Merge branch 'Wei-Shaw:main' into main

backend/internal/handler/admin/account_handler.go

@@ -84,6 +84,7 @@ type CreateAccountRequest struct {
 	ProxyID                 *int64         `json:"proxy_id"`
 	Concurrency             int            `json:"concurrency"`
 	Priority                int            `json:"priority"`
+	RateMultiplier          *float64       `json:"rate_multiplier"`
 	GroupIDs                []int64        `json:"group_ids"`
 	ExpiresAt               *int64         `json:"expires_at"`
 	AutoPauseOnExpired      *bool          `json:"auto_pause_on_expired"`
@@ -101,6 +102,7 @@ type UpdateAccountRequest struct {
 	ProxyID                 *int64         `json:"proxy_id"`
 	Concurrency             *int           `json:"concurrency"`
 	Priority                *int           `json:"priority"`
+	RateMultiplier          *float64       `json:"rate_multiplier"`
 	Status                  string         `json:"status" binding:"omitempty,oneof=active inactive"`
 	GroupIDs                *[]int64       `json:"group_ids"`
 	ExpiresAt               *int64         `json:"expires_at"`
@@ -115,6 +117,7 @@ type BulkUpdateAccountsRequest struct {
 	ProxyID                 *int64         `json:"proxy_id"`
 	Concurrency             *int           `json:"concurrency"`
 	Priority                *int           `json:"priority"`
+	RateMultiplier          *float64       `json:"rate_multiplier"`
 	Status                  string         `json:"status" binding:"omitempty,oneof=active inactive error"`
 	Schedulable             *bool          `json:"schedulable"`
 	GroupIDs                *[]int64       `json:"group_ids"`
@@ -199,6 +202,10 @@ func (h *AccountHandler) Create(c *gin.Context) {
 		response.BadRequest(c, "Invalid request: "+err.Error())
 		return
 	}
+	if req.RateMultiplier != nil && *req.RateMultiplier < 0 {
+		response.BadRequest(c, "rate_multiplier must be >= 0")
+		return
+	}
 
 	// 确定是否跳过混合渠道检查
 	skipCheck := req.ConfirmMixedChannelRisk != nil && *req.ConfirmMixedChannelRisk
@@ -213,6 +220,7 @@ func (h *AccountHandler) Create(c *gin.Context) {
 		ProxyID:               req.ProxyID,
 		Concurrency:           req.Concurrency,
 		Priority:              req.Priority,
+		RateMultiplier:        req.RateMultiplier,
 		GroupIDs:              req.GroupIDs,
 		ExpiresAt:             req.ExpiresAt,
 		AutoPauseOnExpired:    req.AutoPauseOnExpired,
@@ -258,6 +266,10 @@ func (h *AccountHandler) Update(c *gin.Context) {
 		response.BadRequest(c, "Invalid request: "+err.Error())
 		return
 	}
+	if req.RateMultiplier != nil && *req.RateMultiplier < 0 {
+		response.BadRequest(c, "rate_multiplier must be >= 0")
+		return
+	}
 
 	// 确定是否跳过混合渠道检查
 	skipCheck := req.ConfirmMixedChannelRisk != nil && *req.ConfirmMixedChannelRisk
@@ -271,6 +283,7 @@ func (h *AccountHandler) Update(c *gin.Context) {
 		ProxyID:               req.ProxyID,
 		Concurrency:           req.Concurrency, // 指针类型，nil 表示未提供
 		Priority:              req.Priority,    // 指针类型，nil 表示未提供
+		RateMultiplier:        req.RateMultiplier,
 		Status:                req.Status,
 		GroupIDs:              req.GroupIDs,
 		ExpiresAt:             req.ExpiresAt,
@@ -652,6 +665,10 @@ func (h *AccountHandler) BulkUpdate(c *gin.Context) {
 		response.BadRequest(c, "Invalid request: "+err.Error())
 		return
 	}
+	if req.RateMultiplier != nil && *req.RateMultiplier < 0 {
+		response.BadRequest(c, "rate_multiplier must be >= 0")
+		return
+	}
 
 	// 确定是否跳过混合渠道检查
 	skipCheck := req.ConfirmMixedChannelRisk != nil && *req.ConfirmMixedChannelRisk
@@ -660,6 +677,7 @@ func (h *AccountHandler) BulkUpdate(c *gin.Context) {
 		req.ProxyID != nil ||
 		req.Concurrency != nil ||
 		req.Priority != nil ||
+		req.RateMultiplier != nil ||
 		req.Status != "" ||
 		req.Schedulable != nil ||
 		req.GroupIDs != nil ||
@@ -677,6 +695,7 @@ func (h *AccountHandler) BulkUpdate(c *gin.Context) {
 		ProxyID:               req.ProxyID,
 		Concurrency:           req.Concurrency,
 		Priority:              req.Priority,
+		RateMultiplier:        req.RateMultiplier,
 		Status:                req.Status,
 		Schedulable:           req.Schedulable,
 		GroupIDs:              req.GroupIDs,

backend/internal/handler/admin/dashboard_handler.go

@@ -186,13 +186,16 @@ func (h *DashboardHandler) GetRealtimeMetrics(c *gin.Context) {
 
 // GetUsageTrend handles getting usage trend data
 // GET /api/v1/admin/dashboard/trend
-// Query params: start_date, end_date (YYYY-MM-DD), granularity (day/hour), user_id, api_key_id
+// Query params: start_date, end_date (YYYY-MM-DD), granularity (day/hour), user_id, api_key_id, model, account_id, group_id, stream
 func (h *DashboardHandler) GetUsageTrend(c *gin.Context) {
 	startTime, endTime := parseTimeRange(c)
 	granularity := c.DefaultQuery("granularity", "day")
 
 	// Parse optional filter params
-	var userID, apiKeyID int64
+	var userID, apiKeyID, accountID, groupID int64
+	var model string
+	var stream *bool
+
 	if userIDStr := c.Query("user_id"); userIDStr != "" {
 		if id, err := strconv.ParseInt(userIDStr, 10, 64); err == nil {
 			userID = id
@@ -203,8 +206,26 @@ func (h *DashboardHandler) GetUsageTrend(c *gin.Context) {
 			apiKeyID = id
 		}
 	}
+	if accountIDStr := c.Query("account_id"); accountIDStr != "" {
+		if id, err := strconv.ParseInt(accountIDStr, 10, 64); err == nil {
+			accountID = id
+		}
+	}
+	if groupIDStr := c.Query("group_id"); groupIDStr != "" {
+		if id, err := strconv.ParseInt(groupIDStr, 10, 64); err == nil {
+			groupID = id
+		}
+	}
+	if modelStr := c.Query("model"); modelStr != "" {
+		model = modelStr
+	}
+	if streamStr := c.Query("stream"); streamStr != "" {
+		if streamVal, err := strconv.ParseBool(streamStr); err == nil {
+			stream = &streamVal
+		}
+	}
 
-	trend, err := h.dashboardService.GetUsageTrendWithFilters(c.Request.Context(), startTime, endTime, granularity, userID, apiKeyID)
+	trend, err := h.dashboardService.GetUsageTrendWithFilters(c.Request.Context(), startTime, endTime, granularity, userID, apiKeyID, accountID, groupID, model, stream)
 	if err != nil {
 		response.Error(c, 500, "Failed to get usage trend")
 		return
@@ -220,12 +241,14 @@ func (h *DashboardHandler) GetUsageTrend(c *gin.Context) {
 
 // GetModelStats handles getting model usage statistics
 // GET /api/v1/admin/dashboard/models
-// Query params: start_date, end_date (YYYY-MM-DD), user_id, api_key_id
+// Query params: start_date, end_date (YYYY-MM-DD), user_id, api_key_id, account_id, group_id, stream
 func (h *DashboardHandler) GetModelStats(c *gin.Context) {
 	startTime, endTime := parseTimeRange(c)
 
 	// Parse optional filter params
-	var userID, apiKeyID int64
+	var userID, apiKeyID, accountID, groupID int64
+	var stream *bool
+
 	if userIDStr := c.Query("user_id"); userIDStr != "" {
 		if id, err := strconv.ParseInt(userIDStr, 10, 64); err == nil {
 			userID = id
@@ -236,8 +259,23 @@ func (h *DashboardHandler) GetModelStats(c *gin.Context) {
 			apiKeyID = id
 		}
 	}
+	if accountIDStr := c.Query("account_id"); accountIDStr != "" {
+		if id, err := strconv.ParseInt(accountIDStr, 10, 64); err == nil {
+			accountID = id
+		}
+	}
+	if groupIDStr := c.Query("group_id"); groupIDStr != "" {
+		if id, err := strconv.ParseInt(groupIDStr, 10, 64); err == nil {
+			groupID = id
+		}
+	}
+	if streamStr := c.Query("stream"); streamStr != "" {
+		if streamVal, err := strconv.ParseBool(streamStr); err == nil {
+			stream = &streamVal
+		}
+	}
 
-	stats, err := h.dashboardService.GetModelStatsWithFilters(c.Request.Context(), startTime, endTime, userID, apiKeyID)
+	stats, err := h.dashboardService.GetModelStatsWithFilters(c.Request.Context(), startTime, endTime, userID, apiKeyID, accountID, groupID, stream)
 	if err != nil {
 		response.Error(c, 500, "Failed to get model statistics")
 		return

backend/internal/handler/admin/ops_alerts_handler.go

@@ -7,8 +7,10 @@ import (
 	"net/http"
 	"strconv"
 	"strings"
+	"time"
 
 	"github.com/Wei-Shaw/sub2api/internal/pkg/response"
+	"github.com/Wei-Shaw/sub2api/internal/server/middleware"
 	"github.com/Wei-Shaw/sub2api/internal/service"
 	"github.com/gin-gonic/gin"
 	"github.com/gin-gonic/gin/binding"
@@ -18,8 +20,6 @@ var validOpsAlertMetricTypes = []string{
 	"success_rate",
 	"error_rate",
 	"upstream_error_rate",
-	"p95_latency_ms",
-	"p99_latency_ms",
 	"cpu_usage_percent",
 	"memory_usage_percent",
 	"concurrency_queue_depth",
@@ -372,8 +372,135 @@ func (h *OpsHandler) DeleteAlertRule(c *gin.Context) {
 	response.Success(c, gin.H{"deleted": true})
 }
 
+// GetAlertEvent returns a single ops alert event.
+// GET /api/v1/admin/ops/alert-events/:id
+func (h *OpsHandler) GetAlertEvent(c *gin.Context) {
+	if h.opsService == nil {
+		response.Error(c, http.StatusServiceUnavailable, "Ops service not available")
+		return
+	}
+	if err := h.opsService.RequireMonitoringEnabled(c.Request.Context()); err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+
+	id, err := strconv.ParseInt(c.Param("id"), 10, 64)
+	if err != nil || id <= 0 {
+		response.BadRequest(c, "Invalid event ID")
+		return
+	}
+
+	ev, err := h.opsService.GetAlertEventByID(c.Request.Context(), id)
+	if err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+	response.Success(c, ev)
+}
+
+// UpdateAlertEventStatus updates an ops alert event status.
+// PUT /api/v1/admin/ops/alert-events/:id/status
+func (h *OpsHandler) UpdateAlertEventStatus(c *gin.Context) {
+	if h.opsService == nil {
+		response.Error(c, http.StatusServiceUnavailable, "Ops service not available")
+		return
+	}
+	if err := h.opsService.RequireMonitoringEnabled(c.Request.Context()); err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+
+	id, err := strconv.ParseInt(c.Param("id"), 10, 64)
+	if err != nil || id <= 0 {
+		response.BadRequest(c, "Invalid event ID")
+		return
+	}
+
+	var payload struct {
+		Status string `json:"status"`
+	}
+	if err := c.ShouldBindJSON(&payload); err != nil {
+		response.BadRequest(c, "Invalid request body")
+		return
+	}
+	payload.Status = strings.TrimSpace(payload.Status)
+	if payload.Status == "" {
+		response.BadRequest(c, "Invalid status")
+		return
+	}
+	if payload.Status != service.OpsAlertStatusResolved && payload.Status != service.OpsAlertStatusManualResolved {
+		response.BadRequest(c, "Invalid status")
+		return
+	}
+
+	var resolvedAt *time.Time
+	if payload.Status == service.OpsAlertStatusResolved || payload.Status == service.OpsAlertStatusManualResolved {
+		now := time.Now().UTC()
+		resolvedAt = &now
+	}
+	if err := h.opsService.UpdateAlertEventStatus(c.Request.Context(), id, payload.Status, resolvedAt); err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+	response.Success(c, gin.H{"updated": true})
+}
+
 // ListAlertEvents lists recent ops alert events.
 // GET /api/v1/admin/ops/alert-events
+// CreateAlertSilence creates a scoped silence for ops alerts.
+// POST /api/v1/admin/ops/alert-silences
+func (h *OpsHandler) CreateAlertSilence(c *gin.Context) {
+	if h.opsService == nil {
+		response.Error(c, http.StatusServiceUnavailable, "Ops service not available")
+		return
+	}
+	if err := h.opsService.RequireMonitoringEnabled(c.Request.Context()); err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+
+	var payload struct {
+		RuleID   int64   `json:"rule_id"`
+		Platform string  `json:"platform"`
+		GroupID  *int64  `json:"group_id"`
+		Region   *string `json:"region"`
+		Until    string  `json:"until"`
+		Reason   string  `json:"reason"`
+	}
+	if err := c.ShouldBindJSON(&payload); err != nil {
+		response.BadRequest(c, "Invalid request body")
+		return
+	}
+	until, err := time.Parse(time.RFC3339, strings.TrimSpace(payload.Until))
+	if err != nil {
+		response.BadRequest(c, "Invalid until")
+		return
+	}
+
+	createdBy := (*int64)(nil)
+	if subject, ok := middleware.GetAuthSubjectFromContext(c); ok {
+		uid := subject.UserID
+		createdBy = &uid
+	}
+
+	silence := &service.OpsAlertSilence{
+		RuleID:    payload.RuleID,
+		Platform:  strings.TrimSpace(payload.Platform),
+		GroupID:   payload.GroupID,
+		Region:    payload.Region,
+		Until:     until,
+		Reason:    strings.TrimSpace(payload.Reason),
+		CreatedBy: createdBy,
+	}
+
+	created, err := h.opsService.CreateAlertSilence(c.Request.Context(), silence)
+	if err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+	response.Success(c, created)
+}
+
 func (h *OpsHandler) ListAlertEvents(c *gin.Context) {
 	if h.opsService == nil {
 		response.Error(c, http.StatusServiceUnavailable, "Ops service not available")
@@ -384,7 +511,7 @@ func (h *OpsHandler) ListAlertEvents(c *gin.Context) {
 		return
 	}
 
-	limit := 100
+	limit := 20
 	if raw := strings.TrimSpace(c.Query("limit")); raw != "" {
 		n, err := strconv.Atoi(raw)
 		if err != nil || n <= 0 {
@@ -400,6 +527,49 @@ func (h *OpsHandler) ListAlertEvents(c *gin.Context) {
 		Severity: strings.TrimSpace(c.Query("severity")),
 	}
 
+	if v := strings.TrimSpace(c.Query("email_sent")); v != "" {
+		vv := strings.ToLower(v)
+		switch vv {
+		case "true", "1":
+			b := true
+			filter.EmailSent = &b
+		case "false", "0":
+			b := false
+			filter.EmailSent = &b
+		default:
+			response.BadRequest(c, "Invalid email_sent")
+			return
+		}
+	}
+
+	// Cursor pagination: both params must be provided together.
+	rawTS := strings.TrimSpace(c.Query("before_fired_at"))
+	rawID := strings.TrimSpace(c.Query("before_id"))
+	if (rawTS == "") != (rawID == "") {
+		response.BadRequest(c, "before_fired_at and before_id must be provided together")
+		return
+	}
+	if rawTS != "" {
+		ts, err := time.Parse(time.RFC3339Nano, rawTS)
+		if err != nil {
+			if t2, err2 := time.Parse(time.RFC3339, rawTS); err2 == nil {
+				ts = t2
+			} else {
+				response.BadRequest(c, "Invalid before_fired_at")
+				return
+			}
+		}
+		filter.BeforeFiredAt = &ts
+	}
+	if rawID != "" {
+		id, err := strconv.ParseInt(rawID, 10, 64)
+		if err != nil || id <= 0 {
+			response.BadRequest(c, "Invalid before_id")
+			return
+		}
+		filter.BeforeID = &id
+	}
+
 	// Optional global filter support (platform/group/time range).
 	if platform := strings.TrimSpace(c.Query("platform")); platform != "" {
 		filter.Platform = platform

backend/internal/handler/admin/ops_handler.go

@@ -19,6 +19,57 @@ type OpsHandler struct {
 	opsService *service.OpsService
 }
 
+// GetErrorLogByID returns ops error log detail.
+// GET /api/v1/admin/ops/errors/:id
+func (h *OpsHandler) GetErrorLogByID(c *gin.Context) {
+	if h.opsService == nil {
+		response.Error(c, http.StatusServiceUnavailable, "Ops service not available")
+		return
+	}
+	if err := h.opsService.RequireMonitoringEnabled(c.Request.Context()); err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+
+	idStr := strings.TrimSpace(c.Param("id"))
+	id, err := strconv.ParseInt(idStr, 10, 64)
+	if err != nil || id <= 0 {
+		response.BadRequest(c, "Invalid error id")
+		return
+	}
+
+	detail, err := h.opsService.GetErrorLogByID(c.Request.Context(), id)
+	if err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+
+	response.Success(c, detail)
+}
+
+const (
+	opsListViewErrors   = "errors"
+	opsListViewExcluded = "excluded"
+	opsListViewAll      = "all"
+)
+
+func parseOpsViewParam(c *gin.Context) string {
+	if c == nil {
+		return ""
+	}
+	v := strings.ToLower(strings.TrimSpace(c.Query("view")))
+	switch v {
+	case "", opsListViewErrors:
+		return opsListViewErrors
+	case opsListViewExcluded:
+		return opsListViewExcluded
+	case opsListViewAll:
+		return opsListViewAll
+	default:
+		return opsListViewErrors
+	}
+}
+
 func NewOpsHandler(opsService *service.OpsService) *OpsHandler {
 	return &OpsHandler{opsService: opsService}
 }
@@ -47,16 +98,26 @@ func (h *OpsHandler) GetErrorLogs(c *gin.Context) {
 		return
 	}
 
-	filter := &service.OpsErrorLogFilter{
-		Page:     page,
-		PageSize: pageSize,
-	}
+	filter := &service.OpsErrorLogFilter{Page: page, PageSize: pageSize}
+
 	if !startTime.IsZero() {
 		filter.StartTime = &startTime
 	}
 	if !endTime.IsZero() {
 		filter.EndTime = &endTime
 	}
+	filter.View = parseOpsViewParam(c)
+	filter.Phase = strings.TrimSpace(c.Query("phase"))
+	filter.Owner = strings.TrimSpace(c.Query("error_owner"))
+	filter.Source = strings.TrimSpace(c.Query("error_source"))
+	filter.Query = strings.TrimSpace(c.Query("q"))
+	filter.UserQuery = strings.TrimSpace(c.Query("user_query"))
+
+	// 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 = ""
+	}
 
 	if platform := strings.TrimSpace(c.Query("platform")); platform != "" {
 		filter.Platform = platform
@@ -77,11 +138,19 @@ func (h *OpsHandler) GetErrorLogs(c *gin.Context) {
 		}
 		filter.AccountID = &id
 	}
-	if phase := strings.TrimSpace(c.Query("phase")); phase != "" {
-		filter.Phase = phase
-	}
-	if q := strings.TrimSpace(c.Query("q")); q != "" {
-		filter.Query = q
+
+	if v := strings.TrimSpace(c.Query("resolved")); v != "" {
+		switch strings.ToLower(v) {
+		case "1", "true", "yes":
+			b := true
+			filter.Resolved = &b
+		case "0", "false", "no":
+			b := false
+			filter.Resolved = &b
+		default:
+			response.BadRequest(c, "Invalid resolved")
+			return
+		}
 	}
 	if statusCodesStr := strings.TrimSpace(c.Query("status_codes")); statusCodesStr != "" {
 		parts := strings.Split(statusCodesStr, ",")
@@ -106,13 +175,120 @@ func (h *OpsHandler) GetErrorLogs(c *gin.Context) {
 		response.ErrorFrom(c, err)
 		return
 	}
-
 	response.Paginated(c, result.Errors, int64(result.Total), result.Page, result.PageSize)
 }
 
-// GetErrorLogByID returns a single error log detail.
-// GET /api/v1/admin/ops/errors/:id
-func (h *OpsHandler) GetErrorLogByID(c *gin.Context) {
+// ListRequestErrors lists client-visible request errors.
+// GET /api/v1/admin/ops/request-errors
+func (h *OpsHandler) ListRequestErrors(c *gin.Context) {
+	if h.opsService == nil {
+		response.Error(c, http.StatusServiceUnavailable, "Ops service not available")
+		return
+	}
+	if err := h.opsService.RequireMonitoringEnabled(c.Request.Context()); err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+
+	page, pageSize := response.ParsePagination(c)
+	if pageSize > 500 {
+		pageSize = 500
+	}
+	startTime, endTime, err := parseOpsTimeRange(c, "1h")
+	if err != nil {
+		response.BadRequest(c, err.Error())
+		return
+	}
+
+	filter := &service.OpsErrorLogFilter{Page: page, PageSize: pageSize}
+	if !startTime.IsZero() {
+		filter.StartTime = &startTime
+	}
+	if !endTime.IsZero() {
+		filter.EndTime = &endTime
+	}
+	filter.View = parseOpsViewParam(c)
+	filter.Phase = strings.TrimSpace(c.Query("phase"))
+	filter.Owner = strings.TrimSpace(c.Query("error_owner"))
+	filter.Source = strings.TrimSpace(c.Query("error_source"))
+	filter.Query = strings.TrimSpace(c.Query("q"))
+	filter.UserQuery = strings.TrimSpace(c.Query("user_query"))
+
+	// 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 = ""
+	}
+
+	if platform := strings.TrimSpace(c.Query("platform")); platform != "" {
+		filter.Platform = platform
+	}
+	if v := strings.TrimSpace(c.Query("group_id")); v != "" {
+		id, err := strconv.ParseInt(v, 10, 64)
+		if err != nil || id <= 0 {
+			response.BadRequest(c, "Invalid group_id")
+			return
+		}
+		filter.GroupID = &id
+	}
+	if v := strings.TrimSpace(c.Query("account_id")); v != "" {
+		id, err := strconv.ParseInt(v, 10, 64)
+		if err != nil || id <= 0 {
+			response.BadRequest(c, "Invalid account_id")
+			return
+		}
+		filter.AccountID = &id
+	}
+
+	if v := strings.TrimSpace(c.Query("resolved")); v != "" {
+		switch strings.ToLower(v) {
+		case "1", "true", "yes":
+			b := true
+			filter.Resolved = &b
+		case "0", "false", "no":
+			b := false
+			filter.Resolved = &b
+		default:
+			response.BadRequest(c, "Invalid resolved")
+			return
+		}
+	}
+	if statusCodesStr := strings.TrimSpace(c.Query("status_codes")); statusCodesStr != "" {
+		parts := strings.Split(statusCodesStr, ",")
+		out := make([]int, 0, len(parts))
+		for _, part := range parts {
+			p := strings.TrimSpace(part)
+			if p == "" {
+				continue
+			}
+			n, err := strconv.Atoi(p)
+			if err != nil || n < 0 {
+				response.BadRequest(c, "Invalid status_codes")
+				return
+			}
+			out = append(out, n)
+		}
+		filter.StatusCodes = out
+	}
+
+	result, err := h.opsService.GetErrorLogs(c.Request.Context(), filter)
+	if err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+	response.Paginated(c, result.Errors, int64(result.Total), result.Page, result.PageSize)
+}
+
+// GetRequestError returns request error detail.
+// GET /api/v1/admin/ops/request-errors/:id
+func (h *OpsHandler) GetRequestError(c *gin.Context) {
+	// same storage; just proxy to existing detail
+	h.GetErrorLogByID(c)
+}
+
+// ListRequestErrorUpstreamErrors lists upstream error logs correlated to a request error.
+// GET /api/v1/admin/ops/request-errors/:id/upstream-errors
+func (h *OpsHandler) ListRequestErrorUpstreamErrors(c *gin.Context) {
 	if h.opsService == nil {
 		response.Error(c, http.StatusServiceUnavailable, "Ops service not available")
 		return
@@ -129,15 +305,306 @@ func (h *OpsHandler) GetErrorLogByID(c *gin.Context) {
 		return
 	}
 
+	// Load request error to get correlation keys.
 	detail, err := h.opsService.GetErrorLogByID(c.Request.Context(), id)
 	if err != nil {
 		response.ErrorFrom(c, err)
 		return
 	}
 
-	response.Success(c, detail)
+	// Correlate by request_id/client_request_id.
+	requestID := strings.TrimSpace(detail.RequestID)
+	clientRequestID := strings.TrimSpace(detail.ClientRequestID)
+	if requestID == "" && clientRequestID == "" {
+		response.Paginated(c, []*service.OpsErrorLog{}, 0, 1, 10)
+		return
+	}
+
+	page, pageSize := response.ParsePagination(c)
+	if pageSize > 500 {
+		pageSize = 500
+	}
+
+	// Keep correlation window wide enough so linked upstream errors
+	// are discoverable even when UI defaults to 1h elsewhere.
+	startTime, endTime, err := parseOpsTimeRange(c, "30d")
+	if err != nil {
+		response.BadRequest(c, err.Error())
+		return
+	}
+
+	filter := &service.OpsErrorLogFilter{Page: page, PageSize: pageSize}
+	if !startTime.IsZero() {
+		filter.StartTime = &startTime
+	}
+	if !endTime.IsZero() {
+		filter.EndTime = &endTime
+	}
+	filter.View = "all"
+	filter.Phase = "upstream"
+	filter.Owner = "provider"
+	filter.Source = strings.TrimSpace(c.Query("error_source"))
+	filter.Query = strings.TrimSpace(c.Query("q"))
+
+	if platform := strings.TrimSpace(c.Query("platform")); platform != "" {
+		filter.Platform = platform
+	}
+
+	// Prefer exact match on request_id; if missing, fall back to client_request_id.
+	if requestID != "" {
+		filter.RequestID = requestID
+	} else {
+		filter.ClientRequestID = clientRequestID
+	}
+
+	result, err := h.opsService.GetErrorLogs(c.Request.Context(), filter)
+	if err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+
+	// If client asks for details, expand each upstream error log to include upstream response fields.
+	includeDetail := strings.TrimSpace(c.Query("include_detail"))
+	if includeDetail == "1" || strings.EqualFold(includeDetail, "true") || strings.EqualFold(includeDetail, "yes") {
+		details := make([]*service.OpsErrorLogDetail, 0, len(result.Errors))
+		for _, item := range result.Errors {
+			if item == nil {
+				continue
+			}
+			d, err := h.opsService.GetErrorLogByID(c.Request.Context(), item.ID)
+			if err != nil || d == nil {
+				continue
+			}
+			details = append(details, d)
+		}
+		response.Paginated(c, details, int64(result.Total), result.Page, result.PageSize)
+		return
+	}
+
+	response.Paginated(c, result.Errors, int64(result.Total), result.Page, result.PageSize)
 }
 
+// RetryRequestErrorClient retries the client request based on stored request body.
+// POST /api/v1/admin/ops/request-errors/:id/retry-client
+func (h *OpsHandler) RetryRequestErrorClient(c *gin.Context) {
+	if h.opsService == nil {
+		response.Error(c, http.StatusServiceUnavailable, "Ops service not available")
+		return
+	}
+	if err := h.opsService.RequireMonitoringEnabled(c.Request.Context()); err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+
+	subject, ok := middleware.GetAuthSubjectFromContext(c)
+	if !ok || subject.UserID <= 0 {
+		response.Error(c, http.StatusUnauthorized, "Unauthorized")
+		return
+	}
+
+	idStr := strings.TrimSpace(c.Param("id"))
+	id, err := strconv.ParseInt(idStr, 10, 64)
+	if err != nil || id <= 0 {
+		response.BadRequest(c, "Invalid error id")
+		return
+	}
+
+	result, err := h.opsService.RetryError(c.Request.Context(), subject.UserID, id, service.OpsRetryModeClient, nil)
+	if err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+	response.Success(c, result)
+}
+
+// RetryRequestErrorUpstreamEvent retries a specific upstream attempt using captured upstream_request_body.
+// POST /api/v1/admin/ops/request-errors/:id/upstream-errors/:idx/retry
+func (h *OpsHandler) RetryRequestErrorUpstreamEvent(c *gin.Context) {
+	if h.opsService == nil {
+		response.Error(c, http.StatusServiceUnavailable, "Ops service not available")
+		return
+	}
+	if err := h.opsService.RequireMonitoringEnabled(c.Request.Context()); err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+
+	subject, ok := middleware.GetAuthSubjectFromContext(c)
+	if !ok || subject.UserID <= 0 {
+		response.Error(c, http.StatusUnauthorized, "Unauthorized")
+		return
+	}
+
+	idStr := strings.TrimSpace(c.Param("id"))
+	id, err := strconv.ParseInt(idStr, 10, 64)
+	if err != nil || id <= 0 {
+		response.BadRequest(c, "Invalid error id")
+		return
+	}
+
+	idxStr := strings.TrimSpace(c.Param("idx"))
+	idx, err := strconv.Atoi(idxStr)
+	if err != nil || idx < 0 {
+		response.BadRequest(c, "Invalid upstream idx")
+		return
+	}
+
+	result, err := h.opsService.RetryUpstreamEvent(c.Request.Context(), subject.UserID, id, idx)
+	if err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+	response.Success(c, result)
+}
+
+// ResolveRequestError toggles resolved status.
+// PUT /api/v1/admin/ops/request-errors/:id/resolve
+func (h *OpsHandler) ResolveRequestError(c *gin.Context) {
+	h.UpdateErrorResolution(c)
+}
+
+// ListUpstreamErrors lists independent upstream errors.
+// GET /api/v1/admin/ops/upstream-errors
+func (h *OpsHandler) ListUpstreamErrors(c *gin.Context) {
+	if h.opsService == nil {
+		response.Error(c, http.StatusServiceUnavailable, "Ops service not available")
+		return
+	}
+	if err := h.opsService.RequireMonitoringEnabled(c.Request.Context()); err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+
+	page, pageSize := response.ParsePagination(c)
+	if pageSize > 500 {
+		pageSize = 500
+	}
+	startTime, endTime, err := parseOpsTimeRange(c, "1h")
+	if err != nil {
+		response.BadRequest(c, err.Error())
+		return
+	}
+
+	filter := &service.OpsErrorLogFilter{Page: page, PageSize: pageSize}
+	if !startTime.IsZero() {
+		filter.StartTime = &startTime
+	}
+	if !endTime.IsZero() {
+		filter.EndTime = &endTime
+	}
+
+	filter.View = parseOpsViewParam(c)
+	filter.Phase = "upstream"
+	filter.Owner = "provider"
+	filter.Source = strings.TrimSpace(c.Query("error_source"))
+	filter.Query = strings.TrimSpace(c.Query("q"))
+
+	if platform := strings.TrimSpace(c.Query("platform")); platform != "" {
+		filter.Platform = platform
+	}
+	if v := strings.TrimSpace(c.Query("group_id")); v != "" {
+		id, err := strconv.ParseInt(v, 10, 64)
+		if err != nil || id <= 0 {
+			response.BadRequest(c, "Invalid group_id")
+			return
+		}
+		filter.GroupID = &id
+	}
+	if v := strings.TrimSpace(c.Query("account_id")); v != "" {
+		id, err := strconv.ParseInt(v, 10, 64)
+		if err != nil || id <= 0 {
+			response.BadRequest(c, "Invalid account_id")
+			return
+		}
+		filter.AccountID = &id
+	}
+
+	if v := strings.TrimSpace(c.Query("resolved")); v != "" {
+		switch strings.ToLower(v) {
+		case "1", "true", "yes":
+			b := true
+			filter.Resolved = &b
+		case "0", "false", "no":
+			b := false
+			filter.Resolved = &b
+		default:
+			response.BadRequest(c, "Invalid resolved")
+			return
+		}
+	}
+	if statusCodesStr := strings.TrimSpace(c.Query("status_codes")); statusCodesStr != "" {
+		parts := strings.Split(statusCodesStr, ",")
+		out := make([]int, 0, len(parts))
+		for _, part := range parts {
+			p := strings.TrimSpace(part)
+			if p == "" {
+				continue
+			}
+			n, err := strconv.Atoi(p)
+			if err != nil || n < 0 {
+				response.BadRequest(c, "Invalid status_codes")
+				return
+			}
+			out = append(out, n)
+		}
+		filter.StatusCodes = out
+	}
+
+	result, err := h.opsService.GetErrorLogs(c.Request.Context(), filter)
+	if err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+	response.Paginated(c, result.Errors, int64(result.Total), result.Page, result.PageSize)
+}
+
+// GetUpstreamError returns upstream error detail.
+// GET /api/v1/admin/ops/upstream-errors/:id
+func (h *OpsHandler) GetUpstreamError(c *gin.Context) {
+	h.GetErrorLogByID(c)
+}
+
+// RetryUpstreamError retries upstream error using the original account_id.
+// POST /api/v1/admin/ops/upstream-errors/:id/retry
+func (h *OpsHandler) RetryUpstreamError(c *gin.Context) {
+	if h.opsService == nil {
+		response.Error(c, http.StatusServiceUnavailable, "Ops service not available")
+		return
+	}
+	if err := h.opsService.RequireMonitoringEnabled(c.Request.Context()); err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+
+	subject, ok := middleware.GetAuthSubjectFromContext(c)
+	if !ok || subject.UserID <= 0 {
+		response.Error(c, http.StatusUnauthorized, "Unauthorized")
+		return
+	}
+
+	idStr := strings.TrimSpace(c.Param("id"))
+	id, err := strconv.ParseInt(idStr, 10, 64)
+	if err != nil || id <= 0 {
+		response.BadRequest(c, "Invalid error id")
+		return
+	}
+
+	result, err := h.opsService.RetryError(c.Request.Context(), subject.UserID, id, service.OpsRetryModeUpstream, nil)
+	if err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+	response.Success(c, result)
+}
+
+// ResolveUpstreamError toggles resolved status.
+// PUT /api/v1/admin/ops/upstream-errors/:id/resolve
+func (h *OpsHandler) ResolveUpstreamError(c *gin.Context) {
+	h.UpdateErrorResolution(c)
+}
+
+// ==================== Existing endpoints ====================
+
 // ListRequestDetails returns a request-level list (success + error) for drill-down.
 // GET /api/v1/admin/ops/requests
 func (h *OpsHandler) ListRequestDetails(c *gin.Context) {
@@ -242,6 +709,11 @@ func (h *OpsHandler) ListRequestDetails(c *gin.Context) {
 type opsRetryRequest struct {
 	Mode            string `json:"mode"`
 	PinnedAccountID *int64 `json:"pinned_account_id"`
+	Force           bool   `json:"force"`
+}
+
+type opsResolveRequest struct {
+	Resolved bool `json:"resolved"`
 }
 
 // RetryErrorRequest retries a failed request using stored request_body.
@@ -278,6 +750,16 @@ func (h *OpsHandler) RetryErrorRequest(c *gin.Context) {
 		req.Mode = service.OpsRetryModeClient
 	}
 
+	// Force flag is currently a UI-level acknowledgement. Server may still enforce safety constraints.
+	_ = req.Force
+
+	// Legacy endpoint safety: only allow retrying the client request here.
+	// Upstream retries must go through the split endpoints.
+	if strings.EqualFold(strings.TrimSpace(req.Mode), service.OpsRetryModeUpstream) {
+		response.BadRequest(c, "upstream retry is not supported on this endpoint")
+		return
+	}
+
 	result, err := h.opsService.RetryError(c.Request.Context(), subject.UserID, id, req.Mode, req.PinnedAccountID)
 	if err != nil {
 		response.ErrorFrom(c, err)
@@ -287,6 +769,81 @@ func (h *OpsHandler) RetryErrorRequest(c *gin.Context) {
 	response.Success(c, result)
 }
 
+// ListRetryAttempts lists retry attempts for an error log.
+// GET /api/v1/admin/ops/errors/:id/retries
+func (h *OpsHandler) ListRetryAttempts(c *gin.Context) {
+	if h.opsService == nil {
+		response.Error(c, http.StatusServiceUnavailable, "Ops service not available")
+		return
+	}
+	if err := h.opsService.RequireMonitoringEnabled(c.Request.Context()); err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+
+	idStr := strings.TrimSpace(c.Param("id"))
+	id, err := strconv.ParseInt(idStr, 10, 64)
+	if err != nil || id <= 0 {
+		response.BadRequest(c, "Invalid error id")
+		return
+	}
+
+	limit := 50
+	if v := strings.TrimSpace(c.Query("limit")); v != "" {
+		n, err := strconv.Atoi(v)
+		if err != nil || n <= 0 {
+			response.BadRequest(c, "Invalid limit")
+			return
+		}
+		limit = n
+	}
+
+	items, err := h.opsService.ListRetryAttemptsByErrorID(c.Request.Context(), id, limit)
+	if err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+	response.Success(c, items)
+}
+
+// UpdateErrorResolution allows manual resolve/unresolve.
+// PUT /api/v1/admin/ops/errors/:id/resolve
+func (h *OpsHandler) UpdateErrorResolution(c *gin.Context) {
+	if h.opsService == nil {
+		response.Error(c, http.StatusServiceUnavailable, "Ops service not available")
+		return
+	}
+	if err := h.opsService.RequireMonitoringEnabled(c.Request.Context()); err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+
+	subject, ok := middleware.GetAuthSubjectFromContext(c)
+	if !ok || subject.UserID <= 0 {
+		response.Error(c, http.StatusUnauthorized, "Unauthorized")
+		return
+	}
+
+	idStr := strings.TrimSpace(c.Param("id"))
+	id, err := strconv.ParseInt(idStr, 10, 64)
+	if err != nil || id <= 0 {
+		response.BadRequest(c, "Invalid error id")
+		return
+	}
+
+	var req opsResolveRequest
+	if err := c.ShouldBindJSON(&req); err != nil {
+		response.BadRequest(c, "Invalid request: "+err.Error())
+		return
+	}
+	uid := subject.UserID
+	if err := h.opsService.UpdateErrorResolution(c.Request.Context(), id, req.Resolved, &uid, nil); err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+	response.Success(c, gin.H{"ok": true})
+}
+
 func parseOpsTimeRange(c *gin.Context, defaultRange string) (time.Time, time.Time, error) {
 	startStr := strings.TrimSpace(c.Query("start_time"))
 	endStr := strings.TrimSpace(c.Query("end_time"))
@@ -358,6 +915,10 @@ func parseOpsDuration(v string) (time.Duration, bool) {
 		return 6 * time.Hour, true
 	case "24h":
 		return 24 * time.Hour, true
+	case "7d":
+		return 7 * 24 * time.Hour, true
+	case "30d":
+		return 30 * 24 * time.Hour, true
 	default:
 		return 0, false
 	}

backend/internal/handler/admin/proxy_handler.go

@@ -196,6 +196,28 @@ func (h *ProxyHandler) Delete(c *gin.Context) {
 	response.Success(c, gin.H{"message": "Proxy deleted successfully"})
 }
 
+// BatchDelete handles batch deleting proxies
+// POST /api/v1/admin/proxies/batch-delete
+func (h *ProxyHandler) BatchDelete(c *gin.Context) {
+	type BatchDeleteRequest struct {
+		IDs []int64 `json:"ids" binding:"required,min=1"`
+	}
+
+	var req BatchDeleteRequest
+	if err := c.ShouldBindJSON(&req); err != nil {
+		response.BadRequest(c, "Invalid request: "+err.Error())
+		return
+	}
+
+	result, err := h.adminService.BatchDeleteProxies(c.Request.Context(), req.IDs)
+	if err != nil {
+		response.ErrorFrom(c, err)
+		return
+	}
+
+	response.Success(c, result)
+}
+
 // Test handles testing proxy connectivity
 // POST /api/v1/admin/proxies/:id/test
 func (h *ProxyHandler) Test(c *gin.Context) {
@@ -243,19 +265,17 @@ func (h *ProxyHandler) GetProxyAccounts(c *gin.Context) {
 		return
 	}
 
-	page, pageSize := response.ParsePagination(c)
-
-	accounts, total, err := h.adminService.GetProxyAccounts(c.Request.Context(), proxyID, page, pageSize)
+	accounts, err := h.adminService.GetProxyAccounts(c.Request.Context(), proxyID)
 	if err != nil {
 		response.ErrorFrom(c, err)
 		return
 	}
 
-	out := make([]dto.Account, 0, len(accounts))
+	out := make([]dto.ProxyAccountSummary, 0, len(accounts))
 	for i := range accounts {
-		out = append(out, *dto.AccountFromService(&accounts[i]))
+		out = append(out, *dto.ProxyAccountSummaryFromService(&accounts[i]))
 	}
-	response.Paginated(c, out, total, page, pageSize)
+	response.Success(c, out)
 }
 
 // BatchCreateProxyItem represents a single proxy in batch create request

backend/internal/handler/dto/mappers.go

@@ -125,6 +125,7 @@ func AccountFromServiceShallow(a *service.Account) *Account {
 		ProxyID:                 a.ProxyID,
 		Concurrency:             a.Concurrency,
 		Priority:                a.Priority,
+		RateMultiplier:          a.BillingRateMultiplier(),
 		Status:                  a.Status,
 		ErrorMessage:            a.ErrorMessage,
 		LastUsedAt:              a.LastUsedAt,
@@ -212,8 +213,24 @@ func ProxyWithAccountCountFromService(p *service.ProxyWithAccountCount) *ProxyWi
 		return nil
 	}
 	return &ProxyWithAccountCount{
-		Proxy:        *ProxyFromService(&p.Proxy),
-		AccountCount: p.AccountCount,
+		Proxy:          *ProxyFromService(&p.Proxy),
+		AccountCount:   p.AccountCount,
+		LatencyMs:      p.LatencyMs,
+		LatencyStatus:  p.LatencyStatus,
+		LatencyMessage: p.LatencyMessage,
+	}
+}
+
+func ProxyAccountSummaryFromService(a *service.ProxyAccountSummary) *ProxyAccountSummary {
+	if a == nil {
+		return nil
+	}
+	return &ProxyAccountSummary{
+		ID:       a.ID,
+		Name:     a.Name,
+		Platform: a.Platform,
+		Type:     a.Type,
+		Notes:    a.Notes,
 	}
 }
 
@@ -279,6 +296,7 @@ func usageLogFromServiceBase(l *service.UsageLog, account *AccountSummary, inclu
 		TotalCost:             l.TotalCost,
 		ActualCost:            l.ActualCost,
 		RateMultiplier:        l.RateMultiplier,
+		AccountRateMultiplier: l.AccountRateMultiplier,
 		BillingType:           l.BillingType,
 		Stream:                l.Stream,
 		DurationMs:            l.DurationMs,

backend/internal/handler/dto/types.go

@@ -76,6 +76,7 @@ type Account struct {
 	ProxyID            *int64         `json:"proxy_id"`
 	Concurrency        int            `json:"concurrency"`
 	Priority           int            `json:"priority"`
+	RateMultiplier     float64        `json:"rate_multiplier"`
 	Status             string         `json:"status"`
 	ErrorMessage       string         `json:"error_message"`
 	LastUsedAt         *time.Time     `json:"last_used_at"`
@@ -129,7 +130,18 @@ type Proxy struct {
 
 type ProxyWithAccountCount struct {
 	Proxy
-	AccountCount int64 `json:"account_count"`
+	AccountCount   int64  `json:"account_count"`
+	LatencyMs      *int64 `json:"latency_ms,omitempty"`
+	LatencyStatus  string `json:"latency_status,omitempty"`
+	LatencyMessage string `json:"latency_message,omitempty"`
+}
+
+type ProxyAccountSummary struct {
+	ID       int64   `json:"id"`
+	Name     string  `json:"name"`
+	Platform string  `json:"platform"`
+	Type     string  `json:"type"`
+	Notes    *string `json:"notes,omitempty"`
 }
 
 type RedeemCode struct {
@@ -169,13 +181,14 @@ type UsageLog struct {
 	CacheCreation5mTokens int `json:"cache_creation_5m_tokens"`
 	CacheCreation1hTokens int `json:"cache_creation_1h_tokens"`
 
-	InputCost         float64 `json:"input_cost"`
-	OutputCost        float64 `json:"output_cost"`
-	CacheCreationCost float64 `json:"cache_creation_cost"`
-	CacheReadCost     float64 `json:"cache_read_cost"`
-	TotalCost         float64 `json:"total_cost"`
-	ActualCost        float64 `json:"actual_cost"`
-	RateMultiplier    float64 `json:"rate_multiplier"`
+	InputCost             float64  `json:"input_cost"`
+	OutputCost            float64  `json:"output_cost"`
+	CacheCreationCost     float64  `json:"cache_creation_cost"`
+	CacheReadCost         float64  `json:"cache_read_cost"`
+	TotalCost             float64  `json:"total_cost"`
+	ActualCost            float64  `json:"actual_cost"`
+	RateMultiplier        float64  `json:"rate_multiplier"`
+	AccountRateMultiplier *float64 `json:"account_rate_multiplier"`
 
 	BillingType  int8 `json:"billing_type"`
 	Stream       bool `json:"stream"`

backend/internal/handler/ops_error_logger.go

@@ -544,6 +544,11 @@ func OpsErrorLoggerMiddleware(ops *service.OpsService) gin.HandlerFunc {
 		body := w.buf.Bytes()
 		parsed := parseOpsErrorResponse(body)
 
+		// Skip logging if the error should be filtered based on settings
+		if shouldSkipOpsErrorLog(c.Request.Context(), ops, parsed.Message, string(body), c.Request.URL.Path) {
+			return
+		}
+
 		apiKey, _ := middleware2.GetAPIKeyFromContext(c)
 
 		clientRequestID, _ := c.Request.Context().Value(ctxkey.ClientRequestID).(string)
@@ -832,28 +837,30 @@ func normalizeOpsErrorType(errType string, code string) string {
 
 func classifyOpsPhase(errType, message, code string) string {
 	msg := strings.ToLower(message)
+	// Standardized phases: request|auth|routing|upstream|network|internal
+	// Map billing/concurrency/response => request; scheduling => routing.
 	switch strings.TrimSpace(code) {
 	case "INSUFFICIENT_BALANCE", "USAGE_LIMIT_EXCEEDED", "SUBSCRIPTION_NOT_FOUND", "SUBSCRIPTION_INVALID":
-		return "billing"
+		return "request"
 	}
 
 	switch errType {
 	case "authentication_error":
 		return "auth"
 	case "billing_error", "subscription_error":
-		return "billing"
+		return "request"
 	case "rate_limit_error":
 		if strings.Contains(msg, "concurrency") || strings.Contains(msg, "pending") || strings.Contains(msg, "queue") {
-			return "concurrency"
+			return "request"
 		}
 		return "upstream"
 	case "invalid_request_error":
-		return "response"
+		return "request"
 	case "upstream_error", "overloaded_error":
 		return "upstream"
 	case "api_error":
 		if strings.Contains(msg, "no available accounts") {
-			return "scheduling"
+			return "routing"
 		}
 		return "internal"
 	default:
@@ -914,34 +921,38 @@ func classifyOpsIsBusinessLimited(errType, phase, code string, status int, messa
 }
 
 func classifyOpsErrorOwner(phase string, message string) string {
+	// Standardized owners: client|provider|platform
 	switch phase {
 	case "upstream", "network":
 		return "provider"
-	case "billing", "concurrency", "auth", "response":
+	case "request", "auth":
 		return "client"
+	case "routing", "internal":
+		return "platform"
 	default:
 		if strings.Contains(strings.ToLower(message), "upstream") {
 			return "provider"
 		}
-		return "sub2api"
+		return "platform"
 	}
 }
 
 func classifyOpsErrorSource(phase string, message string) string {
+	// Standardized sources: client_request|upstream_http|gateway
 	switch phase {
 	case "upstream":
 		return "upstream_http"
 	case "network":
-		return "upstream_network"
-	case "billing":
-		return "billing"
-	case "concurrency":
-		return "concurrency"
+		return "gateway"
+	case "request", "auth":
+		return "client_request"
+	case "routing", "internal":
+		return "gateway"
 	default:
 		if strings.Contains(strings.ToLower(message), "upstream") {
 			return "upstream_http"
 		}
-		return "internal"
+		return "gateway"
 	}
 }
 
@@ -963,3 +974,42 @@ func truncateString(s string, max int) string {
 func strconvItoa(v int) string {
 	return strconv.Itoa(v)
 }
+
+// shouldSkipOpsErrorLog determines if an error should be skipped from logging based on settings.
+// Returns true for errors that should be filtered according to OpsAdvancedSettings.
+func shouldSkipOpsErrorLog(ctx context.Context, ops *service.OpsService, message, body, requestPath string) bool {
+	if ops == nil {
+		return false
+	}
+
+	// Get advanced settings to check filter configuration
+	settings, err := ops.GetOpsAdvancedSettings(ctx)
+	if err != nil || settings == nil {
+		// If we can't get settings, don't skip (fail open)
+		return false
+	}
+
+	msgLower := strings.ToLower(message)
+	bodyLower := strings.ToLower(body)
+
+	// Check if count_tokens errors should be ignored
+	if settings.IgnoreCountTokensErrors && strings.Contains(requestPath, "/count_tokens") {
+		return true
+	}
+
+	// Check if context canceled errors should be ignored (client disconnects)
+	if settings.IgnoreContextCanceled {
+		if strings.Contains(msgLower, "context canceled") || strings.Contains(bodyLower, "context canceled") {
+			return true
+		}
+	}
+
+	// Check if "no available accounts" errors should be ignored
+	if settings.IgnoreNoAvailableAccounts {
+		if strings.Contains(msgLower, "no available accounts") || strings.Contains(bodyLower, "no available accounts") {
+			return true
+		}
+	}
+
+	return false
+}

backend/internal/pkg/usagestats/account_stats.go

@@ -1,8 +1,14 @@
 package usagestats
 
 // AccountStats 账号使用统计
+//
+// cost: 账号口径费用（使用 total_cost * account_rate_multiplier）
+// standard_cost: 标准费用（使用 total_cost，不含倍率）
+// user_cost: 用户/API Key 口径费用（使用 actual_cost，受分组倍率影响）
 type AccountStats struct {
-	Requests int64   `json:"requests"`
-	Tokens   int64   `json:"tokens"`
-	Cost     float64 `json:"cost"`
+	Requests     int64   `json:"requests"`
+	Tokens       int64   `json:"tokens"`
+	Cost         float64 `json:"cost"`
+	StandardCost float64 `json:"standard_cost"`
+	UserCost     float64 `json:"user_cost"`
 }

backend/internal/pkg/usagestats/usage_log_types.go

@@ -147,14 +147,15 @@ type UsageLogFilters struct {
 
 // UsageStats represents usage statistics
 type UsageStats struct {
-	TotalRequests     int64   `json:"total_requests"`
-	TotalInputTokens  int64   `json:"total_input_tokens"`
-	TotalOutputTokens int64   `json:"total_output_tokens"`
-	TotalCacheTokens  int64   `json:"total_cache_tokens"`
-	TotalTokens       int64   `json:"total_tokens"`
-	TotalCost         float64 `json:"total_cost"`
-	TotalActualCost   float64 `json:"total_actual_cost"`
-	AverageDurationMs float64 `json:"average_duration_ms"`
+	TotalRequests     int64    `json:"total_requests"`
+	TotalInputTokens  int64    `json:"total_input_tokens"`
+	TotalOutputTokens int64    `json:"total_output_tokens"`
+	TotalCacheTokens  int64    `json:"total_cache_tokens"`
+	TotalTokens       int64    `json:"total_tokens"`
+	TotalCost         float64  `json:"total_cost"`
+	TotalActualCost   float64  `json:"total_actual_cost"`
+	TotalAccountCost  *float64 `json:"total_account_cost,omitempty"`
+	AverageDurationMs float64  `json:"average_duration_ms"`
 }
 
 // BatchUserUsageStats represents usage stats for a single user
@@ -177,25 +178,29 @@ type AccountUsageHistory struct {
 	Label      string  `json:"label"`
 	Requests   int64   `json:"requests"`
 	Tokens     int64   `json:"tokens"`
-	Cost       float64 `json:"cost"`
-	ActualCost float64 `json:"actual_cost"`
+	Cost       float64 `json:"cost"`        // 标准计费（total_cost）
+	ActualCost float64 `json:"actual_cost"` // 账号口径费用（total_cost * account_rate_multiplier）
+	UserCost   float64 `json:"user_cost"`   // 用户口径费用（actual_cost，受分组倍率影响）
 }
 
 // AccountUsageSummary represents summary statistics for an account
 type AccountUsageSummary struct {
 	Days              int     `json:"days"`
 	ActualDaysUsed    int     `json:"actual_days_used"`
-	TotalCost         float64 `json:"total_cost"`
+	TotalCost         float64 `json:"total_cost"`      // 账号口径费用
+	TotalUserCost     float64 `json:"total_user_cost"` // 用户口径费用
 	TotalStandardCost float64 `json:"total_standard_cost"`
 	TotalRequests     int64   `json:"total_requests"`
 	TotalTokens       int64   `json:"total_tokens"`
-	AvgDailyCost      float64 `json:"avg_daily_cost"`
+	AvgDailyCost      float64 `json:"avg_daily_cost"` // 账号口径日均
+	AvgDailyUserCost  float64 `json:"avg_daily_user_cost"`
 	AvgDailyRequests  float64 `json:"avg_daily_requests"`
 	AvgDailyTokens    float64 `json:"avg_daily_tokens"`
 	AvgDurationMs     float64 `json:"avg_duration_ms"`
 	Today             *struct {
 		Date     string  `json:"date"`
 		Cost     float64 `json:"cost"`
+		UserCost float64 `json:"user_cost"`
 		Requests int64   `json:"requests"`
 		Tokens   int64   `json:"tokens"`
 	} `json:"today"`
@@ -203,6 +208,7 @@ type AccountUsageSummary struct {
 		Date     string  `json:"date"`
 		Label    string  `json:"label"`
 		Cost     float64 `json:"cost"`
+		UserCost float64 `json:"user_cost"`
 		Requests int64   `json:"requests"`
 	} `json:"highest_cost_day"`
 	HighestRequestDay *struct {
@@ -210,6 +216,7 @@ type AccountUsageSummary struct {
 		Label    string  `json:"label"`
 		Requests int64   `json:"requests"`
 		Cost     float64 `json:"cost"`
+		UserCost float64 `json:"user_cost"`
 	} `json:"highest_request_day"`
 }
 

backend/internal/repository/account_repo.go

@@ -80,6 +80,10 @@ func (r *accountRepository) Create(ctx context.Context, account *service.Account
 		SetSchedulable(account.Schedulable).
 		SetAutoPauseOnExpired(account.AutoPauseOnExpired)
 
+	if account.RateMultiplier != nil {
+		builder.SetRateMultiplier(*account.RateMultiplier)
+	}
+
 	if account.ProxyID != nil {
 		builder.SetProxyID(*account.ProxyID)
 	}
@@ -291,6 +295,10 @@ func (r *accountRepository) Update(ctx context.Context, account *service.Account
 		SetSchedulable(account.Schedulable).
 		SetAutoPauseOnExpired(account.AutoPauseOnExpired)
 
+	if account.RateMultiplier != nil {
+		builder.SetRateMultiplier(*account.RateMultiplier)
+	}
+
 	if account.ProxyID != nil {
 		builder.SetProxyID(*account.ProxyID)
 	} else {
@@ -999,6 +1007,11 @@ func (r *accountRepository) BulkUpdate(ctx context.Context, ids []int64, updates
 		args = append(args, *updates.Priority)
 		idx++
 	}
+	if updates.RateMultiplier != nil {
+		setClauses = append(setClauses, "rate_multiplier = $"+itoa(idx))
+		args = append(args, *updates.RateMultiplier)
+		idx++
+	}
 	if updates.Status != nil {
 		setClauses = append(setClauses, "status = $"+itoa(idx))
 		args = append(args, *updates.Status)
@@ -1347,6 +1360,8 @@ func accountEntityToService(m *dbent.Account) *service.Account {
 		return nil
 	}
 
+	rateMultiplier := m.RateMultiplier
+
 	return &service.Account{
 		ID:                  m.ID,
 		Name:                m.Name,
@@ -1358,6 +1373,7 @@ func accountEntityToService(m *dbent.Account) *service.Account {
 		ProxyID:             m.ProxyID,
 		Concurrency:         m.Concurrency,
 		Priority:            m.Priority,
+		RateMultiplier:      &rateMultiplier,
 		Status:              m.Status,
 		ErrorMessage:        derefString(m.ErrorMessage),
 		LastUsedAt:          m.LastUsedAt,

backend/internal/repository/dashboard_aggregation_repo.go

@@ -8,6 +8,7 @@ import (
 	"strings"
 	"time"
 
+	"github.com/Wei-Shaw/sub2api/internal/pkg/timezone"
 	"github.com/Wei-Shaw/sub2api/internal/service"
 	"github.com/lib/pq"
 )
@@ -41,21 +42,22 @@ func isPostgresDriver(db *sql.DB) bool {
 }
 
 func (r *dashboardAggregationRepository) AggregateRange(ctx context.Context, start, end time.Time) error {
-	startUTC := start.UTC()
-	endUTC := end.UTC()
-	if !endUTC.After(startUTC) {
+	loc := timezone.Location()
+	startLocal := start.In(loc)
+	endLocal := end.In(loc)
+	if !endLocal.After(startLocal) {
 		return nil
 	}
 
-	hourStart := startUTC.Truncate(time.Hour)
-	hourEnd := endUTC.Truncate(time.Hour)
-	if endUTC.After(hourEnd) {
+	hourStart := startLocal.Truncate(time.Hour)
+	hourEnd := endLocal.Truncate(time.Hour)
+	if endLocal.After(hourEnd) {
 		hourEnd = hourEnd.Add(time.Hour)
 	}
 
-	dayStart := truncateToDayUTC(startUTC)
-	dayEnd := truncateToDayUTC(endUTC)
-	if endUTC.After(dayEnd) {
+	dayStart := truncateToDay(startLocal)
+	dayEnd := truncateToDay(endLocal)
+	if endLocal.After(dayEnd) {
 		dayEnd = dayEnd.Add(24 * time.Hour)
 	}
 
@@ -146,38 +148,41 @@ func (r *dashboardAggregationRepository) EnsureUsageLogsPartitions(ctx context.C
 }
 
 func (r *dashboardAggregationRepository) insertHourlyActiveUsers(ctx context.Context, start, end time.Time) error {
+	tzName := timezone.Name()
 	query := `
 		INSERT INTO usage_dashboard_hourly_users (bucket_start, user_id)
 		SELECT DISTINCT
-			date_trunc('hour', created_at AT TIME ZONE 'UTC') AT TIME ZONE 'UTC' AS bucket_start,
+			date_trunc('hour', created_at AT TIME ZONE $3) AT TIME ZONE $3 AS bucket_start,
 			user_id
 		FROM usage_logs
 		WHERE created_at >= $1 AND created_at < $2
 		ON CONFLICT DO NOTHING
 	`
-	_, err := r.sql.ExecContext(ctx, query, start.UTC(), end.UTC())
+	_, err := r.sql.ExecContext(ctx, query, start, end, tzName)
 	return err
 }
 
 func (r *dashboardAggregationRepository) insertDailyActiveUsers(ctx context.Context, start, end time.Time) error {
+	tzName := timezone.Name()
 	query := `
 		INSERT INTO usage_dashboard_daily_users (bucket_date, user_id)
 		SELECT DISTINCT
-			(bucket_start AT TIME ZONE 'UTC')::date AS bucket_date,
+			(bucket_start AT TIME ZONE $3)::date AS bucket_date,
 			user_id
 		FROM usage_dashboard_hourly_users
 		WHERE bucket_start >= $1 AND bucket_start < $2
 		ON CONFLICT DO NOTHING
 	`
-	_, err := r.sql.ExecContext(ctx, query, start.UTC(), end.UTC())
+	_, err := r.sql.ExecContext(ctx, query, start, end, tzName)
 	return err
 }
 
 func (r *dashboardAggregationRepository) upsertHourlyAggregates(ctx context.Context, start, end time.Time) error {
+	tzName := timezone.Name()
 	query := `
 		WITH hourly AS (
 			SELECT
-				date_trunc('hour', created_at AT TIME ZONE 'UTC') AT TIME ZONE 'UTC' AS bucket_start,
+				date_trunc('hour', created_at AT TIME ZONE $3) AT TIME ZONE $3 AS bucket_start,
 				COUNT(*) AS total_requests,
 				COALESCE(SUM(input_tokens), 0) AS input_tokens,
 				COALESCE(SUM(output_tokens), 0) AS output_tokens,
@@ -236,15 +241,16 @@ func (r *dashboardAggregationRepository) upsertHourlyAggregates(ctx context.Cont
 			active_users = EXCLUDED.active_users,
 			computed_at = EXCLUDED.computed_at
 	`
-	_, err := r.sql.ExecContext(ctx, query, start.UTC(), end.UTC())
+	_, err := r.sql.ExecContext(ctx, query, start, end, tzName)
 	return err
 }
 
 func (r *dashboardAggregationRepository) upsertDailyAggregates(ctx context.Context, start, end time.Time) error {
+	tzName := timezone.Name()
 	query := `
 		WITH daily AS (
 			SELECT
-				(bucket_start AT TIME ZONE 'UTC')::date AS bucket_date,
+				(bucket_start AT TIME ZONE $5)::date AS bucket_date,
 				COALESCE(SUM(total_requests), 0) AS total_requests,
 				COALESCE(SUM(input_tokens), 0) AS input_tokens,
 				COALESCE(SUM(output_tokens), 0) AS output_tokens,
@@ -255,7 +261,7 @@ func (r *dashboardAggregationRepository) upsertDailyAggregates(ctx context.Conte
 				COALESCE(SUM(total_duration_ms), 0) AS total_duration_ms
 			FROM usage_dashboard_hourly
 			WHERE bucket_start >= $1 AND bucket_start < $2
-			GROUP BY (bucket_start AT TIME ZONE 'UTC')::date
+			GROUP BY (bucket_start AT TIME ZONE $5)::date
 		),
 		user_counts AS (
 			SELECT bucket_date, COUNT(*) AS active_users
@@ -303,7 +309,7 @@ func (r *dashboardAggregationRepository) upsertDailyAggregates(ctx context.Conte
 			active_users = EXCLUDED.active_users,
 			computed_at = EXCLUDED.computed_at
 	`
-	_, err := r.sql.ExecContext(ctx, query, start.UTC(), end.UTC(), start.UTC(), end.UTC())
+	_, err := r.sql.ExecContext(ctx, query, start, end, start, end, tzName)
 	return err
 }
 
@@ -376,9 +382,8 @@ func (r *dashboardAggregationRepository) createUsageLogsPartition(ctx context.Co
 	return err
 }
 
-func truncateToDayUTC(t time.Time) time.Time {
-	t = t.UTC()
-	return time.Date(t.Year(), t.Month(), t.Day(), 0, 0, 0, 0, time.UTC)
+func truncateToDay(t time.Time) time.Time {
+	return timezone.StartOfDay(t)
 }
 
 func truncateToMonthUTC(t time.Time) time.Time {

backend/internal/repository/ops_repo.go

@@ -55,7 +55,6 @@ INSERT INTO ops_error_logs (
   upstream_error_message,
   upstream_error_detail,
   upstream_errors,
-  duration_ms,
   time_to_first_token_ms,
   request_body,
   request_body_truncated,
@@ -65,7 +64,7 @@ INSERT INTO ops_error_logs (
   retry_count,
   created_at
 ) VALUES (
-  $1,$2,$3,$4,$5,$6,$7,$8,$9,$10,$11,$12,$13,$14,$15,$16,$17,$18,$19,$20,$21,$22,$23,$24,$25,$26,$27,$28,$29,$30,$31,$32,$33,$34,$35
+  $1,$2,$3,$4,$5,$6,$7,$8,$9,$10,$11,$12,$13,$14,$15,$16,$17,$18,$19,$20,$21,$22,$23,$24,$25,$26,$27,$28,$29,$30,$31,$32,$33,$34
 ) RETURNING id`
 
 	var id int64
@@ -98,7 +97,6 @@ INSERT INTO ops_error_logs (
 		opsNullString(input.UpstreamErrorMessage),
 		opsNullString(input.UpstreamErrorDetail),
 		opsNullString(input.UpstreamErrorsJSON),
-		opsNullInt(input.DurationMs),
 		opsNullInt64(input.TimeToFirstTokenMs),
 		opsNullString(input.RequestBodyJSON),
 		input.RequestBodyTruncated,
@@ -135,7 +133,7 @@ func (r *opsRepository) ListErrorLogs(ctx context.Context, filter *service.OpsEr
 	}
 
 	where, args := buildOpsErrorLogsWhere(filter)
-	countSQL := "SELECT COUNT(*) FROM ops_error_logs " + where
+	countSQL := "SELECT COUNT(*) FROM ops_error_logs e " + where
 
 	var total int
 	if err := r.db.QueryRowContext(ctx, countSQL, args...).Scan(&total); err != nil {
@@ -146,28 +144,43 @@ func (r *opsRepository) ListErrorLogs(ctx context.Context, filter *service.OpsEr
 	argsWithLimit := append(args, pageSize, offset)
 	selectSQL := `
 SELECT
-  id,
-  created_at,
-  error_phase,
-  error_type,
-  severity,
-  COALESCE(upstream_status_code, status_code, 0),
-  COALESCE(platform, ''),
-  COALESCE(model, ''),
-  duration_ms,
-  COALESCE(client_request_id, ''),
-  COALESCE(request_id, ''),
-  COALESCE(error_message, ''),
-  user_id,
-  api_key_id,
-  account_id,
-  group_id,
-  CASE WHEN client_ip IS NULL THEN NULL ELSE client_ip::text END,
-  COALESCE(request_path, ''),
-  stream
-FROM ops_error_logs
+  e.id,
+  e.created_at,
+  e.error_phase,
+  e.error_type,
+  COALESCE(e.error_owner, ''),
+  COALESCE(e.error_source, ''),
+  e.severity,
+  COALESCE(e.upstream_status_code, e.status_code, 0),
+  COALESCE(e.platform, ''),
+  COALESCE(e.model, ''),
+  COALESCE(e.is_retryable, false),
+  COALESCE(e.retry_count, 0),
+  COALESCE(e.resolved, false),
+  e.resolved_at,
+  e.resolved_by_user_id,
+  COALESCE(u2.email, ''),
+  e.resolved_retry_id,
+  COALESCE(e.client_request_id, ''),
+  COALESCE(e.request_id, ''),
+  COALESCE(e.error_message, ''),
+  e.user_id,
+  COALESCE(u.email, ''),
+  e.api_key_id,
+  e.account_id,
+  COALESCE(a.name, ''),
+  e.group_id,
+  COALESCE(g.name, ''),
+  CASE WHEN e.client_ip IS NULL THEN NULL ELSE e.client_ip::text END,
+  COALESCE(e.request_path, ''),
+  e.stream
+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
 ` + where + `
-ORDER BY created_at DESC
+ORDER BY e.created_at DESC
 LIMIT $` + itoa(len(args)+1) + ` OFFSET $` + itoa(len(args)+2)
 
 	rows, err := r.db.QueryContext(ctx, selectSQL, argsWithLimit...)
@@ -179,39 +192,65 @@ LIMIT $` + itoa(len(args)+1) + ` OFFSET $` + itoa(len(args)+2)
 	out := make([]*service.OpsErrorLog, 0, pageSize)
 	for rows.Next() {
 		var item service.OpsErrorLog
-		var latency sql.NullInt64
 		var statusCode sql.NullInt64
 		var clientIP sql.NullString
 		var userID sql.NullInt64
 		var apiKeyID sql.NullInt64
 		var accountID sql.NullInt64
+		var accountName string
 		var groupID sql.NullInt64
+		var groupName string
+		var userEmail string
+		var resolvedAt sql.NullTime
+		var resolvedBy sql.NullInt64
+		var resolvedByName string
+		var resolvedRetryID sql.NullInt64
 		if err := rows.Scan(
 			&item.ID,
 			&item.CreatedAt,
 			&item.Phase,
 			&item.Type,
+			&item.Owner,
+			&item.Source,
 			&item.Severity,
 			&statusCode,
 			&item.Platform,
 			&item.Model,
-			&latency,
+			&item.IsRetryable,
+			&item.RetryCount,
+			&item.Resolved,
+			&resolvedAt,
+			&resolvedBy,
+			&resolvedByName,
+			&resolvedRetryID,
 			&item.ClientRequestID,
 			&item.RequestID,
 			&item.Message,
 			&userID,
+			&userEmail,
 			&apiKeyID,
 			&accountID,
+			&accountName,
 			&groupID,
+			&groupName,
 			&clientIP,
 			&item.RequestPath,
 			&item.Stream,
 		); err != nil {
 			return nil, err
 		}
-		if latency.Valid {
-			v := int(latency.Int64)
-			item.LatencyMs = &v
+		if resolvedAt.Valid {
+			t := resolvedAt.Time
+			item.ResolvedAt = &t
+		}
+		if resolvedBy.Valid {
+			v := resolvedBy.Int64
+			item.ResolvedByUserID = &v
+		}
+		item.ResolvedByUserName = resolvedByName
+		if resolvedRetryID.Valid {
+			v := resolvedRetryID.Int64
+			item.ResolvedRetryID = &v
 		}
 		item.StatusCode = int(statusCode.Int64)
 		if clientIP.Valid {
@@ -222,6 +261,7 @@ LIMIT $` + itoa(len(args)+1) + ` OFFSET $` + itoa(len(args)+2)
 			v := userID.Int64
 			item.UserID = &v
 		}
+		item.UserEmail = userEmail
 		if apiKeyID.Valid {
 			v := apiKeyID.Int64
 			item.APIKeyID = &v
@@ -230,10 +270,12 @@ LIMIT $` + itoa(len(args)+1) + ` OFFSET $` + itoa(len(args)+2)
 			v := accountID.Int64
 			item.AccountID = &v
 		}
+		item.AccountName = accountName
 		if groupID.Valid {
 			v := groupID.Int64
 			item.GroupID = &v
 		}
+		item.GroupName = groupName
 		out = append(out, &item)
 	}
 	if err := rows.Err(); err != nil {
@@ -258,49 +300,64 @@ func (r *opsRepository) GetErrorLogByID(ctx context.Context, id int64) (*service
 
 	q := `
 SELECT
-  id,
-  created_at,
-  error_phase,
-  error_type,
-  severity,
-  COALESCE(upstream_status_code, status_code, 0),
-  COALESCE(platform, ''),
-  COALESCE(model, ''),
-  duration_ms,
-  COALESCE(client_request_id, ''),
-  COALESCE(request_id, ''),
-  COALESCE(error_message, ''),
-  COALESCE(error_body, ''),
-  upstream_status_code,
-  COALESCE(upstream_error_message, ''),
-  COALESCE(upstream_error_detail, ''),
-  COALESCE(upstream_errors::text, ''),
-  is_business_limited,
-  user_id,
-  api_key_id,
-  account_id,
-  group_id,
-  CASE WHEN client_ip IS NULL THEN NULL ELSE client_ip::text END,
-  COALESCE(request_path, ''),
-  stream,
-  COALESCE(user_agent, ''),
-  auth_latency_ms,
-  routing_latency_ms,
-  upstream_latency_ms,
-  response_latency_ms,
-  time_to_first_token_ms,
-  COALESCE(request_body::text, ''),
-  request_body_truncated,
-  request_body_bytes,
-  COALESCE(request_headers::text, '')
-FROM ops_error_logs
-WHERE id = $1
+  e.id,
+  e.created_at,
+  e.error_phase,
+  e.error_type,
+  COALESCE(e.error_owner, ''),
+  COALESCE(e.error_source, ''),
+  e.severity,
+  COALESCE(e.upstream_status_code, e.status_code, 0),
+  COALESCE(e.platform, ''),
+  COALESCE(e.model, ''),
+  COALESCE(e.is_retryable, false),
+  COALESCE(e.retry_count, 0),
+  COALESCE(e.resolved, false),
+  e.resolved_at,
+  e.resolved_by_user_id,
+  e.resolved_retry_id,
+  COALESCE(e.client_request_id, ''),
+  COALESCE(e.request_id, ''),
+  COALESCE(e.error_message, ''),
+  COALESCE(e.error_body, ''),
+  e.upstream_status_code,
+  COALESCE(e.upstream_error_message, ''),
+  COALESCE(e.upstream_error_detail, ''),
+  COALESCE(e.upstream_errors::text, ''),
+  e.is_business_limited,
+  e.user_id,
+  COALESCE(u.email, ''),
+  e.api_key_id,
+  e.account_id,
+  COALESCE(a.name, ''),
+  e.group_id,
+  COALESCE(g.name, ''),
+  CASE WHEN e.client_ip IS NULL THEN NULL ELSE e.client_ip::text END,
+  COALESCE(e.request_path, ''),
+  e.stream,
+  COALESCE(e.user_agent, ''),
+  e.auth_latency_ms,
+  e.routing_latency_ms,
+  e.upstream_latency_ms,
+  e.response_latency_ms,
+  e.time_to_first_token_ms,
+  COALESCE(e.request_body::text, ''),
+  e.request_body_truncated,
+  e.request_body_bytes,
+  COALESCE(e.request_headers::text, '')
+FROM ops_error_logs e
+LEFT JOIN users u ON e.user_id = u.id
+LEFT JOIN accounts a ON e.account_id = a.id
+LEFT JOIN groups g ON e.group_id = g.id
+WHERE e.id = $1
 LIMIT 1`
 
 	var out service.OpsErrorLogDetail
-	var latency sql.NullInt64
 	var statusCode sql.NullInt64
 	var upstreamStatusCode sql.NullInt64
+	var resolvedAt sql.NullTime
+	var resolvedBy sql.NullInt64
+	var resolvedRetryID sql.NullInt64
 	var clientIP sql.NullString
 	var userID sql.NullInt64
 	var apiKeyID sql.NullInt64
@@ -318,11 +375,18 @@ LIMIT 1`
 		&out.CreatedAt,
 		&out.Phase,
 		&out.Type,
+		&out.Owner,
+		&out.Source,
 		&out.Severity,
 		&statusCode,
 		&out.Platform,
 		&out.Model,
-		&latency,
+		&out.IsRetryable,
+		&out.RetryCount,
+		&out.Resolved,
+		&resolvedAt,
+		&resolvedBy,
+		&resolvedRetryID,
 		&out.ClientRequestID,
 		&out.RequestID,
 		&out.Message,
@@ -333,9 +397,12 @@ LIMIT 1`
 		&out.UpstreamErrors,
 		&out.IsBusinessLimited,
 		&userID,
+		&out.UserEmail,
 		&apiKeyID,
 		&accountID,
+		&out.AccountName,
 		&groupID,
+		&out.GroupName,
 		&clientIP,
 		&out.RequestPath,
 		&out.Stream,
@@ -355,9 +422,17 @@ LIMIT 1`
 	}
 
 	out.StatusCode = int(statusCode.Int64)
-	if latency.Valid {
-		v := int(latency.Int64)
-		out.LatencyMs = &v
+	if resolvedAt.Valid {
+		t := resolvedAt.Time
+		out.ResolvedAt = &t
+	}
+	if resolvedBy.Valid {
+		v := resolvedBy.Int64
+		out.ResolvedByUserID = &v
+	}
+	if resolvedRetryID.Valid {
+		v := resolvedRetryID.Int64
+		out.ResolvedRetryID = &v
 	}
 	if clientIP.Valid {
 		s := clientIP.String
@@ -487,9 +562,15 @@ SET
   status = $2,
   finished_at = $3,
   duration_ms = $4,
-  result_request_id = $5,
-  result_error_id = $6,
-  error_message = $7
+  success = $5,
+  http_status_code = $6,
+  upstream_request_id = $7,
+  used_account_id = $8,
+  response_preview = $9,
+  response_truncated = $10,
+  result_request_id = $11,
+  result_error_id = $12,
+  error_message = $13
 WHERE id = $1`
 
 	_, err := r.db.ExecContext(
@@ -499,8 +580,14 @@ WHERE id = $1`
 		strings.TrimSpace(input.Status),
 		nullTime(input.FinishedAt),
 		input.DurationMs,
+		nullBool(input.Success),
+		nullInt(input.HTTPStatusCode),
+		opsNullString(input.UpstreamRequestID),
+		nullInt64(input.UsedAccountID),
+		opsNullString(input.ResponsePreview),
+		nullBool(input.ResponseTruncated),
 		opsNullString(input.ResultRequestID),
-		opsNullInt64(input.ResultErrorID),
+		nullInt64(input.ResultErrorID),
 		opsNullString(input.ErrorMessage),
 	)
 	return err
@@ -526,6 +613,12 @@ SELECT
   started_at,
   finished_at,
   duration_ms,
+  success,
+  http_status_code,
+  upstream_request_id,
+  used_account_id,
+  response_preview,
+  response_truncated,
   result_request_id,
   result_error_id,
   error_message
@@ -540,6 +633,12 @@ LIMIT 1`
 	var startedAt sql.NullTime
 	var finishedAt sql.NullTime
 	var durationMs sql.NullInt64
+	var success sql.NullBool
+	var httpStatusCode sql.NullInt64
+	var upstreamRequestID sql.NullString
+	var usedAccountID sql.NullInt64
+	var responsePreview sql.NullString
+	var responseTruncated sql.NullBool
 	var resultRequestID sql.NullString
 	var resultErrorID sql.NullInt64
 	var errorMessage sql.NullString
@@ -555,6 +654,12 @@ LIMIT 1`
 		&startedAt,
 		&finishedAt,
 		&durationMs,
+		&success,
+		&httpStatusCode,
+		&upstreamRequestID,
+		&usedAccountID,
+		&responsePreview,
+		&responseTruncated,
 		&resultRequestID,
 		&resultErrorID,
 		&errorMessage,
@@ -579,6 +684,30 @@ LIMIT 1`
 		v := durationMs.Int64
 		out.DurationMs = &v
 	}
+	if success.Valid {
+		v := success.Bool
+		out.Success = &v
+	}
+	if httpStatusCode.Valid {
+		v := int(httpStatusCode.Int64)
+		out.HTTPStatusCode = &v
+	}
+	if upstreamRequestID.Valid {
+		s := upstreamRequestID.String
+		out.UpstreamRequestID = &s
+	}
+	if usedAccountID.Valid {
+		v := usedAccountID.Int64
+		out.UsedAccountID = &v
+	}
+	if responsePreview.Valid {
+		s := responsePreview.String
+		out.ResponsePreview = &s
+	}
+	if responseTruncated.Valid {
+		v := responseTruncated.Bool
+		out.ResponseTruncated = &v
+	}
 	if resultRequestID.Valid {
 		s := resultRequestID.String
 		out.ResultRequestID = &s
@@ -602,30 +731,234 @@ func nullTime(t time.Time) sql.NullTime {
 	return sql.NullTime{Time: t, Valid: true}
 }
 
+func nullBool(v *bool) sql.NullBool {
+	if v == nil {
+		return sql.NullBool{}
+	}
+	return sql.NullBool{Bool: *v, Valid: true}
+}
+
+func (r *opsRepository) ListRetryAttemptsByErrorID(ctx context.Context, sourceErrorID int64, limit int) ([]*service.OpsRetryAttempt, error) {
+	if r == nil || r.db == nil {
+		return nil, fmt.Errorf("nil ops repository")
+	}
+	if sourceErrorID <= 0 {
+		return nil, fmt.Errorf("invalid source_error_id")
+	}
+	if limit <= 0 {
+		limit = 50
+	}
+	if limit > 200 {
+		limit = 200
+	}
+
+	q := `
+SELECT
+  r.id,
+  r.created_at,
+  COALESCE(r.requested_by_user_id, 0),
+  r.source_error_id,
+  COALESCE(r.mode, ''),
+  r.pinned_account_id,
+  COALESCE(pa.name, ''),
+  COALESCE(r.status, ''),
+  r.started_at,
+  r.finished_at,
+  r.duration_ms,
+  r.success,
+  r.http_status_code,
+  r.upstream_request_id,
+  r.used_account_id,
+  COALESCE(ua.name, ''),
+  r.response_preview,
+  r.response_truncated,
+  r.result_request_id,
+  r.result_error_id,
+  r.error_message
+FROM ops_retry_attempts r
+LEFT JOIN accounts pa ON r.pinned_account_id = pa.id
+LEFT JOIN accounts ua ON r.used_account_id = ua.id
+WHERE r.source_error_id = $1
+ORDER BY r.created_at DESC
+LIMIT $2`
+
+	rows, err := r.db.QueryContext(ctx, q, sourceErrorID, limit)
+	if err != nil {
+		return nil, err
+	}
+	defer func() { _ = rows.Close() }()
+
+	out := make([]*service.OpsRetryAttempt, 0, 16)
+	for rows.Next() {
+		var item service.OpsRetryAttempt
+		var pinnedAccountID sql.NullInt64
+		var pinnedAccountName string
+		var requestedBy sql.NullInt64
+		var startedAt sql.NullTime
+		var finishedAt sql.NullTime
+		var durationMs sql.NullInt64
+		var success sql.NullBool
+		var httpStatusCode sql.NullInt64
+		var upstreamRequestID sql.NullString
+		var usedAccountID sql.NullInt64
+		var usedAccountName string
+		var responsePreview sql.NullString
+		var responseTruncated sql.NullBool
+		var resultRequestID sql.NullString
+		var resultErrorID sql.NullInt64
+		var errorMessage sql.NullString
+
+		if err := rows.Scan(
+			&item.ID,
+			&item.CreatedAt,
+			&requestedBy,
+			&item.SourceErrorID,
+			&item.Mode,
+			&pinnedAccountID,
+			&pinnedAccountName,
+			&item.Status,
+			&startedAt,
+			&finishedAt,
+			&durationMs,
+			&success,
+			&httpStatusCode,
+			&upstreamRequestID,
+			&usedAccountID,
+			&usedAccountName,
+			&responsePreview,
+			&responseTruncated,
+			&resultRequestID,
+			&resultErrorID,
+			&errorMessage,
+		); err != nil {
+			return nil, err
+		}
+
+		item.RequestedByUserID = requestedBy.Int64
+		if pinnedAccountID.Valid {
+			v := pinnedAccountID.Int64
+			item.PinnedAccountID = &v
+		}
+		item.PinnedAccountName = pinnedAccountName
+		if startedAt.Valid {
+			t := startedAt.Time
+			item.StartedAt = &t
+		}
+		if finishedAt.Valid {
+			t := finishedAt.Time
+			item.FinishedAt = &t
+		}
+		if durationMs.Valid {
+			v := durationMs.Int64
+			item.DurationMs = &v
+		}
+		if success.Valid {
+			v := success.Bool
+			item.Success = &v
+		}
+		if httpStatusCode.Valid {
+			v := int(httpStatusCode.Int64)
+			item.HTTPStatusCode = &v
+		}
+		if upstreamRequestID.Valid {
+			item.UpstreamRequestID = &upstreamRequestID.String
+		}
+		if usedAccountID.Valid {
+			v := usedAccountID.Int64
+			item.UsedAccountID = &v
+		}
+		item.UsedAccountName = usedAccountName
+		if responsePreview.Valid {
+			item.ResponsePreview = &responsePreview.String
+		}
+		if responseTruncated.Valid {
+			v := responseTruncated.Bool
+			item.ResponseTruncated = &v
+		}
+		if resultRequestID.Valid {
+			item.ResultRequestID = &resultRequestID.String
+		}
+		if resultErrorID.Valid {
+			v := resultErrorID.Int64
+			item.ResultErrorID = &v
+		}
+		if errorMessage.Valid {
+			item.ErrorMessage = &errorMessage.String
+		}
+		out = append(out, &item)
+	}
+	if err := rows.Err(); err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
+func (r *opsRepository) UpdateErrorResolution(ctx context.Context, errorID int64, resolved bool, resolvedByUserID *int64, resolvedRetryID *int64, resolvedAt *time.Time) error {
+	if r == nil || r.db == nil {
+		return fmt.Errorf("nil ops repository")
+	}
+	if errorID <= 0 {
+		return fmt.Errorf("invalid error id")
+	}
+
+	q := `
+UPDATE ops_error_logs
+SET
+  resolved = $2,
+  resolved_at = $3,
+  resolved_by_user_id = $4,
+  resolved_retry_id = $5
+WHERE id = $1`
+
+	at := sql.NullTime{}
+	if resolvedAt != nil && !resolvedAt.IsZero() {
+		at = sql.NullTime{Time: resolvedAt.UTC(), Valid: true}
+	} else if resolved {
+		now := time.Now().UTC()
+		at = sql.NullTime{Time: now, Valid: true}
+	}
+
+	_, err := r.db.ExecContext(
+		ctx,
+		q,
+		errorID,
+		resolved,
+		at,
+		nullInt64(resolvedByUserID),
+		nullInt64(resolvedRetryID),
+	)
+	return err
+}
+
 func buildOpsErrorLogsWhere(filter *service.OpsErrorLogFilter) (string, []any) {
-	clauses := make([]string, 0, 8)
-	args := make([]any, 0, 8)
+	clauses := make([]string, 0, 12)
+	args := make([]any, 0, 12)
 	clauses = append(clauses, "1=1")
 
 	phaseFilter := ""
 	if filter != nil {
 		phaseFilter = strings.TrimSpace(strings.ToLower(filter.Phase))
 	}
-	// ops_error_logs primarily stores client-visible error requests (status>=400),
+	// ops_error_logs stores client-visible error requests (status>=400),
 	// but we also persist "recovered" upstream errors (status<400) for upstream health visibility.
-	// By default, keep list endpoints scoped to client errors unless explicitly filtering upstream phase.
+	// If Resolved is not specified, do not filter by resolved state (backward-compatible).
+	resolvedFilter := (*bool)(nil)
+	if filter != nil {
+		resolvedFilter = filter.Resolved
+	}
+	// Keep list endpoints scoped to client errors unless explicitly filtering upstream phase.
 	if phaseFilter != "upstream" {
 		clauses = append(clauses, "COALESCE(status_code, 0) >= 400")
 	}
 
 	if filter.StartTime != nil && !filter.StartTime.IsZero() {
 		args = append(args, filter.StartTime.UTC())
-		clauses = append(clauses, "created_at >= $"+itoa(len(args)))
+		clauses = append(clauses, "e.created_at >= $"+itoa(len(args)))
 	}
 	if filter.EndTime != nil && !filter.EndTime.IsZero() {
 		args = append(args, filter.EndTime.UTC())
 		// Keep time-window semantics consistent with other ops queries: [start, end)
-		clauses = append(clauses, "created_at < $"+itoa(len(args)))
+		clauses = append(clauses, "e.created_at < $"+itoa(len(args)))
 	}
 	if p := strings.TrimSpace(filter.Platform); p != "" {
 		args = append(args, p)
@@ -643,10 +976,59 @@ func buildOpsErrorLogsWhere(filter *service.OpsErrorLogFilter) (string, []any) {
 		args = append(args, phase)
 		clauses = append(clauses, "error_phase = $"+itoa(len(args)))
 	}
+	if filter != nil {
+		if owner := strings.TrimSpace(strings.ToLower(filter.Owner)); owner != "" {
+			args = append(args, owner)
+			clauses = append(clauses, "LOWER(COALESCE(error_owner,'')) = $"+itoa(len(args)))
+		}
+		if source := strings.TrimSpace(strings.ToLower(filter.Source)); source != "" {
+			args = append(args, source)
+			clauses = append(clauses, "LOWER(COALESCE(error_source,'')) = $"+itoa(len(args)))
+		}
+	}
+	if resolvedFilter != nil {
+		args = append(args, *resolvedFilter)
+		clauses = append(clauses, "COALESCE(resolved,false) = $"+itoa(len(args)))
+	}
+
+	// View filter: errors vs excluded vs all.
+	// Excluded = upstream 429/529 and business-limited (quota/concurrency/billing) errors.
+	view := ""
+	if filter != nil {
+		view = strings.ToLower(strings.TrimSpace(filter.View))
+	}
+	switch view {
+	case "", "errors":
+		clauses = append(clauses, "COALESCE(is_business_limited,false) = false")
+		clauses = append(clauses, "COALESCE(upstream_status_code, status_code, 0) NOT IN (429, 529)")
+	case "excluded":
+		clauses = append(clauses, "(COALESCE(is_business_limited,false) = true OR COALESCE(upstream_status_code, status_code, 0) IN (429, 529))")
+	case "all":
+		// no-op
+	default:
+		// treat unknown as default 'errors'
+		clauses = append(clauses, "COALESCE(is_business_limited,false) = false")
+		clauses = append(clauses, "COALESCE(upstream_status_code, status_code, 0) NOT IN (429, 529)")
+	}
 	if len(filter.StatusCodes) > 0 {
 		args = append(args, pq.Array(filter.StatusCodes))
 		clauses = append(clauses, "COALESCE(upstream_status_code, status_code, 0) = ANY($"+itoa(len(args))+")")
+	} else if filter.StatusCodesOther {
+		// "Other" means: status codes not in the common list.
+		known := []int{400, 401, 403, 404, 409, 422, 429, 500, 502, 503, 504, 529}
+		args = append(args, pq.Array(known))
+		clauses = append(clauses, "NOT (COALESCE(upstream_status_code, status_code, 0) = ANY($"+itoa(len(args))+"))")
 	}
+	// Exact correlation keys (preferred for request↔upstream linkage).
+	if rid := strings.TrimSpace(filter.RequestID); rid != "" {
+		args = append(args, rid)
+		clauses = append(clauses, "COALESCE(request_id,'') = $"+itoa(len(args)))
+	}
+	if crid := strings.TrimSpace(filter.ClientRequestID); crid != "" {
+		args = append(args, crid)
+		clauses = append(clauses, "COALESCE(client_request_id,'') = $"+itoa(len(args)))
+	}
+
 	if q := strings.TrimSpace(filter.Query); q != "" {
 		like := "%" + q + "%"
 		args = append(args, like)
@@ -654,6 +1036,13 @@ func buildOpsErrorLogsWhere(filter *service.OpsErrorLogFilter) (string, []any) {
 		clauses = append(clauses, "(request_id ILIKE $"+n+" OR client_request_id ILIKE $"+n+" OR error_message ILIKE $"+n+")")
 	}
 
+	if userQuery := strings.TrimSpace(filter.UserQuery); userQuery != "" {
+		like := "%" + userQuery + "%"
+		args = append(args, like)
+		n := itoa(len(args))
+		clauses = append(clauses, "u.email ILIKE $"+n)
+	}
+
 	return "WHERE " + strings.Join(clauses, " AND "), args
 }
 

backend/internal/repository/ops_repo_alerts.go

@@ -354,7 +354,7 @@ SELECT
   created_at
 FROM ops_alert_events
 ` + where + `
-ORDER BY fired_at DESC
+ORDER BY fired_at DESC, id DESC
 LIMIT ` + limitArg
 
 	rows, err := r.db.QueryContext(ctx, q, args...)
@@ -413,6 +413,43 @@ LIMIT ` + limitArg
 	return out, nil
 }
 
+func (r *opsRepository) GetAlertEventByID(ctx context.Context, eventID int64) (*service.OpsAlertEvent, error) {
+	if r == nil || r.db == nil {
+		return nil, fmt.Errorf("nil ops repository")
+	}
+	if eventID <= 0 {
+		return nil, fmt.Errorf("invalid event id")
+	}
+
+	q := `
+SELECT
+  id,
+  COALESCE(rule_id, 0),
+  COALESCE(severity, ''),
+  COALESCE(status, ''),
+  COALESCE(title, ''),
+  COALESCE(description, ''),
+  metric_value,
+  threshold_value,
+  dimensions,
+  fired_at,
+  resolved_at,
+  email_sent,
+  created_at
+FROM ops_alert_events
+WHERE id = $1`
+
+	row := r.db.QueryRowContext(ctx, q, eventID)
+	ev, err := scanOpsAlertEvent(row)
+	if err != nil {
+		if err == sql.ErrNoRows {
+			return nil, nil
+		}
+		return nil, err
+	}
+	return ev, nil
+}
+
 func (r *opsRepository) GetActiveAlertEvent(ctx context.Context, ruleID int64) (*service.OpsAlertEvent, error) {
 	if r == nil || r.db == nil {
 		return nil, fmt.Errorf("nil ops repository")
@@ -591,6 +628,121 @@ type opsAlertEventRow interface {
 	Scan(dest ...any) error
 }
 
+func (r *opsRepository) CreateAlertSilence(ctx context.Context, input *service.OpsAlertSilence) (*service.OpsAlertSilence, error) {
+	if r == nil || r.db == nil {
+		return nil, fmt.Errorf("nil ops repository")
+	}
+	if input == nil {
+		return nil, fmt.Errorf("nil input")
+	}
+	if input.RuleID <= 0 {
+		return nil, fmt.Errorf("invalid rule_id")
+	}
+	platform := strings.TrimSpace(input.Platform)
+	if platform == "" {
+		return nil, fmt.Errorf("invalid platform")
+	}
+	if input.Until.IsZero() {
+		return nil, fmt.Errorf("invalid until")
+	}
+
+	q := `
+INSERT INTO ops_alert_silences (
+  rule_id,
+  platform,
+  group_id,
+  region,
+  until,
+  reason,
+  created_by,
+  created_at
+) VALUES (
+  $1,$2,$3,$4,$5,$6,$7,NOW()
+)
+RETURNING id, rule_id, platform, group_id, region, until, COALESCE(reason,''), created_by, created_at`
+
+	row := r.db.QueryRowContext(
+		ctx,
+		q,
+		input.RuleID,
+		platform,
+		opsNullInt64(input.GroupID),
+		opsNullString(input.Region),
+		input.Until,
+		opsNullString(input.Reason),
+		opsNullInt64(input.CreatedBy),
+	)
+
+	var out service.OpsAlertSilence
+	var groupID sql.NullInt64
+	var region sql.NullString
+	var createdBy sql.NullInt64
+	if err := row.Scan(
+		&out.ID,
+		&out.RuleID,
+		&out.Platform,
+		&groupID,
+		&region,
+		&out.Until,
+		&out.Reason,
+		&createdBy,
+		&out.CreatedAt,
+	); err != nil {
+		return nil, err
+	}
+	if groupID.Valid {
+		v := groupID.Int64
+		out.GroupID = &v
+	}
+	if region.Valid {
+		v := strings.TrimSpace(region.String)
+		if v != "" {
+			out.Region = &v
+		}
+	}
+	if createdBy.Valid {
+		v := createdBy.Int64
+		out.CreatedBy = &v
+	}
+	return &out, nil
+}
+
+func (r *opsRepository) IsAlertSilenced(ctx context.Context, ruleID int64, platform string, groupID *int64, region *string, now time.Time) (bool, error) {
+	if r == nil || r.db == nil {
+		return false, fmt.Errorf("nil ops repository")
+	}
+	if ruleID <= 0 {
+		return false, fmt.Errorf("invalid rule id")
+	}
+	platform = strings.TrimSpace(platform)
+	if platform == "" {
+		return false, nil
+	}
+	if now.IsZero() {
+		now = time.Now().UTC()
+	}
+
+	q := `
+SELECT 1
+FROM ops_alert_silences
+WHERE rule_id = $1
+  AND platform = $2
+  AND (group_id IS NOT DISTINCT FROM $3)
+  AND (region IS NOT DISTINCT FROM $4)
+  AND until > $5
+LIMIT 1`
+
+	var dummy int
+	err := r.db.QueryRowContext(ctx, q, ruleID, platform, opsNullInt64(groupID), opsNullString(region), now).Scan(&dummy)
+	if err != nil {
+		if err == sql.ErrNoRows {
+			return false, nil
+		}
+		return false, err
+	}
+	return true, nil
+}
+
 func scanOpsAlertEvent(row opsAlertEventRow) (*service.OpsAlertEvent, error) {
 	var ev service.OpsAlertEvent
 	var metricValue sql.NullFloat64
@@ -652,6 +804,10 @@ func buildOpsAlertEventsWhere(filter *service.OpsAlertEventFilter) (string, []an
 		args = append(args, severity)
 		clauses = append(clauses, "severity = $"+itoa(len(args)))
 	}
+	if filter.EmailSent != nil {
+		args = append(args, *filter.EmailSent)
+		clauses = append(clauses, "email_sent = $"+itoa(len(args)))
+	}
 	if filter.StartTime != nil && !filter.StartTime.IsZero() {
 		args = append(args, *filter.StartTime)
 		clauses = append(clauses, "fired_at >= $"+itoa(len(args)))
@@ -661,6 +817,14 @@ func buildOpsAlertEventsWhere(filter *service.OpsAlertEventFilter) (string, []an
 		clauses = append(clauses, "fired_at < $"+itoa(len(args)))
 	}
 
+	// Cursor pagination (descending by fired_at, then id)
+	if filter.BeforeFiredAt != nil && !filter.BeforeFiredAt.IsZero() && filter.BeforeID != nil && *filter.BeforeID > 0 {
+		args = append(args, *filter.BeforeFiredAt)
+		tsArg := "$" + itoa(len(args))
+		args = append(args, *filter.BeforeID)
+		idArg := "$" + itoa(len(args))
+		clauses = append(clauses, fmt.Sprintf("(fired_at < %s OR (fired_at = %s AND id < %s))", tsArg, tsArg, idArg))
+	}
 	// Dimensions are stored in JSONB. We filter best-effort without requiring GIN indexes.
 	if platform := strings.TrimSpace(filter.Platform); platform != "" {
 		args = append(args, platform)

backend/internal/repository/proxy_latency_cache.go

@@ -0,0 +1,74 @@
+package repository
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+
+	"github.com/Wei-Shaw/sub2api/internal/service"
+	"github.com/redis/go-redis/v9"
+)
+
+const proxyLatencyKeyPrefix = "proxy:latency:"
+
+func proxyLatencyKey(proxyID int64) string {
+	return fmt.Sprintf("%s%d", proxyLatencyKeyPrefix, proxyID)
+}
+
+type proxyLatencyCache struct {
+	rdb *redis.Client
+}
+
+func NewProxyLatencyCache(rdb *redis.Client) service.ProxyLatencyCache {
+	return &proxyLatencyCache{rdb: rdb}
+}
+
+func (c *proxyLatencyCache) GetProxyLatencies(ctx context.Context, proxyIDs []int64) (map[int64]*service.ProxyLatencyInfo, error) {
+	results := make(map[int64]*service.ProxyLatencyInfo)
+	if len(proxyIDs) == 0 {
+		return results, nil
+	}
+
+	keys := make([]string, 0, len(proxyIDs))
+	for _, id := range proxyIDs {
+		keys = append(keys, proxyLatencyKey(id))
+	}
+
+	values, err := c.rdb.MGet(ctx, keys...).Result()
+	if err != nil {
+		return results, err
+	}
+
+	for i, raw := range values {
+		if raw == nil {
+			continue
+		}
+		var payload []byte
+		switch v := raw.(type) {
+		case string:
+			payload = []byte(v)
+		case []byte:
+			payload = v
+		default:
+			continue
+		}
+		var info service.ProxyLatencyInfo
+		if err := json.Unmarshal(payload, &info); err != nil {
+			continue
+		}
+		results[proxyIDs[i]] = &info
+	}
+
+	return results, nil
+}
+
+func (c *proxyLatencyCache) SetProxyLatency(ctx context.Context, proxyID int64, info *service.ProxyLatencyInfo) error {
+	if info == nil {
+		return nil
+	}
+	payload, err := json.Marshal(info)
+	if err != nil {
+		return err
+	}
+	return c.rdb.Set(ctx, proxyLatencyKey(proxyID), payload, 0).Err()
+}

backend/internal/repository/proxy_probe_service.go

@@ -34,7 +34,10 @@ func NewProxyExitInfoProber(cfg *config.Config) service.ProxyExitInfoProber {
 	}
 }
 
-const defaultIPInfoURL = "https://ipinfo.io/json"
+const (
+	defaultIPInfoURL         = "https://ipinfo.io/json"
+	defaultProxyProbeTimeout = 30 * time.Second
+)
 
 type proxyProbeService struct {
 	ipInfoURL          string
@@ -46,7 +49,7 @@ type proxyProbeService struct {
 func (s *proxyProbeService) ProbeProxy(ctx context.Context, proxyURL string) (*service.ProxyExitInfo, int64, error) {
 	client, err := httpclient.GetClient(httpclient.Options{
 		ProxyURL:           proxyURL,
-		Timeout:            15 * time.Second,
+		Timeout:            defaultProxyProbeTimeout,
 		InsecureSkipVerify: s.insecureSkipVerify,
 		ProxyStrict:        true,
 		ValidateResolvedIP: s.validateResolvedIP,

backend/internal/repository/proxy_repo.go

@@ -219,12 +219,54 @@ func (r *proxyRepository) ExistsByHostPortAuth(ctx context.Context, host string,
 // CountAccountsByProxyID returns the number of accounts using a specific proxy
 func (r *proxyRepository) CountAccountsByProxyID(ctx context.Context, proxyID int64) (int64, error) {
 	var count int64
-	if err := scanSingleRow(ctx, r.sql, "SELECT COUNT(*) FROM accounts WHERE proxy_id = $1", []any{proxyID}, &count); err != nil {
+	if err := scanSingleRow(ctx, r.sql, "SELECT COUNT(*) FROM accounts WHERE proxy_id = $1 AND deleted_at IS NULL", []any{proxyID}, &count); err != nil {
 		return 0, err
 	}
 	return count, nil
 }
 
+func (r *proxyRepository) ListAccountSummariesByProxyID(ctx context.Context, proxyID int64) ([]service.ProxyAccountSummary, error) {
+	rows, err := r.sql.QueryContext(ctx, `
+		SELECT id, name, platform, type, notes
+		FROM accounts
+		WHERE proxy_id = $1 AND deleted_at IS NULL
+		ORDER BY id DESC
+	`, proxyID)
+	if err != nil {
+		return nil, err
+	}
+	defer func() { _ = rows.Close() }()
+
+	out := make([]service.ProxyAccountSummary, 0)
+	for rows.Next() {
+		var (
+			id       int64
+			name     string
+			platform string
+			accType  string
+			notes    sql.NullString
+		)
+		if err := rows.Scan(&id, &name, &platform, &accType, &notes); err != nil {
+			return nil, err
+		}
+		var notesPtr *string
+		if notes.Valid {
+			notesPtr = &notes.String
+		}
+		out = append(out, service.ProxyAccountSummary{
+			ID:       id,
+			Name:     name,
+			Platform: platform,
+			Type:     accType,
+			Notes:    notesPtr,
+		})
+	}
+	if err := rows.Err(); err != nil {
+		return nil, err
+	}
+	return out, nil
+}
+
 // GetAccountCountsForProxies returns a map of proxy ID to account count for all proxies
 func (r *proxyRepository) GetAccountCountsForProxies(ctx context.Context) (counts map[int64]int64, err error) {
 	rows, err := r.sql.QueryContext(ctx, "SELECT proxy_id, COUNT(*) AS count FROM accounts WHERE proxy_id IS NOT NULL AND deleted_at IS NULL GROUP BY proxy_id")

backend/internal/repository/scheduler_snapshot_outbox_integration_test.go

@@ -27,7 +27,7 @@ func TestSchedulerSnapshotOutboxReplay(t *testing.T) {
 		RunMode: config.RunModeStandard,
 		Gateway: config.GatewayConfig{
 			Scheduling: config.GatewaySchedulingConfig{
-				OutboxPollIntervalSeconds: 1,
+				OutboxPollIntervalSeconds:  1,
 				FullRebuildIntervalSeconds: 0,
 				DbFallbackEnabled:          true,
 			},

backend/internal/repository/usage_log_repo.go

@@ -22,7 +22,7 @@ import (
 	"github.com/lib/pq"
 )
 
-const usageLogSelectColumns = "id, user_id, api_key_id, account_id, request_id, model, group_id, subscription_id, input_tokens, output_tokens, cache_creation_tokens, cache_read_tokens, cache_creation_5m_tokens, cache_creation_1h_tokens, input_cost, output_cost, cache_creation_cost, cache_read_cost, total_cost, actual_cost, rate_multiplier, billing_type, stream, duration_ms, first_token_ms, user_agent, ip_address, image_count, image_size, created_at"
+const usageLogSelectColumns = "id, user_id, api_key_id, account_id, request_id, model, group_id, subscription_id, input_tokens, output_tokens, cache_creation_tokens, cache_read_tokens, cache_creation_5m_tokens, cache_creation_1h_tokens, input_cost, output_cost, cache_creation_cost, cache_read_cost, total_cost, actual_cost, rate_multiplier, account_rate_multiplier, billing_type, stream, duration_ms, first_token_ms, user_agent, ip_address, image_count, image_size, created_at"
 
 type usageLogRepository struct {
 	client *dbent.Client
@@ -105,6 +105,7 @@ func (r *usageLogRepository) Create(ctx context.Context, log *service.UsageLog)
 			total_cost,
 			actual_cost,
 			rate_multiplier,
+			account_rate_multiplier,
 			billing_type,
 			stream,
 			duration_ms,
@@ -120,7 +121,7 @@ func (r *usageLogRepository) Create(ctx context.Context, log *service.UsageLog)
 			$8, $9, $10, $11,
 			$12, $13,
 			$14, $15, $16, $17, $18, $19,
-			$20, $21, $22, $23, $24, $25, $26, $27, $28, $29
+			$20, $21, $22, $23, $24, $25, $26, $27, $28, $29, $30
 		)
 		ON CONFLICT (request_id, api_key_id) DO NOTHING
 		RETURNING id, created_at
@@ -160,6 +161,7 @@ func (r *usageLogRepository) Create(ctx context.Context, log *service.UsageLog)
 		log.TotalCost,
 		log.ActualCost,
 		rateMultiplier,
+		log.AccountRateMultiplier,
 		log.BillingType,
 		log.Stream,
 		duration,
@@ -270,13 +272,13 @@ type DashboardStats = usagestats.DashboardStats
 
 func (r *usageLogRepository) GetDashboardStats(ctx context.Context) (*DashboardStats, error) {
 	stats := &DashboardStats{}
-	now := time.Now().UTC()
-	todayUTC := truncateToDayUTC(now)
+	now := timezone.Now()
+	todayStart := timezone.Today()
 
-	if err := r.fillDashboardEntityStats(ctx, stats, todayUTC, now); err != nil {
+	if err := r.fillDashboardEntityStats(ctx, stats, todayStart, now); err != nil {
 		return nil, err
 	}
-	if err := r.fillDashboardUsageStatsAggregated(ctx, stats, todayUTC, now); err != nil {
+	if err := r.fillDashboardUsageStatsAggregated(ctx, stats, todayStart, now); err != nil {
 		return nil, err
 	}
 
@@ -298,13 +300,13 @@ func (r *usageLogRepository) GetDashboardStatsWithRange(ctx context.Context, sta
 	}
 
 	stats := &DashboardStats{}
-	now := time.Now().UTC()
-	todayUTC := truncateToDayUTC(now)
+	now := timezone.Now()
+	todayStart := timezone.Today()
 
-	if err := r.fillDashboardEntityStats(ctx, stats, todayUTC, now); err != nil {
+	if err := r.fillDashboardEntityStats(ctx, stats, todayStart, now); err != nil {
 		return nil, err
 	}
-	if err := r.fillDashboardUsageStatsFromUsageLogs(ctx, stats, startUTC, endUTC, todayUTC, now); err != nil {
+	if err := r.fillDashboardUsageStatsFromUsageLogs(ctx, stats, startUTC, endUTC, todayStart, now); err != nil {
 		return nil, err
 	}
 
@@ -455,7 +457,7 @@ func (r *usageLogRepository) fillDashboardUsageStatsAggregated(ctx context.Conte
 		FROM usage_dashboard_hourly
 		WHERE bucket_start = $1
 	`
-	hourStart := now.UTC().Truncate(time.Hour)
+	hourStart := now.In(timezone.Location()).Truncate(time.Hour)
 	if err := scanSingleRow(ctx, r.sql, hourlyActiveQuery, []any{hourStart}, &stats.HourlyActiveUsers); err != nil {
 		if err != sql.ErrNoRows {
 			return err
@@ -835,7 +837,9 @@ func (r *usageLogRepository) GetAccountTodayStats(ctx context.Context, accountID
 		SELECT
 			COUNT(*) as requests,
 			COALESCE(SUM(input_tokens + output_tokens + cache_creation_tokens + cache_read_tokens), 0) as tokens,
-			COALESCE(SUM(actual_cost), 0) as cost
+			COALESCE(SUM(total_cost * COALESCE(account_rate_multiplier, 1)), 0) as cost,
+			COALESCE(SUM(total_cost), 0) as standard_cost,
+			COALESCE(SUM(actual_cost), 0) as user_cost
 		FROM usage_logs
 		WHERE account_id = $1 AND created_at >= $2
 	`
@@ -849,6 +853,8 @@ func (r *usageLogRepository) GetAccountTodayStats(ctx context.Context, accountID
 		&stats.Requests,
 		&stats.Tokens,
 		&stats.Cost,
+		&stats.StandardCost,
+		&stats.UserCost,
 	); err != nil {
 		return nil, err
 	}
@@ -861,7 +867,9 @@ func (r *usageLogRepository) GetAccountWindowStats(ctx context.Context, accountI
 		SELECT
 			COUNT(*) as requests,
 			COALESCE(SUM(input_tokens + output_tokens + cache_creation_tokens + cache_read_tokens), 0) as tokens,
-			COALESCE(SUM(actual_cost), 0) as cost
+			COALESCE(SUM(total_cost * COALESCE(account_rate_multiplier, 1)), 0) as cost,
+			COALESCE(SUM(total_cost), 0) as standard_cost,
+			COALESCE(SUM(actual_cost), 0) as user_cost
 		FROM usage_logs
 		WHERE account_id = $1 AND created_at >= $2
 	`
@@ -875,6 +883,8 @@ func (r *usageLogRepository) GetAccountWindowStats(ctx context.Context, accountI
 		&stats.Requests,
 		&stats.Tokens,
 		&stats.Cost,
+		&stats.StandardCost,
+		&stats.UserCost,
 	); err != nil {
 		return nil, err
 	}
@@ -1400,8 +1410,8 @@ func (r *usageLogRepository) GetBatchAPIKeyUsageStats(ctx context.Context, apiKe
 	return result, nil
 }
 
-// GetUsageTrendWithFilters returns usage trend data with optional user/api_key filters
-func (r *usageLogRepository) GetUsageTrendWithFilters(ctx context.Context, startTime, endTime time.Time, granularity string, userID, apiKeyID int64) (results []TrendDataPoint, err error) {
+// GetUsageTrendWithFilters returns usage trend data with optional filters
+func (r *usageLogRepository) GetUsageTrendWithFilters(ctx context.Context, startTime, endTime time.Time, granularity string, userID, apiKeyID, accountID, groupID int64, model string, stream *bool) (results []TrendDataPoint, err error) {
 	dateFormat := "YYYY-MM-DD"
 	if granularity == "hour" {
 		dateFormat = "YYYY-MM-DD HH24:00"
@@ -1430,6 +1440,22 @@ func (r *usageLogRepository) GetUsageTrendWithFilters(ctx context.Context, start
 		query += fmt.Sprintf(" AND api_key_id = $%d", len(args)+1)
 		args = append(args, apiKeyID)
 	}
+	if accountID > 0 {
+		query += fmt.Sprintf(" AND account_id = $%d", len(args)+1)
+		args = append(args, accountID)
+	}
+	if groupID > 0 {
+		query += fmt.Sprintf(" AND group_id = $%d", len(args)+1)
+		args = append(args, groupID)
+	}
+	if model != "" {
+		query += fmt.Sprintf(" AND model = $%d", len(args)+1)
+		args = append(args, model)
+	}
+	if stream != nil {
+		query += fmt.Sprintf(" AND stream = $%d", len(args)+1)
+		args = append(args, *stream)
+	}
 	query += " GROUP BY date ORDER BY date ASC"
 
 	rows, err := r.sql.QueryContext(ctx, query, args...)
@@ -1452,9 +1478,15 @@ func (r *usageLogRepository) GetUsageTrendWithFilters(ctx context.Context, start
 	return results, nil
 }
 
-// GetModelStatsWithFilters returns model statistics with optional user/api_key filters
-func (r *usageLogRepository) GetModelStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID int64) (results []ModelStat, err error) {
-	query := `
+// GetModelStatsWithFilters returns model statistics with optional filters
+func (r *usageLogRepository) GetModelStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, stream *bool) (results []ModelStat, err error) {
+	actualCostExpr := "COALESCE(SUM(actual_cost), 0) as actual_cost"
+	// 当仅按 account_id 聚合时，实际费用使用账号倍率（total_cost * account_rate_multiplier）。
+	if accountID > 0 && userID == 0 && apiKeyID == 0 {
+		actualCostExpr = "COALESCE(SUM(total_cost * COALESCE(account_rate_multiplier, 1)), 0) as actual_cost"
+	}
+
+	query := fmt.Sprintf(`
 		SELECT
 			model,
 			COUNT(*) as requests,
@@ -1462,10 +1494,10 @@ func (r *usageLogRepository) GetModelStatsWithFilters(ctx context.Context, start
 			COALESCE(SUM(output_tokens), 0) as output_tokens,
 			COALESCE(SUM(input_tokens + output_tokens + cache_creation_tokens + cache_read_tokens), 0) as total_tokens,
 			COALESCE(SUM(total_cost), 0) as cost,
-			COALESCE(SUM(actual_cost), 0) as actual_cost
+			%s
 		FROM usage_logs
 		WHERE created_at >= $1 AND created_at < $2
-	`
+	`, actualCostExpr)
 
 	args := []any{startTime, endTime}
 	if userID > 0 {
@@ -1480,6 +1512,14 @@ func (r *usageLogRepository) GetModelStatsWithFilters(ctx context.Context, start
 		query += fmt.Sprintf(" AND account_id = $%d", len(args)+1)
 		args = append(args, accountID)
 	}
+	if groupID > 0 {
+		query += fmt.Sprintf(" AND group_id = $%d", len(args)+1)
+		args = append(args, groupID)
+	}
+	if stream != nil {
+		query += fmt.Sprintf(" AND stream = $%d", len(args)+1)
+		args = append(args, *stream)
+	}
 	query += " GROUP BY model ORDER BY total_tokens DESC"
 
 	rows, err := r.sql.QueryContext(ctx, query, args...)
@@ -1587,12 +1627,14 @@ func (r *usageLogRepository) GetStatsWithFilters(ctx context.Context, filters Us
 			COALESCE(SUM(cache_creation_tokens + cache_read_tokens), 0) as total_cache_tokens,
 			COALESCE(SUM(total_cost), 0) as total_cost,
 			COALESCE(SUM(actual_cost), 0) as total_actual_cost,
+			COALESCE(SUM(total_cost * COALESCE(account_rate_multiplier, 1)), 0) as total_account_cost,
 			COALESCE(AVG(duration_ms), 0) as avg_duration_ms
 		FROM usage_logs
 		%s
 	`, buildWhere(conditions))
 
 	stats := &UsageStats{}
+	var totalAccountCost float64
 	if err := scanSingleRow(
 		ctx,
 		r.sql,
@@ -1604,10 +1646,14 @@ func (r *usageLogRepository) GetStatsWithFilters(ctx context.Context, filters Us
 		&stats.TotalCacheTokens,
 		&stats.TotalCost,
 		&stats.TotalActualCost,
+		&totalAccountCost,
 		&stats.AverageDurationMs,
 	); err != nil {
 		return nil, err
 	}
+	if filters.AccountID > 0 {
+		stats.TotalAccountCost = &totalAccountCost
+	}
 	stats.TotalTokens = stats.TotalInputTokens + stats.TotalOutputTokens + stats.TotalCacheTokens
 	return stats, nil
 }
@@ -1634,7 +1680,8 @@ func (r *usageLogRepository) GetAccountUsageStats(ctx context.Context, accountID
 			COUNT(*) as requests,
 			COALESCE(SUM(input_tokens + output_tokens + cache_creation_tokens + cache_read_tokens), 0) as tokens,
 			COALESCE(SUM(total_cost), 0) as cost,
-			COALESCE(SUM(actual_cost), 0) as actual_cost
+			COALESCE(SUM(total_cost * COALESCE(account_rate_multiplier, 1)), 0) as actual_cost,
+			COALESCE(SUM(actual_cost), 0) as user_cost
 		FROM usage_logs
 		WHERE account_id = $1 AND created_at >= $2 AND created_at < $3
 		GROUP BY date
@@ -1661,7 +1708,8 @@ func (r *usageLogRepository) GetAccountUsageStats(ctx context.Context, accountID
 		var tokens int64
 		var cost float64
 		var actualCost float64
-		if err = rows.Scan(&date, &requests, &tokens, &cost, &actualCost); err != nil {
+		var userCost float64
+		if err = rows.Scan(&date, &requests, &tokens, &cost, &actualCost, &userCost); err != nil {
 			return nil, err
 		}
 		t, _ := time.Parse("2006-01-02", date)
@@ -1672,19 +1720,21 @@ func (r *usageLogRepository) GetAccountUsageStats(ctx context.Context, accountID
 			Tokens:     tokens,
 			Cost:       cost,
 			ActualCost: actualCost,
+			UserCost:   userCost,
 		})
 	}
 	if err = rows.Err(); err != nil {
 		return nil, err
 	}
 
-	var totalActualCost, totalStandardCost float64
+	var totalAccountCost, totalUserCost, totalStandardCost float64
 	var totalRequests, totalTokens int64
 	var highestCostDay, highestRequestDay *AccountUsageHistory
 
 	for i := range history {
 		h := &history[i]
-		totalActualCost += h.ActualCost
+		totalAccountCost += h.ActualCost
+		totalUserCost += h.UserCost
 		totalStandardCost += h.Cost
 		totalRequests += h.Requests
 		totalTokens += h.Tokens
@@ -1711,11 +1761,13 @@ func (r *usageLogRepository) GetAccountUsageStats(ctx context.Context, accountID
 	summary := AccountUsageSummary{
 		Days:              daysCount,
 		ActualDaysUsed:    actualDaysUsed,
-		TotalCost:         totalActualCost,
+		TotalCost:         totalAccountCost,
+		TotalUserCost:     totalUserCost,
 		TotalStandardCost: totalStandardCost,
 		TotalRequests:     totalRequests,
 		TotalTokens:       totalTokens,
-		AvgDailyCost:      totalActualCost / float64(actualDaysUsed),
+		AvgDailyCost:      totalAccountCost / float64(actualDaysUsed),
+		AvgDailyUserCost:  totalUserCost / float64(actualDaysUsed),
 		AvgDailyRequests:  float64(totalRequests) / float64(actualDaysUsed),
 		AvgDailyTokens:    float64(totalTokens) / float64(actualDaysUsed),
 		AvgDurationMs:     avgDuration,
@@ -1727,11 +1779,13 @@ func (r *usageLogRepository) GetAccountUsageStats(ctx context.Context, accountID
 			summary.Today = &struct {
 				Date     string  `json:"date"`
 				Cost     float64 `json:"cost"`
+				UserCost float64 `json:"user_cost"`
 				Requests int64   `json:"requests"`
 				Tokens   int64   `json:"tokens"`
 			}{
 				Date:     history[i].Date,
 				Cost:     history[i].ActualCost,
+				UserCost: history[i].UserCost,
 				Requests: history[i].Requests,
 				Tokens:   history[i].Tokens,
 			}
@@ -1744,11 +1798,13 @@ func (r *usageLogRepository) GetAccountUsageStats(ctx context.Context, accountID
 			Date     string  `json:"date"`
 			Label    string  `json:"label"`
 			Cost     float64 `json:"cost"`
+			UserCost float64 `json:"user_cost"`
 			Requests int64   `json:"requests"`
 		}{
 			Date:     highestCostDay.Date,
 			Label:    highestCostDay.Label,
 			Cost:     highestCostDay.ActualCost,
+			UserCost: highestCostDay.UserCost,
 			Requests: highestCostDay.Requests,
 		}
 	}
@@ -1759,15 +1815,17 @@ func (r *usageLogRepository) GetAccountUsageStats(ctx context.Context, accountID
 			Label    string  `json:"label"`
 			Requests int64   `json:"requests"`
 			Cost     float64 `json:"cost"`
+			UserCost float64 `json:"user_cost"`
 		}{
 			Date:     highestRequestDay.Date,
 			Label:    highestRequestDay.Label,
 			Requests: highestRequestDay.Requests,
 			Cost:     highestRequestDay.ActualCost,
+			UserCost: highestRequestDay.UserCost,
 		}
 	}
 
-	models, err := r.GetModelStatsWithFilters(ctx, startTime, endTime, 0, 0, accountID)
+	models, err := r.GetModelStatsWithFilters(ctx, startTime, endTime, 0, 0, accountID, 0, nil)
 	if err != nil {
 		models = []ModelStat{}
 	}
@@ -1994,36 +2052,37 @@ func (r *usageLogRepository) loadSubscriptions(ctx context.Context, ids []int64)
 
 func scanUsageLog(scanner interface{ Scan(...any) error }) (*service.UsageLog, error) {
 	var (
-		id                  int64
-		userID              int64
-		apiKeyID            int64
-		accountID           int64
-		requestID           sql.NullString
-		model               string
-		groupID             sql.NullInt64
-		subscriptionID      sql.NullInt64
-		inputTokens         int
-		outputTokens        int
-		cacheCreationTokens int
-		cacheReadTokens     int
-		cacheCreation5m     int
-		cacheCreation1h     int
-		inputCost           float64
-		outputCost          float64
-		cacheCreationCost   float64
-		cacheReadCost       float64
-		totalCost           float64
-		actualCost          float64
-		rateMultiplier      float64
-		billingType         int16
-		stream              bool
-		durationMs          sql.NullInt64
-		firstTokenMs        sql.NullInt64
-		userAgent           sql.NullString
-		ipAddress           sql.NullString
-		imageCount          int
-		imageSize           sql.NullString
-		createdAt           time.Time
+		id                    int64
+		userID                int64
+		apiKeyID              int64
+		accountID             int64
+		requestID             sql.NullString
+		model                 string
+		groupID               sql.NullInt64
+		subscriptionID        sql.NullInt64
+		inputTokens           int
+		outputTokens          int
+		cacheCreationTokens   int
+		cacheReadTokens       int
+		cacheCreation5m       int
+		cacheCreation1h       int
+		inputCost             float64
+		outputCost            float64
+		cacheCreationCost     float64
+		cacheReadCost         float64
+		totalCost             float64
+		actualCost            float64
+		rateMultiplier        float64
+		accountRateMultiplier sql.NullFloat64
+		billingType           int16
+		stream                bool
+		durationMs            sql.NullInt64
+		firstTokenMs          sql.NullInt64
+		userAgent             sql.NullString
+		ipAddress             sql.NullString
+		imageCount            int
+		imageSize             sql.NullString
+		createdAt             time.Time
 	)
 
 	if err := scanner.Scan(
@@ -2048,6 +2107,7 @@ func scanUsageLog(scanner interface{ Scan(...any) error }) (*service.UsageLog, e
 		&totalCost,
 		&actualCost,
 		&rateMultiplier,
+		&accountRateMultiplier,
 		&billingType,
 		&stream,
 		&durationMs,
@@ -2080,6 +2140,7 @@ func scanUsageLog(scanner interface{ Scan(...any) error }) (*service.UsageLog, e
 		TotalCost:             totalCost,
 		ActualCost:            actualCost,
 		RateMultiplier:        rateMultiplier,
+		AccountRateMultiplier: nullFloat64Ptr(accountRateMultiplier),
 		BillingType:           int8(billingType),
 		Stream:                stream,
 		ImageCount:            imageCount,
@@ -2186,6 +2247,14 @@ func nullInt(v *int) sql.NullInt64 {
 	return sql.NullInt64{Int64: int64(*v), Valid: true}
 }
 
+func nullFloat64Ptr(v sql.NullFloat64) *float64 {
+	if !v.Valid {
+		return nil
+	}
+	out := v.Float64
+	return &out
+}
+
 func nullString(v *string) sql.NullString {
 	if v == nil || *v == "" {
 		return sql.NullString{}

backend/internal/repository/usage_log_repo_integration_test.go

@@ -11,6 +11,7 @@ import (
 
 	dbent "github.com/Wei-Shaw/sub2api/ent"
 	"github.com/Wei-Shaw/sub2api/internal/pkg/pagination"
+	"github.com/Wei-Shaw/sub2api/internal/pkg/timezone"
 	"github.com/Wei-Shaw/sub2api/internal/pkg/usagestats"
 	"github.com/Wei-Shaw/sub2api/internal/service"
 	"github.com/stretchr/testify/suite"
@@ -36,6 +37,12 @@ func TestUsageLogRepoSuite(t *testing.T) {
 	suite.Run(t, new(UsageLogRepoSuite))
 }
 
+// truncateToDayUTC 截断到 UTC 日期边界（测试辅助函数）
+func truncateToDayUTC(t time.Time) time.Time {
+	t = t.UTC()
+	return time.Date(t.Year(), t.Month(), t.Day(), 0, 0, 0, 0, time.UTC)
+}
+
 func (s *UsageLogRepoSuite) createUsageLog(user *service.User, apiKey *service.APIKey, account *service.Account, inputTokens, outputTokens int, cost float64, createdAt time.Time) *service.UsageLog {
 	log := &service.UsageLog{
 		UserID:       user.ID,
@@ -95,6 +102,34 @@ func (s *UsageLogRepoSuite) TestGetByID_NotFound() {
 	s.Require().Error(err, "expected error for non-existent ID")
 }
 
+func (s *UsageLogRepoSuite) TestGetByID_ReturnsAccountRateMultiplier() {
+	user := mustCreateUser(s.T(), s.client, &service.User{Email: "getbyid-mult@test.com"})
+	apiKey := mustCreateApiKey(s.T(), s.client, &service.APIKey{UserID: user.ID, Key: "sk-getbyid-mult", Name: "k"})
+	account := mustCreateAccount(s.T(), s.client, &service.Account{Name: "acc-getbyid-mult"})
+
+	m := 0.5
+	log := &service.UsageLog{
+		UserID:                user.ID,
+		APIKeyID:              apiKey.ID,
+		AccountID:             account.ID,
+		RequestID:             uuid.New().String(),
+		Model:                 "claude-3",
+		InputTokens:           10,
+		OutputTokens:          20,
+		TotalCost:             1.0,
+		ActualCost:            2.0,
+		AccountRateMultiplier: &m,
+		CreatedAt:             timezone.Today().Add(2 * time.Hour),
+	}
+	_, err := s.repo.Create(s.ctx, log)
+	s.Require().NoError(err)
+
+	got, err := s.repo.GetByID(s.ctx, log.ID)
+	s.Require().NoError(err)
+	s.Require().NotNil(got.AccountRateMultiplier)
+	s.Require().InEpsilon(0.5, *got.AccountRateMultiplier, 0.0001)
+}
+
 // --- Delete ---
 
 func (s *UsageLogRepoSuite) TestDelete() {
@@ -403,12 +438,49 @@ func (s *UsageLogRepoSuite) TestGetAccountTodayStats() {
 	apiKey := mustCreateApiKey(s.T(), s.client, &service.APIKey{UserID: user.ID, Key: "sk-acctoday", Name: "k"})
 	account := mustCreateAccount(s.T(), s.client, &service.Account{Name: "acc-today"})
 
-	s.createUsageLog(user, apiKey, account, 10, 20, 0.5, time.Now())
+	createdAt := timezone.Today().Add(1 * time.Hour)
+
+	m1 := 1.5
+	m2 := 0.0
+	_, err := s.repo.Create(s.ctx, &service.UsageLog{
+		UserID:                user.ID,
+		APIKeyID:              apiKey.ID,
+		AccountID:             account.ID,
+		RequestID:             uuid.New().String(),
+		Model:                 "claude-3",
+		InputTokens:           10,
+		OutputTokens:          20,
+		TotalCost:             1.0,
+		ActualCost:            2.0,
+		AccountRateMultiplier: &m1,
+		CreatedAt:             createdAt,
+	})
+	s.Require().NoError(err)
+	_, err = s.repo.Create(s.ctx, &service.UsageLog{
+		UserID:                user.ID,
+		APIKeyID:              apiKey.ID,
+		AccountID:             account.ID,
+		RequestID:             uuid.New().String(),
+		Model:                 "claude-3",
+		InputTokens:           5,
+		OutputTokens:          5,
+		TotalCost:             0.5,
+		ActualCost:            1.0,
+		AccountRateMultiplier: &m2,
+		CreatedAt:             createdAt,
+	})
+	s.Require().NoError(err)
 
 	stats, err := s.repo.GetAccountTodayStats(s.ctx, account.ID)
 	s.Require().NoError(err, "GetAccountTodayStats")
-	s.Require().Equal(int64(1), stats.Requests)
-	s.Require().Equal(int64(30), stats.Tokens)
+	s.Require().Equal(int64(2), stats.Requests)
+	s.Require().Equal(int64(40), stats.Tokens)
+	// account cost = SUM(total_cost * account_rate_multiplier)
+	s.Require().InEpsilon(1.5, stats.Cost, 0.0001)
+	// standard cost = SUM(total_cost)
+	s.Require().InEpsilon(1.5, stats.StandardCost, 0.0001)
+	// user cost = SUM(actual_cost)
+	s.Require().InEpsilon(3.0, stats.UserCost, 0.0001)
 }
 
 func (s *UsageLogRepoSuite) TestDashboardAggregationConsistency() {
@@ -416,8 +488,8 @@ func (s *UsageLogRepoSuite) TestDashboardAggregationConsistency() {
 	// 使用固定的时间偏移确保 hour1 和 hour2 在同一天且都在过去
 	// 选择当天 02:00 和 03:00 作为测试时间点（基于 now 的日期）
 	dayStart := truncateToDayUTC(now)
-	hour1 := dayStart.Add(2 * time.Hour)  // 当天 02:00
-	hour2 := dayStart.Add(3 * time.Hour)  // 当天 03:00
+	hour1 := dayStart.Add(2 * time.Hour) // 当天 02:00
+	hour2 := dayStart.Add(3 * time.Hour) // 当天 03:00
 	// 如果当前时间早于 hour2，则使用昨天的时间
 	if now.Before(hour2.Add(time.Hour)) {
 		dayStart = dayStart.Add(-24 * time.Hour)
@@ -872,17 +944,17 @@ func (s *UsageLogRepoSuite) TestGetUsageTrendWithFilters() {
 	endTime := base.Add(48 * time.Hour)
 
 	// Test with user filter
-	trend, err := s.repo.GetUsageTrendWithFilters(s.ctx, startTime, endTime, "day", user.ID, 0)
+	trend, err := s.repo.GetUsageTrendWithFilters(s.ctx, startTime, endTime, "day", user.ID, 0, 0, 0, "", nil)
 	s.Require().NoError(err, "GetUsageTrendWithFilters user filter")
 	s.Require().Len(trend, 2)
 
 	// Test with apiKey filter
-	trend, err = s.repo.GetUsageTrendWithFilters(s.ctx, startTime, endTime, "day", 0, apiKey.ID)
+	trend, err = s.repo.GetUsageTrendWithFilters(s.ctx, startTime, endTime, "day", 0, apiKey.ID, 0, 0, "", nil)
 	s.Require().NoError(err, "GetUsageTrendWithFilters apiKey filter")
 	s.Require().Len(trend, 2)
 
 	// Test with both filters
-	trend, err = s.repo.GetUsageTrendWithFilters(s.ctx, startTime, endTime, "day", user.ID, apiKey.ID)
+	trend, err = s.repo.GetUsageTrendWithFilters(s.ctx, startTime, endTime, "day", user.ID, apiKey.ID, 0, 0, "", nil)
 	s.Require().NoError(err, "GetUsageTrendWithFilters both filters")
 	s.Require().Len(trend, 2)
 }
@@ -899,7 +971,7 @@ func (s *UsageLogRepoSuite) TestGetUsageTrendWithFilters_HourlyGranularity() {
 	startTime := base.Add(-1 * time.Hour)
 	endTime := base.Add(3 * time.Hour)
 
-	trend, err := s.repo.GetUsageTrendWithFilters(s.ctx, startTime, endTime, "hour", user.ID, 0)
+	trend, err := s.repo.GetUsageTrendWithFilters(s.ctx, startTime, endTime, "hour", user.ID, 0, 0, 0, "", nil)
 	s.Require().NoError(err, "GetUsageTrendWithFilters hourly")
 	s.Require().Len(trend, 2)
 }
@@ -945,17 +1017,17 @@ func (s *UsageLogRepoSuite) TestGetModelStatsWithFilters() {
 	endTime := base.Add(2 * time.Hour)
 
 	// Test with user filter
-	stats, err := s.repo.GetModelStatsWithFilters(s.ctx, startTime, endTime, user.ID, 0, 0)
+	stats, err := s.repo.GetModelStatsWithFilters(s.ctx, startTime, endTime, user.ID, 0, 0, 0, nil)
 	s.Require().NoError(err, "GetModelStatsWithFilters user filter")
 	s.Require().Len(stats, 2)
 
 	// Test with apiKey filter
-	stats, err = s.repo.GetModelStatsWithFilters(s.ctx, startTime, endTime, 0, apiKey.ID, 0)
+	stats, err = s.repo.GetModelStatsWithFilters(s.ctx, startTime, endTime, 0, apiKey.ID, 0, 0, nil)
 	s.Require().NoError(err, "GetModelStatsWithFilters apiKey filter")
 	s.Require().Len(stats, 2)
 
 	// Test with account filter
-	stats, err = s.repo.GetModelStatsWithFilters(s.ctx, startTime, endTime, 0, 0, account.ID)
+	stats, err = s.repo.GetModelStatsWithFilters(s.ctx, startTime, endTime, 0, 0, account.ID, 0, nil)
 	s.Require().NoError(err, "GetModelStatsWithFilters account filter")
 	s.Require().Len(stats, 2)
 }

backend/internal/repository/wire.go

@@ -69,6 +69,7 @@ var ProviderSet = wire.NewSet(
 	NewGeminiTokenCache,
 	NewSchedulerCache,
 	NewSchedulerOutboxRepository,
+	NewProxyLatencyCache,
 
 	// HTTP service ports (DI Strategy A: return interface directly)
 	NewTurnstileVerifier,

backend/internal/server/api_contract_test.go

@@ -239,9 +239,10 @@ func TestAPIContracts(t *testing.T) {
 							"cache_creation_cost": 0,
 							"cache_read_cost": 0,
 							"total_cost": 0.5,
-							"actual_cost": 0.5,
-							"rate_multiplier": 1,
-							"billing_type": 0,
+						"actual_cost": 0.5,
+						"rate_multiplier": 1,
+						"account_rate_multiplier": null,
+						"billing_type": 0,
 							"stream": true,
 							"duration_ms": 100,
 							"first_token_ms": 50,
@@ -262,11 +263,11 @@ func TestAPIContracts(t *testing.T) {
 			name: "GET /api/v1/admin/settings",
 			setup: func(t *testing.T, deps *contractDeps) {
 				t.Helper()
-					deps.settingRepo.SetAll(map[string]string{
-						service.SettingKeyRegistrationEnabled: "true",
-						service.SettingKeyEmailVerifyEnabled:  "false",
+				deps.settingRepo.SetAll(map[string]string{
+					service.SettingKeyRegistrationEnabled: "true",
+					service.SettingKeyEmailVerifyEnabled:  "false",
 
-						service.SettingKeySMTPHost:     "smtp.example.com",
+					service.SettingKeySMTPHost:     "smtp.example.com",
 					service.SettingKeySMTPPort:     "587",
 					service.SettingKeySMTPUsername: "user",
 					service.SettingKeySMTPPassword: "secret",
@@ -285,15 +286,15 @@ func TestAPIContracts(t *testing.T) {
 					service.SettingKeyContactInfo:  "support",
 					service.SettingKeyDocURL:       "https://docs.example.com",
 
-						service.SettingKeyDefaultConcurrency: "5",
-						service.SettingKeyDefaultBalance:     "1.25",
+					service.SettingKeyDefaultConcurrency: "5",
+					service.SettingKeyDefaultBalance:     "1.25",
 
-						service.SettingKeyOpsMonitoringEnabled:         "false",
-						service.SettingKeyOpsRealtimeMonitoringEnabled: "true",
-						service.SettingKeyOpsQueryModeDefault:          "auto",
-						service.SettingKeyOpsMetricsIntervalSeconds:    "60",
-					})
-				},
+					service.SettingKeyOpsMonitoringEnabled:         "false",
+					service.SettingKeyOpsRealtimeMonitoringEnabled: "true",
+					service.SettingKeyOpsQueryModeDefault:          "auto",
+					service.SettingKeyOpsMetricsIntervalSeconds:    "60",
+				})
+			},
 			method:     http.MethodGet,
 			path:       "/api/v1/admin/settings",
 			wantStatus: http.StatusOK,
@@ -435,7 +436,7 @@ func newContractDeps(t *testing.T) *contractDeps {
 	settingRepo := newStubSettingRepo()
 	settingService := service.NewSettingService(settingRepo, cfg)
 
-	adminService := service.NewAdminService(userRepo, groupRepo, &accountRepo, proxyRepo, apiKeyRepo, redeemRepo, nil, nil, nil)
+	adminService := service.NewAdminService(userRepo, groupRepo, &accountRepo, proxyRepo, apiKeyRepo, redeemRepo, nil, nil, nil, nil)
 	authHandler := handler.NewAuthHandler(cfg, nil, userService, settingService, nil)
 	apiKeyHandler := handler.NewAPIKeyHandler(apiKeyService)
 	usageHandler := handler.NewUsageHandler(usageService, apiKeyService)
@@ -858,6 +859,10 @@ func (stubProxyRepo) CountAccountsByProxyID(ctx context.Context, proxyID int64)
 	return 0, errors.New("not implemented")
 }
 
+func (stubProxyRepo) ListAccountSummariesByProxyID(ctx context.Context, proxyID int64) ([]service.ProxyAccountSummary, error) {
+	return nil, errors.New("not implemented")
+}
+
 type stubRedeemCodeRepo struct{}
 
 func (stubRedeemCodeRepo) Create(ctx context.Context, code *service.RedeemCode) error {
@@ -1229,11 +1234,11 @@ func (r *stubUsageLogRepo) GetDashboardStats(ctx context.Context) (*usagestats.D
 	return nil, errors.New("not implemented")
 }
 
-func (r *stubUsageLogRepo) GetUsageTrendWithFilters(ctx context.Context, startTime, endTime time.Time, granularity string, userID, apiKeyID int64) ([]usagestats.TrendDataPoint, error) {
+func (r *stubUsageLogRepo) GetUsageTrendWithFilters(ctx context.Context, startTime, endTime time.Time, granularity string, userID, apiKeyID, accountID, groupID int64, model string, stream *bool) ([]usagestats.TrendDataPoint, error) {
 	return nil, errors.New("not implemented")
 }
 
-func (r *stubUsageLogRepo) GetModelStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID int64) ([]usagestats.ModelStat, error) {
+func (r *stubUsageLogRepo) GetModelStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, stream *bool) ([]usagestats.ModelStat, error) {
 	return nil, errors.New("not implemented")
 }
 

backend/internal/server/routes/admin.go

@@ -81,6 +81,9 @@ func registerOpsRoutes(admin *gin.RouterGroup, h *handler.Handlers) {
 		ops.PUT("/alert-rules/:id", h.Admin.Ops.UpdateAlertRule)
 		ops.DELETE("/alert-rules/:id", h.Admin.Ops.DeleteAlertRule)
 		ops.GET("/alert-events", h.Admin.Ops.ListAlertEvents)
+		ops.GET("/alert-events/:id", h.Admin.Ops.GetAlertEvent)
+		ops.PUT("/alert-events/:id/status", h.Admin.Ops.UpdateAlertEventStatus)
+		ops.POST("/alert-silences", h.Admin.Ops.CreateAlertSilence)
 
 		// Email notification config (DB-backed)
 		ops.GET("/email-notification/config", h.Admin.Ops.GetEmailNotificationConfig)
@@ -110,10 +113,26 @@ func registerOpsRoutes(admin *gin.RouterGroup, h *handler.Handlers) {
 			ws.GET("/qps", h.Admin.Ops.QPSWSHandler)
 		}
 
-		// Error logs (MVP-1)
+		// Error logs (legacy)
 		ops.GET("/errors", h.Admin.Ops.GetErrorLogs)
 		ops.GET("/errors/:id", h.Admin.Ops.GetErrorLogByID)
+		ops.GET("/errors/:id/retries", h.Admin.Ops.ListRetryAttempts)
 		ops.POST("/errors/:id/retry", h.Admin.Ops.RetryErrorRequest)
+		ops.PUT("/errors/:id/resolve", h.Admin.Ops.UpdateErrorResolution)
+
+		// Request errors (client-visible failures)
+		ops.GET("/request-errors", h.Admin.Ops.ListRequestErrors)
+		ops.GET("/request-errors/:id", h.Admin.Ops.GetRequestError)
+		ops.GET("/request-errors/:id/upstream-errors", h.Admin.Ops.ListRequestErrorUpstreamErrors)
+		ops.POST("/request-errors/:id/retry-client", h.Admin.Ops.RetryRequestErrorClient)
+		ops.POST("/request-errors/:id/upstream-errors/:idx/retry", h.Admin.Ops.RetryRequestErrorUpstreamEvent)
+		ops.PUT("/request-errors/:id/resolve", h.Admin.Ops.ResolveRequestError)
+
+		// Upstream errors (independent upstream failures)
+		ops.GET("/upstream-errors", h.Admin.Ops.ListUpstreamErrors)
+		ops.GET("/upstream-errors/:id", h.Admin.Ops.GetUpstreamError)
+		ops.POST("/upstream-errors/:id/retry", h.Admin.Ops.RetryUpstreamError)
+		ops.PUT("/upstream-errors/:id/resolve", h.Admin.Ops.ResolveUpstreamError)
 
 		// Request drilldown (success + error)
 		ops.GET("/requests", h.Admin.Ops.ListRequestDetails)
@@ -250,6 +269,7 @@ func registerProxyRoutes(admin *gin.RouterGroup, h *handler.Handlers) {
 		proxies.POST("/:id/test", h.Admin.Proxy.Test)
 		proxies.GET("/:id/stats", h.Admin.Proxy.GetStats)
 		proxies.GET("/:id/accounts", h.Admin.Proxy.GetProxyAccounts)
+		proxies.POST("/batch-delete", h.Admin.Proxy.BatchDelete)
 		proxies.POST("/batch", h.Admin.Proxy.BatchCreate)
 	}
 }

backend/internal/service/account.go

@@ -9,16 +9,19 @@ import (
 )
 
 type Account struct {
-	ID                 int64
-	Name               string
-	Notes              *string
-	Platform           string
-	Type               string
-	Credentials        map[string]any
-	Extra              map[string]any
-	ProxyID            *int64
-	Concurrency        int
-	Priority           int
+	ID          int64
+	Name        string
+	Notes       *string
+	Platform    string
+	Type        string
+	Credentials map[string]any
+	Extra       map[string]any
+	ProxyID     *int64
+	Concurrency int
+	Priority    int
+	// RateMultiplier 账号计费倍率（>=0，允许 0 表示该账号计费为 0）。
+	// 使用指针用于兼容旧版本调度缓存（Redis）中缺字段的情况：nil 表示按 1.0 处理。
+	RateMultiplier     *float64
 	Status             string
 	ErrorMessage       string
 	LastUsedAt         *time.Time
@@ -57,6 +60,20 @@ func (a *Account) IsActive() bool {
 	return a.Status == StatusActive
 }
 
+// BillingRateMultiplier 返回账号计费倍率。
+// - nil 表示未配置/旧缓存缺字段，按 1.0 处理
+// - 允许 0，表示该账号计费为 0
+// - 负数属于非法数据，出于安全考虑按 1.0 处理
+func (a *Account) BillingRateMultiplier() float64 {
+	if a == nil || a.RateMultiplier == nil {
+		return 1.0
+	}
+	if *a.RateMultiplier < 0 {
+		return 1.0
+	}
+	return *a.RateMultiplier
+}
+
 func (a *Account) IsSchedulable() bool {
 	if !a.IsActive() || !a.Schedulable {
 		return false

backend/internal/service/account_billing_rate_multiplier_test.go

@@ -0,0 +1,27 @@
+package service
+
+import (
+	"encoding/json"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+)
+
+func TestAccount_BillingRateMultiplier_DefaultsToOneWhenNil(t *testing.T) {
+	var a Account
+	require.NoError(t, json.Unmarshal([]byte(`{"id":1,"name":"acc","status":"active"}`), &a))
+	require.Nil(t, a.RateMultiplier)
+	require.Equal(t, 1.0, a.BillingRateMultiplier())
+}
+
+func TestAccount_BillingRateMultiplier_AllowsZero(t *testing.T) {
+	v := 0.0
+	a := Account{RateMultiplier: &v}
+	require.Equal(t, 0.0, a.BillingRateMultiplier())
+}
+
+func TestAccount_BillingRateMultiplier_NegativeFallsBackToOne(t *testing.T) {
+	v := -1.0
+	a := Account{RateMultiplier: &v}
+	require.Equal(t, 1.0, a.BillingRateMultiplier())
+}

backend/internal/service/account_service.go

@@ -63,14 +63,15 @@ type AccountRepository interface {
 // AccountBulkUpdate describes the fields that can be updated in a bulk operation.
 // Nil pointers mean "do not change".
 type AccountBulkUpdate struct {
-	Name        *string
-	ProxyID     *int64
-	Concurrency *int
-	Priority    *int
-	Status      *string
-	Schedulable *bool
-	Credentials map[string]any
-	Extra       map[string]any
+	Name           *string
+	ProxyID        *int64
+	Concurrency    *int
+	Priority       *int
+	RateMultiplier *float64
+	Status         *string
+	Schedulable    *bool
+	Credentials    map[string]any
+	Extra          map[string]any
 }
 
 // CreateAccountRequest 创建账号请求

backend/internal/service/account_usage_service.go

@@ -32,8 +32,8 @@ type UsageLogRepository interface {
 
 	// Admin dashboard stats
 	GetDashboardStats(ctx context.Context) (*usagestats.DashboardStats, error)
-	GetUsageTrendWithFilters(ctx context.Context, startTime, endTime time.Time, granularity string, userID, apiKeyID int64) ([]usagestats.TrendDataPoint, error)
-	GetModelStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID int64) ([]usagestats.ModelStat, error)
+	GetUsageTrendWithFilters(ctx context.Context, startTime, endTime time.Time, granularity string, userID, apiKeyID, accountID, groupID int64, model string, stream *bool) ([]usagestats.TrendDataPoint, error)
+	GetModelStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, stream *bool) ([]usagestats.ModelStat, error)
 	GetAPIKeyUsageTrend(ctx context.Context, startTime, endTime time.Time, granularity string, limit int) ([]usagestats.APIKeyUsageTrendPoint, error)
 	GetUserUsageTrend(ctx context.Context, startTime, endTime time.Time, granularity string, limit int) ([]usagestats.UserUsageTrendPoint, error)
 	GetBatchUserUsageStats(ctx context.Context, userIDs []int64) (map[int64]*usagestats.BatchUserUsageStats, error)
@@ -96,10 +96,16 @@ func NewUsageCache() *UsageCache {
 }
 
 // WindowStats 窗口期统计
+//
+// cost: 账号口径费用（total_cost * account_rate_multiplier）
+// standard_cost: 标准费用（total_cost，不含倍率）
+// user_cost: 用户/API Key 口径费用（actual_cost，受分组倍率影响）
 type WindowStats struct {
-	Requests int64   `json:"requests"`
-	Tokens   int64   `json:"tokens"`
-	Cost     float64 `json:"cost"`
+	Requests     int64   `json:"requests"`
+	Tokens       int64   `json:"tokens"`
+	Cost         float64 `json:"cost"`
+	StandardCost float64 `json:"standard_cost"`
+	UserCost     float64 `json:"user_cost"`
 }
 
 // UsageProgress 使用量进度
@@ -266,7 +272,7 @@ func (s *AccountUsageService) getGeminiUsage(ctx context.Context, account *Accou
 	}
 
 	dayStart := geminiDailyWindowStart(now)
-	stats, err := s.usageLogRepo.GetModelStatsWithFilters(ctx, dayStart, now, 0, 0, account.ID)
+	stats, err := s.usageLogRepo.GetModelStatsWithFilters(ctx, dayStart, now, 0, 0, account.ID, 0, nil)
 	if err != nil {
 		return nil, fmt.Errorf("get gemini usage stats failed: %w", err)
 	}
@@ -288,7 +294,7 @@ func (s *AccountUsageService) getGeminiUsage(ctx context.Context, account *Accou
 	// Minute window (RPM) - fixed-window approximation: current minute [truncate(now), truncate(now)+1m)
 	minuteStart := now.Truncate(time.Minute)
 	minuteResetAt := minuteStart.Add(time.Minute)
-	minuteStats, err := s.usageLogRepo.GetModelStatsWithFilters(ctx, minuteStart, now, 0, 0, account.ID)
+	minuteStats, err := s.usageLogRepo.GetModelStatsWithFilters(ctx, minuteStart, now, 0, 0, account.ID, 0, nil)
 	if err != nil {
 		return nil, fmt.Errorf("get gemini minute usage stats failed: %w", err)
 	}
@@ -377,9 +383,11 @@ func (s *AccountUsageService) addWindowStats(ctx context.Context, account *Accou
 		}
 
 		windowStats = &WindowStats{
-			Requests: stats.Requests,
-			Tokens:   stats.Tokens,
-			Cost:     stats.Cost,
+			Requests:     stats.Requests,
+			Tokens:       stats.Tokens,
+			Cost:         stats.Cost,
+			StandardCost: stats.StandardCost,
+			UserCost:     stats.UserCost,
 		}
 
 		// 缓存窗口统计（1 分钟）
@@ -403,9 +411,11 @@ func (s *AccountUsageService) GetTodayStats(ctx context.Context, accountID int64
 	}
 
 	return &WindowStats{
-		Requests: stats.Requests,
-		Tokens:   stats.Tokens,
-		Cost:     stats.Cost,
+		Requests:     stats.Requests,
+		Tokens:       stats.Tokens,
+		Cost:         stats.Cost,
+		StandardCost: stats.StandardCost,
+		UserCost:     stats.UserCost,
 	}, nil
 }
 

backend/internal/service/admin_service.go

@@ -54,7 +54,8 @@ type AdminService interface {
 	CreateProxy(ctx context.Context, input *CreateProxyInput) (*Proxy, error)
 	UpdateProxy(ctx context.Context, id int64, input *UpdateProxyInput) (*Proxy, error)
 	DeleteProxy(ctx context.Context, id int64) error
-	GetProxyAccounts(ctx context.Context, proxyID int64, page, pageSize int) ([]Account, int64, error)
+	BatchDeleteProxies(ctx context.Context, ids []int64) (*ProxyBatchDeleteResult, error)
+	GetProxyAccounts(ctx context.Context, proxyID int64) ([]ProxyAccountSummary, error)
 	CheckProxyExists(ctx context.Context, host string, port int, username, password string) (bool, error)
 	TestProxy(ctx context.Context, id int64) (*ProxyTestResult, error)
 
@@ -136,6 +137,7 @@ type CreateAccountInput struct {
 	ProxyID            *int64
 	Concurrency        int
 	Priority           int
+	RateMultiplier     *float64 // 账号计费倍率（>=0，允许 0）
 	GroupIDs           []int64
 	ExpiresAt          *int64
 	AutoPauseOnExpired *bool
@@ -151,8 +153,9 @@ type UpdateAccountInput struct {
 	Credentials           map[string]any
 	Extra                 map[string]any
 	ProxyID               *int64
-	Concurrency           *int // 使用指针区分"未提供"和"设置为0"
-	Priority              *int // 使用指针区分"未提供"和"设置为0"
+	Concurrency           *int     // 使用指针区分"未提供"和"设置为0"
+	Priority              *int     // 使用指针区分"未提供"和"设置为0"
+	RateMultiplier        *float64 // 账号计费倍率（>=0，允许 0）
 	Status                string
 	GroupIDs              *[]int64
 	ExpiresAt             *int64
@@ -162,16 +165,17 @@ type UpdateAccountInput struct {
 
 // BulkUpdateAccountsInput describes the payload for bulk updating accounts.
 type BulkUpdateAccountsInput struct {
-	AccountIDs  []int64
-	Name        string
-	ProxyID     *int64
-	Concurrency *int
-	Priority    *int
-	Status      string
-	Schedulable *bool
-	GroupIDs    *[]int64
-	Credentials map[string]any
-	Extra       map[string]any
+	AccountIDs     []int64
+	Name           string
+	ProxyID        *int64
+	Concurrency    *int
+	Priority       *int
+	RateMultiplier *float64 // 账号计费倍率（>=0，允许 0）
+	Status         string
+	Schedulable    *bool
+	GroupIDs       *[]int64
+	Credentials    map[string]any
+	Extra          map[string]any
 	// SkipMixedChannelCheck skips the mixed channel risk check when binding groups.
 	// This should only be set when the caller has explicitly confirmed the risk.
 	SkipMixedChannelCheck bool
@@ -220,6 +224,16 @@ type GenerateRedeemCodesInput struct {
 	ValidityDays int    // 订阅类型专用：有效天数
 }
 
+type ProxyBatchDeleteResult struct {
+	DeletedIDs []int64                   `json:"deleted_ids"`
+	Skipped    []ProxyBatchDeleteSkipped `json:"skipped"`
+}
+
+type ProxyBatchDeleteSkipped struct {
+	ID     int64  `json:"id"`
+	Reason string `json:"reason"`
+}
+
 // ProxyTestResult represents the result of testing a proxy
 type ProxyTestResult struct {
 	Success   bool   `json:"success"`
@@ -254,6 +268,7 @@ type adminServiceImpl struct {
 	redeemCodeRepo       RedeemCodeRepository
 	billingCacheService  *BillingCacheService
 	proxyProber          ProxyExitInfoProber
+	proxyLatencyCache    ProxyLatencyCache
 	authCacheInvalidator APIKeyAuthCacheInvalidator
 }
 
@@ -267,6 +282,7 @@ func NewAdminService(
 	redeemCodeRepo RedeemCodeRepository,
 	billingCacheService *BillingCacheService,
 	proxyProber ProxyExitInfoProber,
+	proxyLatencyCache ProxyLatencyCache,
 	authCacheInvalidator APIKeyAuthCacheInvalidator,
 ) AdminService {
 	return &adminServiceImpl{
@@ -278,6 +294,7 @@ func NewAdminService(
 		redeemCodeRepo:       redeemCodeRepo,
 		billingCacheService:  billingCacheService,
 		proxyProber:          proxyProber,
+		proxyLatencyCache:    proxyLatencyCache,
 		authCacheInvalidator: authCacheInvalidator,
 	}
 }
@@ -817,6 +834,12 @@ func (s *adminServiceImpl) CreateAccount(ctx context.Context, input *CreateAccou
 	} else {
 		account.AutoPauseOnExpired = true
 	}
+	if input.RateMultiplier != nil {
+		if *input.RateMultiplier < 0 {
+			return nil, errors.New("rate_multiplier must be >= 0")
+		}
+		account.RateMultiplier = input.RateMultiplier
+	}
 	if err := s.accountRepo.Create(ctx, account); err != nil {
 		return nil, err
 	}
@@ -869,6 +892,12 @@ func (s *adminServiceImpl) UpdateAccount(ctx context.Context, id int64, input *U
 	if input.Priority != nil {
 		account.Priority = *input.Priority
 	}
+	if input.RateMultiplier != nil {
+		if *input.RateMultiplier < 0 {
+			return nil, errors.New("rate_multiplier must be >= 0")
+		}
+		account.RateMultiplier = input.RateMultiplier
+	}
 	if input.Status != "" {
 		account.Status = input.Status
 	}
@@ -942,6 +971,12 @@ func (s *adminServiceImpl) BulkUpdateAccounts(ctx context.Context, input *BulkUp
 		}
 	}
 
+	if input.RateMultiplier != nil {
+		if *input.RateMultiplier < 0 {
+			return nil, errors.New("rate_multiplier must be >= 0")
+		}
+	}
+
 	// Prepare bulk updates for columns and JSONB fields.
 	repoUpdates := AccountBulkUpdate{
 		Credentials: input.Credentials,
@@ -959,6 +994,9 @@ func (s *adminServiceImpl) BulkUpdateAccounts(ctx context.Context, input *BulkUp
 	if input.Priority != nil {
 		repoUpdates.Priority = input.Priority
 	}
+	if input.RateMultiplier != nil {
+		repoUpdates.RateMultiplier = input.RateMultiplier
+	}
 	if input.Status != "" {
 		repoUpdates.Status = &input.Status
 	}
@@ -1069,6 +1107,7 @@ func (s *adminServiceImpl) ListProxiesWithAccountCount(ctx context.Context, page
 	if err != nil {
 		return nil, 0, err
 	}
+	s.attachProxyLatency(ctx, proxies)
 	return proxies, result.Total, nil
 }
 
@@ -1077,7 +1116,12 @@ func (s *adminServiceImpl) GetAllProxies(ctx context.Context) ([]Proxy, error) {
 }
 
 func (s *adminServiceImpl) GetAllProxiesWithAccountCount(ctx context.Context) ([]ProxyWithAccountCount, error) {
-	return s.proxyRepo.ListActiveWithAccountCount(ctx)
+	proxies, err := s.proxyRepo.ListActiveWithAccountCount(ctx)
+	if err != nil {
+		return nil, err
+	}
+	s.attachProxyLatency(ctx, proxies)
+	return proxies, nil
 }
 
 func (s *adminServiceImpl) GetProxy(ctx context.Context, id int64) (*Proxy, error) {
@@ -1097,6 +1141,8 @@ func (s *adminServiceImpl) CreateProxy(ctx context.Context, input *CreateProxyIn
 	if err := s.proxyRepo.Create(ctx, proxy); err != nil {
 		return nil, err
 	}
+	// Probe latency asynchronously so creation isn't blocked by network timeout.
+	go s.probeProxyLatency(context.Background(), proxy)
 	return proxy, nil
 }
 
@@ -1135,12 +1181,53 @@ func (s *adminServiceImpl) UpdateProxy(ctx context.Context, id int64, input *Upd
 }
 
 func (s *adminServiceImpl) DeleteProxy(ctx context.Context, id int64) error {
+	count, err := s.proxyRepo.CountAccountsByProxyID(ctx, id)
+	if err != nil {
+		return err
+	}
+	if count > 0 {
+		return ErrProxyInUse
+	}
 	return s.proxyRepo.Delete(ctx, id)
 }
 
-func (s *adminServiceImpl) GetProxyAccounts(ctx context.Context, proxyID int64, page, pageSize int) ([]Account, int64, error) {
-	// Return mock data for now - would need a dedicated repository method
-	return []Account{}, 0, nil
+func (s *adminServiceImpl) BatchDeleteProxies(ctx context.Context, ids []int64) (*ProxyBatchDeleteResult, error) {
+	result := &ProxyBatchDeleteResult{}
+	if len(ids) == 0 {
+		return result, nil
+	}
+
+	for _, id := range ids {
+		count, err := s.proxyRepo.CountAccountsByProxyID(ctx, id)
+		if err != nil {
+			result.Skipped = append(result.Skipped, ProxyBatchDeleteSkipped{
+				ID:     id,
+				Reason: err.Error(),
+			})
+			continue
+		}
+		if count > 0 {
+			result.Skipped = append(result.Skipped, ProxyBatchDeleteSkipped{
+				ID:     id,
+				Reason: ErrProxyInUse.Error(),
+			})
+			continue
+		}
+		if err := s.proxyRepo.Delete(ctx, id); err != nil {
+			result.Skipped = append(result.Skipped, ProxyBatchDeleteSkipped{
+				ID:     id,
+				Reason: err.Error(),
+			})
+			continue
+		}
+		result.DeletedIDs = append(result.DeletedIDs, id)
+	}
+
+	return result, nil
+}
+
+func (s *adminServiceImpl) GetProxyAccounts(ctx context.Context, proxyID int64) ([]ProxyAccountSummary, error) {
+	return s.proxyRepo.ListAccountSummariesByProxyID(ctx, proxyID)
 }
 
 func (s *adminServiceImpl) CheckProxyExists(ctx context.Context, host string, port int, username, password string) (bool, error) {
@@ -1240,12 +1327,24 @@ func (s *adminServiceImpl) TestProxy(ctx context.Context, id int64) (*ProxyTestR
 	proxyURL := proxy.URL()
 	exitInfo, latencyMs, err := s.proxyProber.ProbeProxy(ctx, proxyURL)
 	if err != nil {
+		s.saveProxyLatency(ctx, id, &ProxyLatencyInfo{
+			Success:   false,
+			Message:   err.Error(),
+			UpdatedAt: time.Now(),
+		})
 		return &ProxyTestResult{
 			Success: false,
 			Message: err.Error(),
 		}, nil
 	}
 
+	latency := latencyMs
+	s.saveProxyLatency(ctx, id, &ProxyLatencyInfo{
+		Success:   true,
+		LatencyMs: &latency,
+		Message:   "Proxy is accessible",
+		UpdatedAt: time.Now(),
+	})
 	return &ProxyTestResult{
 		Success:   true,
 		Message:   "Proxy is accessible",
@@ -1257,6 +1356,29 @@ func (s *adminServiceImpl) TestProxy(ctx context.Context, id int64) (*ProxyTestR
 	}, nil
 }
 
+func (s *adminServiceImpl) probeProxyLatency(ctx context.Context, proxy *Proxy) {
+	if s.proxyProber == nil || proxy == nil {
+		return
+	}
+	_, latencyMs, err := s.proxyProber.ProbeProxy(ctx, proxy.URL())
+	if err != nil {
+		s.saveProxyLatency(ctx, proxy.ID, &ProxyLatencyInfo{
+			Success:   false,
+			Message:   err.Error(),
+			UpdatedAt: time.Now(),
+		})
+		return
+	}
+
+	latency := latencyMs
+	s.saveProxyLatency(ctx, proxy.ID, &ProxyLatencyInfo{
+		Success:   true,
+		LatencyMs: &latency,
+		Message:   "Proxy is accessible",
+		UpdatedAt: time.Now(),
+	})
+}
+
 // checkMixedChannelRisk 检查分组中是否存在混合渠道（Antigravity + Anthropic）
 // 如果存在混合，返回错误提示用户确认
 func (s *adminServiceImpl) checkMixedChannelRisk(ctx context.Context, currentAccountID int64, currentAccountPlatform string, groupIDs []int64) error {
@@ -1306,6 +1428,46 @@ func (s *adminServiceImpl) checkMixedChannelRisk(ctx context.Context, currentAcc
 	return nil
 }
 
+func (s *adminServiceImpl) attachProxyLatency(ctx context.Context, proxies []ProxyWithAccountCount) {
+	if s.proxyLatencyCache == nil || len(proxies) == 0 {
+		return
+	}
+
+	ids := make([]int64, 0, len(proxies))
+	for i := range proxies {
+		ids = append(ids, proxies[i].ID)
+	}
+
+	latencies, err := s.proxyLatencyCache.GetProxyLatencies(ctx, ids)
+	if err != nil {
+		log.Printf("Warning: load proxy latency cache failed: %v", err)
+		return
+	}
+
+	for i := range proxies {
+		info := latencies[proxies[i].ID]
+		if info == nil {
+			continue
+		}
+		if info.Success {
+			proxies[i].LatencyStatus = "success"
+			proxies[i].LatencyMs = info.LatencyMs
+		} else {
+			proxies[i].LatencyStatus = "failed"
+		}
+		proxies[i].LatencyMessage = info.Message
+	}
+}
+
+func (s *adminServiceImpl) saveProxyLatency(ctx context.Context, proxyID int64, info *ProxyLatencyInfo) {
+	if s.proxyLatencyCache == nil || info == nil {
+		return
+	}
+	if err := s.proxyLatencyCache.SetProxyLatency(ctx, proxyID, info); err != nil {
+		log.Printf("Warning: store proxy latency cache failed: %v", err)
+	}
+}
+
 // getAccountPlatform 根据账号 platform 判断混合渠道检查用的平台标识
 func getAccountPlatform(accountPlatform string) string {
 	switch strings.ToLower(strings.TrimSpace(accountPlatform)) {

backend/internal/service/admin_service_bulk_update_test.go

@@ -12,9 +12,9 @@ import (
 
 type accountRepoStubForBulkUpdate struct {
 	accountRepoStub
-	bulkUpdateErr     error
-	bulkUpdateIDs     []int64
-	bindGroupErrByID  map[int64]error
+	bulkUpdateErr    error
+	bulkUpdateIDs    []int64
+	bindGroupErrByID map[int64]error
 }
 
 func (s *accountRepoStubForBulkUpdate) BulkUpdate(_ context.Context, ids []int64, _ AccountBulkUpdate) (int64, error) {

backend/internal/service/admin_service_delete_test.go

@@ -153,8 +153,10 @@ func (s *groupRepoStub) DeleteAccountGroupsByGroupID(ctx context.Context, groupI
 }
 
 type proxyRepoStub struct {
-	deleteErr  error
-	deletedIDs []int64
+	deleteErr    error
+	countErr     error
+	accountCount int64
+	deletedIDs   []int64
 }
 
 func (s *proxyRepoStub) Create(ctx context.Context, proxy *Proxy) error {
@@ -199,7 +201,14 @@ func (s *proxyRepoStub) ExistsByHostPortAuth(ctx context.Context, host string, p
 }
 
 func (s *proxyRepoStub) CountAccountsByProxyID(ctx context.Context, proxyID int64) (int64, error) {
-	panic("unexpected CountAccountsByProxyID call")
+	if s.countErr != nil {
+		return 0, s.countErr
+	}
+	return s.accountCount, nil
+}
+
+func (s *proxyRepoStub) ListAccountSummariesByProxyID(ctx context.Context, proxyID int64) ([]ProxyAccountSummary, error) {
+	panic("unexpected ListAccountSummariesByProxyID call")
 }
 
 type redeemRepoStub struct {
@@ -409,6 +418,15 @@ func TestAdminService_DeleteProxy_Idempotent(t *testing.T) {
 	require.Equal(t, []int64{404}, repo.deletedIDs)
 }
 
+func TestAdminService_DeleteProxy_InUse(t *testing.T) {
+	repo := &proxyRepoStub{accountCount: 2}
+	svc := &adminServiceImpl{proxyRepo: repo}
+
+	err := svc.DeleteProxy(context.Background(), 77)
+	require.ErrorIs(t, err, ErrProxyInUse)
+	require.Empty(t, repo.deletedIDs)
+}
+
 func TestAdminService_DeleteProxy_Error(t *testing.T) {
 	deleteErr := errors.New("delete failed")
 	repo := &proxyRepoStub{deleteErr: deleteErr}

backend/internal/service/antigravity_gateway_service.go

@@ -564,6 +564,10 @@ urlFallbackLoop:
 			}
 
 			upstreamReq, err := antigravity.NewAPIRequestWithURL(ctx, baseURL, action, accessToken, geminiBody)
+			// Capture upstream request body for ops retry of this attempt.
+			if c != nil {
+				c.Set(OpsUpstreamRequestBodyKey, string(geminiBody))
+			}
 			if err != nil {
 				return nil, err
 			}
@@ -574,6 +578,7 @@ urlFallbackLoop:
 				appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 					Platform:           account.Platform,
 					AccountID:          account.ID,
+					AccountName:        account.Name,
 					UpstreamStatusCode: 0,
 					Kind:               "request_error",
 					Message:            safeErr,
@@ -615,6 +620,7 @@ urlFallbackLoop:
 				appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 					Platform:           account.Platform,
 					AccountID:          account.ID,
+					AccountName:        account.Name,
 					UpstreamStatusCode: resp.StatusCode,
 					UpstreamRequestID:  resp.Header.Get("x-request-id"),
 					Kind:               "retry",
@@ -645,6 +651,7 @@ urlFallbackLoop:
 					appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 						Platform:           account.Platform,
 						AccountID:          account.ID,
+						AccountName:        account.Name,
 						UpstreamStatusCode: resp.StatusCode,
 						UpstreamRequestID:  resp.Header.Get("x-request-id"),
 						Kind:               "retry",
@@ -697,6 +704,7 @@ urlFallbackLoop:
 			appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 				Platform:           account.Platform,
 				AccountID:          account.ID,
+				AccountName:        account.Name,
 				UpstreamStatusCode: resp.StatusCode,
 				UpstreamRequestID:  resp.Header.Get("x-request-id"),
 				Kind:               "signature_error",
@@ -740,6 +748,7 @@ urlFallbackLoop:
 					appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 						Platform:           account.Platform,
 						AccountID:          account.ID,
+						AccountName:        account.Name,
 						UpstreamStatusCode: 0,
 						Kind:               "signature_retry_request_error",
 						Message:            sanitizeUpstreamErrorMessage(retryErr.Error()),
@@ -770,6 +779,7 @@ urlFallbackLoop:
 				appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 					Platform:           account.Platform,
 					AccountID:          account.ID,
+					AccountName:        account.Name,
 					UpstreamStatusCode: retryResp.StatusCode,
 					UpstreamRequestID:  retryResp.Header.Get("x-request-id"),
 					Kind:               kind,
@@ -817,6 +827,7 @@ urlFallbackLoop:
 				appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 					Platform:           account.Platform,
 					AccountID:          account.ID,
+					AccountName:        account.Name,
 					UpstreamStatusCode: resp.StatusCode,
 					UpstreamRequestID:  resp.Header.Get("x-request-id"),
 					Kind:               "failover",
@@ -1371,6 +1382,7 @@ urlFallbackLoop:
 				appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 					Platform:           account.Platform,
 					AccountID:          account.ID,
+					AccountName:        account.Name,
 					UpstreamStatusCode: 0,
 					Kind:               "request_error",
 					Message:            safeErr,
@@ -1412,6 +1424,7 @@ urlFallbackLoop:
 				appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 					Platform:           account.Platform,
 					AccountID:          account.ID,
+					AccountName:        account.Name,
 					UpstreamStatusCode: resp.StatusCode,
 					UpstreamRequestID:  resp.Header.Get("x-request-id"),
 					Kind:               "retry",
@@ -1442,6 +1455,7 @@ urlFallbackLoop:
 					appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 						Platform:           account.Platform,
 						AccountID:          account.ID,
+						AccountName:        account.Name,
 						UpstreamStatusCode: resp.StatusCode,
 						UpstreamRequestID:  resp.Header.Get("x-request-id"),
 						Kind:               "retry",
@@ -1543,6 +1557,7 @@ urlFallbackLoop:
 			appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 				Platform:           account.Platform,
 				AccountID:          account.ID,
+				AccountName:        account.Name,
 				UpstreamStatusCode: resp.StatusCode,
 				UpstreamRequestID:  requestID,
 				Kind:               "failover",
@@ -1559,6 +1574,7 @@ urlFallbackLoop:
 		appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 			Platform:           account.Platform,
 			AccountID:          account.ID,
+			AccountName:        account.Name,
 			UpstreamStatusCode: resp.StatusCode,
 			UpstreamRequestID:  requestID,
 			Kind:               "http_error",
@@ -2039,6 +2055,7 @@ func (s *AntigravityGatewayService) writeMappedClaudeError(c *gin.Context, accou
 	appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 		Platform:           account.Platform,
 		AccountID:          account.ID,
+		AccountName:        account.Name,
 		UpstreamStatusCode: upstreamStatus,
 		UpstreamRequestID:  upstreamRequestID,
 		Kind:               "http_error",

backend/internal/service/dashboard_service.go

@@ -124,16 +124,16 @@ func (s *DashboardService) GetDashboardStats(ctx context.Context) (*usagestats.D
 	return stats, nil
 }
 
-func (s *DashboardService) GetUsageTrendWithFilters(ctx context.Context, startTime, endTime time.Time, granularity string, userID, apiKeyID int64) ([]usagestats.TrendDataPoint, error) {
-	trend, err := s.usageRepo.GetUsageTrendWithFilters(ctx, startTime, endTime, granularity, userID, apiKeyID)
+func (s *DashboardService) GetUsageTrendWithFilters(ctx context.Context, startTime, endTime time.Time, granularity string, userID, apiKeyID, accountID, groupID int64, model string, stream *bool) ([]usagestats.TrendDataPoint, error) {
+	trend, err := s.usageRepo.GetUsageTrendWithFilters(ctx, startTime, endTime, granularity, userID, apiKeyID, accountID, groupID, model, stream)
 	if err != nil {
 		return nil, fmt.Errorf("get usage trend with filters: %w", err)
 	}
 	return trend, nil
 }
 
-func (s *DashboardService) GetModelStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID int64) ([]usagestats.ModelStat, error) {
-	stats, err := s.usageRepo.GetModelStatsWithFilters(ctx, startTime, endTime, userID, apiKeyID, 0)
+func (s *DashboardService) GetModelStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, stream *bool) ([]usagestats.ModelStat, error) {
+	stats, err := s.usageRepo.GetModelStatsWithFilters(ctx, startTime, endTime, userID, apiKeyID, accountID, groupID, stream)
 	if err != nil {
 		return nil, fmt.Errorf("get model stats with filters: %w", err)
 	}

backend/internal/service/gateway_service.go

@@ -1466,6 +1466,9 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A
 	for attempt := 1; attempt <= maxRetryAttempts; attempt++ {
 		// 构建上游请求（每次重试需要重新构建，因为请求体需要重新读取）
 		upstreamReq, err := s.buildUpstreamRequest(ctx, c, account, body, token, tokenType, reqModel)
+		// Capture upstream request body for ops retry of this attempt.
+		c.Set(OpsUpstreamRequestBodyKey, string(body))
+
 		if err != nil {
 			return nil, err
 		}
@@ -1482,6 +1485,7 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A
 			appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 				Platform:           account.Platform,
 				AccountID:          account.ID,
+				AccountName:        account.Name,
 				UpstreamStatusCode: 0,
 				Kind:               "request_error",
 				Message:            safeErr,
@@ -1506,6 +1510,7 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A
 					appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 						Platform:           account.Platform,
 						AccountID:          account.ID,
+						AccountName:        account.Name,
 						UpstreamStatusCode: resp.StatusCode,
 						UpstreamRequestID:  resp.Header.Get("x-request-id"),
 						Kind:               "signature_error",
@@ -1557,6 +1562,7 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A
 								appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 									Platform:           account.Platform,
 									AccountID:          account.ID,
+									AccountName:        account.Name,
 									UpstreamStatusCode: retryResp.StatusCode,
 									UpstreamRequestID:  retryResp.Header.Get("x-request-id"),
 									Kind:               "signature_retry_thinking",
@@ -1585,6 +1591,7 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A
 										appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 											Platform:           account.Platform,
 											AccountID:          account.ID,
+											AccountName:        account.Name,
 											UpstreamStatusCode: 0,
 											Kind:               "signature_retry_tools_request_error",
 											Message:            sanitizeUpstreamErrorMessage(retryErr2.Error()),
@@ -1643,6 +1650,7 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A
 				appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 					Platform:           account.Platform,
 					AccountID:          account.ID,
+					AccountName:        account.Name,
 					UpstreamStatusCode: resp.StatusCode,
 					UpstreamRequestID:  resp.Header.Get("x-request-id"),
 					Kind:               "retry",
@@ -1691,6 +1699,7 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A
 			appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 				Platform:           account.Platform,
 				AccountID:          account.ID,
+				AccountName:        account.Name,
 				UpstreamStatusCode: resp.StatusCode,
 				UpstreamRequestID:  resp.Header.Get("x-request-id"),
 				Kind:               "retry_exhausted_failover",
@@ -1757,6 +1766,7 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A
 				appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 					Platform:           account.Platform,
 					AccountID:          account.ID,
+					AccountName:        account.Name,
 					UpstreamStatusCode: resp.StatusCode,
 					UpstreamRequestID:  resp.Header.Get("x-request-id"),
 					Kind:               "failover_on_400",
@@ -2634,30 +2644,32 @@ func (s *GatewayService) RecordUsage(ctx context.Context, input *RecordUsageInpu
 	if result.ImageSize != "" {
 		imageSize = &result.ImageSize
 	}
+	accountRateMultiplier := account.BillingRateMultiplier()
 	usageLog := &UsageLog{
-		UserID:              user.ID,
-		APIKeyID:            apiKey.ID,
-		AccountID:           account.ID,
-		RequestID:           result.RequestID,
-		Model:               result.Model,
-		InputTokens:         result.Usage.InputTokens,
-		OutputTokens:        result.Usage.OutputTokens,
-		CacheCreationTokens: result.Usage.CacheCreationInputTokens,
-		CacheReadTokens:     result.Usage.CacheReadInputTokens,
-		InputCost:           cost.InputCost,
-		OutputCost:          cost.OutputCost,
-		CacheCreationCost:   cost.CacheCreationCost,
-		CacheReadCost:       cost.CacheReadCost,
-		TotalCost:           cost.TotalCost,
-		ActualCost:          cost.ActualCost,
-		RateMultiplier:      multiplier,
-		BillingType:         billingType,
-		Stream:              result.Stream,
-		DurationMs:          &durationMs,
-		FirstTokenMs:        result.FirstTokenMs,
-		ImageCount:          result.ImageCount,
-		ImageSize:           imageSize,
-		CreatedAt:           time.Now(),
+		UserID:                user.ID,
+		APIKeyID:              apiKey.ID,
+		AccountID:             account.ID,
+		RequestID:             result.RequestID,
+		Model:                 result.Model,
+		InputTokens:           result.Usage.InputTokens,
+		OutputTokens:          result.Usage.OutputTokens,
+		CacheCreationTokens:   result.Usage.CacheCreationInputTokens,
+		CacheReadTokens:       result.Usage.CacheReadInputTokens,
+		InputCost:             cost.InputCost,
+		OutputCost:            cost.OutputCost,
+		CacheCreationCost:     cost.CacheCreationCost,
+		CacheReadCost:         cost.CacheReadCost,
+		TotalCost:             cost.TotalCost,
+		ActualCost:            cost.ActualCost,
+		RateMultiplier:        multiplier,
+		AccountRateMultiplier: &accountRateMultiplier,
+		BillingType:           billingType,
+		Stream:                result.Stream,
+		DurationMs:            &durationMs,
+		FirstTokenMs:          result.FirstTokenMs,
+		ImageCount:            result.ImageCount,
+		ImageSize:             imageSize,
+		CreatedAt:             time.Now(),
 	}
 
 	// 添加 UserAgent

backend/internal/service/gemini_messages_compat_service.go

@@ -545,12 +545,19 @@ func (s *GeminiMessagesCompatService) Forward(ctx context.Context, c *gin.Contex
 		}
 		requestIDHeader = idHeader
 
+		// Capture upstream request body for ops retry of this attempt.
+		if c != nil {
+			// In this code path `body` is already the JSON sent to upstream.
+			c.Set(OpsUpstreamRequestBodyKey, string(body))
+		}
+
 		resp, err = s.httpUpstream.Do(upstreamReq, proxyURL, account.ID, account.Concurrency)
 		if err != nil {
 			safeErr := sanitizeUpstreamErrorMessage(err.Error())
 			appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 				Platform:           account.Platform,
 				AccountID:          account.ID,
+				AccountName:        account.Name,
 				UpstreamStatusCode: 0,
 				Kind:               "request_error",
 				Message:            safeErr,
@@ -588,6 +595,7 @@ func (s *GeminiMessagesCompatService) Forward(ctx context.Context, c *gin.Contex
 				appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 					Platform:           account.Platform,
 					AccountID:          account.ID,
+					AccountName:        account.Name,
 					UpstreamStatusCode: resp.StatusCode,
 					UpstreamRequestID:  upstreamReqID,
 					Kind:               "signature_error",
@@ -662,6 +670,7 @@ func (s *GeminiMessagesCompatService) Forward(ctx context.Context, c *gin.Contex
 				appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 					Platform:           account.Platform,
 					AccountID:          account.ID,
+					AccountName:        account.Name,
 					UpstreamStatusCode: resp.StatusCode,
 					UpstreamRequestID:  upstreamReqID,
 					Kind:               "retry",
@@ -711,6 +720,7 @@ func (s *GeminiMessagesCompatService) Forward(ctx context.Context, c *gin.Contex
 			appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 				Platform:           account.Platform,
 				AccountID:          account.ID,
+				AccountName:        account.Name,
 				UpstreamStatusCode: resp.StatusCode,
 				UpstreamRequestID:  upstreamReqID,
 				Kind:               "failover",
@@ -737,6 +747,7 @@ func (s *GeminiMessagesCompatService) Forward(ctx context.Context, c *gin.Contex
 			appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 				Platform:           account.Platform,
 				AccountID:          account.ID,
+				AccountName:        account.Name,
 				UpstreamStatusCode: resp.StatusCode,
 				UpstreamRequestID:  upstreamReqID,
 				Kind:               "failover",
@@ -972,12 +983,19 @@ func (s *GeminiMessagesCompatService) ForwardNative(ctx context.Context, c *gin.
 		}
 		requestIDHeader = idHeader
 
+		// Capture upstream request body for ops retry of this attempt.
+		if c != nil {
+			// In this code path `body` is already the JSON sent to upstream.
+			c.Set(OpsUpstreamRequestBodyKey, string(body))
+		}
+
 		resp, err = s.httpUpstream.Do(upstreamReq, proxyURL, account.ID, account.Concurrency)
 		if err != nil {
 			safeErr := sanitizeUpstreamErrorMessage(err.Error())
 			appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 				Platform:           account.Platform,
 				AccountID:          account.ID,
+				AccountName:        account.Name,
 				UpstreamStatusCode: 0,
 				Kind:               "request_error",
 				Message:            safeErr,
@@ -1036,6 +1054,7 @@ func (s *GeminiMessagesCompatService) ForwardNative(ctx context.Context, c *gin.
 				appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 					Platform:           account.Platform,
 					AccountID:          account.ID,
+					AccountName:        account.Name,
 					UpstreamStatusCode: resp.StatusCode,
 					UpstreamRequestID:  upstreamReqID,
 					Kind:               "retry",
@@ -1120,6 +1139,7 @@ func (s *GeminiMessagesCompatService) ForwardNative(ctx context.Context, c *gin.
 			appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 				Platform:           account.Platform,
 				AccountID:          account.ID,
+				AccountName:        account.Name,
 				UpstreamStatusCode: resp.StatusCode,
 				UpstreamRequestID:  requestID,
 				Kind:               "failover",
@@ -1143,6 +1163,7 @@ func (s *GeminiMessagesCompatService) ForwardNative(ctx context.Context, c *gin.
 			appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 				Platform:           account.Platform,
 				AccountID:          account.ID,
+				AccountName:        account.Name,
 				UpstreamStatusCode: resp.StatusCode,
 				UpstreamRequestID:  requestID,
 				Kind:               "failover",
@@ -1168,6 +1189,7 @@ func (s *GeminiMessagesCompatService) ForwardNative(ctx context.Context, c *gin.
 		appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 			Platform:           account.Platform,
 			AccountID:          account.ID,
+			AccountName:        account.Name,
 			UpstreamStatusCode: resp.StatusCode,
 			UpstreamRequestID:  requestID,
 			Kind:               "http_error",
@@ -1300,6 +1322,7 @@ func (s *GeminiMessagesCompatService) writeGeminiMappedError(c *gin.Context, acc
 	appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 		Platform:           account.Platform,
 		AccountID:          account.ID,
+		AccountName:        account.Name,
 		UpstreamStatusCode: upstreamStatus,
 		UpstreamRequestID:  upstreamRequestID,
 		Kind:               "http_error",

backend/internal/service/openai_gateway_service.go

@@ -664,6 +664,11 @@ func (s *OpenAIGatewayService) Forward(ctx context.Context, c *gin.Context, acco
 		proxyURL = account.Proxy.URL()
 	}
 
+	// Capture upstream request body for ops retry of this attempt.
+	if c != nil {
+		c.Set(OpsUpstreamRequestBodyKey, string(body))
+	}
+
 	// Send request
 	resp, err := s.httpUpstream.Do(upstreamReq, proxyURL, account.ID, account.Concurrency)
 	if err != nil {
@@ -673,6 +678,7 @@ func (s *OpenAIGatewayService) Forward(ctx context.Context, c *gin.Context, acco
 		appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 			Platform:           account.Platform,
 			AccountID:          account.ID,
+			AccountName:        account.Name,
 			UpstreamStatusCode: 0,
 			Kind:               "request_error",
 			Message:            safeErr,
@@ -707,6 +713,7 @@ func (s *OpenAIGatewayService) Forward(ctx context.Context, c *gin.Context, acco
 			appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 				Platform:           account.Platform,
 				AccountID:          account.ID,
+				AccountName:        account.Name,
 				UpstreamStatusCode: resp.StatusCode,
 				UpstreamRequestID:  resp.Header.Get("x-request-id"),
 				Kind:               "failover",
@@ -864,6 +871,7 @@ func (s *OpenAIGatewayService) handleErrorResponse(ctx context.Context, resp *ht
 		appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 			Platform:           account.Platform,
 			AccountID:          account.ID,
+			AccountName:        account.Name,
 			UpstreamStatusCode: resp.StatusCode,
 			UpstreamRequestID:  resp.Header.Get("x-request-id"),
 			Kind:               "http_error",
@@ -894,6 +902,7 @@ func (s *OpenAIGatewayService) handleErrorResponse(ctx context.Context, resp *ht
 	appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
 		Platform:           account.Platform,
 		AccountID:          account.ID,
+		AccountName:        account.Name,
 		UpstreamStatusCode: resp.StatusCode,
 		UpstreamRequestID:  resp.Header.Get("x-request-id"),
 		Kind:               kind,
@@ -1443,28 +1452,30 @@ func (s *OpenAIGatewayService) RecordUsage(ctx context.Context, input *OpenAIRec
 
 	// Create usage log
 	durationMs := int(result.Duration.Milliseconds())
+	accountRateMultiplier := account.BillingRateMultiplier()
 	usageLog := &UsageLog{
-		UserID:              user.ID,
-		APIKeyID:            apiKey.ID,
-		AccountID:           account.ID,
-		RequestID:           result.RequestID,
-		Model:               result.Model,
-		InputTokens:         actualInputTokens,
-		OutputTokens:        result.Usage.OutputTokens,
-		CacheCreationTokens: result.Usage.CacheCreationInputTokens,
-		CacheReadTokens:     result.Usage.CacheReadInputTokens,
-		InputCost:           cost.InputCost,
-		OutputCost:          cost.OutputCost,
-		CacheCreationCost:   cost.CacheCreationCost,
-		CacheReadCost:       cost.CacheReadCost,
-		TotalCost:           cost.TotalCost,
-		ActualCost:          cost.ActualCost,
-		RateMultiplier:      multiplier,
-		BillingType:         billingType,
-		Stream:              result.Stream,
-		DurationMs:          &durationMs,
-		FirstTokenMs:        result.FirstTokenMs,
-		CreatedAt:           time.Now(),
+		UserID:                user.ID,
+		APIKeyID:              apiKey.ID,
+		AccountID:             account.ID,
+		RequestID:             result.RequestID,
+		Model:                 result.Model,
+		InputTokens:           actualInputTokens,
+		OutputTokens:          result.Usage.OutputTokens,
+		CacheCreationTokens:   result.Usage.CacheCreationInputTokens,
+		CacheReadTokens:       result.Usage.CacheReadInputTokens,
+		InputCost:             cost.InputCost,
+		OutputCost:            cost.OutputCost,
+		CacheCreationCost:     cost.CacheCreationCost,
+		CacheReadCost:         cost.CacheReadCost,
+		TotalCost:             cost.TotalCost,
+		ActualCost:            cost.ActualCost,
+		RateMultiplier:        multiplier,
+		AccountRateMultiplier: &accountRateMultiplier,
+		BillingType:           billingType,
+		Stream:                result.Stream,
+		DurationMs:            &durationMs,
+		FirstTokenMs:          result.FirstTokenMs,
+		CreatedAt:             time.Now(),
 	}
 
 	// 添加 UserAgent

backend/internal/service/ops_alert_evaluator_service.go

@@ -206,7 +206,7 @@ func (s *OpsAlertEvaluatorService) evaluateOnce(interval time.Duration) {
 			continue
 		}
 
-		scopePlatform, scopeGroupID := parseOpsAlertRuleScope(rule.Filters)
+		scopePlatform, scopeGroupID, scopeRegion := parseOpsAlertRuleScope(rule.Filters)
 
 		windowMinutes := rule.WindowMinutes
 		if windowMinutes <= 0 {
@@ -236,6 +236,17 @@ func (s *OpsAlertEvaluatorService) evaluateOnce(interval time.Duration) {
 				continue
 			}
 
+			// Scoped silencing: if a matching silence exists, skip creating a firing event.
+			if s.opsService != nil {
+				platform := strings.TrimSpace(scopePlatform)
+				region := scopeRegion
+				if platform != "" {
+					if ok, err := s.opsService.IsAlertSilenced(ctx, rule.ID, platform, scopeGroupID, region, now); err == nil && ok {
+						continue
+					}
+				}
+			}
+
 			latestEvent, err := s.opsRepo.GetLatestAlertEvent(ctx, rule.ID)
 			if err != nil {
 				log.Printf("[OpsAlertEvaluator] get latest event failed (rule=%d): %v", rule.ID, err)
@@ -359,9 +370,9 @@ func requiredSustainedBreaches(sustainedMinutes int, interval time.Duration) int
 	return required
 }
 
-func parseOpsAlertRuleScope(filters map[string]any) (platform string, groupID *int64) {
+func parseOpsAlertRuleScope(filters map[string]any) (platform string, groupID *int64, region *string) {
 	if filters == nil {
-		return "", nil
+		return "", nil, nil
 	}
 	if v, ok := filters["platform"]; ok {
 		if s, ok := v.(string); ok {
@@ -392,7 +403,15 @@ func parseOpsAlertRuleScope(filters map[string]any) (platform string, groupID *i
 			}
 		}
 	}
-	return platform, groupID
+	if v, ok := filters["region"]; ok {
+		if s, ok := v.(string); ok {
+			vv := strings.TrimSpace(s)
+			if vv != "" {
+				region = &vv
+			}
+		}
+	}
+	return platform, groupID, region
 }
 
 func (s *OpsAlertEvaluatorService) computeRuleMetric(
@@ -504,16 +523,6 @@ func (s *OpsAlertEvaluatorService) computeRuleMetric(
 			return 0, false
 		}
 		return overview.UpstreamErrorRate * 100, true
-	case "p95_latency_ms":
-		if overview.Duration.P95 == nil {
-			return 0, false
-		}
-		return float64(*overview.Duration.P95), true
-	case "p99_latency_ms":
-		if overview.Duration.P99 == nil {
-			return 0, false
-		}
-		return float64(*overview.Duration.P99), true
 	default:
 		return 0, false
 	}

backend/internal/service/ops_alert_models.go

@@ -8,8 +8,9 @@ import "time"
 // with the existing ops dashboard frontend (backup style).
 
 const (
-	OpsAlertStatusFiring   = "firing"
-	OpsAlertStatusResolved = "resolved"
+	OpsAlertStatusFiring         = "firing"
+	OpsAlertStatusResolved       = "resolved"
+	OpsAlertStatusManualResolved = "manual_resolved"
 )
 
 type OpsAlertRule struct {
@@ -58,12 +59,32 @@ type OpsAlertEvent struct {
 	CreatedAt time.Time `json:"created_at"`
 }
 
+type OpsAlertSilence struct {
+	ID int64 `json:"id"`
+
+	RuleID   int64   `json:"rule_id"`
+	Platform string  `json:"platform"`
+	GroupID  *int64  `json:"group_id,omitempty"`
+	Region   *string `json:"region,omitempty"`
+
+	Until  time.Time `json:"until"`
+	Reason string    `json:"reason"`
+
+	CreatedBy *int64    `json:"created_by,omitempty"`
+	CreatedAt time.Time `json:"created_at"`
+}
+
 type OpsAlertEventFilter struct {
 	Limit int
 
+	// Cursor pagination (descending by fired_at, then id).
+	BeforeFiredAt *time.Time
+	BeforeID      *int64
+
 	// Optional filters.
-	Status   string
-	Severity string
+	Status    string
+	Severity  string
+	EmailSent *bool
 
 	StartTime *time.Time
 	EndTime   *time.Time

backend/internal/service/ops_alerts.go

@@ -88,6 +88,29 @@ func (s *OpsService) ListAlertEvents(ctx context.Context, filter *OpsAlertEventF
 	return s.opsRepo.ListAlertEvents(ctx, filter)
 }
 
+func (s *OpsService) GetAlertEventByID(ctx context.Context, eventID int64) (*OpsAlertEvent, error) {
+	if err := s.RequireMonitoringEnabled(ctx); err != nil {
+		return nil, err
+	}
+	if s.opsRepo == nil {
+		return nil, infraerrors.ServiceUnavailable("OPS_REPO_UNAVAILABLE", "Ops repository not available")
+	}
+	if eventID <= 0 {
+		return nil, infraerrors.BadRequest("INVALID_EVENT_ID", "invalid event id")
+	}
+	ev, err := s.opsRepo.GetAlertEventByID(ctx, eventID)
+	if err != nil {
+		if errors.Is(err, sql.ErrNoRows) {
+			return nil, infraerrors.NotFound("OPS_ALERT_EVENT_NOT_FOUND", "alert event not found")
+		}
+		return nil, err
+	}
+	if ev == nil {
+		return nil, infraerrors.NotFound("OPS_ALERT_EVENT_NOT_FOUND", "alert event not found")
+	}
+	return ev, nil
+}
+
 func (s *OpsService) GetActiveAlertEvent(ctx context.Context, ruleID int64) (*OpsAlertEvent, error) {
 	if err := s.RequireMonitoringEnabled(ctx); err != nil {
 		return nil, err
@@ -101,6 +124,49 @@ func (s *OpsService) GetActiveAlertEvent(ctx context.Context, ruleID int64) (*Op
 	return s.opsRepo.GetActiveAlertEvent(ctx, ruleID)
 }
 
+func (s *OpsService) CreateAlertSilence(ctx context.Context, input *OpsAlertSilence) (*OpsAlertSilence, error) {
+	if err := s.RequireMonitoringEnabled(ctx); err != nil {
+		return nil, err
+	}
+	if s.opsRepo == nil {
+		return nil, infraerrors.ServiceUnavailable("OPS_REPO_UNAVAILABLE", "Ops repository not available")
+	}
+	if input == nil {
+		return nil, infraerrors.BadRequest("INVALID_SILENCE", "invalid silence")
+	}
+	if input.RuleID <= 0 {
+		return nil, infraerrors.BadRequest("INVALID_RULE_ID", "invalid rule id")
+	}
+	if strings.TrimSpace(input.Platform) == "" {
+		return nil, infraerrors.BadRequest("INVALID_PLATFORM", "invalid platform")
+	}
+	if input.Until.IsZero() {
+		return nil, infraerrors.BadRequest("INVALID_UNTIL", "invalid until")
+	}
+
+	created, err := s.opsRepo.CreateAlertSilence(ctx, input)
+	if err != nil {
+		return nil, err
+	}
+	return created, nil
+}
+
+func (s *OpsService) IsAlertSilenced(ctx context.Context, ruleID int64, platform string, groupID *int64, region *string, now time.Time) (bool, error) {
+	if err := s.RequireMonitoringEnabled(ctx); err != nil {
+		return false, err
+	}
+	if s.opsRepo == nil {
+		return false, infraerrors.ServiceUnavailable("OPS_REPO_UNAVAILABLE", "Ops repository not available")
+	}
+	if ruleID <= 0 {
+		return false, infraerrors.BadRequest("INVALID_RULE_ID", "invalid rule id")
+	}
+	if strings.TrimSpace(platform) == "" {
+		return false, nil
+	}
+	return s.opsRepo.IsAlertSilenced(ctx, ruleID, platform, groupID, region, now)
+}
+
 func (s *OpsService) GetLatestAlertEvent(ctx context.Context, ruleID int64) (*OpsAlertEvent, error) {
 	if err := s.RequireMonitoringEnabled(ctx); err != nil {
 		return nil, err
@@ -142,7 +208,11 @@ func (s *OpsService) UpdateAlertEventStatus(ctx context.Context, eventID int64,
 	if eventID <= 0 {
 		return infraerrors.BadRequest("INVALID_EVENT_ID", "invalid event id")
 	}
-	if strings.TrimSpace(status) == "" {
+	status = strings.TrimSpace(status)
+	if status == "" {
+		return infraerrors.BadRequest("INVALID_STATUS", "invalid status")
+	}
+	if status != OpsAlertStatusResolved && status != OpsAlertStatusManualResolved {
 		return infraerrors.BadRequest("INVALID_STATUS", "invalid status")
 	}
 	return s.opsRepo.UpdateAlertEventStatus(ctx, eventID, status, resolvedAt)

backend/internal/service/ops_health_score.go

@@ -32,49 +32,38 @@ func computeDashboardHealthScore(now time.Time, overview *OpsDashboardOverview)
 }
 
 // computeBusinessHealth calculates business health score (0-100)
-// Components: SLA (50%) + Error Rate (30%) + Latency (20%)
+// Components: Error Rate (50%) + TTFT (50%)
 func computeBusinessHealth(overview *OpsDashboardOverview) float64 {
-	// SLA score: 99.5% → 100, 95% → 0 (linear)
-	slaScore := 100.0
-	slaPct := clampFloat64(overview.SLA*100, 0, 100)
-	if slaPct < 99.5 {
-		if slaPct >= 95 {
-			slaScore = (slaPct - 95) / 4.5 * 100
-		} else {
-			slaScore = 0
-		}
-	}
-
-	// Error rate score: 0.5% → 100, 5% → 0 (linear)
+	// Error rate score: 1% → 100, 10% → 0 (linear)
 	// Combines request errors and upstream errors
 	errorScore := 100.0
 	errorPct := clampFloat64(overview.ErrorRate*100, 0, 100)
 	upstreamPct := clampFloat64(overview.UpstreamErrorRate*100, 0, 100)
 	combinedErrorPct := math.Max(errorPct, upstreamPct) // Use worst case
-	if combinedErrorPct > 0.5 {
-		if combinedErrorPct <= 5 {
-			errorScore = (5 - combinedErrorPct) / 4.5 * 100
+	if combinedErrorPct > 1.0 {
+		if combinedErrorPct <= 10.0 {
+			errorScore = (10.0 - combinedErrorPct) / 9.0 * 100
 		} else {
 			errorScore = 0
 		}
 	}
 
-	// Latency score: 1s → 100, 10s → 0 (linear)
-	// Uses P99 of duration (TTFT is less critical for overall health)
-	latencyScore := 100.0
-	if overview.Duration.P99 != nil {
-		p99 := float64(*overview.Duration.P99)
+	// TTFT score: 1s → 100, 3s → 0 (linear)
+	// Time to first token is critical for user experience
+	ttftScore := 100.0
+	if overview.TTFT.P99 != nil {
+		p99 := float64(*overview.TTFT.P99)
 		if p99 > 1000 {
-			if p99 <= 10000 {
-				latencyScore = (10000 - p99) / 9000 * 100
+			if p99 <= 3000 {
+				ttftScore = (3000 - p99) / 2000 * 100
 			} else {
-				latencyScore = 0
+				ttftScore = 0
 			}
 		}
 	}
 
-	// Weighted combination
-	return slaScore*0.5 + errorScore*0.3 + latencyScore*0.2
+	// Weighted combination: 50% error rate + 50% TTFT
+	return errorScore*0.5 + ttftScore*0.5
 }
 
 // computeInfraHealth calculates infrastructure health score (0-100)

backend/internal/service/ops_health_score_test.go

@@ -127,8 +127,8 @@ func TestComputeDashboardHealthScore_Comprehensive(t *testing.T) {
 					MemoryUsagePercent: float64Ptr(75),
 				},
 			},
-			wantMin: 60,
-			wantMax: 85,
+			wantMin: 96,
+			wantMax: 97,
 		},
 		{
 			name: "DB failure",
@@ -203,8 +203,8 @@ func TestComputeDashboardHealthScore_Comprehensive(t *testing.T) {
 					MemoryUsagePercent: float64Ptr(30),
 				},
 			},
-			wantMin: 25,
-			wantMax: 50,
+			wantMin: 84,
+			wantMax: 85,
 		},
 		{
 			name: "combined failures - business healthy + infra degraded",
@@ -277,30 +277,41 @@ func TestComputeBusinessHealth(t *testing.T) {
 				UpstreamErrorRate: 0,
 				Duration:          OpsPercentiles{P99: intPtr(500)},
 			},
-			wantMin: 50,
-			wantMax: 60,
+			wantMin: 100,
+			wantMax: 100,
 		},
 		{
-			name: "error rate boundary 0.5%",
+			name: "error rate boundary 1%",
 			overview: &OpsDashboardOverview{
-				SLA:               0.995,
-				ErrorRate:         0.005,
+				SLA:               0.99,
+				ErrorRate:         0.01,
 				UpstreamErrorRate: 0,
 				Duration:          OpsPercentiles{P99: intPtr(500)},
 			},
-			wantMin: 95,
+			wantMin: 100,
 			wantMax: 100,
 		},
 		{
-			name: "latency boundary 1000ms",
+			name: "error rate 5%",
 			overview: &OpsDashboardOverview{
-				SLA:               0.995,
+				SLA:               0.95,
+				ErrorRate:         0.05,
+				UpstreamErrorRate: 0,
+				Duration:          OpsPercentiles{P99: intPtr(500)},
+			},
+			wantMin: 77,
+			wantMax: 78,
+		},
+		{
+			name: "TTFT boundary 2s",
+			overview: &OpsDashboardOverview{
+				SLA:               0.99,
 				ErrorRate:         0,
 				UpstreamErrorRate: 0,
-				Duration:          OpsPercentiles{P99: intPtr(1000)},
+				TTFT:              OpsPercentiles{P99: intPtr(2000)},
 			},
-			wantMin: 95,
-			wantMax: 100,
+			wantMin: 75,
+			wantMax: 75,
 		},
 		{
 			name: "upstream error dominates",
@@ -310,7 +321,7 @@ func TestComputeBusinessHealth(t *testing.T) {
 				UpstreamErrorRate: 0.03,
 				Duration:          OpsPercentiles{P99: intPtr(500)},
 			},
-			wantMin: 75,
+			wantMin: 88,
 			wantMax: 90,
 		},
 	}

backend/internal/service/ops_models.go

@@ -6,24 +6,43 @@ type OpsErrorLog struct {
 	ID        int64     `json:"id"`
 	CreatedAt time.Time `json:"created_at"`
 
-	Phase    string `json:"phase"`
-	Type     string `json:"type"`
+	// Standardized classification
+	// - phase: request|auth|routing|upstream|network|internal
+	// - owner: client|provider|platform
+	// - source: client_request|upstream_http|gateway
+	Phase string `json:"phase"`
+	Type  string `json:"type"`
+
+	Owner  string `json:"error_owner"`
+	Source string `json:"error_source"`
+
 	Severity string `json:"severity"`
 
 	StatusCode int    `json:"status_code"`
 	Platform   string `json:"platform"`
 	Model      string `json:"model"`
 
-	LatencyMs *int `json:"latency_ms"`
+	IsRetryable bool `json:"is_retryable"`
+	RetryCount  int  `json:"retry_count"`
+
+	Resolved           bool       `json:"resolved"`
+	ResolvedAt         *time.Time `json:"resolved_at"`
+	ResolvedByUserID   *int64     `json:"resolved_by_user_id"`
+	ResolvedByUserName string     `json:"resolved_by_user_name"`
+	ResolvedRetryID    *int64     `json:"resolved_retry_id"`
+	ResolvedStatusRaw  string     `json:"-"`
 
 	ClientRequestID string `json:"client_request_id"`
 	RequestID       string `json:"request_id"`
 	Message         string `json:"message"`
 
-	UserID    *int64 `json:"user_id"`
-	APIKeyID  *int64 `json:"api_key_id"`
-	AccountID *int64 `json:"account_id"`
-	GroupID   *int64 `json:"group_id"`
+	UserID      *int64 `json:"user_id"`
+	UserEmail   string `json:"user_email"`
+	APIKeyID    *int64 `json:"api_key_id"`
+	AccountID   *int64 `json:"account_id"`
+	AccountName string `json:"account_name"`
+	GroupID     *int64 `json:"group_id"`
+	GroupName   string `json:"group_name"`
 
 	ClientIP    *string `json:"client_ip"`
 	RequestPath string  `json:"request_path"`
@@ -67,9 +86,24 @@ type OpsErrorLogFilter struct {
 	GroupID   *int64
 	AccountID *int64
 
-	StatusCodes []int
-	Phase       string
-	Query       string
+	StatusCodes      []int
+	StatusCodesOther bool
+	Phase            string
+	Owner            string
+	Source           string
+	Resolved         *bool
+	Query            string
+	UserQuery        string // Search by user email
+
+	// Optional correlation keys for exact matching.
+	RequestID       string
+	ClientRequestID string
+
+	// View controls error categorization for list endpoints.
+	// - errors: show actionable errors (exclude business-limited / 429 / 529)
+	// - excluded: only show excluded errors
+	// - all: show everything
+	View string
 
 	Page     int
 	PageSize int
@@ -90,12 +124,23 @@ type OpsRetryAttempt struct {
 	SourceErrorID     int64  `json:"source_error_id"`
 	Mode              string `json:"mode"`
 	PinnedAccountID   *int64 `json:"pinned_account_id"`
+	PinnedAccountName string `json:"pinned_account_name"`
 
 	Status     string     `json:"status"`
 	StartedAt  *time.Time `json:"started_at"`
 	FinishedAt *time.Time `json:"finished_at"`
 	DurationMs *int64     `json:"duration_ms"`
 
+	// Persisted execution results (best-effort)
+	Success           *bool   `json:"success"`
+	HTTPStatusCode    *int    `json:"http_status_code"`
+	UpstreamRequestID *string `json:"upstream_request_id"`
+	UsedAccountID     *int64  `json:"used_account_id"`
+	UsedAccountName   string  `json:"used_account_name"`
+	ResponsePreview   *string `json:"response_preview"`
+	ResponseTruncated *bool   `json:"response_truncated"`
+
+	// Optional correlation
 	ResultRequestID *string `json:"result_request_id"`
 	ResultErrorID   *int64  `json:"result_error_id"`
 

backend/internal/service/ops_port.go

@@ -14,6 +14,8 @@ type OpsRepository interface {
 	InsertRetryAttempt(ctx context.Context, input *OpsInsertRetryAttemptInput) (int64, error)
 	UpdateRetryAttempt(ctx context.Context, input *OpsUpdateRetryAttemptInput) error
 	GetLatestRetryAttemptForError(ctx context.Context, sourceErrorID int64) (*OpsRetryAttempt, error)
+	ListRetryAttemptsByErrorID(ctx context.Context, sourceErrorID int64, limit int) ([]*OpsRetryAttempt, error)
+	UpdateErrorResolution(ctx context.Context, errorID int64, resolved bool, resolvedByUserID *int64, resolvedRetryID *int64, resolvedAt *time.Time) error
 
 	// Lightweight window stats (for realtime WS / quick sampling).
 	GetWindowStats(ctx context.Context, filter *OpsDashboardFilter) (*OpsWindowStats, error)
@@ -39,12 +41,17 @@ type OpsRepository interface {
 	DeleteAlertRule(ctx context.Context, id int64) error
 
 	ListAlertEvents(ctx context.Context, filter *OpsAlertEventFilter) ([]*OpsAlertEvent, error)
+	GetAlertEventByID(ctx context.Context, eventID int64) (*OpsAlertEvent, error)
 	GetActiveAlertEvent(ctx context.Context, ruleID int64) (*OpsAlertEvent, error)
 	GetLatestAlertEvent(ctx context.Context, ruleID int64) (*OpsAlertEvent, error)
 	CreateAlertEvent(ctx context.Context, event *OpsAlertEvent) (*OpsAlertEvent, error)
 	UpdateAlertEventStatus(ctx context.Context, eventID int64, status string, resolvedAt *time.Time) error
 	UpdateAlertEventEmailSent(ctx context.Context, eventID int64, emailSent bool) error
 
+	// Alert silences
+	CreateAlertSilence(ctx context.Context, input *OpsAlertSilence) (*OpsAlertSilence, error)
+	IsAlertSilenced(ctx context.Context, ruleID int64, platform string, groupID *int64, region *string, now time.Time) (bool, error)
+
 	// Pre-aggregation (hourly/daily) used for long-window dashboard performance.
 	UpsertHourlyMetrics(ctx context.Context, startTime, endTime time.Time) error
 	UpsertDailyMetrics(ctx context.Context, startTime, endTime time.Time) error
@@ -91,7 +98,6 @@ type OpsInsertErrorLogInput struct {
 	// It is set by OpsService.RecordError before persisting.
 	UpstreamErrorsJSON *string
 
-	DurationMs         *int
 	TimeToFirstTokenMs *int64
 
 	RequestBodyJSON      *string // sanitized json string (not raw bytes)
@@ -124,7 +130,15 @@ type OpsUpdateRetryAttemptInput struct {
 	FinishedAt time.Time
 	DurationMs int64
 
-	// Optional correlation
+	// Persisted execution results (best-effort)
+	Success           *bool
+	HTTPStatusCode    *int
+	UpstreamRequestID *string
+	UsedAccountID     *int64
+	ResponsePreview   *string
+	ResponseTruncated *bool
+
+	// Optional correlation (legacy fields kept)
 	ResultRequestID *string
 	ResultErrorID   *int64
 

backend/internal/service/ops_retry.go

@@ -108,6 +108,10 @@ func (w *limitedResponseWriter) truncated() bool {
 	return w.totalWritten > int64(w.limit)
 }
 
+const (
+	OpsRetryModeUpstreamEvent = "upstream_event"
+)
+
 func (s *OpsService) RetryError(ctx context.Context, requestedByUserID int64, errorID int64, mode string, pinnedAccountID *int64) (*OpsRetryResult, error) {
 	if err := s.RequireMonitoringEnabled(ctx); err != nil {
 		return nil, err
@@ -123,6 +127,81 @@ func (s *OpsService) RetryError(ctx context.Context, requestedByUserID int64, er
 		return nil, infraerrors.BadRequest("OPS_RETRY_INVALID_MODE", "mode must be client or upstream")
 	}
 
+	errorLog, err := s.GetErrorLogByID(ctx, errorID)
+	if err != nil {
+		return nil, err
+	}
+	if errorLog == nil {
+		return nil, infraerrors.NotFound("OPS_ERROR_NOT_FOUND", "ops error log not found")
+	}
+	if strings.TrimSpace(errorLog.RequestBody) == "" {
+		return nil, infraerrors.BadRequest("OPS_RETRY_NO_REQUEST_BODY", "No request body found to retry")
+	}
+
+	var pinned *int64
+	if mode == OpsRetryModeUpstream {
+		if pinnedAccountID != nil && *pinnedAccountID > 0 {
+			pinned = pinnedAccountID
+		} else if errorLog.AccountID != nil && *errorLog.AccountID > 0 {
+			pinned = errorLog.AccountID
+		} else {
+			return nil, infraerrors.BadRequest("OPS_RETRY_PINNED_ACCOUNT_REQUIRED", "pinned_account_id is required for upstream retry")
+		}
+	}
+
+	return s.retryWithErrorLog(ctx, requestedByUserID, errorID, mode, mode, pinned, errorLog)
+}
+
+// RetryUpstreamEvent retries a specific upstream attempt captured inside ops_error_logs.upstream_errors.
+// idx is 0-based. It always pins the original event account_id.
+func (s *OpsService) RetryUpstreamEvent(ctx context.Context, requestedByUserID int64, errorID int64, idx int) (*OpsRetryResult, error) {
+	if err := s.RequireMonitoringEnabled(ctx); err != nil {
+		return nil, err
+	}
+	if s.opsRepo == nil {
+		return nil, infraerrors.ServiceUnavailable("OPS_REPO_UNAVAILABLE", "Ops repository not available")
+	}
+	if idx < 0 {
+		return nil, infraerrors.BadRequest("OPS_RETRY_INVALID_UPSTREAM_IDX", "invalid upstream idx")
+	}
+
+	errorLog, err := s.GetErrorLogByID(ctx, errorID)
+	if err != nil {
+		return nil, err
+	}
+	if errorLog == nil {
+		return nil, infraerrors.NotFound("OPS_ERROR_NOT_FOUND", "ops error log not found")
+	}
+
+	events, err := ParseOpsUpstreamErrors(errorLog.UpstreamErrors)
+	if err != nil {
+		return nil, infraerrors.BadRequest("OPS_RETRY_UPSTREAM_EVENTS_INVALID", "invalid upstream_errors")
+	}
+	if idx >= len(events) {
+		return nil, infraerrors.BadRequest("OPS_RETRY_UPSTREAM_IDX_OOB", "upstream idx out of range")
+	}
+	ev := events[idx]
+	if ev == nil {
+		return nil, infraerrors.BadRequest("OPS_RETRY_UPSTREAM_EVENT_MISSING", "upstream event missing")
+	}
+	if ev.AccountID <= 0 {
+		return nil, infraerrors.BadRequest("OPS_RETRY_PINNED_ACCOUNT_REQUIRED", "account_id is required for upstream retry")
+	}
+
+	upstreamBody := strings.TrimSpace(ev.UpstreamRequestBody)
+	if upstreamBody == "" {
+		return nil, infraerrors.BadRequest("OPS_RETRY_UPSTREAM_NO_REQUEST_BODY", "No upstream request body found to retry")
+	}
+
+	override := *errorLog
+	override.RequestBody = upstreamBody
+	pinned := ev.AccountID
+
+	// Persist as upstream_event, execute as upstream pinned retry.
+	return s.retryWithErrorLog(ctx, requestedByUserID, errorID, OpsRetryModeUpstreamEvent, OpsRetryModeUpstream, &pinned, &override)
+}
+
+func (s *OpsService) retryWithErrorLog(ctx context.Context, requestedByUserID int64, errorID int64, mode string, execMode string, pinnedAccountID *int64, errorLog *OpsErrorLogDetail) (*OpsRetryResult, error) {
 	latest, err := s.opsRepo.GetLatestRetryAttemptForError(ctx, errorID)
 	if err != nil && !errors.Is(err, sql.ErrNoRows) {
 		return nil, infraerrors.InternalServer("OPS_RETRY_LOAD_LATEST_FAILED", "Failed to check retry status").WithCause(err)
@@ -144,22 +223,18 @@ func (s *OpsService) RetryError(ctx context.Context, requestedByUserID int64, er
 		}
 	}
 
-	errorLog, err := s.GetErrorLogByID(ctx, errorID)
-	if err != nil {
-		return nil, err
-	}
-	if strings.TrimSpace(errorLog.RequestBody) == "" {
+	if errorLog == nil || strings.TrimSpace(errorLog.RequestBody) == "" {
 		return nil, infraerrors.BadRequest("OPS_RETRY_NO_REQUEST_BODY", "No request body found to retry")
 	}
 
 	var pinned *int64
-	if mode == OpsRetryModeUpstream {
+	if execMode == OpsRetryModeUpstream {
 		if pinnedAccountID != nil && *pinnedAccountID > 0 {
 			pinned = pinnedAccountID
 		} else if errorLog.AccountID != nil && *errorLog.AccountID > 0 {
 			pinned = errorLog.AccountID
 		} else {
-			return nil, infraerrors.BadRequest("OPS_RETRY_PINNED_ACCOUNT_REQUIRED", "pinned_account_id is required for upstream retry")
+			return nil, infraerrors.BadRequest("OPS_RETRY_PINNED_ACCOUNT_REQUIRED", "account_id is required for upstream retry")
 		}
 	}
 
@@ -196,7 +271,7 @@ func (s *OpsService) RetryError(ctx context.Context, requestedByUserID int64, er
 	execCtx, cancel := context.WithTimeout(ctx, opsRetryTimeout)
 	defer cancel()
 
-	execRes := s.executeRetry(execCtx, errorLog, mode, pinned)
+	execRes := s.executeRetry(execCtx, errorLog, execMode, pinned)
 
 	finishedAt := time.Now()
 	result.FinishedAt = finishedAt
@@ -220,27 +295,40 @@ func (s *OpsService) RetryError(ctx context.Context, requestedByUserID int64, er
 		msg := result.ErrorMessage
 		updateErrMsg = &msg
 	}
+	// Keep legacy result_request_id empty; use upstream_request_id instead.
 	var resultRequestID *string
-	if strings.TrimSpace(result.UpstreamRequestID) != "" {
-		v := result.UpstreamRequestID
-		resultRequestID = &v
-	}
 
 	finalStatus := result.Status
 	if strings.TrimSpace(finalStatus) == "" {
 		finalStatus = opsRetryStatusFailed
 	}
 
+	success := strings.EqualFold(finalStatus, opsRetryStatusSucceeded)
+	httpStatus := result.HTTPStatusCode
+	upstreamReqID := result.UpstreamRequestID
+	usedAccountID := result.UsedAccountID
+	preview := result.ResponsePreview
+	truncated := result.ResponseTruncated
+
 	if err := s.opsRepo.UpdateRetryAttempt(updateCtx, &OpsUpdateRetryAttemptInput{
-		ID:              attemptID,
-		Status:          finalStatus,
-		FinishedAt:      finishedAt,
-		DurationMs:      result.DurationMs,
-		ResultRequestID: resultRequestID,
-		ErrorMessage:    updateErrMsg,
+		ID:                attemptID,
+		Status:            finalStatus,
+		FinishedAt:        finishedAt,
+		DurationMs:        result.DurationMs,
+		Success:           &success,
+		HTTPStatusCode:    &httpStatus,
+		UpstreamRequestID: &upstreamReqID,
+		UsedAccountID:     usedAccountID,
+		ResponsePreview:   &preview,
+		ResponseTruncated: &truncated,
+		ResultRequestID:   resultRequestID,
+		ErrorMessage:      updateErrMsg,
 	}); err != nil {
-		// Best-effort: retry itself already executed; do not fail the API response.
 		log.Printf("[Ops] UpdateRetryAttempt failed: %v", err)
+	} else if success {
+		if err := s.opsRepo.UpdateErrorResolution(updateCtx, errorID, true, &requestedByUserID, &attemptID, &finishedAt); err != nil {
+			log.Printf("[Ops] UpdateErrorResolution failed: %v", err)
+		}
 	}
 
 	return result, nil

backend/internal/service/ops_service.go

@@ -208,6 +208,25 @@ func (s *OpsService) RecordError(ctx context.Context, entry *OpsInsertErrorLogIn
 				out.Detail = ""
 			}
 
+			out.UpstreamRequestBody = strings.TrimSpace(out.UpstreamRequestBody)
+			if out.UpstreamRequestBody != "" {
+				// Reuse the same sanitization/trimming strategy as request body storage.
+				// Keep it small so it is safe to persist in ops_error_logs JSON.
+				sanitized, truncated, _ := sanitizeAndTrimRequestBody([]byte(out.UpstreamRequestBody), 10*1024)
+				if sanitized != "" {
+					out.UpstreamRequestBody = sanitized
+					if truncated {
+						out.Kind = strings.TrimSpace(out.Kind)
+						if out.Kind == "" {
+							out.Kind = "upstream"
+						}
+						out.Kind = out.Kind + ":request_body_truncated"
+					}
+				} else {
+					out.UpstreamRequestBody = ""
+				}
+			}
+
 			// Drop fully-empty events (can happen if only status code was known).
 			if out.UpstreamStatusCode == 0 && out.Message == "" && out.Detail == "" {
 				continue
@@ -236,7 +255,13 @@ func (s *OpsService) GetErrorLogs(ctx context.Context, filter *OpsErrorLogFilter
 	if s.opsRepo == nil {
 		return &OpsErrorLogList{Errors: []*OpsErrorLog{}, Total: 0, Page: 1, PageSize: 20}, nil
 	}
-	return s.opsRepo.ListErrorLogs(ctx, filter)
+	result, err := s.opsRepo.ListErrorLogs(ctx, filter)
+	if err != nil {
+		log.Printf("[Ops] GetErrorLogs failed: %v", err)
+		return nil, err
+	}
+
+	return result, nil
 }
 
 func (s *OpsService) GetErrorLogByID(ctx context.Context, id int64) (*OpsErrorLogDetail, error) {
@@ -256,6 +281,46 @@ func (s *OpsService) GetErrorLogByID(ctx context.Context, id int64) (*OpsErrorLo
 	return detail, nil
 }
 
+func (s *OpsService) ListRetryAttemptsByErrorID(ctx context.Context, errorID int64, limit int) ([]*OpsRetryAttempt, error) {
+	if err := s.RequireMonitoringEnabled(ctx); err != nil {
+		return nil, err
+	}
+	if s.opsRepo == nil {
+		return nil, infraerrors.ServiceUnavailable("OPS_REPO_UNAVAILABLE", "Ops repository not available")
+	}
+	if errorID <= 0 {
+		return nil, infraerrors.BadRequest("OPS_ERROR_INVALID_ID", "invalid error id")
+	}
+	items, err := s.opsRepo.ListRetryAttemptsByErrorID(ctx, errorID, limit)
+	if err != nil {
+		if errors.Is(err, sql.ErrNoRows) {
+			return []*OpsRetryAttempt{}, nil
+		}
+		return nil, infraerrors.InternalServer("OPS_RETRY_LIST_FAILED", "Failed to list retry attempts").WithCause(err)
+	}
+	return items, nil
+}
+
+func (s *OpsService) UpdateErrorResolution(ctx context.Context, errorID int64, resolved bool, resolvedByUserID *int64, resolvedRetryID *int64) error {
+	if err := s.RequireMonitoringEnabled(ctx); err != nil {
+		return err
+	}
+	if s.opsRepo == nil {
+		return infraerrors.ServiceUnavailable("OPS_REPO_UNAVAILABLE", "Ops repository not available")
+	}
+	if errorID <= 0 {
+		return infraerrors.BadRequest("OPS_ERROR_INVALID_ID", "invalid error id")
+	}
+	// Best-effort ensure the error exists
+	if _, err := s.opsRepo.GetErrorLogByID(ctx, errorID); err != nil {
+		if errors.Is(err, sql.ErrNoRows) {
+			return infraerrors.NotFound("OPS_ERROR_NOT_FOUND", "ops error log not found")
+		}
+		return infraerrors.InternalServer("OPS_ERROR_LOAD_FAILED", "Failed to load ops error log").WithCause(err)
+	}
+	return s.opsRepo.UpdateErrorResolution(ctx, errorID, resolved, resolvedByUserID, resolvedRetryID, nil)
+}
+
 func sanitizeAndTrimRequestBody(raw []byte, maxBytes int) (jsonString string, truncated bool, bytesLen int) {
 	bytesLen = len(raw)
 	if len(raw) == 0 {
@@ -296,14 +361,34 @@ func sanitizeAndTrimRequestBody(raw []byte, maxBytes int) (jsonString string, tr
 		}
 	}
 
-	// Last resort: store a minimal placeholder (still valid JSON).
-	placeholder := map[string]any{
-		"request_body_truncated": true,
+	// Last resort: keep JSON shape but drop big fields.
+	// This avoids downstream code that expects certain top-level keys from crashing.
+	if root, ok := decoded.(map[string]any); ok {
+		placeholder := shallowCopyMap(root)
+		placeholder["request_body_truncated"] = true
+
+		// Replace potentially huge arrays/strings, but keep the keys present.
+		for _, k := range []string{"messages", "contents", "input", "prompt"} {
+			if _, exists := placeholder[k]; exists {
+				placeholder[k] = []any{}
+			}
+		}
+		for _, k := range []string{"text"} {
+			if _, exists := placeholder[k]; exists {
+				placeholder[k] = ""
+			}
+		}
+
+		encoded4, err4 := json.Marshal(placeholder)
+		if err4 == nil {
+			if len(encoded4) <= maxBytes {
+				return string(encoded4), true, bytesLen
+			}
+		}
 	}
-	if model := extractString(decoded, "model"); model != "" {
-		placeholder["model"] = model
-	}
-	encoded4, err4 := json.Marshal(placeholder)
+
+	// Final fallback: minimal valid JSON.
+	encoded4, err4 := json.Marshal(map[string]any{"request_body_truncated": true})
 	if err4 != nil {
 		return "", true, bytesLen
 	}
@@ -526,12 +611,3 @@ func sanitizeErrorBodyForStorage(raw string, maxBytes int) (sanitized string, tr
 	}
 	return raw, false
 }
-
-func extractString(v any, key string) string {
-	root, ok := v.(map[string]any)
-	if !ok {
-		return ""
-	}
-	s, _ := root[key].(string)
-	return strings.TrimSpace(s)
-}

backend/internal/service/ops_settings.go

@@ -368,9 +368,11 @@ func defaultOpsAdvancedSettings() *OpsAdvancedSettings {
 		Aggregation: OpsAggregationSettings{
 			AggregationEnabled: false,
 		},
-		IgnoreCountTokensErrors: false,
-		AutoRefreshEnabled:      false,
-		AutoRefreshIntervalSec:  30,
+		IgnoreCountTokensErrors:   false,
+		IgnoreContextCanceled:     true,  // Default to true - client disconnects are not errors
+		IgnoreNoAvailableAccounts: false, // Default to false - this is a real routing issue
+		AutoRefreshEnabled:        false,
+		AutoRefreshIntervalSec:    30,
 	}
 }
 
@@ -482,13 +484,11 @@ const SettingKeyOpsMetricThresholds = "ops_metric_thresholds"
 
 func defaultOpsMetricThresholds() *OpsMetricThresholds {
 	slaMin := 99.5
-	latencyMax := 2000.0
 	ttftMax := 500.0
 	reqErrMax := 5.0
 	upstreamErrMax := 5.0
 	return &OpsMetricThresholds{
 		SLAPercentMin:               &slaMin,
-		LatencyP99MsMax:             &latencyMax,
 		TTFTp99MsMax:                &ttftMax,
 		RequestErrorRatePercentMax:  &reqErrMax,
 		UpstreamErrorRatePercentMax: &upstreamErrMax,
@@ -538,9 +538,6 @@ func (s *OpsService) UpdateMetricThresholds(ctx context.Context, cfg *OpsMetricT
 	if cfg.SLAPercentMin != nil && (*cfg.SLAPercentMin < 0 || *cfg.SLAPercentMin > 100) {
 		return nil, errors.New("sla_percent_min must be between 0 and 100")
 	}
-	if cfg.LatencyP99MsMax != nil && *cfg.LatencyP99MsMax < 0 {
-		return nil, errors.New("latency_p99_ms_max must be >= 0")
-	}
 	if cfg.TTFTp99MsMax != nil && *cfg.TTFTp99MsMax < 0 {
 		return nil, errors.New("ttft_p99_ms_max must be >= 0")
 	}

backend/internal/service/ops_settings_models.go

@@ -63,7 +63,6 @@ type OpsAlertSilencingSettings struct {
 
 type OpsMetricThresholds struct {
 	SLAPercentMin               *float64 `json:"sla_percent_min,omitempty"`                 // SLA低于此值变红
-	LatencyP99MsMax             *float64 `json:"latency_p99_ms_max,omitempty"`              // 延迟P99高于此值变红
 	TTFTp99MsMax                *float64 `json:"ttft_p99_ms_max,omitempty"`                 // TTFT P99高于此值变红
 	RequestErrorRatePercentMax  *float64 `json:"request_error_rate_percent_max,omitempty"`  // 请求错误率高于此值变红
 	UpstreamErrorRatePercentMax *float64 `json:"upstream_error_rate_percent_max,omitempty"` // 上游错误率高于此值变红
@@ -79,11 +78,13 @@ type OpsAlertRuntimeSettings struct {
 
 // OpsAdvancedSettings stores advanced ops configuration (data retention, aggregation).
 type OpsAdvancedSettings struct {
-	DataRetention           OpsDataRetentionSettings `json:"data_retention"`
-	Aggregation             OpsAggregationSettings   `json:"aggregation"`
-	IgnoreCountTokensErrors bool                     `json:"ignore_count_tokens_errors"`
-	AutoRefreshEnabled      bool                     `json:"auto_refresh_enabled"`
-	AutoRefreshIntervalSec  int                      `json:"auto_refresh_interval_seconds"`
+	DataRetention             OpsDataRetentionSettings `json:"data_retention"`
+	Aggregation               OpsAggregationSettings   `json:"aggregation"`
+	IgnoreCountTokensErrors   bool                     `json:"ignore_count_tokens_errors"`
+	IgnoreContextCanceled     bool                     `json:"ignore_context_canceled"`
+	IgnoreNoAvailableAccounts bool                     `json:"ignore_no_available_accounts"`
+	AutoRefreshEnabled        bool                     `json:"auto_refresh_enabled"`
+	AutoRefreshIntervalSec    int                      `json:"auto_refresh_interval_seconds"`
 }
 
 type OpsDataRetentionSettings struct {

backend/internal/service/ops_upstream_context.go

@@ -15,6 +15,11 @@ const (
 	OpsUpstreamErrorMessageKey = "ops_upstream_error_message"
 	OpsUpstreamErrorDetailKey  = "ops_upstream_error_detail"
 	OpsUpstreamErrorsKey       = "ops_upstream_errors"
+
+	// Best-effort capture of the current upstream request body so ops can
+	// retry the specific upstream attempt (not just the client request).
+	// This value is sanitized+trimmed before being persisted.
+	OpsUpstreamRequestBodyKey = "ops_upstream_request_body"
 )
 
 func setOpsUpstreamError(c *gin.Context, upstreamStatusCode int, upstreamMessage, upstreamDetail string) {
@@ -38,13 +43,21 @@ type OpsUpstreamErrorEvent struct {
 	AtUnixMs int64 `json:"at_unix_ms,omitempty"`
 
 	// Context
-	Platform  string `json:"platform,omitempty"`
-	AccountID int64  `json:"account_id,omitempty"`
+	Platform    string `json:"platform,omitempty"`
+	AccountID   int64  `json:"account_id,omitempty"`
+	AccountName string `json:"account_name,omitempty"`
 
 	// Outcome
 	UpstreamStatusCode int    `json:"upstream_status_code,omitempty"`
 	UpstreamRequestID  string `json:"upstream_request_id,omitempty"`
 
+	// Best-effort upstream request capture (sanitized+trimmed).
+	// Required for retrying a specific upstream attempt.
+	UpstreamRequestBody string `json:"upstream_request_body,omitempty"`
+
+	// Best-effort upstream response capture (sanitized+trimmed).
+	UpstreamResponseBody string `json:"upstream_response_body,omitempty"`
+
 	// Kind: http_error | request_error | retry_exhausted | failover
 	Kind string `json:"kind,omitempty"`
 
@@ -61,6 +74,8 @@ func appendOpsUpstreamError(c *gin.Context, ev OpsUpstreamErrorEvent) {
 	}
 	ev.Platform = strings.TrimSpace(ev.Platform)
 	ev.UpstreamRequestID = strings.TrimSpace(ev.UpstreamRequestID)
+	ev.UpstreamRequestBody = strings.TrimSpace(ev.UpstreamRequestBody)
+	ev.UpstreamResponseBody = strings.TrimSpace(ev.UpstreamResponseBody)
 	ev.Kind = strings.TrimSpace(ev.Kind)
 	ev.Message = strings.TrimSpace(ev.Message)
 	ev.Detail = strings.TrimSpace(ev.Detail)
@@ -68,6 +83,16 @@ func appendOpsUpstreamError(c *gin.Context, ev OpsUpstreamErrorEvent) {
 		ev.Message = sanitizeUpstreamErrorMessage(ev.Message)
 	}
 
+	// If the caller didn't explicitly pass upstream request body but the gateway
+	// stored it on the context, attach it so ops can retry this specific attempt.
+	if ev.UpstreamRequestBody == "" {
+		if v, ok := c.Get(OpsUpstreamRequestBodyKey); ok {
+			if s, ok := v.(string); ok {
+				ev.UpstreamRequestBody = strings.TrimSpace(s)
+			}
+		}
+	}
+
 	var existing []*OpsUpstreamErrorEvent
 	if v, ok := c.Get(OpsUpstreamErrorsKey); ok {
 		if arr, ok := v.([]*OpsUpstreamErrorEvent); ok {
@@ -92,3 +117,15 @@ func marshalOpsUpstreamErrors(events []*OpsUpstreamErrorEvent) *string {
 	s := string(raw)
 	return &s
 }
+
+func ParseOpsUpstreamErrors(raw string) ([]*OpsUpstreamErrorEvent, error) {
+	raw = strings.TrimSpace(raw)
+	if raw == "" {
+		return []*OpsUpstreamErrorEvent{}, nil
+	}
+	var out []*OpsUpstreamErrorEvent
+	if err := json.Unmarshal([]byte(raw), &out); err != nil {
+		return nil, err
+	}
+	return out, nil
+}

backend/internal/service/proxy.go

@@ -31,5 +31,16 @@ func (p *Proxy) URL() string {
 
 type ProxyWithAccountCount struct {
 	Proxy
-	AccountCount int64
+	AccountCount   int64
+	LatencyMs      *int64
+	LatencyStatus  string
+	LatencyMessage string
+}
+
+type ProxyAccountSummary struct {
+	ID       int64
+	Name     string
+	Platform string
+	Type     string
+	Notes    *string
 }

backend/internal/service/proxy_latency_cache.go

@@ -0,0 +1,18 @@
+package service
+
+import (
+	"context"
+	"time"
+)
+
+type ProxyLatencyInfo struct {
+	Success   bool      `json:"success"`
+	LatencyMs *int64    `json:"latency_ms,omitempty"`
+	Message   string    `json:"message,omitempty"`
+	UpdatedAt time.Time `json:"updated_at"`
+}
+
+type ProxyLatencyCache interface {
+	GetProxyLatencies(ctx context.Context, proxyIDs []int64) (map[int64]*ProxyLatencyInfo, error)
+	SetProxyLatency(ctx context.Context, proxyID int64, info *ProxyLatencyInfo) error
+}

backend/internal/service/proxy_service.go

@@ -10,6 +10,7 @@ import (
 
 var (
 	ErrProxyNotFound = infraerrors.NotFound("PROXY_NOT_FOUND", "proxy not found")
+	ErrProxyInUse    = infraerrors.Conflict("PROXY_IN_USE", "proxy is in use by accounts")
 )
 
 type ProxyRepository interface {
@@ -26,6 +27,7 @@ type ProxyRepository interface {
 
 	ExistsByHostPortAuth(ctx context.Context, host string, port int, username, password string) (bool, error)
 	CountAccountsByProxyID(ctx context.Context, proxyID int64) (int64, error)
+	ListAccountSummariesByProxyID(ctx context.Context, proxyID int64) ([]ProxyAccountSummary, error)
 }
 
 // CreateProxyRequest 创建代理请求

backend/internal/service/ratelimit_service.go

@@ -243,7 +243,7 @@ func (s *RateLimitService) PreCheckUsage(ctx context.Context, account *Account,
 			start := geminiDailyWindowStart(now)
 			totals, ok := s.getGeminiUsageTotals(account.ID, start, now)
 			if !ok {
-				stats, err := s.usageRepo.GetModelStatsWithFilters(ctx, start, now, 0, 0, account.ID)
+				stats, err := s.usageRepo.GetModelStatsWithFilters(ctx, start, now, 0, 0, account.ID, 0, nil)
 				if err != nil {
 					return true, err
 				}
@@ -290,7 +290,7 @@ func (s *RateLimitService) PreCheckUsage(ctx context.Context, account *Account,
 
 		if limit > 0 {
 			start := now.Truncate(time.Minute)
-			stats, err := s.usageRepo.GetModelStatsWithFilters(ctx, start, now, 0, 0, account.ID)
+			stats, err := s.usageRepo.GetModelStatsWithFilters(ctx, start, now, 0, 0, account.ID, 0, nil)
 			if err != nil {
 				return true, err
 			}

backend/internal/service/usage_log.go

@@ -33,6 +33,8 @@ type UsageLog struct {
 	TotalCost         float64
 	ActualCost        float64
 	RateMultiplier    float64
+	// AccountRateMultiplier 账号计费倍率快照（nil 表示历史数据，按 1.0 处理）
+	AccountRateMultiplier *float64
 
 	BillingType  int8
 	Stream       bool