feat: 完善使用记录端点可观测性与分布统计

将入站、上游与路径三类端点分布统一到使用记录页的一致化卡片交互中，并补齐端点元数据与统计链路，提升排障与流量分析效率。

backend/internal/handler/dto/mappers.go

@@ -523,6 +523,8 @@ func usageLogFromServiceUser(l *service.UsageLog) UsageLog {
 		Model:                 l.Model,
 		ServiceTier:           l.ServiceTier,
 		ReasoningEffort:       l.ReasoningEffort,
+		InboundEndpoint:       l.InboundEndpoint,
+		UpstreamEndpoint:      l.UpstreamEndpoint,
 		GroupID:               l.GroupID,
 		SubscriptionID:        l.SubscriptionID,
 		InputTokens:           l.InputTokens,

backend/internal/handler/dto/mappers_usage_test.go

@@ -76,10 +76,14 @@ func TestUsageLogFromService_IncludesServiceTierForUserAndAdmin(t *testing.T) {
 	t.Parallel()
 
 	serviceTier := "priority"
+	inboundEndpoint := "/v1/chat/completions"
+	upstreamEndpoint := "/v1/responses"
 	log := &service.UsageLog{
 		RequestID:             "req_3",
 		Model:                 "gpt-5.4",
 		ServiceTier:           &serviceTier,
+		InboundEndpoint:       &inboundEndpoint,
+		UpstreamEndpoint:      &upstreamEndpoint,
 		AccountRateMultiplier: f64Ptr(1.5),
 	}
 
@@ -88,8 +92,16 @@ func TestUsageLogFromService_IncludesServiceTierForUserAndAdmin(t *testing.T) {
 
 	require.NotNil(t, userDTO.ServiceTier)
 	require.Equal(t, serviceTier, *userDTO.ServiceTier)
+	require.NotNil(t, userDTO.InboundEndpoint)
+	require.Equal(t, inboundEndpoint, *userDTO.InboundEndpoint)
+	require.NotNil(t, userDTO.UpstreamEndpoint)
+	require.Equal(t, upstreamEndpoint, *userDTO.UpstreamEndpoint)
 	require.NotNil(t, adminDTO.ServiceTier)
 	require.Equal(t, serviceTier, *adminDTO.ServiceTier)
+	require.NotNil(t, adminDTO.InboundEndpoint)
+	require.Equal(t, inboundEndpoint, *adminDTO.InboundEndpoint)
+	require.NotNil(t, adminDTO.UpstreamEndpoint)
+	require.Equal(t, upstreamEndpoint, *adminDTO.UpstreamEndpoint)
 	require.NotNil(t, adminDTO.AccountRateMultiplier)
 	require.InDelta(t, 1.5, *adminDTO.AccountRateMultiplier, 1e-12)
 }

backend/internal/handler/dto/types.go

@@ -337,6 +337,10 @@ type UsageLog struct {
 	// ReasoningEffort is the request's reasoning effort level (OpenAI Responses API).
 	// nil means not provided / not applicable.
 	ReasoningEffort *string `json:"reasoning_effort,omitempty"`
+	// InboundEndpoint is the client-facing API endpoint path, e.g. /v1/chat/completions.
+	InboundEndpoint *string `json:"inbound_endpoint,omitempty"`
+	// UpstreamEndpoint is the normalized upstream endpoint path, e.g. /v1/responses.
+	UpstreamEndpoint *string `json:"upstream_endpoint,omitempty"`
 
 	GroupID        *int64 `json:"group_id"`
 	SubscriptionID *int64 `json:"subscription_id"`