if (stored !== null) {
+        const parsed = Number(stored)
+        if (Number.isFinite(parsed)) {
+          return normalizeTablePageSize(parsed)
+        }
+      }
+    } catch (error) {
+      console.warn('Failed to read persisted page size:', error)
+    }
+  }
  return normalizeTablePageSize(getConfiguredTableDefaultPageSize() || fallback)
 }
+
+export function setPersistedPageSize(size: number): void {
+  if (typeof window === 'undefined') return
+  try {
+    window.localStorage.setItem(STORAGE_KEY, String(size))
+  } catch (error) {
+    console.warn('Failed to persist page size:', error)
+  }
+}
--- a/frontend/src/composables/useTableLoader.ts
+++ b/frontend/src/composables/useTableLoader.ts
@@ -1,7 +1,7 @@
 import { ref, reactive, onUnmounted, toRaw } from 'vue'
 import { useDebounceFn } from '@vueuse/core'
 import type { BasePaginationResponse, FetchOptions } from '@/types'
-import { getPersistedPageSize } from './usePersistedPageSize'
+import { getPersistedPageSize, setPersistedPageSize } from './usePersistedPageSize'

 interface PaginationState {
  page: number
@@ -88,6 +88,7 @@ export function useTableLoader<T, P extends Record<string, any>>(options: TableL
  const handlePageSizeChange = (size: number) => {
    pagination.page_size = size
    pagination.page = 1
+    setPersistedPageSize(size)
    load()
  }

--- a/frontend/src/i18n/locales/en.ts
+++ b/frontend/src/i18n/locales/en.ts
@@ -5019,6 +5019,8 @@ export default {
        metadataPassthroughHint: 'Pass through client\'s original metadata.user_id without rewriting. May improve upstream cache hit rates.',
        cchSigning: 'CCH Signing',
        cchSigningHint: 'Sign the billing header in forwarded requests with CCH hash. When disabled, the placeholder is preserved.',
+        anthropicCacheTTL1hInjection: 'Anthropic Cache TTL Injection',
+        anthropicCacheTTL1hInjectionHint: 'When enabled, existing ephemeral cache_control blocks in Anthropic OAuth/Setup Token request bodies are forced to 1h; response usage is billed back as 5m by default, with account-level TTL billing override taking priority.',
      },
      webSearchEmulation: {
        title: 'Web Search Emulation',
--- a/frontend/src/i18n/locales/zh.ts
+++ b/frontend/src/i18n/locales/zh.ts
@@ -5178,6 +5178,8 @@ export default {
        metadataPassthroughHint: '透传客户端原始 metadata.user_id，不进行重写。可能提高上游缓存命中率。',
        cchSigning: 'CCH 签名',
        cchSigningHint: '对转发请求的 billing header 进行 CCH 哈希签名。关闭时保留原始占位符。',
+        anthropicCacheTTL1hInjection: 'Anthropic 缓存 TTL 注入',
+        anthropicCacheTTL1hInjectionHint: '开启后，对 Anthropic OAuth/Setup Token 请求体中已有的 ephemeral 缓存块强制写入 1h；响应 usage 默认按 5m 回写计费，账号级 TTL 计费设置优先。',
      },
      webSearchEmulation: {
        title: 'Web Search 模拟',
--- a/frontend/src/views/admin/SettingsView.vue
+++ b/frontend/src/views/admin/SettingsView.vue
@@ -3057,6 +3057,31 @@
                </div>
                <Toggle v-model="form.enable_cch_signing" />
              </div>
+
+              <!-- Anthropic Cache TTL 1h Injection -->
+              <div class="flex items-center justify-between">
+                <div>
+                  <label
+                    class="text-sm font-medium text-gray-700 dark:text-gray-300"
+                  >
+                    {{
+                      t(
+                        "admin.settings.gatewayForwarding.anthropicCacheTTL1hInjection",
+                      )
+                    }}
+                  </label>
+                  <p class="mt-0.5 text-xs text-gray-500 dark:text-gray-400">
+                    {{
+                      t(
+                        "admin.settings.gatewayForwarding.anthropicCacheTTL1hInjectionHint",
+                      )
+                    }}
+                  </p>
+                </div>
+                <Toggle
+                  v-model="form.enable_anthropic_cache_ttl_1h_injection"
+                />
+              </div>
            </div>
          </div>
          <!-- Web Search Emulation -->
@@ -5810,6 +5835,7 @@ const form = reactive<SettingsForm>({
  enable_fingerprint_unification: true,
  enable_metadata_passthrough: false,
  enable_cch_signing: false,
+  enable_anthropic_cache_ttl_1h_injection: false,
  // Balance & quota notification
  balance_low_notify_enabled: false,
  balance_low_notify_threshold: 0,
@@ -6718,6 +6744,8 @@ async function saveSettings() {
      enable_fingerprint_unification: form.enable_fingerprint_unification,
      enable_metadata_passthrough: form.enable_metadata_passthrough,
      enable_cch_signing: form.enable_cch_signing,
+      enable_anthropic_cache_ttl_1h_injection:
+        form.enable_anthropic_cache_ttl_1h_injection,
      // Payment configuration
      payment_enabled: form.payment_enabled,
      payment_min_amount: Number(form.payment_min_amount) || 0,
--- a/frontend/src/views/admin/tests/SettingsView.spec.ts
+++ b/frontend/src/views/admin/tests/SettingsView.spec.ts
@@ -362,6 +362,7 @@ const baseSettingsResponse = {
  enable_fingerprint_unification: true,
  enable_metadata_passthrough: false,
  enable_cch_signing: false,
+  enable_anthropic_cache_ttl_1h_injection: false,
  payment_enabled: true,
  payment_min_amount: 1,
  payment_max_amount: 10000,
@@ -567,6 +568,26 @@ describe("admin SettingsView payment visible method controls", () => {
    expect(payload).not.toHaveProperty("payment_visible_method_wxpay_enabled");
  });

+  it("submits Anthropic cache TTL injection gateway setting", async () => {
+    getSettings.mockResolvedValueOnce({
+      ...baseSettingsResponse,
+      enable_anthropic_cache_ttl_1h_injection: true,
+    });
+
+    const wrapper = mountView();
+
+    await flushPromises();
+    await wrapper.find("form").trigger("submit.prevent");
+    await flushPromises();
+
+    expect(updateSettings).toHaveBeenCalledTimes(1);
+    expect(updateSettings).toHaveBeenCalledWith(
+      expect.objectContaining({
+        enable_anthropic_cache_ttl_1h_injection: true,
+      }),
+    );
+  });
+
  it("updates provider enablement immediately and reloads providers", async () => {
    const provider = {
      id: 7,
Author	SHA1	Message	Date
shaw	9d801595c9	test: 更新管理员设置契约字段	2026-04-30 13:48:27 +08:00
Wesley Liddick	9c448f89a8	Merge pull request #2118 from DaydreamCoding/fix/restore-pagination-localStorage fix: 恢复表格分页大小 localStorage 持久化	2026-04-30 13:42:18 +08:00
shaw	73b872998e	feat: 添加 Anthropic 缓存 TTL 注入开关	2026-04-30 13:38:22 +08:00
shaw	094e1171ef	fix(openai): infer previous response for item references	2026-04-30 12:02:08 +08:00
shaw	733627cf9d	fix: improve sticky session scheduling	2026-04-30 11:38:11 +08:00
DaydreamCoding	f084d30d65	fix: 恢复表格分页大小 localStorage 持久化 - usePersistedPageSize: 恢复 localStorage 读写，以系统配置为 fallback - useTableLoader: handlePageSizeChange 时写入 localStorage - Pagination.vue: handlePageSizeChange 时写入 localStorage Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>	2026-04-30 10:35:15 +08:00
github-actions[bot]	8ad099baa6	chore: sync VERSION to 0.1.120 [skip ci]	2026-04-29 15:08:59 +00:00
@@ -1 +1 @@
 .1.119
 .1.120