feat: add configurable upstream error message override

controller/midjourney.go

@@ -14,6 +14,7 @@ import (
 	"github.com/QuantumNous/new-api/model"
 	"github.com/QuantumNous/new-api/service"
 	"github.com/QuantumNous/new-api/setting"
+	"github.com/QuantumNous/new-api/setting/operation_setting"
 	"github.com/QuantumNous/new-api/setting/system_setting"
 
 	"github.com/gin-gonic/gin"
@@ -332,11 +333,13 @@ func GetUserMidjourney(c *gin.Context) {
 	items := model.GetAllUserTask(userId, pageInfo.GetStartIdx(), pageInfo.GetPageSize(), queryParams)
 	total := model.CountAllUserTask(userId, queryParams)
 
-	if setting.MjForwardUrlEnabled {
-		for i, midjourney := range items {
+	for _, midjourney := range items {
+		if setting.MjForwardUrlEnabled {
 			midjourney.ImageUrl = system_setting.ServerAddress + "/mj/image/" + midjourney.MjId
-			items[i] = midjourney
 		}
+		// 面向用户的 MJ 任务列表同样是上游失败原因的出口，与 /mj/task 查询
+		// （coverMidjourneyTaskDto）保持一致；管理员列表 GetAllMidjourney 保留原文。
+		midjourney.FailReason = operation_setting.OverrideUpstreamMessage(midjourney.FailReason)
 	}
 	pageInfo.SetTotal(int(total))
 	pageInfo.SetItems(items)

controller/midjourney_fail_reason_override_test.go

@@ -0,0 +1,78 @@
+package controller
+
+import (
+	"net/http"
+	"net/http/httptest"
+	"testing"
+
+	"github.com/QuantumNous/new-api/common"
+	"github.com/QuantumNous/new-api/model"
+	"github.com/QuantumNous/new-api/setting/operation_setting"
+
+	"github.com/gin-gonic/gin"
+	"github.com/glebarez/sqlite"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"gorm.io/gorm"
+)
+
+// The Midjourney task list is a user-facing outlet for the same upstream failure text
+// the /mj/task fetch path already masks (relay.coverMidjourneyTaskDto). The user list
+// must mask it while the admin list keeps the original for diagnosis.
+func TestMidjourneyListFailReasonOverride(t *testing.T) {
+	enableTaskErrorOverride(t)
+	gin.SetMode(gin.TestMode)
+
+	previousDB := model.DB
+	db, err := gorm.Open(sqlite.Open(":memory:"), &gorm.Config{})
+	require.NoError(t, err)
+	require.NoError(t, db.AutoMigrate(&model.Midjourney{}))
+	model.DB = db
+	t.Cleanup(func() { model.DB = previousDB })
+
+	const userID = 7
+	const upstreamReason = "insufficient credits, please top-up your account"
+	const localReason = "获取渠道信息失败，请联系管理员，渠道ID：3"
+
+	require.NoError(t, db.Create(&model.Midjourney{
+		UserId: userID, MjId: "mj-upstream", Status: "FAILURE", FailReason: upstreamReason,
+	}).Error)
+	require.NoError(t, db.Create(&model.Midjourney{
+		UserId: userID, MjId: "mj-local", Status: "FAILURE", FailReason: localReason,
+	}).Error)
+
+	failReasons := func(handler gin.HandlerFunc) map[string]string {
+		t.Helper()
+		recorder := httptest.NewRecorder()
+		c, _ := gin.CreateTestContext(recorder)
+		c.Request = httptest.NewRequest(http.MethodGet, "/api/mj/self", nil)
+		c.Set("id", userID)
+
+		handler(c)
+		require.Equal(t, http.StatusOK, recorder.Code)
+
+		var response struct {
+			Data struct {
+				Items []struct {
+					MjId       string `json:"mj_id"`
+					FailReason string `json:"fail_reason"`
+				} `json:"items"`
+			} `json:"data"`
+		}
+		require.NoError(t, common.Unmarshal(recorder.Body.Bytes(), &response))
+		reasons := make(map[string]string, len(response.Data.Items))
+		for _, item := range response.Data.Items {
+			reasons[item.MjId] = item.FailReason
+		}
+		return reasons
+	}
+
+	userReasons := failReasons(GetUserMidjourney)
+	assert.Equal(t, operation_setting.ErrorOverrideMessage, userReasons["mj-upstream"])
+	// This site's own failure text must survive the user-facing path.
+	assert.Equal(t, localReason, userReasons["mj-local"])
+
+	adminReasons := failReasons(GetAllMidjourney)
+	assert.Equal(t, upstreamReason, adminReasons["mj-upstream"])
+	assert.Equal(t, localReason, adminReasons["mj-local"])
+}

controller/playground.go

@@ -8,6 +8,7 @@ import (
 	"github.com/QuantumNous/new-api/model"
 	relaycommon "github.com/QuantumNous/new-api/relay/common"
 	"github.com/QuantumNous/new-api/relaykit/types"
+	"github.com/QuantumNous/new-api/setting/operation_setting"
 
 	"github.com/gin-gonic/gin"
 )
@@ -17,6 +18,7 @@ func Playground(c *gin.Context) {
 
 	defer func() {
 		if newAPIError != nil {
+			operation_setting.OverrideUpstreamError(newAPIError)
 			c.JSON(newAPIError.StatusCode, gin.H{
 				"error": newAPIError.ToOpenAIError(),
 			})

controller/relay.go

@@ -92,7 +92,14 @@ func Relay(c *gin.Context, relayFormat types.RelayFormat) {
 	defer func() {
 		if newAPIError != nil {
 			logger.LogError(c, fmt.Sprintf("relay error: %s", common.LocalLogPreview(newAPIError.Error())))
-			newAPIError.SetMessage(common.MessageWithRequestId(newAPIError.Error(), requestId))
+			// 必须在错误日志之后执行：日志与渠道禁用判定始终使用原始上游文案。
+			// 覆写文案与 request id 一次写入：上游错误的 ToOpenAIError() 直接返回 RelayError，
+			// 不读 Err，SetMessage 改不到响应体，必须走 ReplaceMessage
+			if operation_setting.ShouldOverrideUpstreamError(newAPIError) {
+				newAPIError.ReplaceMessage(common.MessageWithRequestId(operation_setting.ErrorOverrideMessage, requestId))
+			} else {
+				newAPIError.SetMessage(common.MessageWithRequestId(newAPIError.Error(), requestId))
+			}
 			switch relayFormat {
 			case types.RelayFormatOpenAIRealtime:
 				helper.WssError(c, ws, newAPIError.ToOpenAIError())
@@ -431,13 +438,26 @@ func processChannelError(c *gin.Context, channelError types.ChannelError, err *t
 			adminInfo["multi_key_index"] = common.GetContextKeyInt(c, constant.ContextKeyChannelMultiKeyIndex)
 		}
 		service.AppendChannelAffinityAdminInfo(c, adminInfo)
+		// 错误日志的 Content 会通过 /api/log/self 回显给发起请求的用户，因此它和 HTTP 响应
+		// 一样是对外出口。覆写生效时这里必须同步覆写，否则客户端在响应里看到
+		// Service Unavailable，转头在日志页仍能读到上游账务原文。原文改记到
+		// admin_info，model.formatUserLogs 会为普通用户剥离整个 admin_info。
+		logContent := err.MaskSensitiveErrorWithStatusCode()
+		if operation_setting.ShouldOverrideUpstreamError(err) {
+			adminInfo["original_error"] = logContent
+			logContent = operation_setting.ErrorOverrideMessage
+			// 与 MaskSensitiveErrorWithStatusCode 保持一致：无状态码时不写 status_code=0
+			if err.StatusCode != 0 {
+				logContent = fmt.Sprintf("status_code=%d, %s", err.StatusCode, logContent)
+			}
+		}
 		other["admin_info"] = adminInfo
 		startTime := common.GetContextKeyTime(c, constant.ContextKeyRequestStartTime)
 		if startTime.IsZero() {
 			startTime = time.Now()
 		}
 		useTimeSeconds := int(time.Since(startTime).Seconds())
-		model.RecordErrorLog(c, userId, channelId, modelName, tokenName, err.MaskSensitiveErrorWithStatusCode(), tokenId, useTimeSeconds, common.GetContextKeyBool(c, constant.ContextKeyIsStream), userGroup, other)
+		model.RecordErrorLog(c, userId, channelId, modelName, tokenName, logContent, tokenId, useTimeSeconds, common.GetContextKeyBool(c, constant.ContextKeyIsStream), userGroup, other)
 	}
 
 }
@@ -475,13 +495,23 @@ func RelayMidjourney(c *gin.Context) {
 			mjErr.Result = "当前分组负载已饱和，请稍后再试，或升级账户以提升服务质量。"
 			statusCode = http.StatusTooManyRequests
 		}
+		description := fmt.Sprintf("%s %s", mjErr.Description, mjErr.Result)
+		channelId := c.GetInt("channel_id")
+		logger.LogError(c, fmt.Sprintf("relay error (channel #%d, status code %d): %s", channelId, statusCode, description))
+		// 这里一律不覆写：能走到 mjErr 的 MidjourneyResponse 全部是本站自产的（参数校验、
+		// 额度不足、DB/IO 失败）。mj-proxy 各 handler 拿到上游响应后是把上游 body 原样
+		// io.Copy 给客户端再 return nil 的，上游错误文案根本不经过这个分支。
+		// 之前在此处按关键词覆写会把本站的 quota_not_enough（mjproxy_handler.go 的
+		// RelaySwapFace / RelayMidjourneySubmit）误伤成 Service Unavailable，
+		// 让用户看不到自己额度不足的真实原因。
+		// MJ 上游文案的出口是被代理的响应体本身与任务的 FailReason，前者要改写就得重写
+		// 上游 JSON，会破坏 mj-proxy 协议兼容性，故不在本功能范围内；后者已在
+		// TaskModel2Dto / coverMidjourneyTaskDto 的读取边界处理。
 		c.JSON(statusCode, gin.H{
-			"description": fmt.Sprintf("%s %s", mjErr.Description, mjErr.Result),
+			"description": description,
 			"type":        "upstream_error",
 			"code":        mjErr.Code,
 		})
-		channelId := c.GetInt("channel_id")
-		logger.LogError(c, fmt.Sprintf("relay error (channel #%d, status code %d): %s", channelId, statusCode, fmt.Sprintf("%s %s", mjErr.Description, mjErr.Result)))
 	}
 }
 
@@ -617,7 +647,7 @@ func RelayTask(c *gin.Context) {
 			processChannelError(c,
 				*types.NewChannelError(channel.Id, channel.Type, channel.Name, channel.ChannelInfo.IsMultiKey,
 					common.GetContextKeyString(c, constant.ContextKeyChannelKey), channel.GetAutoBan()),
-				types.NewOpenAIError(taskErr.Error, types.ErrorCodeBadResponseStatusCode, taskErr.StatusCode))
+				service.APIErrorFromTaskError(taskErr))
 		}
 
 		if taskFailoverEnabled {
@@ -674,10 +704,24 @@ func respondTaskError(c *gin.Context, taskErr *taskdto.TaskError) {
 	if taskErr.StatusCode == http.StatusTooManyRequests {
 		taskErr.Message = "当前分组上游负载已饱和，请稍后再试"
 	}
+	// 仅覆写文案确实取自上游任务平台的错误。这里必须用正向标记 FromUpstream 判定：
+	// 本站自产错误（读 body 失败、解析失败、预扣费额度不足）默认 LocalError == false，
+	// 用 !LocalError 反推会把它们一并掩盖。
+	if taskErr.FromUpstream {
+		if overridden := operation_setting.OverrideUpstreamMessage(taskErr.Message); overridden != taskErr.Message {
+			// 覆写前先记录原始上游文案，后台日志与排障始终可见全文
+			logger.LogError(c, fmt.Sprintf("task upstream error overridden: %s", common.LocalLogPreview(taskErr.Message)))
+			taskErr.Message = overridden
+			// Data 带 json:"data" 会一并返回客户端。当前 task 适配器不往里塞上游 body，
+			// 但覆写掉 Message 却留着一个可能承载上游原文的字段是自相矛盾的，
+			// 与 NewAPIError.ReplaceMessage 清 Metadata/Param 保持一致。
+			taskErr.Data = nil
+		}
+	}
 	c.JSON(taskErr.StatusCode, taskErr)
 }
 
-func shouldRetryTaskRelay(c *gin.Context, channelId int, taskErr *taskdto.TaskError, retryTimes int) bool {
+func shouldRetryTaskRelay(c *gin.Context, channelId int, taskErr *taskdto.TaskError, retryTimes int, failoverEnabled bool) bool {
 	if taskErr == nil {
 		return false
 	}

controller/task.go

@@ -32,7 +32,8 @@ func GetAllTask(c *gin.Context) {
 	items := model.TaskGetAllTasks(pageInfo.GetStartIdx(), pageInfo.GetPageSize(), queryParams)
 	total := model.TaskCountAllTasks(queryParams)
 	pageInfo.SetTotal(int(total))
-	pageInfo.SetItems(tasksToDto(items, true))
+	// 管理员视图不掩盖上游失败原因，排障需要原文
+	pageInfo.SetItems(tasksToDto(items, true, false))
 	common.ApiSuccess(c, pageInfo)
 }
 
@@ -56,11 +57,13 @@ func GetUserTask(c *gin.Context) {
 	items := model.TaskGetAllUserTask(userId, pageInfo.GetStartIdx(), pageInfo.GetPageSize(), queryParams)
 	total := model.TaskCountAllUserTask(userId, queryParams)
 	pageInfo.SetTotal(int(total))
-	pageInfo.SetItems(tasksToDto(items, false))
+	pageInfo.SetItems(tasksToDto(items, false, true))
 	common.ApiSuccess(c, pageInfo)
 }
 
-func tasksToDto(tasks []*model.Task, fillUser bool) []*dto.TaskDto {
+// tasksToDto 转换任务列表。maskUpstreamFailReason 单独传参而不复用 fillUser：填充用户名
+// 与是否掩盖上游失败原因是两个无关维度，绑在一起会让后续加入新调用方时选错默认值。
+func tasksToDto(tasks []*model.Task, fillUser bool, maskUpstreamFailReason bool) []*dto.TaskDto {
 	var userIdMap map[int]*model.UserBase
 	if fillUser {
 		userIdMap = make(map[int]*model.UserBase)
@@ -82,7 +85,7 @@ func tasksToDto(tasks []*model.Task, fillUser bool) []*dto.TaskDto {
 				task.Username = user.Username
 			}
 		}
-		result[i] = relay.TaskModel2Dto(task)
+		result[i] = relay.TaskModel2Dto(task, maskUpstreamFailReason)
 	}
 	return result
 }