Merge branch 'Wei-Shaw:main' into fix/open-issues-cleanup

2026-04-03 06:52:13 +08:00 · 2026-03-15 17:49:20 +08:00
parent 90b3838173 8321e4a647
commit 3718d6dcd4
38 changed files with 1219 additions and 68 deletions
--- a/backend/internal/handler/dto/mappers.go
+++ b/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,
--- a/backend/internal/handler/dto/mappers_usage_test.go
+++ b/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)
 }
--- a/backend/internal/handler/dto/types.go
+++ b/backend/internal/handler/dto/types.go
@@ -334,9 +334,13 @@ type UsageLog struct {
 	Model     string `json:"model"`
 	// ServiceTier records the OpenAI service tier used for billing, e.g. "priority" / "flex".
 	ServiceTier *string `json:"service_tier,omitempty"`
-	// ReasoningEffort is the request's reasoning effort level (OpenAI Responses API).
-	// nil means not provided / not applicable.
+	// ReasoningEffort is the request's reasoning effort level.
+	// OpenAI: "low"/"medium"/"high"/"xhigh"; Claude: "low"/"medium"/"high"/"max".
 	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"`
--- a/backend/internal/handler/gateway_handler.go
+++ b/backend/internal/handler/gateway_handler.go
@@ -391,6 +391,8 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 			if fs.SwitchCount > 0 {
 				requestCtx = service.WithAccountSwitchCount(requestCtx, fs.SwitchCount, h.metadataBridgeEnabled())
 			}
+			// 记录 Forward 前已写入字节数，Forward 后若增加则说明 SSE 内容已发，禁止 failover
+			writerSizeBeforeForward := c.Writer.Size()
 			if account.Platform == service.PlatformAntigravity {
 				result, err = h.antigravityGatewayService.ForwardGemini(requestCtx, c, account, reqModel, "generateContent", reqStream, body, hasBoundSession)
 			} else {
@@ -402,6 +404,11 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 			if err != nil {
 				var failoverErr *service.UpstreamFailoverError
 				if errors.As(err, &failoverErr) {
+					// 流式内容已写入客户端，无法撤销，禁止 failover 以防止流拼接腐化
+					if c.Writer.Size() != writerSizeBeforeForward {
+						h.handleFailoverExhausted(c, failoverErr, service.PlatformGemini, true)
+						return
+					}
 					action := fs.HandleFailoverError(c.Request.Context(), h.gatewayService, account.ID, account.Platform, failoverErr)
 					switch action {
 					case FailoverContinue:
@@ -436,6 +443,10 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 			clientIP := ip.GetClientIP(c)
 			requestPayloadHash := service.HashUsageRequestPayload(body)

+			if result.ReasoningEffort == nil {
+				result.ReasoningEffort = service.NormalizeClaudeOutputEffort(parsedReq.OutputEffort)
+			}
+
 			// 使用量记录通过有界 worker 池提交，避免请求热路径创建无界 goroutine。
 			h.submitUsageRecordTask(func(ctx context.Context) {
 				if err := h.gatewayService.RecordUsage(ctx, &service.RecordUsageInput{
@@ -637,6 +648,8 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 			if fs.SwitchCount > 0 {
 				requestCtx = service.WithAccountSwitchCount(requestCtx, fs.SwitchCount, h.metadataBridgeEnabled())
 			}
+			// 记录 Forward 前已写入字节数，Forward 后若增加则说明 SSE 内容已发，禁止 failover
+			writerSizeBeforeForward := c.Writer.Size()
 			if account.Platform == service.PlatformAntigravity && account.Type != service.AccountTypeAPIKey {
 				result, err = h.antigravityGatewayService.Forward(requestCtx, c, account, body, hasBoundSession)
 			} else {
@@ -706,6 +719,11 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 				}
 				var failoverErr *service.UpstreamFailoverError
 				if errors.As(err, &failoverErr) {
+					// 流式内容已写入客户端，无法撤销，禁止 failover 以防止流拼接腐化
+					if c.Writer.Size() != writerSizeBeforeForward {
+						h.handleFailoverExhausted(c, failoverErr, account.Platform, true)
+						return
+					}
 					action := fs.HandleFailoverError(c.Request.Context(), h.gatewayService, account.ID, account.Platform, failoverErr)
 					switch action {
 					case FailoverContinue:
@@ -740,6 +758,10 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 			clientIP := ip.GetClientIP(c)
 			requestPayloadHash := service.HashUsageRequestPayload(body)

+			if result.ReasoningEffort == nil {
+				result.ReasoningEffort = service.NormalizeClaudeOutputEffort(parsedReq.OutputEffort)
+			}
+
 			// 使用量记录通过有界 worker 池提交，避免请求热路径创建无界 goroutine。
 			h.submitUsageRecordTask(func(ctx context.Context) {
 				if err := h.gatewayService.RecordUsage(ctx, &service.RecordUsageInput{
--- a/backend/internal/handler/gateway_handler_stream_failover_test.go
+++ b/backend/internal/handler/gateway_handler_stream_failover_test.go
@@ -0,0 +1,122 @@
+package handler
+
+import (
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+
+	"github.com/Wei-Shaw/sub2api/internal/service"
+	"github.com/gin-gonic/gin"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// partialMessageStartSSE 模拟 handleStreamingResponse 已写入的首批 SSE 事件。
+const partialMessageStartSSE = "event: message_start\ndata: {\"type\":\"message_start\",\"message\":{\"id\":\"msg_01\",\"type\":\"message\",\"role\":\"assistant\",\"content\":[],\"model\":\"claude-sonnet-4-5\",\"stop_reason\":null,\"stop_sequence\":null,\"usage\":{\"input_tokens\":10,\"output_tokens\":1}}}\n\n" +
+	"event: content_block_start\ndata: {\"type\":\"content_block_start\",\"index\":0,\"content_block\":{\"type\":\"text\",\"text\":\"\"}}\n\n"
+
+// TestStreamWrittenGuard_MessagesPath_AbortFailoverOnSSEContentWritten 验证：
+// 当 Forward 在返回 UpstreamFailoverError 前已向客户端写入 SSE 内容时，
+// 故障转移保护逻辑必须终止循环并发送 SSE 错误事件，而不是进行下一次 Forward。
+// 具体验证：
+//  1. c.Writer.Size() 检测条件正确触发（字节数已增加）
+//  2. handleFailoverExhausted 以 streamStarted=true 调用后，响应体以 SSE 错误事件结尾
+//  3. 响应体中只出现一个 message_start，不存在第二个（防止流拼接腐化）
+func TestStreamWrittenGuard_MessagesPath_AbortFailoverOnSSEContentWritten(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	w := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(w)
+	c.Request = httptest.NewRequest(http.MethodPost, "/v1/messages", nil)
+
+	// 步骤 1：记录 Forward 前的 writer size（模拟 writerSizeBeforeForward := c.Writer.Size()）
+	sizeBeforeForward := c.Writer.Size()
+	require.Equal(t, -1, sizeBeforeForward, "gin writer 初始 Size 应为 -1（未写入任何字节）")
+
+	// 步骤 2：模拟 Forward 已向客户端写入部分 SSE 内容（message_start + content_block_start）
+	_, err := c.Writer.Write([]byte(partialMessageStartSSE))
+	require.NoError(t, err)
+
+	// 步骤 3：验证守卫条件成立（c.Writer.Size() != sizeBeforeForward）
+	require.NotEqual(t, sizeBeforeForward, c.Writer.Size(),
+		"写入 SSE 内容后 writer size 必须增加，守卫条件应为 true")
+
+	// 步骤 4：模拟 UpstreamFailoverError（上游在流中途返回 403）
+	failoverErr := &service.UpstreamFailoverError{
+		StatusCode:   http.StatusForbidden,
+		ResponseBody: []byte(`{"error":{"type":"permission_error","message":"forbidden"}}`),
+	}
+
+	// 步骤 5：守卫触发 → 调用 handleFailoverExhausted，streamStarted=true
+	h := &GatewayHandler{}
+	h.handleFailoverExhausted(c, failoverErr, service.PlatformAnthropic, true)
+
+	body := w.Body.String()
+
+	// 断言 A：响应体中包含最初写入的 message_start SSE 事件行
+	require.Contains(t, body, "event: message_start", "响应体应包含已写入的 message_start SSE 事件")
+
+	// 断言 B：响应体以 SSE 错误事件结尾（data: {"type":"error",...}\n\n）
+	require.True(t, strings.HasSuffix(strings.TrimRight(body, "\n"), "}"),
+		"响应体应以 JSON 对象结尾（SSE error event 的 data 字段）")
+	require.Contains(t, body, `"type":"error"`, "响应体末尾必须包含 SSE 错误事件")
+
+	// 断言 C：SSE event 行 "event: message_start" 只出现一次（防止双 message_start 拼接腐化）
+	firstIdx := strings.Index(body, "event: message_start")
+	lastIdx := strings.LastIndex(body, "event: message_start")
+	assert.Equal(t, firstIdx, lastIdx,
+		"响应体中 'event: message_start' 必须只出现一次，不得因 failover 拼接导致两次")
+}
+
+// TestStreamWrittenGuard_GeminiPath_AbortFailoverOnSSEContentWritten 与上述测试相同，
+// 验证 Gemini 路径使用 service.PlatformGemini（而非 account.Platform）时行为一致。
+func TestStreamWrittenGuard_GeminiPath_AbortFailoverOnSSEContentWritten(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	w := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(w)
+	c.Request = httptest.NewRequest(http.MethodPost, "/v1beta/models/gemini-2.0-flash:streamGenerateContent", nil)
+
+	sizeBeforeForward := c.Writer.Size()
+
+	_, err := c.Writer.Write([]byte(partialMessageStartSSE))
+	require.NoError(t, err)
+
+	require.NotEqual(t, sizeBeforeForward, c.Writer.Size())
+
+	failoverErr := &service.UpstreamFailoverError{
+		StatusCode: http.StatusForbidden,
+	}
+
+	h := &GatewayHandler{}
+	h.handleFailoverExhausted(c, failoverErr, service.PlatformGemini, true)
+
+	body := w.Body.String()
+
+	require.Contains(t, body, "event: message_start")
+	require.Contains(t, body, `"type":"error"`)
+
+	firstIdx := strings.Index(body, "event: message_start")
+	lastIdx := strings.LastIndex(body, "event: message_start")
+	assert.Equal(t, firstIdx, lastIdx, "Gemini 路径不得出现双 message_start")
+}
+
+// TestStreamWrittenGuard_NoByteWritten_GuardNotTriggered 验证反向场景：
+// 当 Forward 返回 UpstreamFailoverError 时若未向客户端写入任何 SSE 内容，
+// 守卫条件（c.Writer.Size() != sizeBeforeForward）为 false，不应中止 failover。
+func TestStreamWrittenGuard_NoByteWritten_GuardNotTriggered(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	w := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(w)
+	c.Request = httptest.NewRequest(http.MethodPost, "/v1/messages", nil)
+
+	// 模拟 writerSizeBeforeForward：初始为 -1
+	sizeBeforeForward := c.Writer.Size()
+
+	// Forward 未写入任何字节直接返回错误（例如 401 发生在连接建立前）
+	// c.Writer.Size() 仍为 -1
+
+	// 守卫条件：sizeBeforeForward == c.Writer.Size() → 不触发
+	guardTriggered := c.Writer.Size() != sizeBeforeForward
+	require.False(t, guardTriggered,
+		"未写入任何字节时，守卫条件必须为 false，应允许正常 failover 继续")
+}
--- a/backend/internal/handler/openai_chat_completions.go
+++ b/backend/internal/handler/openai_chat_completions.go
@@ -256,14 +256,16 @@ func (h *OpenAIGatewayHandler) ChatCompletions(c *gin.Context) {

 		h.submitUsageRecordTask(func(ctx context.Context) {
 			if err := h.gatewayService.RecordUsage(ctx, &service.OpenAIRecordUsageInput{
-				Result:        result,
-				APIKey:        apiKey,
-				User:          apiKey.User,
-				Account:       account,
-				Subscription:  subscription,
-				UserAgent:     userAgent,
-				IPAddress:     clientIP,
-				APIKeyService: h.apiKeyService,
+				Result:           result,
+				APIKey:           apiKey,
+				User:             apiKey.User,
+				Account:          account,
+				Subscription:     subscription,
+				InboundEndpoint:  normalizedOpenAIInboundEndpoint(c, openAIInboundEndpointChatCompletions),
+				UpstreamEndpoint: normalizedOpenAIUpstreamEndpoint(c, openAIUpstreamEndpointResponses),
+				UserAgent:        userAgent,
+				IPAddress:        clientIP,
+				APIKeyService:    h.apiKeyService,
 			}); err != nil {
 				logger.L().With(
 					zap.String("component", "handler.openai_gateway.chat_completions"),
--- a/backend/internal/handler/openai_gateway_endpoint_normalization_test.go
+++ b/backend/internal/handler/openai_gateway_endpoint_normalization_test.go
@@ -0,0 +1,57 @@
+package handler
+
+import (
+	"net/http"
+	"net/http/httptest"
+	"testing"
+
+	"github.com/gin-gonic/gin"
+	"github.com/stretchr/testify/require"
+)
+
+func TestNormalizedOpenAIUpstreamEndpoint(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+
+	tests := []struct {
+		name     string
+		path     string
+		fallback string
+		want     string
+	}{
+		{
+			name:     "responses root maps to responses upstream",
+			path:     "/v1/responses",
+			fallback: openAIUpstreamEndpointResponses,
+			want:     "/v1/responses",
+		},
+		{
+			name:     "responses compact keeps compact suffix",
+			path:     "/openai/v1/responses/compact",
+			fallback: openAIUpstreamEndpointResponses,
+			want:     "/v1/responses/compact",
+		},
+		{
+			name:     "responses nested suffix preserved",
+			path:     "/openai/v1/responses/compact/detail",
+			fallback: openAIUpstreamEndpointResponses,
+			want:     "/v1/responses/compact/detail",
+		},
+		{
+			name:     "non responses path uses fallback",
+			path:     "/v1/messages",
+			fallback: openAIUpstreamEndpointResponses,
+			want:     "/v1/responses",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			rec := httptest.NewRecorder()
+			c, _ := gin.CreateTestContext(rec)
+			c.Request = httptest.NewRequest(http.MethodPost, tt.path, nil)
+
+			got := normalizedOpenAIUpstreamEndpoint(c, tt.fallback)
+			require.Equal(t, tt.want, got)
+		})
+	}
+}
--- a/backend/internal/handler/openai_gateway_handler.go
+++ b/backend/internal/handler/openai_gateway_handler.go
@@ -37,6 +37,13 @@ type OpenAIGatewayHandler struct {
 	cfg                     *config.Config
 }

+const (
+	openAIInboundEndpointResponses       = "/v1/responses"
+	openAIInboundEndpointMessages        = "/v1/messages"
+	openAIInboundEndpointChatCompletions = "/v1/chat/completions"
+	openAIUpstreamEndpointResponses      = "/v1/responses"
+)
+
 // NewOpenAIGatewayHandler creates a new OpenAIGatewayHandler
 func NewOpenAIGatewayHandler(
 	gatewayService *service.OpenAIGatewayService,
@@ -362,6 +369,8 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 				User:               apiKey.User,
 				Account:            account,
 				Subscription:       subscription,
+				InboundEndpoint:    normalizedOpenAIInboundEndpoint(c, openAIInboundEndpointResponses),
+				UpstreamEndpoint:   normalizedOpenAIUpstreamEndpoint(c, openAIUpstreamEndpointResponses),
 				UserAgent:          userAgent,
 				IPAddress:          clientIP,
 				RequestPayloadHash: requestPayloadHash,
@@ -738,6 +747,8 @@ func (h *OpenAIGatewayHandler) Messages(c *gin.Context) {
 				User:               apiKey.User,
 				Account:            account,
 				Subscription:       subscription,
+				InboundEndpoint:    normalizedOpenAIInboundEndpoint(c, openAIInboundEndpointMessages),
+				UpstreamEndpoint:   normalizedOpenAIUpstreamEndpoint(c, openAIUpstreamEndpointResponses),
 				UserAgent:          userAgent,
 				IPAddress:          clientIP,
 				RequestPayloadHash: requestPayloadHash,
@@ -1235,6 +1246,8 @@ func (h *OpenAIGatewayHandler) ResponsesWebSocket(c *gin.Context) {
 					User:               apiKey.User,
 					Account:            account,
 					Subscription:       subscription,
+					InboundEndpoint:    normalizedOpenAIInboundEndpoint(c, openAIInboundEndpointResponses),
+					UpstreamEndpoint:   normalizedOpenAIUpstreamEndpoint(c, openAIUpstreamEndpointResponses),
 					UserAgent:          userAgent,
 					IPAddress:          clientIP,
 					RequestPayloadHash: service.HashUsageRequestPayload(firstMessage),
@@ -1530,6 +1543,62 @@ func openAIWSIngressFallbackSessionSeed(userID, apiKeyID int64, groupID *int64)
 	return fmt.Sprintf("openai_ws_ingress:%d:%d:%d", gid, userID, apiKeyID)
 }

+func normalizedOpenAIInboundEndpoint(c *gin.Context, fallback string) string {
+	path := strings.TrimSpace(fallback)
+	if c != nil {
+		if fullPath := strings.TrimSpace(c.FullPath()); fullPath != "" {
+			path = fullPath
+		} else if c.Request != nil && c.Request.URL != nil {
+			if requestPath := strings.TrimSpace(c.Request.URL.Path); requestPath != "" {
+				path = requestPath
+			}
+		}
+	}
+
+	switch {
+	case strings.Contains(path, openAIInboundEndpointChatCompletions):
+		return openAIInboundEndpointChatCompletions
+	case strings.Contains(path, openAIInboundEndpointMessages):
+		return openAIInboundEndpointMessages
+	case strings.Contains(path, openAIInboundEndpointResponses):
+		return openAIInboundEndpointResponses
+	default:
+		return path
+	}
+}
+
+func normalizedOpenAIUpstreamEndpoint(c *gin.Context, fallback string) string {
+	base := strings.TrimSpace(fallback)
+	if base == "" {
+		base = openAIUpstreamEndpointResponses
+	}
+	base = strings.TrimRight(base, "/")
+
+	if c == nil || c.Request == nil || c.Request.URL == nil {
+		return base
+	}
+
+	path := strings.TrimRight(strings.TrimSpace(c.Request.URL.Path), "/")
+	if path == "" {
+		return base
+	}
+
+	idx := strings.LastIndex(path, "/responses")
+	if idx < 0 {
+		return base
+	}
+
+	suffix := strings.TrimSpace(path[idx+len("/responses"):])
+	if suffix == "" || suffix == "/" {
+		return base
+	}
+	if !strings.HasPrefix(suffix, "/") {
+		return base
+	}
+
+	return base + suffix
+}
+
 func isOpenAIWSUpgradeRequest(r *http.Request) bool {
 	if r == nil {
 		return false
--- a/backend/internal/handler/sora_gateway_handler_test.go
+++ b/backend/internal/handler/sora_gateway_handler_test.go
@@ -334,6 +334,14 @@ func (s *stubUsageLogRepo) GetUsageTrendWithFilters(ctx context.Context, startTi
 func (s *stubUsageLogRepo) GetModelStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, requestType *int16, stream *bool, billingType *int8) ([]usagestats.ModelStat, error) {
 	return nil, nil
 }
+
+func (s *stubUsageLogRepo) GetEndpointStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, model string, requestType *int16, stream *bool, billingType *int8) ([]usagestats.EndpointStat, error) {
+	return []usagestats.EndpointStat{}, nil
+}
+
+func (s *stubUsageLogRepo) GetUpstreamEndpointStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, model string, requestType *int16, stream *bool, billingType *int8) ([]usagestats.EndpointStat, error) {
+	return []usagestats.EndpointStat{}, nil
+}
 func (s *stubUsageLogRepo) GetGroupStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, requestType *int16, stream *bool, billingType *int8) ([]usagestats.GroupStat, error) {
 	return nil, nil
 }
--- a/backend/internal/pkg/usagestats/usage_log_types.go
+++ b/backend/internal/pkg/usagestats/usage_log_types.go
@@ -81,6 +81,15 @@ type ModelStat struct {
 	ActualCost          float64 `json:"actual_cost"` // 实际扣除
 }

+// EndpointStat represents usage statistics for a single request endpoint.
+type EndpointStat struct {
+	Endpoint    string  `json:"endpoint"`
+	Requests    int64   `json:"requests"`
+	TotalTokens int64   `json:"total_tokens"`
+	Cost        float64 `json:"cost"`        // 标准计费
+	ActualCost  float64 `json:"actual_cost"` // 实际扣除
+}
+
 // GroupStat represents usage statistics for a single group
 type GroupStat struct {
 	GroupID     int64   `json:"group_id"`
@@ -179,15 +188,18 @@ 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"`
-	TotalAccountCost  *float64 `json:"total_account_cost,omitempty"`
-	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"`
+	Endpoints         []EndpointStat `json:"endpoints,omitempty"`
+	UpstreamEndpoints []EndpointStat `json:"upstream_endpoints,omitempty"`
+	EndpointPaths     []EndpointStat `json:"endpoint_paths,omitempty"`
 }

 // BatchUserUsageStats represents usage stats for a single user
@@ -254,7 +266,9 @@ type AccountUsageSummary struct {

 // AccountUsageStatsResponse represents the full usage statistics response for an account
 type AccountUsageStatsResponse struct {
-	History []AccountUsageHistory `json:"history"`
-	Summary AccountUsageSummary   `json:"summary"`
-	Models  []ModelStat           `json:"models"`
+	History           []AccountUsageHistory `json:"history"`
+	Summary           AccountUsageSummary   `json:"summary"`
+	Models            []ModelStat           `json:"models"`
+	Endpoints         []EndpointStat        `json:"endpoints"`
+	UpstreamEndpoints []EndpointStat        `json:"upstream_endpoints"`
 }
--- a/backend/internal/repository/usage_billing_repo.go
+++ b/backend/internal/repository/usage_billing_repo.go
@@ -132,7 +132,7 @@ func (r *usageBillingRepository) applyUsageBillingEffects(ctx context.Context, t
 		}
 	}

-	if cmd.AccountQuotaCost > 0 && strings.EqualFold(cmd.AccountType, service.AccountTypeAPIKey) {
+	if cmd.AccountQuotaCost > 0 && (strings.EqualFold(cmd.AccountType, service.AccountTypeAPIKey) || strings.EqualFold(cmd.AccountType, service.AccountTypeBedrock)) {
 		if err := incrementUsageBillingAccountQuota(ctx, tx, cmd.AccountID, cmd.AccountQuotaCost); err != nil {
 			return err
 		}
--- a/backend/internal/repository/usage_log_repo.go
+++ b/backend/internal/repository/usage_log_repo.go
@@ -28,7 +28,7 @@ import (
 	gocache "github.com/patrickmn/go-cache"
 )

-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, request_type, stream, openai_ws_mode, duration_ms, first_token_ms, user_agent, ip_address, image_count, image_size, media_type, service_tier, reasoning_effort, cache_ttl_overridden, 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, request_type, stream, openai_ws_mode, duration_ms, first_token_ms, user_agent, ip_address, image_count, image_size, media_type, service_tier, reasoning_effort, inbound_endpoint, upstream_endpoint, cache_ttl_overridden, created_at"

 var usageLogInsertArgTypes = [...]string{
 	"bigint",
@@ -65,6 +65,8 @@ var usageLogInsertArgTypes = [...]string{
 	"text",
 	"text",
 	"text",
+	"text",
+	"text",
 	"boolean",
 	"timestamptz",
 }
@@ -304,6 +306,8 @@ func (r *usageLogRepository) createSingle(ctx context.Context, sqlq sqlExecutor,
 			media_type,
 			service_tier,
 			reasoning_effort,
+			inbound_endpoint,
+			upstream_endpoint,
 			cache_ttl_overridden,
 			created_at
 		) VALUES (
@@ -312,7 +316,7 @@ func (r *usageLogRepository) createSingle(ctx context.Context, sqlq sqlExecutor,
 			$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, $36
+			$20, $21, $22, $23, $24, $25, $26, $27, $28, $29, $30, $31, $32, $33, $34, $35, $36, $37, $38
 		)
 		ON CONFLICT (request_id, api_key_id) DO NOTHING
 		RETURNING id, created_at
@@ -732,11 +736,13 @@ func buildUsageLogBatchInsertQuery(keys []string, preparedByKey map[string]usage
 			media_type,
 			service_tier,
 			reasoning_effort,
+			inbound_endpoint,
+			upstream_endpoint,
 			cache_ttl_overridden,
 			created_at
 		) AS (VALUES `)

-	args := make([]any, 0, len(keys)*37)
+	args := make([]any, 0, len(keys)*38)
 	argPos := 1
 	for idx, key := range keys {
 		if idx > 0 {
@@ -799,6 +805,8 @@ func buildUsageLogBatchInsertQuery(keys []string, preparedByKey map[string]usage
 				media_type,
 				service_tier,
 				reasoning_effort,
+				inbound_endpoint,
+				upstream_endpoint,
 				cache_ttl_overridden,
 				created_at
 			)
@@ -837,6 +845,8 @@ func buildUsageLogBatchInsertQuery(keys []string, preparedByKey map[string]usage
 				media_type,
 				service_tier,
 				reasoning_effort,
+				inbound_endpoint,
+				upstream_endpoint,
 				cache_ttl_overridden,
 				created_at
 			FROM input
@@ -915,11 +925,13 @@ func buildUsageLogBestEffortInsertQuery(preparedList []usageLogInsertPrepared) (
 			media_type,
 			service_tier,
 			reasoning_effort,
+			inbound_endpoint,
+			upstream_endpoint,
 			cache_ttl_overridden,
 			created_at
 		) AS (VALUES `)

-	args := make([]any, 0, len(preparedList)*36)
+	args := make([]any, 0, len(preparedList)*38)
 	argPos := 1
 	for idx, prepared := range preparedList {
 		if idx > 0 {
@@ -979,6 +991,8 @@ func buildUsageLogBestEffortInsertQuery(preparedList []usageLogInsertPrepared) (
 			media_type,
 			service_tier,
 			reasoning_effort,
+			inbound_endpoint,
+			upstream_endpoint,
 			cache_ttl_overridden,
 			created_at
 		)
@@ -1017,6 +1031,8 @@ func buildUsageLogBestEffortInsertQuery(preparedList []usageLogInsertPrepared) (
 			media_type,
 			service_tier,
 			reasoning_effort,
+			inbound_endpoint,
+			upstream_endpoint,
 			cache_ttl_overridden,
 			created_at
 		FROM input
@@ -1063,6 +1079,8 @@ func execUsageLogInsertNoResult(ctx context.Context, sqlq sqlExecutor, prepared
 			media_type,
 			service_tier,
 			reasoning_effort,
+			inbound_endpoint,
+			upstream_endpoint,
 			cache_ttl_overridden,
 			created_at
 		) VALUES (
@@ -1071,7 +1089,7 @@ func execUsageLogInsertNoResult(ctx context.Context, sqlq sqlExecutor, prepared
 			$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, $36
+			$20, $21, $22, $23, $24, $25, $26, $27, $28, $29, $30, $31, $32, $33, $34, $35, $36, $37, $38
 		)
 		ON CONFLICT (request_id, api_key_id) DO NOTHING
 	`, prepared.args...)
@@ -1101,6 +1119,8 @@ func prepareUsageLogInsert(log *service.UsageLog) usageLogInsertPrepared {
 	mediaType := nullString(log.MediaType)
 	serviceTier := nullString(log.ServiceTier)
 	reasoningEffort := nullString(log.ReasoningEffort)
+	inboundEndpoint := nullString(log.InboundEndpoint)
+	upstreamEndpoint := nullString(log.UpstreamEndpoint)

 	var requestIDArg any
 	if requestID != "" {
@@ -1147,6 +1167,8 @@ func prepareUsageLogInsert(log *service.UsageLog) usageLogInsertPrepared {
 			mediaType,
 			serviceTier,
 			reasoningEffort,
+			inboundEndpoint,
+			upstreamEndpoint,
 			log.CacheTTLOverridden,
 			createdAt,
 		},
@@ -2505,7 +2527,7 @@ func (r *usageLogRepository) ListWithFilters(ctx context.Context, params paginat
 		args = append(args, *filters.StartTime)
 	}
 	if filters.EndTime != nil {
-		conditions = append(conditions, fmt.Sprintf("created_at <= $%d", len(args)+1))
+		conditions = append(conditions, fmt.Sprintf("created_at < $%d", len(args)+1))
 		args = append(args, *filters.EndTime)
 	}

@@ -3040,7 +3062,7 @@ func (r *usageLogRepository) GetStatsWithFilters(ctx context.Context, filters Us
 		args = append(args, *filters.StartTime)
 	}
 	if filters.EndTime != nil {
-		conditions = append(conditions, fmt.Sprintf("created_at <= $%d", len(args)+1))
+		conditions = append(conditions, fmt.Sprintf("created_at < $%d", len(args)+1))
 		args = append(args, *filters.EndTime)
 	}

@@ -3080,6 +3102,35 @@ func (r *usageLogRepository) GetStatsWithFilters(ctx context.Context, filters Us
 		stats.TotalAccountCost = &totalAccountCost
 	}
 	stats.TotalTokens = stats.TotalInputTokens + stats.TotalOutputTokens + stats.TotalCacheTokens
+
+	start := time.Unix(0, 0).UTC()
+	if filters.StartTime != nil {
+		start = *filters.StartTime
+	}
+	end := time.Now().UTC()
+	if filters.EndTime != nil {
+		end = *filters.EndTime
+	}
+
+	endpoints, endpointErr := r.GetEndpointStatsWithFilters(ctx, start, end, filters.UserID, filters.APIKeyID, filters.AccountID, filters.GroupID, filters.Model, filters.RequestType, filters.Stream, filters.BillingType)
+	if endpointErr != nil {
+		logger.LegacyPrintf("repository.usage_log", "GetEndpointStatsWithFilters failed in GetStatsWithFilters: %v", endpointErr)
+		endpoints = []EndpointStat{}
+	}
+	upstreamEndpoints, upstreamEndpointErr := r.GetUpstreamEndpointStatsWithFilters(ctx, start, end, filters.UserID, filters.APIKeyID, filters.AccountID, filters.GroupID, filters.Model, filters.RequestType, filters.Stream, filters.BillingType)
+	if upstreamEndpointErr != nil {
+		logger.LegacyPrintf("repository.usage_log", "GetUpstreamEndpointStatsWithFilters failed in GetStatsWithFilters: %v", upstreamEndpointErr)
+		upstreamEndpoints = []EndpointStat{}
+	}
+	endpointPaths, endpointPathErr := r.getEndpointPathStatsWithFilters(ctx, start, end, filters.UserID, filters.APIKeyID, filters.AccountID, filters.GroupID, filters.Model, filters.RequestType, filters.Stream, filters.BillingType)
+	if endpointPathErr != nil {
+		logger.LegacyPrintf("repository.usage_log", "getEndpointPathStatsWithFilters failed in GetStatsWithFilters: %v", endpointPathErr)
+		endpointPaths = []EndpointStat{}
+	}
+	stats.Endpoints = endpoints
+	stats.UpstreamEndpoints = upstreamEndpoints
+	stats.EndpointPaths = endpointPaths
+
 	return stats, nil
 }

@@ -3092,6 +3143,163 @@ type AccountUsageSummary = usagestats.AccountUsageSummary
 // AccountUsageStatsResponse represents the full usage statistics response for an account
 type AccountUsageStatsResponse = usagestats.AccountUsageStatsResponse

+// EndpointStat represents endpoint usage statistics row.
+type EndpointStat = usagestats.EndpointStat
+
+func (r *usageLogRepository) getEndpointStatsByColumnWithFilters(ctx context.Context, endpointColumn string, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, model string, requestType *int16, stream *bool, billingType *int8) (results []EndpointStat, err error) {
+	actualCostExpr := "COALESCE(SUM(actual_cost), 0) as actual_cost"
+	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
+			COALESCE(NULLIF(TRIM(%s), ''), 'unknown') AS endpoint,
+			COUNT(*) AS requests,
+			COALESCE(SUM(input_tokens + output_tokens + cache_creation_tokens + cache_read_tokens), 0) AS total_tokens,
+			COALESCE(SUM(total_cost), 0) as cost,
+			%s
+		FROM usage_logs
+		WHERE created_at >= $1 AND created_at < $2
+	`, endpointColumn, actualCostExpr)
+
+	args := []any{startTime, endTime}
+	if userID > 0 {
+		query += fmt.Sprintf(" AND user_id = $%d", len(args)+1)
+		args = append(args, userID)
+	}
+	if apiKeyID > 0 {
+		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)
+	}
+	query, args = appendRequestTypeOrStreamQueryFilter(query, args, requestType, stream)
+	if billingType != nil {
+		query += fmt.Sprintf(" AND billing_type = $%d", len(args)+1)
+		args = append(args, int16(*billingType))
+	}
+	query += " GROUP BY endpoint ORDER BY requests DESC"
+
+	rows, err := r.sql.QueryContext(ctx, query, args...)
+	if err != nil {
+		return nil, err
+	}
+	defer func() {
+		if closeErr := rows.Close(); closeErr != nil && err == nil {
+			err = closeErr
+			results = nil
+		}
+	}()
+
+	results = make([]EndpointStat, 0)
+	for rows.Next() {
+		var row EndpointStat
+		if err := rows.Scan(&row.Endpoint, &row.Requests, &row.TotalTokens, &row.Cost, &row.ActualCost); err != nil {
+			return nil, err
+		}
+		results = append(results, row)
+	}
+	if err := rows.Err(); err != nil {
+		return nil, err
+	}
+	return results, nil
+}
+
+func (r *usageLogRepository) getEndpointPathStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, model string, requestType *int16, stream *bool, billingType *int8) (results []EndpointStat, err error) {
+	actualCostExpr := "COALESCE(SUM(actual_cost), 0) as actual_cost"
+	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
+			CONCAT(
+				COALESCE(NULLIF(TRIM(inbound_endpoint), ''), 'unknown'),
+				' -> ',
+				COALESCE(NULLIF(TRIM(upstream_endpoint), ''), 'unknown')
+			) AS endpoint,
+			COUNT(*) AS requests,
+			COALESCE(SUM(input_tokens + output_tokens + cache_creation_tokens + cache_read_tokens), 0) AS total_tokens,
+			COALESCE(SUM(total_cost), 0) as cost,
+			%s
+		FROM usage_logs
+		WHERE created_at >= $1 AND created_at < $2
+	`, actualCostExpr)
+
+	args := []any{startTime, endTime}
+	if userID > 0 {
+		query += fmt.Sprintf(" AND user_id = $%d", len(args)+1)
+		args = append(args, userID)
+	}
+	if apiKeyID > 0 {
+		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)
+	}
+	query, args = appendRequestTypeOrStreamQueryFilter(query, args, requestType, stream)
+	if billingType != nil {
+		query += fmt.Sprintf(" AND billing_type = $%d", len(args)+1)
+		args = append(args, int16(*billingType))
+	}
+	query += " GROUP BY endpoint ORDER BY requests DESC"
+
+	rows, err := r.sql.QueryContext(ctx, query, args...)
+	if err != nil {
+		return nil, err
+	}
+	defer func() {
+		if closeErr := rows.Close(); closeErr != nil && err == nil {
+			err = closeErr
+			results = nil
+		}
+	}()
+
+	results = make([]EndpointStat, 0)
+	for rows.Next() {
+		var row EndpointStat
+		if err := rows.Scan(&row.Endpoint, &row.Requests, &row.TotalTokens, &row.Cost, &row.ActualCost); err != nil {
+			return nil, err
+		}
+		results = append(results, row)
+	}
+	if err := rows.Err(); err != nil {
+		return nil, err
+	}
+	return results, nil
+}
+
+// GetEndpointStatsWithFilters returns inbound endpoint statistics with optional filters.
+func (r *usageLogRepository) GetEndpointStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, model string, requestType *int16, stream *bool, billingType *int8) ([]EndpointStat, error) {
+	return r.getEndpointStatsByColumnWithFilters(ctx, "inbound_endpoint", startTime, endTime, userID, apiKeyID, accountID, groupID, model, requestType, stream, billingType)
+}
+
+// GetUpstreamEndpointStatsWithFilters returns upstream endpoint statistics with optional filters.
+func (r *usageLogRepository) GetUpstreamEndpointStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, model string, requestType *int16, stream *bool, billingType *int8) ([]EndpointStat, error) {
+	return r.getEndpointStatsByColumnWithFilters(ctx, "upstream_endpoint", startTime, endTime, userID, apiKeyID, accountID, groupID, model, requestType, stream, billingType)
+}
+
 // GetAccountUsageStats returns comprehensive usage statistics for an account over a time range
 func (r *usageLogRepository) GetAccountUsageStats(ctx context.Context, accountID int64, startTime, endTime time.Time) (resp *AccountUsageStatsResponse, err error) {
 	daysCount := int(endTime.Sub(startTime).Hours()/24) + 1
@@ -3254,11 +3462,23 @@ func (r *usageLogRepository) GetAccountUsageStats(ctx context.Context, accountID
 	if err != nil {
 		models = []ModelStat{}
 	}
+	endpoints, endpointErr := r.GetEndpointStatsWithFilters(ctx, startTime, endTime, 0, 0, accountID, 0, "", nil, nil, nil)
+	if endpointErr != nil {
+		logger.LegacyPrintf("repository.usage_log", "GetEndpointStatsWithFilters failed in GetAccountUsageStats: %v", endpointErr)
+		endpoints = []EndpointStat{}
+	}
+	upstreamEndpoints, upstreamEndpointErr := r.GetUpstreamEndpointStatsWithFilters(ctx, startTime, endTime, 0, 0, accountID, 0, "", nil, nil, nil)
+	if upstreamEndpointErr != nil {
+		logger.LegacyPrintf("repository.usage_log", "GetUpstreamEndpointStatsWithFilters failed in GetAccountUsageStats: %v", upstreamEndpointErr)
+		upstreamEndpoints = []EndpointStat{}
+	}

 	resp = &AccountUsageStatsResponse{
-		History: history,
-		Summary: summary,
-		Models:  models,
+		History:           history,
+		Summary:           summary,
+		Models:            models,
+		Endpoints:         endpoints,
+		UpstreamEndpoints: upstreamEndpoints,
 	}
 	return resp, nil
 }
@@ -3541,6 +3761,8 @@ func scanUsageLog(scanner interface{ Scan(...any) error }) (*service.UsageLog, e
 		mediaType             sql.NullString
 		serviceTier           sql.NullString
 		reasoningEffort       sql.NullString
+		inboundEndpoint       sql.NullString
+		upstreamEndpoint      sql.NullString
 		cacheTTLOverridden    bool
 		createdAt             time.Time
 	)
@@ -3581,6 +3803,8 @@ func scanUsageLog(scanner interface{ Scan(...any) error }) (*service.UsageLog, e
 		&mediaType,
 		&serviceTier,
 		&reasoningEffort,
+		&inboundEndpoint,
+		&upstreamEndpoint,
 		&cacheTTLOverridden,
 		&createdAt,
 	); err != nil {
@@ -3656,6 +3880,12 @@ func scanUsageLog(scanner interface{ Scan(...any) error }) (*service.UsageLog, e
 	if reasoningEffort.Valid {
 		log.ReasoningEffort = &reasoningEffort.String
 	}
+	if inboundEndpoint.Valid {
+		log.InboundEndpoint = &inboundEndpoint.String
+	}
+	if upstreamEndpoint.Valid {
+		log.UpstreamEndpoint = &upstreamEndpoint.String
+	}

 	return log, nil
 }
--- a/backend/internal/repository/usage_log_repo_request_type_test.go
+++ b/backend/internal/repository/usage_log_repo_request_type_test.go
@@ -73,6 +73,8 @@ func TestUsageLogRepositoryCreateSyncRequestTypeAndLegacyFields(t *testing.T) {
 			sqlmock.AnyArg(), // media_type
 			sqlmock.AnyArg(), // service_tier
 			sqlmock.AnyArg(), // reasoning_effort
+			sqlmock.AnyArg(), // inbound_endpoint
+			sqlmock.AnyArg(), // upstream_endpoint
 			log.CacheTTLOverridden,
 			createdAt,
 		).
@@ -141,6 +143,8 @@ func TestUsageLogRepositoryCreate_PersistsServiceTier(t *testing.T) {
 			sqlmock.AnyArg(),
 			serviceTier,
 			sqlmock.AnyArg(),
+			sqlmock.AnyArg(),
+			sqlmock.AnyArg(),
 			log.CacheTTLOverridden,
 			createdAt,
 		).
@@ -376,6 +380,8 @@ func TestScanUsageLogRequestTypeAndLegacyFallback(t *testing.T) {
 			sql.NullString{},
 			sql.NullString{Valid: true, String: "priority"},
 			sql.NullString{},
+			sql.NullString{},
+			sql.NullString{},
 			false,
 			now,
 		}})
@@ -415,6 +421,8 @@ func TestScanUsageLogRequestTypeAndLegacyFallback(t *testing.T) {
 			sql.NullString{},
 			sql.NullString{Valid: true, String: "flex"},
 			sql.NullString{},
+			sql.NullString{},
+			sql.NullString{},
 			false,
 			now,
 		}})
@@ -454,6 +462,8 @@ func TestScanUsageLogRequestTypeAndLegacyFallback(t *testing.T) {
 			sql.NullString{},
 			sql.NullString{Valid: true, String: "priority"},
 			sql.NullString{},
+			sql.NullString{},
+			sql.NullString{},
 			false,
 			now,
 		}})
--- a/backend/internal/server/api_contract_test.go
+++ b/backend/internal/server/api_contract_test.go
@@ -1624,6 +1624,14 @@ func (r *stubUsageLogRepo) GetModelStatsWithFilters(ctx context.Context, startTi
 	return nil, errors.New("not implemented")
 }

+func (r *stubUsageLogRepo) GetEndpointStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, model string, requestType *int16, stream *bool, billingType *int8) ([]usagestats.EndpointStat, error) {
+	return nil, errors.New("not implemented")
+}
+
+func (r *stubUsageLogRepo) GetUpstreamEndpointStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, model string, requestType *int16, stream *bool, billingType *int8) ([]usagestats.EndpointStat, error) {
+	return nil, errors.New("not implemented")
+}
+
 func (r *stubUsageLogRepo) GetGroupStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, requestType *int16, stream *bool, billingType *int8) ([]usagestats.GroupStat, error) {
 	return nil, errors.New("not implemented")
 }
--- a/backend/internal/service/account_usage_service.go
+++ b/backend/internal/service/account_usage_service.go
@@ -45,6 +45,8 @@ type UsageLogRepository interface {
 	GetDashboardStats(ctx context.Context) (*usagestats.DashboardStats, error)
 	GetUsageTrendWithFilters(ctx context.Context, startTime, endTime time.Time, granularity string, userID, apiKeyID, accountID, groupID int64, model string, requestType *int16, stream *bool, billingType *int8) ([]usagestats.TrendDataPoint, error)
 	GetModelStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, requestType *int16, stream *bool, billingType *int8) ([]usagestats.ModelStat, error)
+	GetEndpointStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, model string, requestType *int16, stream *bool, billingType *int8) ([]usagestats.EndpointStat, error)
+	GetUpstreamEndpointStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, model string, requestType *int16, stream *bool, billingType *int8) ([]usagestats.EndpointStat, error)
 	GetGroupStatsWithFilters(ctx context.Context, startTime, endTime time.Time, userID, apiKeyID, accountID, groupID int64, requestType *int16, stream *bool, billingType *int8) ([]usagestats.GroupStat, 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)
--- a/backend/internal/service/admin_service.go
+++ b/backend/internal/service/admin_service.go
@@ -832,7 +832,7 @@ func (s *adminServiceImpl) CreateGroup(ctx context.Context, input *CreateGroupIn
 		subscriptionType = SubscriptionTypeStandard
 	}

-	// 限额字段：0 和 nil 都表示"无限制"
+	// 限额字段：nil/负数 表示"无限制"，0 表示"不允许用量"，正数表示具体限额
 	dailyLimit := normalizeLimit(input.DailyLimitUSD)
 	weeklyLimit := normalizeLimit(input.WeeklyLimitUSD)
 	monthlyLimit := normalizeLimit(input.MonthlyLimitUSD)
@@ -944,9 +944,9 @@ func (s *adminServiceImpl) CreateGroup(ctx context.Context, input *CreateGroupIn
 	return group, nil
 }

-// normalizeLimit 将 0 或负数转换为 nil（表示无限制）
+// normalizeLimit 将负数转换为 nil（表示无限制），0 保留（表示限额为零）
 func normalizeLimit(limit *float64) *float64 {
-	if limit == nil || *limit <= 0 {
+	if limit == nil || *limit < 0 {
 		return nil
 	}
 	return limit
@@ -1058,16 +1058,11 @@ func (s *adminServiceImpl) UpdateGroup(ctx context.Context, id int64, input *Upd
 	if input.SubscriptionType != "" {
 		group.SubscriptionType = input.SubscriptionType
 	}
-	// 限额字段：0 和 nil 都表示"无限制"，正数表示具体限额
-	if input.DailyLimitUSD != nil {
-		group.DailyLimitUSD = normalizeLimit(input.DailyLimitUSD)
-	}
-	if input.WeeklyLimitUSD != nil {
-		group.WeeklyLimitUSD = normalizeLimit(input.WeeklyLimitUSD)
-	}
-	if input.MonthlyLimitUSD != nil {
-		group.MonthlyLimitUSD = normalizeLimit(input.MonthlyLimitUSD)
-	}
+	// 限额字段：nil/负数 表示"无限制"，0 表示"不允许用量"，正数表示具体限额
+	// 前端始终发送这三个字段，无需 nil 守卫
+	group.DailyLimitUSD = normalizeLimit(input.DailyLimitUSD)
+	group.WeeklyLimitUSD = normalizeLimit(input.WeeklyLimitUSD)
+	group.MonthlyLimitUSD = normalizeLimit(input.MonthlyLimitUSD)
 	// 图片生成计费配置：负数表示清除（使用默认价格）
 	if input.ImagePrice1K != nil {
 		group.ImagePrice1K = normalizePrice(input.ImagePrice1K)
--- a/backend/internal/service/api_key.go
+++ b/backend/internal/service/api_key.go
@@ -22,8 +22,9 @@ const (
 )

 // IsWindowExpired returns true if the window starting at windowStart has exceeded the given duration.
+// A nil windowStart is treated as expired — no initialized window means any accumulated usage is stale.
 func IsWindowExpired(windowStart *time.Time, duration time.Duration) bool {
-	return windowStart != nil && time.Since(*windowStart) >= duration
+	return windowStart == nil || time.Since(*windowStart) >= duration
 }

 type APIKey struct {
--- a/backend/internal/service/api_key_rate_limit_test.go
+++ b/backend/internal/service/api_key_rate_limit_test.go
@@ -15,10 +15,10 @@ func TestIsWindowExpired(t *testing.T) {
 		want     bool
 	}{
 		{
-			name:     "nil window start",
+			name:     "nil window start (treated as expired)",
 			start:    nil,
 			duration: RateLimitWindow5h,
-			want:     false,
+			want:     true,
 		},
 		{
 			name:     "active window (started 1h ago, 5h window)",
@@ -113,7 +113,7 @@ func TestAPIKey_EffectiveUsage(t *testing.T) {
 			want7d: 0,
 		},
 		{
-			name: "nil window starts return raw usage",
+			name: "nil window starts return 0 (stale usage reset)",
 			key: APIKey{
 				Usage5h:       5.0,
 				Usage1d:       10.0,
@@ -122,9 +122,9 @@ func TestAPIKey_EffectiveUsage(t *testing.T) {
 				Window1dStart: nil,
 				Window7dStart: nil,
 			},
-			want5h: 5.0,
-			want1d: 10.0,
-			want7d: 50.0,
+			want5h: 0,
+			want1d: 0,
+			want7d: 0,
 		},
 		{
 			name: "mixed: 5h expired, 1d active, 7d nil",
@@ -138,7 +138,7 @@ func TestAPIKey_EffectiveUsage(t *testing.T) {
 			},
 			want5h: 0,
 			want1d: 10.0,
-			want7d: 50.0,
+			want7d: 0,
 		},
 		{
 			name: "zero usage with active windows",
@@ -210,7 +210,7 @@ func TestAPIKeyRateLimitData_EffectiveUsage(t *testing.T) {
 			want7d: 0,
 		},
 		{
-			name: "nil window starts return raw usage",
+			name: "nil window starts return 0 (stale usage reset)",
 			data: APIKeyRateLimitData{
 				Usage5h:       3.0,
 				Usage1d:       8.0,
@@ -219,9 +219,9 @@ func TestAPIKeyRateLimitData_EffectiveUsage(t *testing.T) {
 				Window1dStart: nil,
 				Window7dStart: nil,
 			},
-			want5h: 3.0,
-			want1d: 8.0,
-			want7d: 40.0,
+			want5h: 0,
+			want1d: 0,
+			want7d: 0,
 		},
 	}

--- a/backend/internal/service/gateway_record_usage_test.go
+++ b/backend/internal/service/gateway_record_usage_test.go
@@ -369,3 +369,54 @@ func TestGatewayServiceRecordUsage_BillingErrorSkipsUsageLogWrite(t *testing.T)
 	require.Equal(t, 1, billingRepo.calls)
 	require.Equal(t, 0, usageRepo.calls)
 }
+
+func TestGatewayServiceRecordUsage_ReasoningEffortPersisted(t *testing.T) {
+	usageRepo := &openAIRecordUsageBestEffortLogRepoStub{}
+	svc := newGatewayRecordUsageServiceForTest(usageRepo, &openAIRecordUsageUserRepoStub{}, &openAIRecordUsageSubRepoStub{})
+
+	effort := "max"
+	err := svc.RecordUsage(context.Background(), &RecordUsageInput{
+		Result: &ForwardResult{
+			RequestID: "effort_test",
+			Usage: ClaudeUsage{
+				InputTokens:  10,
+				OutputTokens: 5,
+			},
+			Model:           "claude-opus-4-6",
+			Duration:        time.Second,
+			ReasoningEffort: &effort,
+		},
+		APIKey:  &APIKey{ID: 1},
+		User:    &User{ID: 1},
+		Account: &Account{ID: 1},
+	})
+
+	require.NoError(t, err)
+	require.NotNil(t, usageRepo.lastLog)
+	require.NotNil(t, usageRepo.lastLog.ReasoningEffort)
+	require.Equal(t, "max", *usageRepo.lastLog.ReasoningEffort)
+}
+
+func TestGatewayServiceRecordUsage_ReasoningEffortNil(t *testing.T) {
+	usageRepo := &openAIRecordUsageBestEffortLogRepoStub{}
+	svc := newGatewayRecordUsageServiceForTest(usageRepo, &openAIRecordUsageUserRepoStub{}, &openAIRecordUsageSubRepoStub{})
+
+	err := svc.RecordUsage(context.Background(), &RecordUsageInput{
+		Result: &ForwardResult{
+			RequestID: "no_effort_test",
+			Usage: ClaudeUsage{
+				InputTokens:  10,
+				OutputTokens: 5,
+			},
+			Model:    "claude-sonnet-4",
+			Duration: time.Second,
+		},
+		APIKey:  &APIKey{ID: 1},
+		User:    &User{ID: 1},
+		Account: &Account{ID: 1},
+	})
+
+	require.NoError(t, err)
+	require.NotNil(t, usageRepo.lastLog)
+	require.Nil(t, usageRepo.lastLog.ReasoningEffort)
+}
--- a/backend/internal/service/gateway_request.go
+++ b/backend/internal/service/gateway_request.go
@@ -60,6 +60,7 @@ type ParsedRequest struct {
 	Messages        []any           // messages 数组
 	HasSystem       bool            // 是否包含 system 字段（包含 null 也视为显式传入）
 	ThinkingEnabled bool            // 是否开启 thinking（部分平台会影响最终模型名）
+	OutputEffort    string          // output_config.effort（Claude API 的推理强度控制）
 	MaxTokens       int             // max_tokens 值（用于探测请求拦截）
 	SessionContext  *SessionContext // 可选：请求上下文区分因子（nil 时行为不变）

@@ -116,6 +117,9 @@ func ParseGatewayRequest(body []byte, protocol string) (*ParsedRequest, error) {
 		parsed.ThinkingEnabled = true
 	}

+	// output_config.effort: Claude API 的推理强度控制参数
+	parsed.OutputEffort = strings.TrimSpace(gjson.Get(jsonStr, "output_config.effort").String())
+
 	// max_tokens: 仅接受整数值
 	maxTokensResult := gjson.Get(jsonStr, "max_tokens")
 	if maxTokensResult.Exists() && maxTokensResult.Type == gjson.Number {
@@ -747,6 +751,21 @@ func filterThinkingBlocksInternal(body []byte, _ bool) []byte {
 	return newBody
 }

+// NormalizeClaudeOutputEffort normalizes Claude's output_config.effort value.
+// Returns nil for empty or unrecognized values.
+func NormalizeClaudeOutputEffort(raw string) *string {
+	value := strings.ToLower(strings.TrimSpace(raw))
+	if value == "" {
+		return nil
+	}
+	switch value {
+	case "low", "medium", "high", "max":
+		return &value
+	default:
+		return nil
+	}
+}
+
 // =========================
 // Thinking Budget Rectifier
 // =========================
--- a/backend/internal/service/gateway_request_test.go
+++ b/backend/internal/service/gateway_request_test.go
@@ -972,6 +972,76 @@ func BenchmarkParseGatewayRequest_Old_Large(b *testing.B) {
 	}
 }

+func TestParseGatewayRequest_OutputEffort(t *testing.T) {
+	tests := []struct {
+		name       string
+		body       string
+		wantEffort string
+	}{
+		{
+			name:       "output_config.effort present",
+			body:       `{"model":"claude-opus-4-6","output_config":{"effort":"medium"},"messages":[]}`,
+			wantEffort: "medium",
+		},
+		{
+			name:       "output_config.effort max",
+			body:       `{"model":"claude-opus-4-6","output_config":{"effort":"max"},"messages":[]}`,
+			wantEffort: "max",
+		},
+		{
+			name:       "output_config without effort",
+			body:       `{"model":"claude-opus-4-6","output_config":{},"messages":[]}`,
+			wantEffort: "",
+		},
+		{
+			name:       "no output_config",
+			body:       `{"model":"claude-opus-4-6","messages":[]}`,
+			wantEffort: "",
+		},
+		{
+			name:       "effort with whitespace trimmed",
+			body:       `{"model":"claude-opus-4-6","output_config":{"effort":" high "},"messages":[]}`,
+			wantEffort: "high",
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			parsed, err := ParseGatewayRequest([]byte(tt.body), "")
+			require.NoError(t, err)
+			require.Equal(t, tt.wantEffort, parsed.OutputEffort)
+		})
+	}
+}
+
+func TestNormalizeClaudeOutputEffort(t *testing.T) {
+	tests := []struct {
+		input string
+		want  *string
+	}{
+		{"low", strPtr("low")},
+		{"medium", strPtr("medium")},
+		{"high", strPtr("high")},
+		{"max", strPtr("max")},
+		{"LOW", strPtr("low")},
+		{"Max", strPtr("max")},
+		{" medium ", strPtr("medium")},
+		{"", nil},
+		{"unknown", nil},
+		{"xhigh", nil},
+	}
+	for _, tt := range tests {
+		t.Run(tt.input, func(t *testing.T) {
+			got := NormalizeClaudeOutputEffort(tt.input)
+			if tt.want == nil {
+				require.Nil(t, got)
+			} else {
+				require.NotNil(t, got)
+				require.Equal(t, *tt.want, *got)
+			}
+		})
+	}
+}
+
 func BenchmarkParseGatewayRequest_New_Large(b *testing.B) {
 	data := buildLargeJSON()
 	b.SetBytes(int64(len(data)))
--- a/backend/internal/service/gateway_service.go
+++ b/backend/internal/service/gateway_service.go
@@ -492,6 +492,7 @@ type ForwardResult struct {
 	Duration         time.Duration
 	FirstTokenMs     *int // 首字时间（流式请求）
 	ClientDisconnect bool // 客户端是否在流式传输过程中断开
+	ReasoningEffort  *string

 	// 图片生成计费字段（图片生成模型使用）
 	ImageCount int    // 生成的图片数量
@@ -7523,6 +7524,7 @@ func (s *GatewayService) RecordUsage(ctx context.Context, input *RecordUsageInpu
 		AccountID:             account.ID,
 		RequestID:             requestID,
 		Model:                 result.Model,
+		ReasoningEffort:       result.ReasoningEffort,
 		InputTokens:           result.Usage.InputTokens,
 		OutputTokens:          result.Usage.OutputTokens,
 		CacheCreationTokens:   result.Usage.CacheCreationInputTokens,
@@ -7699,6 +7701,7 @@ func (s *GatewayService) RecordUsageWithLongContext(ctx context.Context, input *
 		AccountID:             account.ID,
 		RequestID:             requestID,
 		Model:                 result.Model,
+		ReasoningEffort:       result.ReasoningEffort,
 		InputTokens:           result.Usage.InputTokens,
 		OutputTokens:          result.Usage.OutputTokens,
 		CacheCreationTokens:   result.Usage.CacheCreationInputTokens,
--- a/backend/internal/service/openai_gateway_record_usage_test.go
+++ b/backend/internal/service/openai_gateway_record_usage_test.go
@@ -226,6 +226,41 @@ func TestOpenAIGatewayServiceRecordUsage_UsesUserSpecificGroupRate(t *testing.T)
 	require.Equal(t, 1, userRepo.deductCalls)
 }

+func TestOpenAIGatewayServiceRecordUsage_IncludesEndpointMetadata(t *testing.T) {
+	usageRepo := &openAIRecordUsageLogRepoStub{inserted: true}
+	userRepo := &openAIRecordUsageUserRepoStub{}
+	subRepo := &openAIRecordUsageSubRepoStub{}
+	rateRepo := &openAIUserGroupRateRepoStub{}
+	svc := newOpenAIRecordUsageServiceForTest(usageRepo, userRepo, subRepo, rateRepo)
+
+	err := svc.RecordUsage(context.Background(), &OpenAIRecordUsageInput{
+		Result: &OpenAIForwardResult{
+			RequestID: "resp_endpoint_metadata",
+			Usage: OpenAIUsage{
+				InputTokens:  8,
+				OutputTokens: 2,
+			},
+			Model:    "gpt-5.1",
+			Duration: time.Second,
+		},
+		APIKey: &APIKey{
+			ID:    1002,
+			Group: &Group{RateMultiplier: 1},
+		},
+		User:             &User{ID: 2002},
+		Account:          &Account{ID: 3002},
+		InboundEndpoint:  " /v1/chat/completions ",
+		UpstreamEndpoint: " /v1/responses ",
+	})
+
+	require.NoError(t, err)
+	require.NotNil(t, usageRepo.lastLog)
+	require.NotNil(t, usageRepo.lastLog.InboundEndpoint)
+	require.Equal(t, "/v1/chat/completions", *usageRepo.lastLog.InboundEndpoint)
+	require.NotNil(t, usageRepo.lastLog.UpstreamEndpoint)
+	require.Equal(t, "/v1/responses", *usageRepo.lastLog.UpstreamEndpoint)
+}
+
 func TestOpenAIGatewayServiceRecordUsage_FallsBackToGroupDefaultRateOnResolverError(t *testing.T) {
 	groupID := int64(12)
 	groupRate := 1.6
--- a/backend/internal/service/openai_gateway_service.go
+++ b/backend/internal/service/openai_gateway_service.go
@@ -4028,6 +4028,8 @@ type OpenAIRecordUsageInput struct {
 	User               *User
 	Account            *Account
 	Subscription       *UserSubscription
+	InboundEndpoint    string
+	UpstreamEndpoint   string
 	UserAgent          string // 请求的 User-Agent
 	IPAddress          string // 请求的客户端 IP 地址
 	RequestPayloadHash string
@@ -4106,6 +4108,8 @@ func (s *OpenAIGatewayService) RecordUsage(ctx context.Context, input *OpenAIRec
 		Model:                 billingModel,
 		ServiceTier:           result.ServiceTier,
 		ReasoningEffort:       result.ReasoningEffort,
+		InboundEndpoint:       optionalTrimmedStringPtr(input.InboundEndpoint),
+		UpstreamEndpoint:      optionalTrimmedStringPtr(input.UpstreamEndpoint),
 		InputTokens:           actualInputTokens,
 		OutputTokens:          result.Usage.OutputTokens,
 		CacheCreationTokens:   result.Usage.CacheCreationInputTokens,
@@ -4125,7 +4129,6 @@ func (s *OpenAIGatewayService) RecordUsage(ctx context.Context, input *OpenAIRec
 		FirstTokenMs:          result.FirstTokenMs,
 		CreatedAt:             time.Now(),
 	}
-
 	// 添加 UserAgent
 	if input.UserAgent != "" {
 		usageLog.UserAgent = &input.UserAgent
@@ -4668,3 +4671,11 @@ func normalizeOpenAIReasoningEffort(raw string) string {
 		return ""
 	}
 }
+
+func optionalTrimmedStringPtr(raw string) *string {
+	trimmed := strings.TrimSpace(raw)
+	if trimmed == "" {
+		return nil
+	}
+	return &trimmed
+}
--- a/backend/internal/service/usage_log.go
+++ b/backend/internal/service/usage_log.go
@@ -100,9 +100,14 @@ type UsageLog struct {
 	Model     string
 	// ServiceTier records the OpenAI service tier used for billing, e.g. "priority" / "flex".
 	ServiceTier *string
-	// ReasoningEffort is the request's reasoning effort level (OpenAI Responses API),
-	// e.g. "low" / "medium" / "high" / "xhigh". Nil means not provided / not applicable.
+	// ReasoningEffort is the request's reasoning effort level.
+	// OpenAI: "low" / "medium" / "high" / "xhigh"; Claude: "low" / "medium" / "high" / "max".
+	// Nil means not provided / not applicable.
 	ReasoningEffort *string
+	// InboundEndpoint is the client-facing API endpoint path, e.g. /v1/chat/completions.
+	InboundEndpoint *string
+	// UpstreamEndpoint is the normalized upstream endpoint path, e.g. /v1/responses.
+	UpstreamEndpoint *string

 	GroupID        *int64
 	SubscriptionID *int64
--- a/backend/migrations/074_add_usage_log_endpoints.sql
+++ b/backend/migrations/074_add_usage_log_endpoints.sql
@@ -0,0 +1,5 @@
+-- Add endpoint tracking fields to usage_logs.
+-- inbound_endpoint: client-facing API route (e.g. /v1/chat/completions, /v1/messages, /v1/responses)
+-- upstream_endpoint: normalized upstream route (e.g. /v1/responses)
+ALTER TABLE usage_logs ADD COLUMN IF NOT EXISTS inbound_endpoint VARCHAR(128);
+ALTER TABLE usage_logs ADD COLUMN IF NOT EXISTS upstream_endpoint VARCHAR(128);
--- a/frontend/src/api/admin/usage.ts
+++ b/frontend/src/api/admin/usage.ts
@@ -5,6 +5,7 @@

 import { apiClient } from '../client'
 import type { AdminUsageLog, UsageQueryParams, PaginatedResponse, UsageRequestType } from '@/types'
+import type { EndpointStat } from '@/types'

 // ==================== Types ====================

@@ -18,6 +19,9 @@ export interface AdminUsageStatsResponse {
  total_actual_cost: number
  total_account_cost?: number
  average_duration_ms: number
+  endpoints?: EndpointStat[]
+  upstream_endpoints?: EndpointStat[]
+  endpoint_paths?: EndpointStat[]
 }

 export interface SimpleUser {
--- a/frontend/src/components/account/AccountStatsModal.vue
+++ b/frontend/src/components/account/AccountStatsModal.vue
@@ -446,6 +446,18 @@

        <!-- Model Distribution -->
        <ModelDistributionChart :model-stats="stats.models" :loading="false" />
+
+        <EndpointDistributionChart
+          :endpoint-stats="stats.endpoints || []"
+          :loading="false"
+          :title="t('usage.inboundEndpoint')"
+        />
+
+        <EndpointDistributionChart
+          :endpoint-stats="stats.upstream_endpoints || []"
+          :loading="false"
+          :title="t('usage.upstreamEndpoint')"
+        />
      </template>

      <!-- No Data State -->
@@ -489,6 +501,7 @@ import { Line } from 'vue-chartjs'
 import BaseDialog from '@/components/common/BaseDialog.vue'
 import LoadingSpinner from '@/components/common/LoadingSpinner.vue'
 import ModelDistributionChart from '@/components/charts/ModelDistributionChart.vue'
+import EndpointDistributionChart from '@/components/charts/EndpointDistributionChart.vue'
 import Icon from '@/components/icons/Icon.vue'
 import { adminAPI } from '@/api/admin'
 import type { Account, AccountUsageStatsResponse } from '@/types'
--- a/frontend/src/components/admin/account/AccountStatsModal.vue
+++ b/frontend/src/components/admin/account/AccountStatsModal.vue
@@ -410,6 +410,18 @@

        <!-- Model Distribution -->
        <ModelDistributionChart :model-stats="stats.models" :loading="false" />
+
+        <EndpointDistributionChart
+          :endpoint-stats="stats.endpoints || []"
+          :loading="false"
+          :title="t('usage.inboundEndpoint')"
+        />
+
+        <EndpointDistributionChart
+          :endpoint-stats="stats.upstream_endpoints || []"
+          :loading="false"
+          :title="t('usage.upstreamEndpoint')"
+        />
      </template>

      <!-- No Data State -->
@@ -453,6 +465,7 @@ import { Line } from 'vue-chartjs'
 import BaseDialog from '@/components/common/BaseDialog.vue'
 import LoadingSpinner from '@/components/common/LoadingSpinner.vue'
 import ModelDistributionChart from '@/components/charts/ModelDistributionChart.vue'
+import EndpointDistributionChart from '@/components/charts/EndpointDistributionChart.vue'
 import Icon from '@/components/icons/Icon.vue'
 import { adminAPI } from '@/api/admin'
 import type { Account, AccountUsageStatsResponse } from '@/types'
--- a/frontend/src/components/admin/account/AccountTableFilters.vue
+++ b/frontend/src/components/admin/account/AccountTableFilters.vue
@@ -24,7 +24,7 @@ const updateType = (value: string | number | boolean | null) => { emit('update:f
 const updateStatus = (value: string | number | boolean | null) => { emit('update:filters', { ...props.filters, status: value }) }
 const updateGroup = (value: string | number | boolean | null) => { emit('update:filters', { ...props.filters, group: value }) }
 const pOpts = computed(() => [{ value: '', label: t('admin.accounts.allPlatforms') }, { value: 'anthropic', label: 'Anthropic' }, { value: 'openai', label: 'OpenAI' }, { value: 'gemini', label: 'Gemini' }, { value: 'antigravity', label: 'Antigravity' }, { value: 'sora', label: 'Sora' }])
-const tOpts = computed(() => [{ value: '', label: t('admin.accounts.allTypes') }, { value: 'oauth', label: t('admin.accounts.oauthType') }, { value: 'setup-token', label: t('admin.accounts.setupToken') }, { value: 'apikey', label: t('admin.accounts.apiKey') }])
+const tOpts = computed(() => [{ value: '', label: t('admin.accounts.allTypes') }, { value: 'oauth', label: t('admin.accounts.oauthType') }, { value: 'setup-token', label: t('admin.accounts.setupToken') }, { value: 'apikey', label: t('admin.accounts.apiKey') }, { value: 'bedrock', label: 'AWS Bedrock' }])
 const sOpts = computed(() => [{ value: '', label: t('admin.accounts.allStatus') }, { value: 'active', label: t('admin.accounts.status.active') }, { value: 'inactive', label: t('admin.accounts.status.inactive') }, { value: 'error', label: t('admin.accounts.status.error') }, { value: 'rate_limited', label: t('admin.accounts.status.rateLimited') }, { value: 'temp_unschedulable', label: t('admin.accounts.status.tempUnschedulable') }])
 const gOpts = computed(() => [{ value: '', label: t('admin.accounts.allGroups') }, ...(props.groups || []).map(g => ({ value: String(g.id), label: g.name }))])
 </script>
--- a/frontend/src/components/admin/usage/UsageTable.vue
+++ b/frontend/src/components/admin/usage/UsageTable.vue
@@ -35,6 +35,19 @@
          </span>
        </template>

+        <template #cell-endpoint="{ row }">
+          <div class="max-w-[320px] space-y-1 text-xs">
+            <div class="break-all text-gray-700 dark:text-gray-300">
+              <span class="font-medium text-gray-500 dark:text-gray-400">{{ t('usage.inbound') }}:</span>
+              <span class="ml-1">{{ row.inbound_endpoint?.trim() || '-' }}</span>
+            </div>
+            <div class="break-all text-gray-700 dark:text-gray-300">
+              <span class="font-medium text-gray-500 dark:text-gray-400">{{ t('usage.upstream') }}:</span>
+              <span class="ml-1">{{ row.upstream_endpoint?.trim() || '-' }}</span>
+            </div>
+          </div>
+        </template>
+
        <template #cell-group="{ row }">
          <span v-if="row.group" class="inline-flex items-center rounded px-2 py-0.5 text-xs font-medium bg-indigo-100 text-indigo-800 dark:bg-indigo-900 dark:text-indigo-200">
            {{ row.group.name }}
@@ -328,6 +341,7 @@ const getRequestTypeBadgeClass = (row: AdminUsageLog): string => {
  if (requestType === 'sync') return 'bg-gray-100 text-gray-800 dark:bg-gray-700 dark:text-gray-200'
  return 'bg-amber-100 text-amber-800 dark:bg-amber-900 dark:text-amber-200'
 }
+
 const formatCacheTokens = (tokens: number): string => {
  if (tokens >= 1000000) return `${(tokens / 1000000).toFixed(1)}M`
  if (tokens >= 1000) return `${(tokens / 1000).toFixed(1)}K`
--- a/frontend/src/components/charts/EndpointDistributionChart.vue
+++ b/frontend/src/components/charts/EndpointDistributionChart.vue
@@ -0,0 +1,257 @@
+<template>
+  <div class="card p-4">
+    <div class="mb-4 flex items-start justify-between gap-3">
+      <h3 class="text-sm font-semibold text-gray-900 dark:text-white">
+        {{ title || t('usage.endpointDistribution') }}
+      </h3>
+      <div class="flex flex-col items-end gap-2">
+        <div
+          v-if="showSourceToggle"
+          class="inline-flex rounded-lg border border-gray-200 bg-gray-50 p-0.5 dark:border-gray-700 dark:bg-dark-800"
+        >
+          <button
+            type="button"
+            class="rounded-md px-2.5 py-1 text-xs font-medium transition-colors"
+            :class="source === 'inbound'
+              ? 'bg-white text-gray-900 shadow-sm dark:bg-dark-700 dark:text-white'
+              : 'text-gray-500 hover:text-gray-700 dark:text-gray-400 dark:hover:text-gray-200'"
+            @click="emit('update:source', 'inbound')"
+          >
+            {{ t('usage.inbound') }}
+          </button>
+          <button
+            type="button"
+            class="rounded-md px-2.5 py-1 text-xs font-medium transition-colors"
+            :class="source === 'upstream'
+              ? 'bg-white text-gray-900 shadow-sm dark:bg-dark-700 dark:text-white'
+              : 'text-gray-500 hover:text-gray-700 dark:text-gray-400 dark:hover:text-gray-200'"
+            @click="emit('update:source', 'upstream')"
+          >
+            {{ t('usage.upstream') }}
+          </button>
+          <button
+            type="button"
+            class="rounded-md px-2.5 py-1 text-xs font-medium transition-colors"
+            :class="source === 'path'
+              ? 'bg-white text-gray-900 shadow-sm dark:bg-dark-700 dark:text-white'
+              : 'text-gray-500 hover:text-gray-700 dark:text-gray-400 dark:hover:text-gray-200'"
+            @click="emit('update:source', 'path')"
+          >
+            {{ t('usage.path') }}
+          </button>
+        </div>
+
+        <div
+          v-if="showMetricToggle"
+          class="inline-flex rounded-lg border border-gray-200 bg-gray-50 p-0.5 dark:border-gray-700 dark:bg-dark-800"
+        >
+          <button
+            type="button"
+            class="rounded-md px-2.5 py-1 text-xs font-medium transition-colors"
+            :class="metric === 'tokens'
+              ? 'bg-white text-gray-900 shadow-sm dark:bg-dark-700 dark:text-white'
+              : 'text-gray-500 hover:text-gray-700 dark:text-gray-400 dark:hover:text-gray-200'"
+            @click="emit('update:metric', 'tokens')"
+          >
+            {{ t('admin.dashboard.metricTokens') }}
+          </button>
+          <button
+            type="button"
+            class="rounded-md px-2.5 py-1 text-xs font-medium transition-colors"
+            :class="metric === 'actual_cost'
+              ? 'bg-white text-gray-900 shadow-sm dark:bg-dark-700 dark:text-white'
+              : 'text-gray-500 hover:text-gray-700 dark:text-gray-400 dark:hover:text-gray-200'"
+            @click="emit('update:metric', 'actual_cost')"
+          >
+            {{ t('admin.dashboard.metricActualCost') }}
+          </button>
+        </div>
+      </div>
+    </div>
+    <div v-if="loading" class="flex h-48 items-center justify-center">
+      <LoadingSpinner />
+    </div>
+    <div v-else-if="displayEndpointStats.length > 0 && chartData" class="flex items-center gap-6">
+      <div class="h-48 w-48">
+        <Doughnut :data="chartData" :options="doughnutOptions" />
+      </div>
+      <div class="max-h-48 flex-1 overflow-y-auto">
+        <table class="w-full text-xs">
+          <thead>
+            <tr class="text-gray-500 dark:text-gray-400">
+              <th class="pb-2 text-left">{{ t('usage.endpoint') }}</th>
+              <th class="pb-2 text-right">{{ t('admin.dashboard.requests') }}</th>
+              <th class="pb-2 text-right">{{ t('admin.dashboard.tokens') }}</th>
+              <th class="pb-2 text-right">{{ t('admin.dashboard.actual') }}</th>
+              <th class="pb-2 text-right">{{ t('admin.dashboard.standard') }}</th>
+            </tr>
+          </thead>
+          <tbody>
+            <tr
+              v-for="item in displayEndpointStats"
+              :key="item.endpoint"
+              class="border-t border-gray-100 dark:border-gray-700"
+            >
+              <td class="max-w-[180px] truncate py-1.5 font-medium text-gray-900 dark:text-white" :title="item.endpoint">
+                {{ item.endpoint }}
+              </td>
+              <td class="py-1.5 text-right text-gray-600 dark:text-gray-400">
+                {{ formatNumber(item.requests) }}
+              </td>
+              <td class="py-1.5 text-right text-gray-600 dark:text-gray-400">
+                {{ formatTokens(item.total_tokens) }}
+              </td>
+              <td class="py-1.5 text-right text-green-600 dark:text-green-400">
+                ${{ formatCost(item.actual_cost) }}
+              </td>
+              <td class="py-1.5 text-right text-gray-400 dark:text-gray-500">
+                ${{ formatCost(item.cost) }}
+              </td>
+            </tr>
+          </tbody>
+        </table>
+      </div>
+    </div>
+    <div v-else class="flex h-48 items-center justify-center text-sm text-gray-500 dark:text-gray-400">
+      {{ t('admin.dashboard.noDataAvailable') }}
+    </div>
+  </div>
+</template>
+
+<script setup lang="ts">
+import { computed } from 'vue'
+import { useI18n } from 'vue-i18n'
+import { Chart as ChartJS, ArcElement, Tooltip, Legend } from 'chart.js'
+import { Doughnut } from 'vue-chartjs'
+import LoadingSpinner from '@/components/common/LoadingSpinner.vue'
+import type { EndpointStat } from '@/types'
+
+ChartJS.register(ArcElement, Tooltip, Legend)
+
+const { t } = useI18n()
+
+type DistributionMetric = 'tokens' | 'actual_cost'
+type EndpointSource = 'inbound' | 'upstream' | 'path'
+
+const props = withDefaults(
+  defineProps<{
+    endpointStats: EndpointStat[]
+    upstreamEndpointStats?: EndpointStat[]
+    endpointPathStats?: EndpointStat[]
+    loading?: boolean
+    title?: string
+    metric?: DistributionMetric
+    source?: EndpointSource
+    showMetricToggle?: boolean
+    showSourceToggle?: boolean
+  }>(),
+  {
+    upstreamEndpointStats: () => [],
+    endpointPathStats: () => [],
+    loading: false,
+    title: '',
+    metric: 'tokens',
+    source: 'inbound',
+    showMetricToggle: false,
+    showSourceToggle: false
+  }
+)
+
+const emit = defineEmits<{
+  'update:metric': [value: DistributionMetric]
+  'update:source': [value: EndpointSource]
+}>()
+
+const chartColors = [
+  '#3b82f6',
+  '#10b981',
+  '#f59e0b',
+  '#ef4444',
+  '#8b5cf6',
+  '#ec4899',
+  '#14b8a6',
+  '#f97316',
+  '#6366f1',
+  '#84cc16',
+  '#06b6d4',
+  '#a855f7'
+]
+
+const displayEndpointStats = computed(() => {
+  const sourceStats = props.source === 'upstream'
+    ? props.upstreamEndpointStats
+    : props.source === 'path'
+      ? props.endpointPathStats
+      : props.endpointStats
+  if (!sourceStats?.length) return []
+
+  const metricKey = props.metric === 'actual_cost' ? 'actual_cost' : 'total_tokens'
+  return [...sourceStats].sort((a, b) => b[metricKey] - a[metricKey])
+})
+
+const chartData = computed(() => {
+  if (!displayEndpointStats.value?.length) return null
+
+  return {
+    labels: displayEndpointStats.value.map((item) => item.endpoint),
+    datasets: [
+      {
+        data: displayEndpointStats.value.map((item) =>
+          props.metric === 'actual_cost' ? item.actual_cost : item.total_tokens
+        ),
+        backgroundColor: chartColors.slice(0, displayEndpointStats.value.length),
+        borderWidth: 0
+      }
+    ]
+  }
+})
+
+const doughnutOptions = computed(() => ({
+  responsive: true,
+  maintainAspectRatio: false,
+  plugins: {
+    legend: {
+      display: false
+    },
+    tooltip: {
+      callbacks: {
+        label: (context: any) => {
+          const value = context.raw as number
+          const total = context.dataset.data.reduce((a: number, b: number) => a + b, 0)
+          const percentage = total > 0 ? ((value / total) * 100).toFixed(1) : '0.0'
+          const formattedValue = props.metric === 'actual_cost'
+            ? `$${formatCost(value)}`
+            : formatTokens(value)
+          return `${context.label}: ${formattedValue} (${percentage}%)`
+        }
+      }
+    }
+  }
+}))
+
+const formatTokens = (value: number): string => {
+  if (value >= 1_000_000_000) {
+    return `${(value / 1_000_000_000).toFixed(2)}B`
+  } else if (value >= 1_000_000) {
+    return `${(value / 1_000_000).toFixed(2)}M`
+  } else if (value >= 1_000) {
+    return `${(value / 1_000).toFixed(2)}K`
+  }
+  return value.toLocaleString()
+}
+
+const formatNumber = (value: number): string => {
+  return value.toLocaleString()
+}
+
+const formatCost = (value: number): string => {
+  if (value >= 1000) {
+    return (value / 1000).toFixed(2) + 'K'
+  } else if (value >= 1) {
+    return value.toFixed(2)
+  } else if (value >= 0.01) {
+    return value.toFixed(3)
+  }
+  return value.toFixed(4)
+}
+</script>
--- a/frontend/src/i18n/locales/en.ts
+++ b/frontend/src/i18n/locales/en.ts
@@ -718,6 +718,13 @@ export default {
    preparingExport: 'Preparing export...',
    model: 'Model',
    reasoningEffort: 'Reasoning Effort',
+    endpoint: 'Endpoint',
+    endpointDistribution: 'Endpoint Distribution',
+    inbound: 'Inbound',
+    upstream: 'Upstream',
+    path: 'Path',
+    inboundEndpoint: 'Inbound Endpoint',
+    upstreamEndpoint: 'Upstream Endpoint',
    type: 'Type',
    tokens: 'Tokens',
    cost: 'Cost',
--- a/frontend/src/i18n/locales/zh.ts
+++ b/frontend/src/i18n/locales/zh.ts
@@ -723,6 +723,13 @@ export default {
    preparingExport: '正在准备导出...',
    model: '模型',
    reasoningEffort: '推理强度',
+    endpoint: '端点',
+    endpointDistribution: '端点分布',
+    inbound: '入站',
+    upstream: '上游',
+    path: '路径',
+    inboundEndpoint: '入站端点',
+    upstreamEndpoint: '上游端点',
    type: '类型',
    tokens: 'Token',
    cost: '费用',
--- a/frontend/src/types/index.ts
+++ b/frontend/src/types/index.ts
@@ -962,6 +962,8 @@ export interface UsageLog {
  model: string
  service_tier?: string | null
  reasoning_effort?: string | null
+  inbound_endpoint?: string | null
+  upstream_endpoint?: string | null

  group_id: number | null
  subscription_id: number | null
@@ -1168,6 +1170,14 @@ export interface ModelStat {
  actual_cost: number // 实际扣除
 }

+export interface EndpointStat {
+  endpoint: string
+  requests: number
+  total_tokens: number
+  cost: number
+  actual_cost: number
+}
+
 export interface GroupStat {
  group_id: number
  group_name: string
@@ -1362,6 +1372,8 @@ export interface AccountUsageStatsResponse {
  history: AccountUsageHistory[]
  summary: AccountUsageSummary
  models: ModelStat[]
+  endpoints: EndpointStat[]
+  upstream_endpoints: EndpointStat[]
 }

 // ==================== User Attribute Types ====================
--- a/frontend/src/views/admin/GroupsView.vue
+++ b/frontend/src/views/admin/GroupsView.vue
@@ -2402,6 +2402,11 @@ const handleCreateGroup = async () => {
      sora_storage_quota_bytes: createQuotaGb ? Math.round(createQuotaGb * 1024 * 1024 * 1024) : 0,
      model_routing: convertRoutingRulesToApiFormat(createModelRoutingRules.value)
    }
+    // v-model.number 清空输入框时产生 ""，转为 null 让后端设为无限制
+    const emptyToNull = (v: any) => v === '' ? null : v
+    requestData.daily_limit_usd = emptyToNull(requestData.daily_limit_usd)
+    requestData.weekly_limit_usd = emptyToNull(requestData.weekly_limit_usd)
+    requestData.monthly_limit_usd = emptyToNull(requestData.monthly_limit_usd)
    await adminAPI.groups.create(requestData)
    appStore.showSuccess(t('admin.groups.groupCreated'))
    closeCreateModal()
@@ -2488,6 +2493,11 @@ const handleUpdateGroup = async () => {
          : editForm.fallback_group_id_on_invalid_request,
      model_routing: convertRoutingRulesToApiFormat(editModelRoutingRules.value)
    }
+    // v-model.number 清空输入框时产生 ""，转为 null 让后端设为无限制
+    const emptyToNull = (v: any) => v === '' ? null : v
+    payload.daily_limit_usd = emptyToNull(payload.daily_limit_usd)
+    payload.weekly_limit_usd = emptyToNull(payload.weekly_limit_usd)
+    payload.monthly_limit_usd = emptyToNull(payload.monthly_limit_usd)
    await adminAPI.groups.update(editingGroup.value.id, payload)
    appStore.showSuccess(t('admin.groups.groupUpdated'))
    closeEditModal()
--- a/frontend/src/views/admin/UsageView.vue
+++ b/frontend/src/views/admin/UsageView.vue
@@ -26,7 +26,20 @@
            :show-metric-toggle="true"
          />
        </div>
-        <TokenUsageTrend :trend-data="trendData" :loading="chartsLoading" />
+        <div class="grid grid-cols-1 gap-6 lg:grid-cols-2">
+          <EndpointDistributionChart
+            v-model:source="endpointDistributionSource"
+            v-model:metric="endpointDistributionMetric"
+            :endpoint-stats="inboundEndpointStats"
+            :upstream-endpoint-stats="upstreamEndpointStats"
+            :endpoint-path-stats="endpointPathStats"
+            :loading="endpointStatsLoading"
+            :show-source-toggle="true"
+            :show-metric-toggle="true"
+            :title="t('usage.endpointDistribution')"
+          />
+          <TokenUsageTrend :trend-data="trendData" :loading="chartsLoading" />
+        </div>
      </div>
      <UsageFilters v-model="filters" v-model:startDate="startDate" v-model:endDate="endDate" :exporting="exporting" @change="applyFilters" @refresh="refreshData" @reset="resetFilters" @cleanup="openCleanupDialog" @export="exportToExcel">
        <template #after-reset>
@@ -99,19 +112,28 @@ import UsageTable from '@/components/admin/usage/UsageTable.vue'; import UsageEx
 import UsageCleanupDialog from '@/components/admin/usage/UsageCleanupDialog.vue'
 import UserBalanceHistoryModal from '@/components/admin/user/UserBalanceHistoryModal.vue'
 import ModelDistributionChart from '@/components/charts/ModelDistributionChart.vue'; import GroupDistributionChart from '@/components/charts/GroupDistributionChart.vue'; import TokenUsageTrend from '@/components/charts/TokenUsageTrend.vue'
+import EndpointDistributionChart from '@/components/charts/EndpointDistributionChart.vue'
 import Icon from '@/components/icons/Icon.vue'
-import type { AdminUsageLog, TrendDataPoint, ModelStat, GroupStat, AdminUser } from '@/types'; import type { AdminUsageStatsResponse, AdminUsageQueryParams } from '@/api/admin/usage'
+import type { AdminUsageLog, TrendDataPoint, ModelStat, GroupStat, EndpointStat, AdminUser } from '@/types'; import type { AdminUsageStatsResponse, AdminUsageQueryParams } from '@/api/admin/usage'

 const { t } = useI18n()
 const appStore = useAppStore()
 type DistributionMetric = 'tokens' | 'actual_cost'
+type EndpointSource = 'inbound' | 'upstream' | 'path'
 const route = useRoute()
 const usageStats = ref<AdminUsageStatsResponse | null>(null); const usageLogs = ref<AdminUsageLog[]>([]); const loading = ref(false); const exporting = ref(false)
 const trendData = ref<TrendDataPoint[]>([]); const modelStats = ref<ModelStat[]>([]); const groupStats = ref<GroupStat[]>([]); const chartsLoading = ref(false); const granularity = ref<'day' | 'hour'>('day')
 const modelDistributionMetric = ref<DistributionMetric>('tokens')
 const groupDistributionMetric = ref<DistributionMetric>('tokens')
+const endpointDistributionMetric = ref<DistributionMetric>('tokens')
+const endpointDistributionSource = ref<EndpointSource>('inbound')
+const inboundEndpointStats = ref<EndpointStat[]>([])
+const upstreamEndpointStats = ref<EndpointStat[]>([])
+const endpointPathStats = ref<EndpointStat[]>([])
+const endpointStatsLoading = ref(false)
 let abortController: AbortController | null = null; let exportAbortController: AbortController | null = null
 let chartReqSeq = 0
+let statsReqSeq = 0
 const exportProgress = reactive({ show: false, progress: 0, current: 0, total: 0, estimatedTime: '' })
 const cleanupDialogVisible = ref(false)
 // Balance history modal state
@@ -183,13 +205,25 @@ const loadLogs = async () => {
  } catch (error: any) { if(error?.name !== 'AbortError') console.error('Failed to load usage logs:', error) } finally { if(abortController === c) loading.value = false }
 }
 const loadStats = async () => {
+  const seq = ++statsReqSeq
+  endpointStatsLoading.value = true
  try {
    const requestType = filters.value.request_type
    const legacyStream = requestType ? requestTypeToLegacyStream(requestType) : filters.value.stream
    const s = await adminAPI.usage.getStats({ ...filters.value, stream: legacyStream === null ? undefined : legacyStream })
+    if (seq !== statsReqSeq) return
    usageStats.value = s
+    inboundEndpointStats.value = s.endpoints || []
+    upstreamEndpointStats.value = s.upstream_endpoints || []
+    endpointPathStats.value = s.endpoint_paths || []
  } catch (error) {
+    if (seq !== statsReqSeq) return
    console.error('Failed to load usage stats:', error)
+    inboundEndpointStats.value = []
+    upstreamEndpointStats.value = []
+    endpointPathStats.value = []
+  } finally {
+    if (seq === statsReqSeq) endpointStatsLoading.value = false
  }
 }
 const loadChartData = async () => {
@@ -246,6 +280,7 @@ const exportToExcel = async () => {
    const headers = [
      t('usage.time'), t('admin.usage.user'), t('usage.apiKeyFilter'),
      t('admin.usage.account'), t('usage.model'), t('usage.reasoningEffort'), t('admin.usage.group'),
+      t('usage.inboundEndpoint'), t('usage.upstreamEndpoint'),
      t('usage.type'),
      t('admin.usage.inputTokens'), t('admin.usage.outputTokens'),
      t('admin.usage.cacheReadTokens'), t('admin.usage.cacheCreationTokens'),
@@ -263,7 +298,8 @@ const exportToExcel = async () => {
      if (c.signal.aborted) break; if (p === 1) { total = res.total; exportProgress.total = total }
      const rows = (res.items || []).map((log: AdminUsageLog) => [
        log.created_at, log.user?.email || '', log.api_key?.name || '', log.account?.name || '', log.model,
-        formatReasoningEffort(log.reasoning_effort), log.group?.name || '', getRequestTypeLabel(log),
+        formatReasoningEffort(log.reasoning_effort), log.group?.name || '',
+        log.inbound_endpoint || '', log.upstream_endpoint || '', getRequestTypeLabel(log),
        log.input_tokens, log.output_tokens, log.cache_read_tokens, log.cache_creation_tokens,
        log.input_cost?.toFixed(6) || '0.000000', log.output_cost?.toFixed(6) || '0.000000',
        log.cache_read_cost?.toFixed(6) || '0.000000', log.cache_creation_cost?.toFixed(6) || '0.000000',
@@ -301,6 +337,7 @@ const allColumns = computed(() => [
  { key: 'account', label: t('admin.usage.account'), sortable: false },
  { key: 'model', label: t('usage.model'), sortable: true },
  { key: 'reasoning_effort', label: t('usage.reasoningEffort'), sortable: false },
+  { key: 'endpoint', label: t('usage.endpoint'), sortable: false },
  { key: 'group', label: t('admin.usage.group'), sortable: false },
  { key: 'stream', label: t('usage.type'), sortable: false },
  { key: 'tokens', label: t('usage.tokens'), sortable: false },
@@ -343,12 +380,18 @@ const loadSavedColumns = () => {
  try {
    const saved = localStorage.getItem(HIDDEN_COLUMNS_KEY)
    if (saved) {
-      (JSON.parse(saved) as string[]).forEach(key => hiddenColumns.add(key))
+      (JSON.parse(saved) as string[]).forEach((key) => {
+        hiddenColumns.add(key)
+      })
    } else {
-      DEFAULT_HIDDEN_COLUMNS.forEach(key => hiddenColumns.add(key))
+      DEFAULT_HIDDEN_COLUMNS.forEach((key) => {
+        hiddenColumns.add(key)
+      })
    }
  } catch {
-    DEFAULT_HIDDEN_COLUMNS.forEach(key => hiddenColumns.add(key))
+    DEFAULT_HIDDEN_COLUMNS.forEach((key) => {
+      hiddenColumns.add(key)
+    })
  }
 }

--- a/frontend/src/views/user/UsageView.vue
+++ b/frontend/src/views/user/UsageView.vue
@@ -166,6 +166,12 @@
            </span>
          </template>

+          <template #cell-endpoint="{ row }">
+            <span class="text-sm text-gray-600 dark:text-gray-300 block max-w-[320px] whitespace-normal break-all">
+              {{ formatUsageEndpoints(row) }}
+            </span>
+          </template>
+
          <template #cell-stream="{ row }">
            <span
              class="inline-flex items-center rounded px-2 py-0.5 text-xs font-medium"
@@ -516,6 +522,7 @@ const columns = computed<Column[]>(() => [
  { key: 'api_key', label: t('usage.apiKeyFilter'), sortable: false },
  { key: 'model', label: t('usage.model'), sortable: true },
  { key: 'reasoning_effort', label: t('usage.reasoningEffort'), sortable: false },
+  { key: 'endpoint', label: t('usage.endpoint'), sortable: false },
  { key: 'stream', label: t('usage.type'), sortable: false },
  { key: 'tokens', label: t('usage.tokens'), sortable: false },
  { key: 'cost', label: t('usage.cost'), sortable: false },
@@ -615,6 +622,11 @@ const getRequestTypeExportText = (log: UsageLog): string => {
  return 'Unknown'
 }

+const formatUsageEndpoints = (log: UsageLog): string => {
+  const inbound = log.inbound_endpoint?.trim()
+  return inbound || '-'
+}
+
 const formatTokens = (value: number): string => {
  if (value >= 1_000_000_000) {
    return `${(value / 1_000_000_000).toFixed(2)}B`
@@ -789,6 +801,7 @@ const exportToCSV = async () => {
      'API Key Name',
      'Model',
      'Reasoning Effort',
+      'Inbound Endpoint',
      'Type',
      'Input Tokens',
      'Output Tokens',
@@ -806,6 +819,7 @@ const exportToCSV = async () => {
        log.api_key?.name || '',
        log.model,
        formatReasoningEffort(log.reasoning_effort),
+        log.inbound_endpoint || '',
        getRequestTypeExportText(log),
        log.input_tokens,
        log.output_tokens,