Reduce DB write amplification on quota and account extra updates

backend/internal/service/api_key_service.go

@@ -6,6 +6,7 @@ import (
 	"encoding/hex"
 	"fmt"
 	"strconv"
+	"strings"
 	"sync"
 	"time"
 
@@ -110,6 +111,15 @@ func (d *APIKeyRateLimitData) EffectiveUsage7d() float64 {
 	return d.Usage7d
 }
 
+// APIKeyQuotaUsageState captures the latest quota fields after an atomic quota update.
+// It is intentionally small so repositories can return it from a single SQL statement.
+type APIKeyQuotaUsageState struct {
+	QuotaUsed float64
+	Quota     float64
+	Key       string
+	Status    string
+}
+
 // APIKeyCache defines cache operations for API key service
 type APIKeyCache interface {
 	GetCreateAttemptCount(ctx context.Context, userID int64) (int, error)
@@ -817,6 +827,21 @@ func (s *APIKeyService) UpdateQuotaUsed(ctx context.Context, apiKeyID int64, cos
 		return nil
 	}
 
+	type quotaStateReader interface {
+		IncrementQuotaUsedAndGetState(ctx context.Context, id int64, amount float64) (*APIKeyQuotaUsageState, error)
+	}
+
+	if repo, ok := s.apiKeyRepo.(quotaStateReader); ok {
+		state, err := repo.IncrementQuotaUsedAndGetState(ctx, apiKeyID, cost)
+		if err != nil {
+			return fmt.Errorf("increment quota used: %w", err)
+		}
+		if state != nil && state.Status == StatusAPIKeyQuotaExhausted && strings.TrimSpace(state.Key) != "" {
+			s.InvalidateAuthCacheByKey(ctx, state.Key)
+		}
+		return nil
+	}
+
 	// Use repository to atomically increment quota_used
 	newQuotaUsed, err := s.apiKeyRepo.IncrementQuotaUsed(ctx, apiKeyID, cost)
 	if err != nil {