perf(web): simplify public page hero copy

Merge pull request #5317 from P2K0/feat/profile-show-user-id

.env.example

@@ -56,6 +56,8 @@
 # 对话超时设置
 # 所有请求超时时间，单位秒，默认为0，表示不限制
 # RELAY_TIMEOUT=0
+# Relay HTTP 客户端空闲连接超时时间，单位秒，默认跟随 Go 标准库，设置为0表示不限制
+# RELAY_IDLE_CONN_TIMEOUT=90
 # 流模式无响应超时时间，单位秒，如果出现空补全可以尝试改为更大值
 # STREAMING_TIMEOUT=300
 

.github/ISSUE_TEMPLATE/bug_report.md

@@ -11,6 +11,8 @@ assignees: ''
 
 - 文档：https://docs.newapi.ai/
 - 使用问题先看或先问：https://deepwiki.com/QuantumNous/new-api
+- 开启透传后的转发相关反馈不接受 issue；透传模式会直接转发请求，请自行确认上游行为。
+- 不接受 coding plan、逆向渠道等技术支持类 issue。
 - 警告：删除本模板、删除小节标题或随意清空内容的 issue，可能会被直接关闭；重复恶意提交者可能会被 block。
 
 **您当前的 newapi 版本**
@@ -20,13 +22,18 @@ assignees: ''
 **提交确认**
 
 [//]: # (方框内删除已有的空格，填 x 号)
-+ [ ] 我已确认目前没有类似 issue
-+ [ ] 我已完整查看过文档 https://docs.newapi.ai/ 和项目 README，尤其是常见问题部分
-+ [ ] 我未删除此模板中的任何引导内容或小节标题，并会按要求完整填写
-+ [ ] 我理解项目维护者精力有限，不遵循模板要求的 issue 可能会被无视或直接关闭
+- [ ] **非重复 issue:** 我已搜索现有 [Issues](https://github.com/QuantumNous/new-api/issues?q=is%3Aissue)，确认目前没有类似 issue。
+- [ ] **提交前必读:** 我已完整阅读上方“提交前必读”，并已查看文档 https://docs.newapi.ai/、项目 README 且向 AI 提问，确认这不是使用、配置或接入类问题。
+- [ ] **模板完整:** 我未删除此模板中的任何引导内容或小节标题，并会按要求完整填写。
+- [ ] **维护成本:** 我理解项目维护者精力有限，不遵循模板要求的 issue 可能会被无视或直接关闭。
 
 **问题描述**
 
+请尽可能说明问题现象、影响范围，以及你判断它是程序问题而不是上游行为或使用问题的依据。
+
+- 转发问题请尽可能说明渠道类型、转换格式、上游原生支持依据和服务端日志。
+- 计费问题请尽可能附请求返回的 `usage` 示例。
+
 **复现步骤**
 
 **预期结果**

.github/ISSUE_TEMPLATE/bug_report_en.md

@@ -11,6 +11,8 @@ assignees: ''
 
 - Docs: https://docs.newapi.ai/
 - Usage questions first: https://deepwiki.com/QuantumNous/new-api
+- Issues about forwarding behavior after enabling pass-through mode are not accepted; pass-through mode forwards requests directly, so please verify upstream behavior yourself.
+- Technical support requests such as coding plans or reverse-engineering channels are not accepted as issues.
 - Warning: issues with this template removed, section headings deleted, or content cleared may be closed directly. Repeated abusive submissions may result in a block.
 
 **Your current newapi version**
@@ -20,13 +22,18 @@ Please fill this in, for example: `v1.0.0`
 **Submission Checks**
 
 [//]: # (Remove the space in the box and fill with an x)
-+ [ ] I have confirmed there are no similar issues
-+ [ ] I have thoroughly read the docs at https://docs.newapi.ai/ and the project README, especially the FAQ section
-+ [ ] I have not removed any guidance or section headings from this template and will complete it as requested
-+ [ ] I understand that maintainers have limited time and issues that do not follow this template may be ignored or closed directly
+- [ ] **Non-duplicate issue:** I have searched existing [Issues](https://github.com/QuantumNous/new-api/issues?q=is%3Aissue) and confirmed there are no similar issues.
+- [ ] **Read this first:** I have fully read the section above, reviewed the docs at https://docs.newapi.ai/ and the project README, and asked AI first, confirming this is not a usage, configuration, or integration question.
+- [ ] **Template intact:** I have not removed any guidance or section headings from this template and will complete it as requested.
+- [ ] **Maintainer time:** I understand that maintainers have limited time and issues that do not follow this template may be ignored or closed directly.
 
 **Issue Description**
 
+Describe the symptom, impact scope, and why you believe this is an application issue rather than upstream behavior or a usage question with as much detail as possible.
+
+- For forwarding issues, include the channel type, conversion format, upstream native-support evidence, and server logs when possible.
+- For billing issues, include an example of the returned `usage` when possible.
+
 **Steps to Reproduce**
 
 **Expected Result**

.github/ISSUE_TEMPLATE/feature_request.md

@@ -11,6 +11,8 @@ assignees: ''
 
 - 文档：https://docs.newapi.ai/
 - 使用问题先看或先问：https://deepwiki.com/QuantumNous/new-api
+- 开启透传后的转发相关反馈不接受 issue；透传模式会直接转发请求，请自行确认上游行为。
+- 不接受 coding plan、逆向渠道等技术支持类 issue。
 - 警告：删除本模板、删除小节标题或随意清空内容的 issue，可能会被直接关闭；重复恶意提交者可能会被 block。
 
 **您当前的 newapi 版本**
@@ -20,10 +22,10 @@ assignees: ''
 **提交确认**
 
 [//]: # (方框内删除已有的空格，填 x 号)
-+ [ ] 我已确认目前没有类似 issue
-+ [ ] 我已完整查看过文档 https://docs.newapi.ai/ 和项目 README，已确定现有版本无法满足需求
-+ [ ] 我未删除此模板中的任何引导内容或小节标题，并会按要求完整填写
-+ [ ] 我理解项目维护者精力有限，不遵循模板要求的 issue 可能会被无视或直接关闭
+- [ ] **非重复 issue:** 我已搜索现有 [Issues](https://github.com/QuantumNous/new-api/issues?q=is%3Aissue)，确认目前没有类似 issue。
+- [ ] **提交前必读:** 我已完整阅读上方“提交前必读”，并已查看文档 https://docs.newapi.ai/、项目 README 且向 AI 提问，确认这不是使用、配置或接入类问题，且现有版本无法满足需求。
+- [ ] **模板完整:** 我未删除此模板中的任何引导内容或小节标题，并会按要求完整填写。
+- [ ] **维护成本:** 我理解项目维护者精力有限，不遵循模板要求的 issue 可能会被无视或直接关闭。
 
 **功能描述**
 

.github/ISSUE_TEMPLATE/feature_request_en.md

@@ -11,6 +11,8 @@ assignees: ''
 
 - Docs: https://docs.newapi.ai/
 - Usage questions first: https://deepwiki.com/QuantumNous/new-api
+- Issues about forwarding behavior after enabling pass-through mode are not accepted; pass-through mode forwards requests directly, so please verify upstream behavior yourself.
+- Technical support requests such as coding plans or reverse-engineering channels are not accepted as issues.
 - Warning: issues with this template removed, section headings deleted, or content cleared may be closed directly. Repeated abusive submissions may result in a block.
 
 **Your current newapi version**
@@ -20,10 +22,10 @@ Please fill this in, for example: `v1.0.0`
 **Submission Checks**
 
 [//]: # (Remove the space in the box and fill with an x)
-+ [ ] I have confirmed there are no similar issues
-+ [ ] I have thoroughly read the docs at https://docs.newapi.ai/ and the project README, and confirmed the current version cannot meet my needs
-+ [ ] I have not removed any guidance or section headings from this template and will complete it as requested
-+ [ ] I understand that maintainers have limited time and issues that do not follow this template may be ignored or closed directly
+- [ ] **Non-duplicate issue:** I have searched existing [Issues](https://github.com/QuantumNous/new-api/issues?q=is%3Aissue) and confirmed there are no similar issues.
+- [ ] **Read this first:** I have fully read the section above, reviewed the docs at https://docs.newapi.ai/ and the project README, and asked AI first, confirming this is not a usage, configuration, or integration question, and that the current version cannot meet my needs.
+- [ ] **Template intact:** I have not removed any guidance or section headings from this template and will complete it as requested.
+- [ ] **Maintainer time:** I understand that maintainers have limited time and issues that do not follow this template may be ignored or closed directly.
 
 **Feature Description**
 

README.md

@@ -316,6 +316,7 @@ docker run --name new-api -d --restart always \
 | `CRYPTO_SECRET` | Encryption secret (required for Redis) | - |
 | `SQL_DSN` | Database connection string | - |
 | `REDIS_CONN_STRING` | Redis connection string | - |
+| `RELAY_IDLE_CONN_TIMEOUT` | Idle keep-alive timeout for relay HTTP clients, seconds. Defaults to Go standard library behavior; set `0` to disable | `90` |
 | `STREAMING_TIMEOUT` | Streaming timeout (seconds) | `300` |
 | `STREAM_SCANNER_MAX_BUFFER_MB` | Max per-line buffer (MB) for the stream scanner; increase when upstream sends huge image/base64 payloads | `64` |
 | `MAX_REQUEST_BODY_MB` | Max request body size (MB, counted **after decompression**; prevents huge requests/zip bombs from exhausting memory). Exceeding it returns `413` | `32` |

common/constants.go

@@ -170,6 +170,7 @@ var BatchUpdateInterval int
 
 var RelayTimeout int // unit is second
 
+var RelayIdleConnTimeout int // unit is second
 var RelayMaxIdleConns int
 var RelayMaxIdleConnsPerHost int
 

common/init.go

@@ -102,6 +102,7 @@ func InitEnv() {
 	SyncFrequency = GetEnvOrDefault("SYNC_FREQUENCY", 60)
 	BatchUpdateInterval = GetEnvOrDefault("BATCH_UPDATE_INTERVAL", 5)
 	RelayTimeout = GetEnvOrDefault("RELAY_TIMEOUT", 0)
+	RelayIdleConnTimeout = GetEnvOrDefault("RELAY_IDLE_CONN_TIMEOUT", 90)
 	RelayMaxIdleConns = GetEnvOrDefault("RELAY_MAX_IDLE_CONNS", 500)
 	RelayMaxIdleConnsPerHost = GetEnvOrDefault("RELAY_MAX_IDLE_CONNS_PER_HOST", 100)
 
@@ -135,6 +136,7 @@ func initConstantEnv() {
 	constant.StreamScannerMaxBufferMB = GetEnvOrDefault("STREAM_SCANNER_MAX_BUFFER_MB", 128)
 	// MaxRequestBodyMB 请求体最大大小（解压后），用于防止超大请求/zip bomb导致内存暴涨
 	constant.MaxRequestBodyMB = GetEnvOrDefault("MAX_REQUEST_BODY_MB", 128)
+	constant.AnonymousRequestBodyLimitKB = GetEnvOrDefault("ANONYMOUS_REQUEST_BODY_LIMIT_KB", 512)
 	// ForceStreamOption 覆盖请求参数，强制返回usage信息
 	constant.ForceStreamOption = GetEnvOrDefaultBool("FORCE_STREAM_OPTION", true)
 	constant.CountToken = GetEnvOrDefaultBool("CountToken", true)

common/request_body_limit.go

@@ -0,0 +1,13 @@
+package common
+
+import "github.com/QuantumNous/new-api/constant"
+
+const defaultAnonymousRequestBodyLimitKB = 512
+
+func GetAnonymousRequestBodyLimitBytes() int64 {
+	limitKB := constant.AnonymousRequestBodyLimitKB
+	if limitKB < 0 {
+		limitKB = defaultAnonymousRequestBodyLimitKB
+	}
+	return int64(limitKB) << 10
+}

constant/env.go

@@ -10,6 +10,7 @@ var GetMediaToken bool
 var GetMediaTokenNotStream bool
 var UpdateTask bool
 var MaxRequestBodyMB int
+var AnonymousRequestBodyLimitKB int
 var AzureDefaultAPIVersion string
 var NotifyLimitCount int
 var NotificationLimitDurationMinute int

controller/channel-test.go

@@ -814,7 +814,7 @@ func buildTestRequest(model string, endpointType string, channel *model.Channel,
 		testRequest.StreamOptions = &dto.StreamOptions{IncludeUsage: true}
 	}
 
-	if strings.HasPrefix(model, "o") {
+	if dto.IsOpenAIReasoningOModel(model) {
 		testRequest.MaxCompletionTokens = lo.ToPtr(uint(16))
 	} else if strings.Contains(model, "thinking") {
 		if !strings.Contains(model, "claude") {

docker-compose.yml

@@ -34,6 +34,7 @@ services:
       - BATCH_UPDATE_ENABLED=true  # 是否启用批量更新 (Whether to enable batch update)
       - NODE_NAME=new-api-node-1  # 节点名称，用于审计日志中标识节点身份；多节点/容器部署时建议设置 (Node name used in audit logs; recommended when running multiple instances or in containers)
 #      - STREAMING_TIMEOUT=300  # 流模式无响应超时时间，单位秒，默认120秒，如果出现空补全可以尝试改为更大值 （Streaming timeout in seconds, default is 120s. Increase if experiencing empty completions）
+#      - RELAY_IDLE_CONN_TIMEOUT=90  # Relay HTTP 客户端空闲连接超时时间，单位秒，默认跟随 Go 标准库，设置为0表示不限制 (Relay HTTP client idle keep-alive timeout in seconds, defaults to Go standard library; set 0 to disable)
 #      - SESSION_SECRET=random_string  # 多机部署时设置，必须修改这个随机字符串！！ （multi-node deployment, set this to a random string!!!!!!!）
 #      - SYNC_FREQUENCY=60  # Uncomment if regular database syncing is needed
 #      - GOOGLE_ANALYTICS_ID=G-XXXXXXXXXX  # Google Analytics 的测量 ID (Google Analytics Measurement ID)

dto/openai_request.go

@@ -213,12 +213,22 @@ func (r *GeneralOpenAIRequest) ToMap() map[string]any {
 	return result
 }
 
+func IsOpenAIReasoningOModel(modelName string) bool {
+	return strings.HasPrefix(modelName, "o1") ||
+		strings.HasPrefix(modelName, "o3") ||
+		strings.HasPrefix(modelName, "o4")
+}
+
+func IsOpenAIGPT5Model(modelName string) bool {
+	return strings.HasPrefix(modelName, "gpt-5")
+}
+
 func (r *GeneralOpenAIRequest) GetSystemRoleName() string {
-	if strings.HasPrefix(r.Model, "o") {
+	if IsOpenAIReasoningOModel(r.Model) {
 		if !strings.HasPrefix(r.Model, "o1-mini") && !strings.HasPrefix(r.Model, "o1-preview") {
 			return "developer"
 		}
-	} else if strings.HasPrefix(r.Model, "gpt-5") {
+	} else if IsOpenAIGPT5Model(r.Model) {
 		return "developer"
 	}
 	return "system"

dto/openai_request_zero_value_test.go

@@ -71,3 +71,27 @@ func TestOpenAIResponsesRequestPreserveExplicitZeroValues(t *testing.T) {
 	require.True(t, gjson.GetBytes(encoded, "stream").Exists())
 	require.True(t, gjson.GetBytes(encoded, "top_p").Exists())
 }
+
+func TestGeneralOpenAIRequestGetSystemRoleName(t *testing.T) {
+	tests := []struct {
+		name  string
+		model string
+		want  string
+	}{
+		{name: "o1 uses developer", model: "o1", want: "developer"},
+		{name: "o3 family uses developer", model: "o3-mini-high", want: "developer"},
+		{name: "o4 family uses developer", model: "o4-mini", want: "developer"},
+		{name: "o1 mini stays system", model: "o1-mini", want: "system"},
+		{name: "o1 preview stays system", model: "o1-preview", want: "system"},
+		{name: "gpt 5 uses developer", model: "gpt-5", want: "developer"},
+		{name: "omni is not o series", model: "omni-moderation-latest", want: "system"},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			req := GeneralOpenAIRequest{Model: tt.model}
+
+			require.Equal(t, tt.want, req.GetSystemRoleName())
+		})
+	}
+}

middleware/distributor.go

@@ -102,14 +102,10 @@ func Distribute() func(c *gin.Context) {
 				}
 
 				if preferredChannelID, found := service.GetPreferredChannelByAffinity(c, modelRequest.Model, usingGroup); found {
+					affinityUsable := false
 					preferred, err := model.CacheGetChannel(preferredChannelID)
-					if err == nil && preferred != nil {
-						if preferred.Status != common.ChannelStatusEnabled {
-							if service.ShouldSkipRetryAfterChannelAffinityFailure(c) {
-								abortWithOpenAiMessage(c, http.StatusForbidden, i18n.T(c, i18n.MsgDistributorAffinityChannelDisabled))
-								return
-							}
-						} else if usingGroup == "auto" {
+					if err == nil && preferred != nil && preferred.Status == common.ChannelStatusEnabled {
+						if usingGroup == "auto" {
 							userGroup := common.GetContextKeyString(c, constant.ContextKeyUserGroup)
 							autoGroups := service.GetUserAutoGroup(userGroup)
 							for _, g := range autoGroups {
@@ -117,6 +113,7 @@ func Distribute() func(c *gin.Context) {
 									selectGroup = g
 									common.SetContextKey(c, constant.ContextKeyAutoGroup, g)
 									channel = preferred
+									affinityUsable = true
 									service.MarkChannelAffinityUsed(c, g, preferred.Id)
 									break
 								}
@@ -124,9 +121,13 @@ func Distribute() func(c *gin.Context) {
 						} else if model.IsChannelEnabledForGroupModel(usingGroup, modelRequest.Model, preferred.Id) {
 							channel = preferred
 							selectGroup = usingGroup
+							affinityUsable = true
 							service.MarkChannelAffinityUsed(c, usingGroup, preferred.Id)
 						}
 					}
+					if !affinityUsable && !service.ShouldKeepChannelAffinityOnChannelDisabled() {
+						service.ClearCurrentChannelAffinityCache(c)
+					}
 				}
 
 				if channel == nil {
@@ -298,6 +299,7 @@ func getModelRequest(c *gin.Context) (*ModelRequest, bool, error) {
 		} else if c.Request.Method == http.MethodGet {
 			relayMode = relayconstant.RelayModeVideoFetchByID
 			shouldSelectChannel = false
+			modelRequest.Model = getTaskOriginModelName(c)
 		}
 		c.Set("relay_mode", relayMode)
 	} else if strings.Contains(c.Request.URL.Path, "/v1/video/generations") {
@@ -312,6 +314,7 @@ func getModelRequest(c *gin.Context) (*ModelRequest, bool, error) {
 		} else if c.Request.Method == http.MethodGet {
 			relayMode = relayconstant.RelayModeVideoFetchByID
 			shouldSelectChannel = false
+			modelRequest.Model = getTaskOriginModelName(c)
 		}
 		if _, ok := c.Get("relay_mode"); !ok {
 			c.Set("relay_mode", relayMode)
@@ -396,6 +399,31 @@ func getModelRequest(c *gin.Context) (*ModelRequest, bool, error) {
 	return &modelRequest, shouldSelectChannel, nil
 }
 
+// 修复 #4834: GET /v1/video/generations/:task_id && /v1/video/:task_id 此前不解析 model，
+// 当 token 启用「可用模型限制」时，下游 modelLimitEnable 校验会因
+// modelRequest.Model 为空而误报 "This token has no access to model"。
+// 从已存储的任务记录中回填 OriginModelName 即可让校验走在正确的模型上。
+func getTaskOriginModelName(c *gin.Context) string {
+	if !common.GetContextKeyBool(c, constant.ContextKeyTokenModelLimitEnabled) {
+		return ""
+	}
+
+	taskId := c.Param("task_id")
+	if taskId == "" {
+		// jimeng adapter
+		taskId = c.GetString("task_id")
+	}
+	if taskId == "" {
+		return ""
+	}
+
+	userId := c.GetInt("id")
+	if task, exist, err := model.GetByTaskId(userId, taskId); err == nil && exist && task != nil {
+		return task.Properties.OriginModelName
+	}
+	return ""
+}
+
 func SetupContextForSelectedChannel(c *gin.Context, channel *model.Channel, modelName string) *types.NewAPIError {
 	c.Set("original_model", modelName) // for retry
 	if channel == nil {

middleware/request_body_limit.go

@@ -0,0 +1,47 @@
+package middleware
+
+import (
+	"bytes"
+	"io"
+	"net/http"
+
+	"github.com/QuantumNous/new-api/common"
+	"github.com/gin-gonic/gin"
+)
+
+func AnonymousRequestBodyLimit() gin.HandlerFunc {
+	return func(c *gin.Context) {
+		maxBytes := common.GetAnonymousRequestBodyLimitBytes()
+		if maxBytes <= 0 || c.Request.Body == nil {
+			c.Next()
+			return
+		}
+
+		originalBody := c.Request.Body
+		limitedBody, err := readAnonymousRequestBody(originalBody, maxBytes)
+		_ = originalBody.Close()
+		if err != nil {
+			if common.IsRequestBodyTooLargeError(err) {
+				c.AbortWithStatus(http.StatusRequestEntityTooLarge)
+				return
+			}
+			c.AbortWithStatus(http.StatusBadRequest)
+			return
+		}
+
+		c.Request.Body = io.NopCloser(bytes.NewReader(limitedBody))
+		c.Request.ContentLength = int64(len(limitedBody))
+		c.Next()
+	}
+}
+
+func readAnonymousRequestBody(body io.Reader, maxBytes int64) ([]byte, error) {
+	data, err := io.ReadAll(io.LimitReader(body, maxBytes+1))
+	if err != nil {
+		return nil, err
+	}
+	if int64(len(data)) > maxBytes {
+		return nil, common.ErrRequestBodyTooLarge
+	}
+	return data, nil
+}

relay/channel/cloudflare/relay_cloudflare.go

@@ -30,7 +30,7 @@ func convertCf2CompletionsRequest(textRequest dto.GeneralOpenAIRequest) *CfReque
 }
 
 func cfStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Response) (*types.NewAPIError, *dto.Usage) {
-	scanner := bufio.NewScanner(resp.Body)
+	scanner := helper.NewStreamScanner(resp.Body)
 	scanner.Split(bufio.ScanLines)
 
 	helper.SetEventStreamHeaders(c)

relay/channel/cohere/relay-cohere.go

@@ -1,7 +1,6 @@
 package cohere
 
 import (
-	"bufio"
 	"encoding/json"
 	"io"
 	"net/http"
@@ -86,7 +85,7 @@ func cohereStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http
 	createdTime := common.GetTimestamp()
 	usage := &dto.Usage{}
 	responseText := ""
-	scanner := bufio.NewScanner(resp.Body)
+	scanner := helper.NewStreamScanner(resp.Body)
 	scanner.Split(func(data []byte, atEOF bool) (advance int, token []byte, err error) {
 		if atEOF && len(data) == 0 {
 			return 0, nil, nil
@@ -106,6 +105,9 @@ func cohereStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http
 			data := scanner.Text()
 			dataChan <- data
 		}
+		if err := scanner.Err(); err != nil {
+			common.SysLog("error reading stream: " + err.Error())
+		}
 		stopChan <- true
 	}()
 	helper.SetEventStreamHeaders(c)

relay/channel/coze/relay-coze.go

@@ -98,7 +98,7 @@ func cozeChatHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Res
 }
 
 func cozeChatStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Response) (*dto.Usage, *types.NewAPIError) {
-	scanner := bufio.NewScanner(resp.Body)
+	scanner := helper.NewStreamScanner(resp.Body)
 	scanner.Split(bufio.ScanLines)
 	helper.SetEventStreamHeaders(c)
 	id := helper.GetResponseID(c)

relay/channel/dify/relay-dify.go

@@ -159,9 +159,14 @@ func requestOpenAI2Dify(c *gin.Context, info *relaycommon.RelayInfo, request dto
 					media := mediaContent.GetImageMedia()
 					var file *DifyFile
 					if media.IsRemoteImage() {
-						file.Type = media.MimeType
-						file.TransferMode = "remote_url"
-						file.URL = media.Url
+						// 修复 #2083: 远程图片分支此前未初始化 file，
+						// 导致 file.Type = ... 触发 nil pointer dereference
+						// 而 panic（500: "invalid memory address or nil pointer dereference"）。
+						file = &DifyFile{
+							Type:         media.MimeType,
+							TransferMode: "remote_url",
+							URL:          media.Url,
+						}
 					} else {
 						file = uploadDifyFile(c, info, difyReq.User, mediaContent)
 					}

relay/channel/ollama/relay-ollama.go

@@ -1,7 +1,6 @@
 package ollama
 
 import (
-	"bufio"
 	"encoding/json"
 	"fmt"
 	"io"
@@ -12,6 +11,7 @@ import (
 	"github.com/QuantumNous/new-api/common"
 	"github.com/QuantumNous/new-api/dto"
 	relaycommon "github.com/QuantumNous/new-api/relay/common"
+	"github.com/QuantumNous/new-api/relay/helper"
 	"github.com/QuantumNous/new-api/service"
 	"github.com/QuantumNous/new-api/types"
 
@@ -397,7 +397,7 @@ func PullOllamaModelStream(baseURL, apiKey, modelName string, progressCallback f
 	}
 
 	// 读取流式响应
-	scanner := bufio.NewScanner(response.Body)
+	scanner := helper.NewStreamScanner(response.Body)
 	successful := false
 	for scanner.Scan() {
 		line := scanner.Text()

relay/channel/ollama/stream.go

@@ -1,7 +1,6 @@
 package ollama
 
 import (
-	"bufio"
 	"encoding/json"
 	"fmt"
 	"io"
@@ -70,7 +69,7 @@ func ollamaStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http
 	defer service.CloseResponseBodyGracefully(resp)
 
 	helper.SetEventStreamHeaders(c)
-	scanner := bufio.NewScanner(resp.Body)
+	scanner := helper.NewStreamScanner(resp.Body)
 	usage := &dto.Usage{}
 	var model = info.UpstreamModelName
 	var responseId = common.GetUUID()

relay/channel/openai/adaptor.go

@@ -310,18 +310,20 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
 		}
 
 	}
-	if strings.HasPrefix(info.UpstreamModelName, "o") || strings.HasPrefix(info.UpstreamModelName, "gpt-5") {
+	isOModel := dto.IsOpenAIReasoningOModel(info.UpstreamModelName)
+	isGPT5Model := dto.IsOpenAIGPT5Model(info.UpstreamModelName)
+	if isOModel || isGPT5Model {
 		if lo.FromPtrOr(request.MaxCompletionTokens, uint(0)) == 0 && lo.FromPtrOr(request.MaxTokens, uint(0)) != 0 {
 			request.MaxCompletionTokens = request.MaxTokens
 			request.MaxTokens = nil
 		}
 
-		if strings.HasPrefix(info.UpstreamModelName, "o") {
+		if isOModel {
 			request.Temperature = nil
 		}
 
 		// gpt-5系列模型适配 归零不再支持的参数
-		if strings.HasPrefix(info.UpstreamModelName, "gpt-5") {
+		if isGPT5Model {
 			request.Temperature = nil
 			request.TopP = nil
 			request.LogProbs = nil

relay/channel/tencent/relay-tencent.go

@@ -92,7 +92,7 @@ func streamResponseTencent2OpenAI(TencentResponse *TencentChatResponse) *dto.Cha
 
 func tencentStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Response) (*dto.Usage, *types.NewAPIError) {
 	var responseText string
-	scanner := bufio.NewScanner(resp.Body)
+	scanner := helper.NewStreamScanner(resp.Body)
 	scanner.Split(bufio.ScanLines)
 
 	helper.SetEventStreamHeaders(c)

relay/channel/zhipu/relay-zhipu.go

@@ -157,7 +157,7 @@ func streamMetaResponseZhipu2OpenAI(zhipuResponse *ZhipuStreamMetaResponse) (*dt
 
 func zhipuStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.Response) (*dto.Usage, *types.NewAPIError) {
 	var usage *dto.Usage
-	scanner := bufio.NewScanner(resp.Body)
+	scanner := helper.NewStreamScanner(resp.Body)
 	scanner.Split(bufio.ScanLines)
 	dataChan := make(chan string)
 	metaChan := make(chan string)
@@ -180,6 +180,9 @@ func zhipuStreamHandler(c *gin.Context, info *relaycommon.RelayInfo, resp *http.
 				}
 			}
 		}
+		if err := scanner.Err(); err != nil {
+			common.SysLog("error reading stream: " + err.Error())
+		}
 		stopChan <- true
 	}()
 	helper.SetEventStreamHeaders(c)

relay/claude_handler.go

@@ -155,6 +155,7 @@ func ClaudeHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *typ
 		if err != nil {
 			return types.NewErrorWithStatusCode(err, types.ErrorCodeReadRequestBodyFailed, http.StatusBadRequest, types.ErrOptionWithSkipRetry())
 		}
+		info.UpstreamRequestBodySize = storage.Size()
 		requestBody = common.ReaderOnly(storage)
 	} else {
 		convertedRequest, err := adaptor.ConvertClaudeRequest(c, info, request)

relay/helper/stream_scanner.go

@@ -34,6 +34,12 @@ func getScannerBufferSize() int {
 	return DefaultMaxScannerBufferSize
 }
 
+func NewStreamScanner(reader io.Reader) *bufio.Scanner {
+	scanner := bufio.NewScanner(reader)
+	scanner.Buffer(make([]byte, InitialScannerBufferSize), getScannerBufferSize())
+	return scanner
+}
+
 func StreamScannerHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo, dataHandler func(data string, sr *StreamResult)) {
 
 	if resp == nil || dataHandler == nil {
@@ -54,7 +60,7 @@ func StreamScannerHandler(c *gin.Context, resp *http.Response, info *relaycommon
 
 	var (
 		stopChan   = make(chan bool, 3) // 增加缓冲区避免阻塞
-		scanner    = bufio.NewScanner(resp.Body)
+		scanner    = NewStreamScanner(resp.Body)
 		ticker     = time.NewTicker(streamingTimeout)
 		pingTicker *time.Ticker
 		writeMutex sync.Mutex     // Mutex to protect concurrent writes
@@ -104,7 +110,6 @@ func StreamScannerHandler(c *gin.Context, resp *http.Response, info *relaycommon
 		close(stopChan)
 	}()
 
-	scanner.Buffer(make([]byte, InitialScannerBufferSize), getScannerBufferSize())
 	scanner.Split(bufio.ScanLines)
 	SetEventStreamHeaders(c)
 

relay/helper/stream_scanner_test.go

@@ -1,6 +1,7 @@
 package helper
 
 import (
+	"bufio"
 	"fmt"
 	"io"
 	"net/http"
@@ -81,6 +82,22 @@ func TestStreamScannerHandler_NilInputs(t *testing.T) {
 	StreamScannerHandler(c, &http.Response{Body: io.NopCloser(strings.NewReader(""))}, info, nil)
 }
 
+func TestNewStreamScanner_AllowsLargeStreamLine(t *testing.T) {
+	oldBufferMB := constant.StreamScannerMaxBufferMB
+	constant.StreamScannerMaxBufferMB = 1
+	t.Cleanup(func() {
+		constant.StreamScannerMaxBufferMB = oldBufferMB
+	})
+
+	payload := strings.Repeat("x", 128<<10)
+	scanner := NewStreamScanner(strings.NewReader("data: " + payload + "\n"))
+	scanner.Split(bufio.ScanLines)
+
+	require.True(t, scanner.Scan())
+	assert.Equal(t, "data: "+payload, scanner.Text())
+	require.NoError(t, scanner.Err())
+}
+
 func TestStreamScannerHandler_EmptyBody(t *testing.T) {
 	t.Parallel()
 

router/api-router.go

@@ -17,9 +17,10 @@ func SetApiRouter(router *gin.Engine) {
 	apiRouter.Use(gzip.Gzip(gzip.DefaultCompression))
 	apiRouter.Use(middleware.BodyStorageCleanup()) // 清理请求体存储
 	apiRouter.Use(middleware.GlobalAPIRateLimit())
+	anonymousRequestBodyLimit := middleware.AnonymousRequestBodyLimit()
 	{
 		apiRouter.GET("/setup", controller.GetSetup)
-		apiRouter.POST("/setup", controller.PostSetup)
+		apiRouter.POST("/setup", anonymousRequestBodyLimit, controller.PostSetup)
 		apiRouter.GET("/status", controller.GetStatus)
 		apiRouter.GET("/uptime/status", controller.GetUptimeKumaStatus)
 		apiRouter.GET("/models", middleware.UserAuth(), controller.DashboardListModels)
@@ -40,39 +41,39 @@ func SetApiRouter(router *gin.Engine) {
 		apiRouter.GET("/rankings", middleware.HeaderNavModuleAuth("rankings"), controller.GetRankings)
 		apiRouter.GET("/verification", middleware.EmailVerificationRateLimit(), middleware.TurnstileCheck(), controller.SendEmailVerification)
 		apiRouter.GET("/reset_password", middleware.CriticalRateLimit(), middleware.TurnstileCheck(), controller.SendPasswordResetEmail)
-		apiRouter.POST("/user/reset", middleware.CriticalRateLimit(), controller.ResetPassword)
+		apiRouter.POST("/user/reset", middleware.CriticalRateLimit(), anonymousRequestBodyLimit, controller.ResetPassword)
 		// OAuth routes - specific routes must come before :provider wildcard
 		apiRouter.GET("/oauth/state", middleware.CriticalRateLimit(), controller.GenerateOAuthCode)
-		apiRouter.POST("/oauth/email/bind", middleware.CriticalRateLimit(), controller.EmailBind)
+		apiRouter.POST("/oauth/email/bind", middleware.CriticalRateLimit(), anonymousRequestBodyLimit, controller.EmailBind)
 		// Non-standard OAuth (WeChat, Telegram) - keep original routes
 		apiRouter.GET("/oauth/wechat", middleware.CriticalRateLimit(), controller.WeChatAuth)
-		apiRouter.POST("/oauth/wechat/bind", middleware.CriticalRateLimit(), controller.WeChatBind)
+		apiRouter.POST("/oauth/wechat/bind", middleware.CriticalRateLimit(), anonymousRequestBodyLimit, controller.WeChatBind)
 		apiRouter.GET("/oauth/telegram/login", middleware.CriticalRateLimit(), controller.TelegramLogin)
 		apiRouter.GET("/oauth/telegram/bind", middleware.CriticalRateLimit(), controller.TelegramBind)
 		// Standard OAuth providers (GitHub, Discord, OIDC, LinuxDO) - unified route
 		apiRouter.GET("/oauth/:provider", middleware.CriticalRateLimit(), controller.HandleOAuth)
 		apiRouter.GET("/ratio_config", middleware.CriticalRateLimit(), controller.GetRatioConfig)
 
-		apiRouter.POST("/stripe/webhook", controller.StripeWebhook)
-		apiRouter.POST("/creem/webhook", controller.CreemWebhook)
-		apiRouter.POST("/waffo/webhook", controller.WaffoWebhook)
+		apiRouter.POST("/stripe/webhook", anonymousRequestBodyLimit, controller.StripeWebhook)
+		apiRouter.POST("/creem/webhook", anonymousRequestBodyLimit, controller.CreemWebhook)
+		apiRouter.POST("/waffo/webhook", anonymousRequestBodyLimit, controller.WaffoWebhook)
 		// :env separates test vs prod URLs so the operator can register each
 		// in Pancake's matching webhook slot; handler enforces env match.
-		apiRouter.POST("/waffo-pancake/webhook/:env", controller.WaffoPancakeWebhook)
+		apiRouter.POST("/waffo-pancake/webhook/:env", anonymousRequestBodyLimit, controller.WaffoPancakeWebhook)
 
 		// Universal secure verification routes
 		apiRouter.POST("/verify", middleware.UserAuth(), middleware.CriticalRateLimit(), controller.UniversalVerify)
 
 		userRoute := apiRouter.Group("/user")
 		{
-			userRoute.POST("/register", middleware.CriticalRateLimit(), middleware.TurnstileCheck(), controller.Register)
-			userRoute.POST("/login", middleware.CriticalRateLimit(), middleware.TurnstileCheck(), controller.Login)
-			userRoute.POST("/login/2fa", middleware.CriticalRateLimit(), controller.Verify2FALogin)
-			userRoute.POST("/passkey/login/begin", middleware.CriticalRateLimit(), controller.PasskeyLoginBegin)
-			userRoute.POST("/passkey/login/finish", middleware.CriticalRateLimit(), controller.PasskeyLoginFinish)
+			userRoute.POST("/register", middleware.CriticalRateLimit(), anonymousRequestBodyLimit, middleware.TurnstileCheck(), controller.Register)
+			userRoute.POST("/login", middleware.CriticalRateLimit(), anonymousRequestBodyLimit, middleware.TurnstileCheck(), controller.Login)
+			userRoute.POST("/login/2fa", middleware.CriticalRateLimit(), anonymousRequestBodyLimit, controller.Verify2FALogin)
+			userRoute.POST("/passkey/login/begin", middleware.CriticalRateLimit(), anonymousRequestBodyLimit, controller.PasskeyLoginBegin)
+			userRoute.POST("/passkey/login/finish", middleware.CriticalRateLimit(), anonymousRequestBodyLimit, controller.PasskeyLoginFinish)
 			//userRoute.POST("/tokenlog", middleware.CriticalRateLimit(), controller.TokenLog)
 			userRoute.GET("/logout", controller.Logout)
-			userRoute.POST("/epay/notify", controller.EpayNotify)
+			userRoute.POST("/epay/notify", anonymousRequestBodyLimit, controller.EpayNotify)
 			userRoute.GET("/epay/notify", controller.EpayNotify)
 			userRoute.GET("/groups", controller.GetUserGroups)
 
@@ -176,10 +177,10 @@ func SetApiRouter(router *gin.Engine) {
 		}
 
 		// Subscription payment callbacks (no auth)
-		apiRouter.POST("/subscription/epay/notify", controller.SubscriptionEpayNotify)
+		apiRouter.POST("/subscription/epay/notify", anonymousRequestBodyLimit, controller.SubscriptionEpayNotify)
 		apiRouter.GET("/subscription/epay/notify", controller.SubscriptionEpayNotify)
 		apiRouter.GET("/subscription/epay/return", controller.SubscriptionEpayReturn)
-		apiRouter.POST("/subscription/epay/return", controller.SubscriptionEpayReturn)
+		apiRouter.POST("/subscription/epay/return", anonymousRequestBodyLimit, controller.SubscriptionEpayReturn)
 		optionRoute := apiRouter.Group("/option")
 		optionRoute.Use(middleware.RootAuth())
 		{

service/channel_affinity.go

@@ -641,6 +641,38 @@ func ShouldSkipRetryAfterChannelAffinityFailure(c *gin.Context) bool {
 	return meta.SkipRetry
 }
 
+func ClearCurrentChannelAffinityCache(c *gin.Context) bool {
+	if c == nil {
+		return false
+	}
+	cacheKey, _, ok := getChannelAffinityContext(c)
+	if !ok || cacheKey == "" {
+		return false
+	}
+
+	cache := getChannelAffinityCache()
+	deleted, err := cache.DeleteMany([]string{cacheKey})
+	if err != nil {
+		common.SysError(fmt.Sprintf("channel affinity cache delete current failed: err=%v", err))
+		return false
+	}
+	c.Set(ginKeyChannelAffinitySkipRetry, false)
+	for _, ok := range deleted {
+		if ok {
+			return true
+		}
+	}
+	return false
+}
+
+func ShouldKeepChannelAffinityOnChannelDisabled() bool {
+	setting := operation_setting.GetChannelAffinitySetting()
+	if setting == nil {
+		return false
+	}
+	return setting.KeepOnChannelDisabled
+}
+
 func MarkChannelAffinityUsed(c *gin.Context, selectedGroup string, channelID int) {
 	if c == nil || channelID <= 0 {
 		return

service/channel_affinity_template_test.go

@@ -236,6 +236,33 @@ func TestGetPreferredChannelByAffinity_RequestHeaderKeySource(t *testing.T) {
 	require.Equal(t, buildChannelAffinityKeyHint(affinityValue), meta.KeyHint)
 }
 
+func TestClearCurrentChannelAffinityCache(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+
+	cacheKeySuffix := fmt.Sprintf("codex cli trace:default:clear-current-%d", time.Now().UnixNano())
+	cacheKeyFull := channelAffinityCacheNamespace + ":" + cacheKeySuffix
+	cache := getChannelAffinityCache()
+	require.NoError(t, cache.SetWithTTL(cacheKeySuffix, 9527, time.Minute))
+	t.Cleanup(func() {
+		_, _ = cache.DeleteMany([]string{cacheKeySuffix})
+	})
+
+	ctx := buildChannelAffinityTemplateContextForTest(channelAffinityMeta{
+		CacheKey:   cacheKeyFull,
+		TTLSeconds: 60,
+		RuleName:   "codex cli trace",
+		SkipRetry:  true,
+	})
+	require.True(t, ShouldSkipRetryAfterChannelAffinityFailure(ctx))
+
+	deleted := ClearCurrentChannelAffinityCache(ctx)
+	require.True(t, deleted)
+	_, found, err := cache.Get(cacheKeySuffix)
+	require.NoError(t, err)
+	require.False(t, found)
+	require.False(t, ShouldSkipRetryAfterChannelAffinityFailure(ctx))
+}
+
 func TestChannelAffinityHitCodexTemplatePassHeadersEffective(t *testing.T) {
 	gin.SetMode(gin.TestMode)
 

service/http_client.go

@@ -37,6 +37,7 @@ func InitHttpClient() {
 	transport := &http.Transport{
 		MaxIdleConns:        common.RelayMaxIdleConns,
 		MaxIdleConnsPerHost: common.RelayMaxIdleConnsPerHost,
+		IdleConnTimeout:     time.Duration(common.RelayIdleConnTimeout) * time.Second,
 		ForceAttemptHTTP2:   true,
 		Proxy:               http.ProxyFromEnvironment, // Support HTTP_PROXY, HTTPS_PROXY, NO_PROXY env vars
 	}
@@ -108,6 +109,7 @@ func NewProxyHttpClient(proxyURL string) (*http.Client, error) {
 		transport := &http.Transport{
 			MaxIdleConns:        common.RelayMaxIdleConns,
 			MaxIdleConnsPerHost: common.RelayMaxIdleConnsPerHost,
+			IdleConnTimeout:     time.Duration(common.RelayIdleConnTimeout) * time.Second,
 			ForceAttemptHTTP2:   true,
 			Proxy:               http.ProxyURL(parsedURL),
 		}
@@ -147,6 +149,7 @@ func NewProxyHttpClient(proxyURL string) (*http.Client, error) {
 		transport := &http.Transport{
 			MaxIdleConns:        common.RelayMaxIdleConns,
 			MaxIdleConnsPerHost: common.RelayMaxIdleConnsPerHost,
+			IdleConnTimeout:     time.Duration(common.RelayIdleConnTimeout) * time.Second,
 			ForceAttemptHTTP2:   true,
 			DialContext: func(ctx context.Context, network, addr string) (net.Conn, error) {
 				return dialer.Dial(network, addr)

setting/operation_setting/channel_affinity_setting.go

@@ -28,11 +28,12 @@ type ChannelAffinityRule struct {
 }
 
 type ChannelAffinitySetting struct {
-	Enabled           bool                  `json:"enabled"`
-	SwitchOnSuccess   bool                  `json:"switch_on_success"`
-	MaxEntries        int                   `json:"max_entries"`
-	DefaultTTLSeconds int                   `json:"default_ttl_seconds"`
-	Rules             []ChannelAffinityRule `json:"rules"`
+	Enabled               bool                  `json:"enabled"`
+	SwitchOnSuccess       bool                  `json:"switch_on_success"`
+	KeepOnChannelDisabled bool                  `json:"keep_on_channel_disabled"`
+	MaxEntries            int                   `json:"max_entries"`
+	DefaultTTLSeconds     int                   `json:"default_ttl_seconds"`
+	Rules                 []ChannelAffinityRule `json:"rules"`
 }
 
 var codexCliPassThroughHeaders = []string{
@@ -74,10 +75,11 @@ func buildPassHeaderTemplate(headers []string) map[string]interface{} {
 }
 
 var channelAffinitySetting = ChannelAffinitySetting{
-	Enabled:           true,
-	SwitchOnSuccess:   true,
-	MaxEntries:        100_000,
-	DefaultTTLSeconds: 3600,
+	Enabled:               true,
+	SwitchOnSuccess:       true,
+	KeepOnChannelDisabled: false,
+	MaxEntries:            100_000,
+	DefaultTTLSeconds:     3600,
 	Rules: []ChannelAffinityRule{
 		{
 			Name:       "codex cli trace",

web/classic/src/helpers/render.jsx

@@ -1068,31 +1068,17 @@ export function getQuotaWithUnit(quota, digits = 6) {
   return (quota / quotaPerUnit).toFixed(digits);
 }
 
+// amount 为系统内部的美元值
 export function renderQuotaWithAmount(amount) {
-  const quotaDisplayType = localStorage.getItem('quota_display_type') || 'USD';
-  if (quotaDisplayType === 'TOKENS') {
+  const { symbol, rate, type } = getCurrencyConfig();
+  if (type === 'TOKENS') {
     return renderNumber(renderUnitWithQuota(amount));
   }
-
   const numericAmount = Number(amount);
-  const formattedAmount = Number.isFinite(numericAmount)
-      ? numericAmount.toFixed(2)
-      : amount;
-
-  if (quotaDisplayType === 'CNY') {
-    return '¥' + formattedAmount;
-  } else if (quotaDisplayType === 'CUSTOM') {
-    const statusStr = localStorage.getItem('status');
-    let symbol = '¤';
-    try {
-      if (statusStr) {
-        const s = JSON.parse(statusStr);
-        symbol = s?.custom_currency_symbol || symbol;
-      }
-    } catch (e) {}
-    return symbol + formattedAmount;
+  if (!Number.isFinite(numericAmount)) {
+    return symbol + amount;
   }
-  return '$' + formattedAmount;
+  return symbol + (numericAmount * rate).toFixed(2);
 }
 
 /**

web/classic/src/i18n/locales/en.json

@@ -1197,6 +1197,7 @@
     "套餐的基本信息和定价": "Basic plan info and pricing",
     "如：大带宽批量分析图片推荐": "e.g. Large bandwidth batch analysis of image recommendations",
     "如：香港线路": "e.g. Hong Kong line",
+    "开启后，亲和到的渠道被禁用，或不再适用于当前分组/模型时，仍保留这条亲和；关闭时会删除并重新选择渠道。": "When enabled, keep the affinity entry even if the affinity channel is disabled or no longer usable for the current group/model. When disabled, the entry will be deleted and another channel will be selected.",
     "如果亲和到的渠道失败，重试到其他渠道成功后，将亲和更新到成功的渠道。": "If the affinity channel fails, after a successful retry on another channel, the affinity will be updated to the successful channel.",
     "如果你对接的是上游One API或者New API等转发项目，请使用OpenAI类型，不要使用此类型，除非你知道你在做什么。": "If you are connecting to upstream One API or New API forwarding projects, please use OpenAI type. Do not use this type unless you know what you are doing.",
     "如果用户请求中包含系统提示词，则使用此设置拼接到用户的系统提示词前面": "If the user request contains a system prompt, this setting will be appended to the user's system prompt",
@@ -1579,6 +1580,7 @@
     "成功": "Success",
     "成功兑换额度：": "Successful redemption amount:",
     "成功后切换亲和": "Switch Affinity on Success",
+    "渠道禁用后保留亲和": "Keep Affinity When Channel Is Disabled",
     "成功时自动启用通道": "Enable channel when successful",
     "我已了解禁用两步验证将永久删除所有相关设置和备用码，此操作不可撤销": "I have understood that disabling two-factor authentication will permanently delete all related settings and backup codes, this operation cannot be undone",
     "我已阅读并同意": "I have read and agree to",

web/classic/src/i18n/locales/fr.json

@@ -1193,6 +1193,7 @@
     "套餐的基本信息和定价": "Informations de base et tarification du plan",
     "如：大带宽批量分析图片推荐": "par exemple, Recommandations d'analyse d'images par lots à large bande passante",
     "如：香港线路": "par exemple, Ligne de Hong Kong",
+    "开启后，亲和到的渠道被禁用，或不再适用于当前分组/模型时，仍保留这条亲和；关闭时会删除并重新选择渠道。": "Lorsque cette option est activée, conserver l'entrée d'affinité même si le canal d'affinité est désactivé ou n'est plus utilisable pour le groupe/modèle actuel. Lorsqu'elle est désactivée, l'entrée sera supprimée et un autre canal sera sélectionné.",
     "如果亲和到的渠道失败，重试到其他渠道成功后，将亲和更新到成功的渠道。": "Si le canal d'affinité échoue, après une nouvelle tentative réussie sur un autre canal, l'affinité sera mise à jour vers le canal réussi.",
     "如果你对接的是上游One API或者New API等转发项目，请使用OpenAI类型，不要使用此类型，除非你知道你在做什么。": "Si vous vous connectez à des projets de redirection One API ou New API en amont, veuillez utiliser le type OpenAI. N'utilisez pas ce type, sauf si vous savez ce que vous faites.",
     "如果用户请求中包含系统提示词，则使用此设置拼接到用户的系统提示词前面": "Si la requête de l'utilisateur contient un prompt système, utilisez ce paramètre pour le concaténer avant le prompt système de l'utilisateur",
@@ -1584,6 +1585,7 @@
     "成功": "Succès",
     "成功兑换额度：": "Montant de l'échange réussi :",
     "成功后切换亲和": "Changer l'affinité en cas de succès",
+    "渠道禁用后保留亲和": "Conserver l'affinité lorsque le canal est désactivé",
     "成功时自动启用通道": "Activer le canal en cas de succès",
     "我已了解禁用两步验证将永久删除所有相关设置和备用码，此操作不可撤销": "J'ai compris que la désactivation de l'authentification à deux facteurs supprimera définitivement tous les paramètres et codes de sauvegarde associés, cette opération ne peut pas être annulée",
     "我已阅读并同意": "J'ai lu et j'accepte",

web/classic/src/i18n/locales/ja.json

@@ -1180,6 +1180,7 @@
     "套餐的基本信息和定价": "プランの基本情報と価格",
     "如：大带宽批量分析图片推荐": "例：広帯域での画像一括分析に推奨",
     "如：香港线路": "例：香港回線",
+    "开启后，亲和到的渠道被禁用，或不再适用于当前分组/模型时，仍保留这条亲和；关闭时会删除并重新选择渠道。": "有効にすると、アフィニティチャネルが無効化された、または現在のグループ/モデルで利用できなくなった場合でも、そのアフィニティエントリを保持します。無効にすると、エントリを削除して別のチャネルを選択します。",
     "如果亲和到的渠道失败，重试到其他渠道成功后，将亲和更新到成功的渠道。": "アフィニティチャネルが失敗した場合、別のチャネルでリトライが成功すると、アフィニティが成功したチャネルに更新されます。",
     "如果你对接的是上游One API或者New API等转发项目，请使用OpenAI类型，不要使用此类型，除非你知道你在做什么。": "New APIなどのリレープロジェクトに接続する場合は、OpenAIタイプを利用してください。設定内容を熟知している場合を除き、このタイプは利用しないでください",
     "如果用户请求中包含系统提示词，则使用此设置拼接到用户的系统提示词前面": "ユーザーリクエストにシステムプロンプトが含まれている場合、この設定内容がユーザーのシステムプロンプトの前に追加されます",
@@ -1555,6 +1556,7 @@
     "成功": "成功",
     "成功兑换额度：": "引き換え額：",
     "成功后切换亲和": "成功時にアフィニティを切り替え",
+    "渠道禁用后保留亲和": "チャネル無効時にアフィニティを保持",
     "成功时自动启用通道": "成功時にチャネルを自動的に有効にする",
     "我已了解禁用两步验证将永久删除所有相关设置和备用码，此操作不可撤销": "2要素認証を無効にすると、すべての関連設定とバックアップコードが永久に削除され、この操作は元に戻すことができないことを理解しました",
     "我已阅读并同意": "読んで同意します",

web/classic/src/i18n/locales/ru.json

@@ -1201,6 +1201,7 @@
     "套餐的基本信息和定价": "Основная информация и цена плана",
     "如：大带宽批量分析图片推荐": "Например: рекомендуется для пакетного анализа изображений с большой пропускной способностью",
     "如：香港线路": "Например: Гонконгская линия",
+    "开启后，亲和到的渠道被禁用，或不再适用于当前分组/模型时，仍保留这条亲和；关闭时会删除并重新选择渠道。": "Если включено, запись аффинити сохраняется, даже когда канал аффинити отключён или больше не подходит для текущей группы/модели. Если выключено, запись будет удалена и выбран другой канал.",
     "如果亲和到的渠道失败，重试到其他渠道成功后，将亲和更新到成功的渠道。": "Если канал аффинити не сработал, после успешного повтора на другом канале аффинити будет обновлена на успешный канал.",
     "如果你对接的是上游One API或者New API等转发项目，请使用OpenAI类型，不要使用此类型，除非你知道你在做什么。": "Если вы интегрируетесь с восходящими проектами пересылки, такими как One API или New API, используйте тип OpenAI, не используйте этот тип, если вы не знаете, что делаете.",
     "如果用户请求中包含系统提示词，则使用此设置拼接到用户的系统提示词前面": "Если запрос пользователя содержит системный промпт, используйте эту настройку для добавления перед системным промптом пользователя",
@@ -1602,6 +1603,7 @@
     "成功": "Успешно",
     "成功兑换额度：": "Успешно обменяно квота: ",
     "成功后切换亲和": "Переключить аффинити при успехе",
+    "渠道禁用后保留亲和": "Сохранять аффинити при отключении канала",
     "成功时自动启用通道": "Автоматически включать канал при успехе",
     "我已了解禁用两步验证将永久删除所有相关设置和备用码，此操作不可撤销": "Я понимаю, что отключение двухфакторной аутентификации приведет к постоянному удалению всех связанных настроек и резервных кодов, и эта операция не может быть отменена",
     "我已阅读并同意": "Я прочитал(а) и согласен(на)",

web/classic/src/i18n/locales/vi.json

@@ -1181,6 +1181,7 @@
     "套餐的基本信息和定价": "Thông tin cơ bản và giá của gói",
     "如：大带宽批量分析图片推荐": "ví dụ: Phân tích hàng loạt băng thông lớn đề xuất hình ảnh",
     "如：香港线路": "ví dụ: Tuyến Hồng Kông",
+    "开启后，亲和到的渠道被禁用，或不再适用于当前分组/模型时，仍保留这条亲和；关闭时会删除并重新选择渠道。": "Khi bật, giữ mục ưu ái ngay cả khi kênh ưu ái bị tắt hoặc không còn dùng được cho nhóm/mô hình hiện tại. Khi tắt, mục đó sẽ bị xóa và kênh khác sẽ được chọn.",
     "如果亲和到的渠道失败，重试到其他渠道成功后，将亲和更新到成功的渠道。": "Nếu kênh ưu ái thất bại, sau khi thử lại thành công trên kênh khác, ưu ái sẽ được cập nhật sang kênh thành công.",
     "如果你对接的是上游One API或者New API等转发项目，请使用OpenAI类型，不要使用此类型，除非你知道你在做什么。": "Nếu bạn đang kết nối với các dự án chuyển tiếp One API hoặc New API thượng nguồn, vui lòng sử dụng loại OpenAI. Đừng sử dụng loại này trừ khi bạn biết mình đang làm gì.",
     "如果用户请求中包含系统提示词，则使用此设置拼接到用户的系统提示词前面": "Nếu yêu cầu của người dùng chứa từ nhắc hệ thống, cài đặt này sẽ được nối vào trước từ nhắc hệ thống của người dùng",
@@ -1556,6 +1557,7 @@
     "成功": "Thành công",
     "成功兑换额度：": "Số tiền đổi thành công:",
     "成功后切换亲和": "Chuyển ưu ái khi thành công",
+    "渠道禁用后保留亲和": "Giữ ưu ái khi kênh bị tắt",
     "成功时自动启用通道": "Bật kênh khi thành công",
     "我已了解禁用两步验证将永久删除所有相关设置和备用码，此操作不可撤销": "Tôi đã hiểu rằng việc vô hiệu hóa xác thực hai yếu tố sẽ xóa vĩnh viễn tất cả các cài đặt liên quan và mã dự phòng, thao tác này không thể hoàn tác",
     "我已阅读并同意": "Tôi đã đọc và đồng ý với",

web/classic/src/i18n/locales/zh-CN.json

@@ -1170,6 +1170,7 @@
     "套餐的基本信息和定价": "套餐的基本信息和定价",
     "如：大带宽批量分析图片推荐": "如：大带宽批量分析图片推荐",
     "如：香港线路": "如：香港线路",
+    "开启后，亲和到的渠道被禁用，或不再适用于当前分组/模型时，仍保留这条亲和；关闭时会删除并重新选择渠道。": "开启后，亲和到的渠道被禁用，或不再适用于当前分组/模型时，仍保留这条亲和；关闭时会删除并重新选择渠道。",
     "如果亲和到的渠道失败，重试到其他渠道成功后，将亲和更新到成功的渠道。": "如果亲和到的渠道失败，重试到其他渠道成功后，将亲和更新到成功的渠道。",
     "如果你对接的是上游One API或者New API等转发项目，请使用OpenAI类型，不要使用此类型，除非你知道你在做什么。": "如果你对接的是上游One API或者New API等转发项目，请使用OpenAI类型，不要使用此类型，除非你知道你在做什么。",
     "如果用户请求中包含系统提示词，则使用此设置拼接到用户的系统提示词前面": "如果用户请求中包含系统提示词，则使用此设置拼接到用户的系统提示词前面",
@@ -1541,6 +1542,7 @@
     "成功": "成功",
     "成功兑换额度：": "成功兑换额度：",
     "成功后切换亲和": "成功后切换亲和",
+    "渠道禁用后保留亲和": "渠道禁用后保留亲和",
     "成功时自动启用通道": "成功时自动启用通道",
     "我已了解禁用两步验证将永久删除所有相关设置和备用码，此操作不可撤销": "我已了解禁用两步验证将永久删除所有相关设置和备用码，此操作不可撤销",
     "我已阅读并同意": "我已阅读并同意",

web/classic/src/i18n/locales/zh-TW.json

@@ -1179,6 +1179,7 @@
     "套餐的基本信息和定价": "訂閱的基本資訊和定價",
     "如：大带宽批量分析图片推荐": "如：大頻寬批量分析圖片推薦",
     "如：香港线路": "如：香港線路",
+    "开启后，亲和到的渠道被禁用，或不再适用于当前分组/模型时，仍保留这条亲和；关闭时会删除并重新选择渠道。": "開啟後，親和到的渠道被停用，或不再適用於目前分組/模型時，仍保留這條親和；關閉時會刪除並重新選擇渠道。",
     "如果亲和到的渠道失败，重试到其他渠道成功后，将亲和更新到成功的渠道。": "",
     "如果你对接的是上游One API或者New API等转发项目，请使用OpenAI类型，不要使用此类型，除非你知道你在做什么。": "如果你對接的是上游One API或者New API等轉發項目，請使用OpenAI類型，不要使用此類型，除非你知道你在做什麼。",
     "如果用户请求中包含系统提示词，则使用此设置拼接到用户的系统提示词前面": "如果使用者請求中包含系統提示詞，則使用此設定拼接到使用者的系統提示詞前面",
@@ -1551,6 +1552,7 @@
     "成功": "成功",
     "成功兑换额度：": "成功兌換額度：",
     "成功后切换亲和": "",
+    "渠道禁用后保留亲和": "渠道停用後保留親和",
     "成功时自动启用通道": "成功時自動啟用通道",
     "我已了解禁用两步验证将永久删除所有相关设置和备用码，此操作不可撤销": "我已瞭解禁用兩步驗證將永久刪除所有相關設定和備用碼，此操作不可撤銷",
     "我已阅读并同意": "我已閱讀並同意",

web/classic/src/pages/Setting/Model/SettingGlobalModel.jsx

@@ -208,7 +208,7 @@ export default function SettingGlobalModel(props) {
             <Row>
               <Col span={24}>
                 <Form.TextArea
-                  label={t('禁用思考处理的模型列表')}
+                  label={t('不自动处理思考后缀的模型列表')}
                   field={'global.thinking_model_blacklist'}
                   placeholder={t('例如：') + '\n' + thinkingExample}
                   rows={4}

web/classic/src/pages/Setting/Operation/SettingsChannelAffinity.jsx

@@ -62,6 +62,8 @@ import ParamOverrideEditorModal from '../../../components/table/channels/modals/
 
 const KEY_ENABLED = 'channel_affinity_setting.enabled';
 const KEY_SWITCH_ON_SUCCESS = 'channel_affinity_setting.switch_on_success';
+const KEY_KEEP_ON_CHANNEL_DISABLED =
+  'channel_affinity_setting.keep_on_channel_disabled';
 const KEY_MAX_ENTRIES = 'channel_affinity_setting.max_entries';
 const KEY_DEFAULT_TTL = 'channel_affinity_setting.default_ttl_seconds';
 const KEY_RULES = 'channel_affinity_setting.rules';
@@ -241,6 +243,7 @@ export default function SettingsChannelAffinity(props) {
   const [inputs, setInputs] = useState({
     [KEY_ENABLED]: false,
     [KEY_SWITCH_ON_SUCCESS]: true,
+    [KEY_KEEP_ON_CHANNEL_DISABLED]: false,
     [KEY_MAX_ENTRIES]: 100000,
     [KEY_DEFAULT_TTL]: 3600,
     [KEY_RULES]: '[]',
@@ -858,6 +861,7 @@ export default function SettingsChannelAffinity(props) {
         ![
           KEY_ENABLED,
           KEY_SWITCH_ON_SUCCESS,
+          KEY_KEEP_ON_CHANNEL_DISABLED,
           KEY_MAX_ENTRIES,
           KEY_DEFAULT_TTL,
           KEY_RULES,
@@ -868,6 +872,8 @@ export default function SettingsChannelAffinity(props) {
         currentInputs[key] = toBoolean(props.options[key]);
       else if (key === KEY_SWITCH_ON_SUCCESS)
         currentInputs[key] = toBoolean(props.options[key]);
+      else if (key === KEY_KEEP_ON_CHANNEL_DISABLED)
+        currentInputs[key] = toBoolean(props.options[key]);
       else if (key === KEY_MAX_ENTRIES)
         currentInputs[key] = Number(props.options[key] || 0) || 0;
       else if (key === KEY_DEFAULT_TTL)
@@ -1003,6 +1009,25 @@ export default function SettingsChannelAffinity(props) {
                   )}
                 </Text>
               </Col>
+              <Col xs={24} sm={12} md={8} lg={8} xl={8}>
+                <Form.Switch
+                  field={KEY_KEEP_ON_CHANNEL_DISABLED}
+                  label={t('渠道禁用后保留亲和')}
+                  checkedText='|'
+                  uncheckedText='O'
+                  onChange={(value) =>
+                    setInputs({
+                      ...inputs,
+                      [KEY_KEEP_ON_CHANNEL_DISABLED]: value,
+                    })
+                  }
+                />
+                <Text type='tertiary' size='small'>
+                  {t(
+                    '开启后，亲和到的渠道被禁用，或不再适用于当前分组/模型时，仍保留这条亲和；关闭时会删除并重新选择渠道。',
+                  )}
+                </Text>
+              </Col>
             </Row>
 
             <Divider style={{ marginTop: 12, marginBottom: 12 }} />

web/default/src/features/models/components/drawers/model-mutate-drawer.tsx

@@ -196,6 +196,7 @@ export function ModelMutateDrawer({
       'grok.violation_deduction_amount': 0,
       'channel_affinity_setting.enabled': false,
       'channel_affinity_setting.switch_on_success': true,
+      'channel_affinity_setting.keep_on_channel_disabled': false,
       'channel_affinity_setting.max_entries': 100000,
       'channel_affinity_setting.default_ttl_seconds': 3600,
       'channel_affinity_setting.rules': '[]',

web/default/src/features/pricing/index.tsx

@@ -174,9 +174,6 @@ export function Pricing() {
         />
         <PageTransition className='relative mx-auto w-full max-w-[1800px] px-3 pt-16 pb-8 sm:px-6 sm:pt-20 sm:pb-10 xl:px-8'>
           <header className='mx-auto mb-5 max-w-3xl pt-5 text-center sm:mb-10 sm:pt-10'>
-            <p className='text-muted-foreground mb-3 text-xs font-medium tracking-widest uppercase'>
-              {t('Models Directory')}
-            </p>
             <h1 className='text-[clamp(2rem,5.5vw,3.5rem)] leading-[1.15] font-bold tracking-tight'>
               {t('Model Square')}
             </h1>

web/default/src/features/profile/components/profile-header.tsx

@@ -118,6 +118,11 @@ export function ProfileHeader({ profile, loading }: ProfileHeaderProps) {
                 variant='neutral'
                 copyable={false}
               />
+              <StatusBadge
+                label={`${t('User ID')} ${profile.id}`}
+                variant='info'
+                copyText={String(profile.id)}
+              />
             </div>
 
             <div className='text-muted-foreground flex flex-wrap items-center gap-x-2 gap-y-0.5 text-xs sm:gap-x-4 sm:text-sm'>

web/default/src/features/rankings/components/rankings-hero.tsx

@@ -43,9 +43,6 @@ export function RankingsHero(props: RankingsHeroProps) {
   return (
     <section className='space-y-5'>
       <div className='space-y-2'>
-        <p className='text-muted-foreground text-xs font-medium tracking-widest uppercase'>
-          {t('Leaderboards')}
-        </p>
         <h1 className='text-[clamp(1.75rem,4vw,2.5rem)] leading-[1.15] font-bold tracking-tight'>
           {t('Rankings')}
         </h1>

web/default/src/features/system-settings/general/channel-affinity/index.tsx

@@ -100,6 +100,9 @@ export function ChannelAffinitySection(props: Props) {
   const [switchOnSuccess, setSwitchOnSuccess] = useState(
     props.defaultValues['channel_affinity_setting.switch_on_success']
   )
+  const [keepOnChannelDisabled, setKeepOnChannelDisabled] = useState(
+    props.defaultValues['channel_affinity_setting.keep_on_channel_disabled']
+  )
   const [maxEntries, setMaxEntries] = useState(
     props.defaultValues['channel_affinity_setting.max_entries']
   )
@@ -136,6 +139,9 @@ export function ChannelAffinitySection(props: Props) {
     setSwitchOnSuccess(
       props.defaultValues['channel_affinity_setting.switch_on_success']
     )
+    setKeepOnChannelDisabled(
+      props.defaultValues['channel_affinity_setting.keep_on_channel_disabled']
+    )
     setMaxEntries(props.defaultValues['channel_affinity_setting.max_entries'])
     setDefaultTtl(
       props.defaultValues['channel_affinity_setting.default_ttl_seconds']
@@ -231,6 +237,14 @@ export function ChannelAffinitySection(props: Props) {
           key: 'channel_affinity_setting.switch_on_success',
           value: String(switchOnSuccess),
         })
+      if (
+        keepOnChannelDisabled !==
+        props.defaultValues['channel_affinity_setting.keep_on_channel_disabled']
+      )
+        updates.push({
+          key: 'channel_affinity_setting.keep_on_channel_disabled',
+          value: String(keepOnChannelDisabled),
+        })
       if (
         maxEntries !==
         props.defaultValues['channel_affinity_setting.max_entries']
@@ -397,6 +411,14 @@ export function ChannelAffinitySection(props: Props) {
             'If the affinity channel fails and retry succeeds on another channel, update affinity to the successful channel.'
           )}
         />
+        <SettingsSwitchField
+          checked={keepOnChannelDisabled}
+          onCheckedChange={setKeepOnChannelDisabled}
+          label={t('Keep affinity when channel is disabled')}
+          description={t(
+            'When enabled, keep the affinity entry even if the affinity channel is disabled or no longer usable for the current group/model. Leave it off to delete the entry and select another channel.'
+          )}
+        />
 
         <Separator />
 

web/default/src/features/system-settings/general/channel-affinity/types.ts

@@ -50,6 +50,7 @@ export interface CacheStats {
 export interface ChannelAffinitySettings {
   'channel_affinity_setting.enabled': boolean
   'channel_affinity_setting.switch_on_success': boolean
+  'channel_affinity_setting.keep_on_channel_disabled': boolean
   'channel_affinity_setting.max_entries': number
   'channel_affinity_setting.default_ttl_seconds': number
   'channel_affinity_setting.rules': string

web/default/src/features/system-settings/models/claude-settings-card.tsx

@@ -231,10 +231,10 @@ export function ClaudeSettingsCard({ defaultValues }: ClaudeSettingsCardProps) {
               render={({ field }) => (
                 <SettingsSwitchItem>
                   <SettingsSwitchContent>
-                    <FormLabel>{t('Thinking Adapter')}</FormLabel>
+                    <FormLabel>{t('Thinking Suffix Adapter')}</FormLabel>
                     <FormDescription>
                       {t(
-                        'Translate `-thinking` suffixes into Anthropic native thinking models while keeping pricing predictable.'
+                        'Adapt `-thinking` suffix requests to Anthropic native thinking behavior while keeping billing predictable.'
                       )}
                     </FormDescription>
                   </SettingsSwitchContent>

web/default/src/features/system-settings/models/gemini-settings-card.tsx

@@ -307,7 +307,7 @@ export function GeminiSettingsCard({ defaultValues }: GeminiSettingsCardProps) {
               render={({ field }) => (
                 <SettingsSwitchItem>
                   <SettingsSwitchContent>
-                    <FormLabel>{t('Thinking Adapter')}</FormLabel>
+                    <FormLabel>{t('Thinking Suffix Adapter')}</FormLabel>
                     <FormDescription>
                       {t('Supports `-thinking`, `-thinking-')}
                       {'{{budget}}'}

web/default/src/features/system-settings/models/global-settings-card.tsx

@@ -227,7 +227,9 @@ export function GlobalSettingsCard({ defaultValues }: GlobalSettingsCardProps) {
             name='global.thinking_model_blacklist'
             render={({ field }) => (
               <FormItem>
-                <FormLabel>{t('Disable thinking processing models')}</FormLabel>
+                <FormLabel>
+                  {t('Models that skip thinking suffix processing')}
+                </FormLabel>
                 <FormControl>
                   <Textarea
                     rows={4}

web/default/src/features/system-settings/models/index.tsx

@@ -64,6 +64,7 @@ const defaultModelSettings: ModelSettings = {
   'group_ratio_setting.group_special_usable_group': '{}',
   'channel_affinity_setting.enabled': false,
   'channel_affinity_setting.switch_on_success': true,
+  'channel_affinity_setting.keep_on_channel_disabled': false,
   'channel_affinity_setting.max_entries': 100000,
   'channel_affinity_setting.default_ttl_seconds': 3600,
   'channel_affinity_setting.rules': '[]',

web/default/src/features/system-settings/models/section-registry.tsx

@@ -130,6 +130,8 @@ const MODELS_SECTIONS = [
             settings['channel_affinity_setting.enabled'],
           'channel_affinity_setting.switch_on_success':
             settings['channel_affinity_setting.switch_on_success'],
+          'channel_affinity_setting.keep_on_channel_disabled':
+            settings['channel_affinity_setting.keep_on_channel_disabled'],
           'channel_affinity_setting.max_entries':
             settings['channel_affinity_setting.max_entries'],
           'channel_affinity_setting.default_ttl_seconds':

web/default/src/features/system-settings/types.ts

@@ -176,6 +176,7 @@ export type ModelSettings = {
   'group_ratio_setting.group_special_usable_group': string
   'channel_affinity_setting.enabled': boolean
   'channel_affinity_setting.switch_on_success': boolean
+  'channel_affinity_setting.keep_on_channel_disabled': boolean
   'channel_affinity_setting.max_entries': number
   'channel_affinity_setting.default_ttl_seconds': number
   'channel_affinity_setting.rules': string

web/default/src/features/usage-logs/components/usage-logs-mobile-card.tsx

@@ -183,6 +183,9 @@ function CommonLogsCard<TData>({
 
   const modelCell = cells.get('model_name')
   const quotaCell = cells.get('quota')
+  const rowData = cells.get('created_at')?.row.original as
+    | Record<string, unknown>
+    | undefined
 
   return (
     <div className='space-y-2.5'>
@@ -200,8 +203,8 @@ function CommonLogsCard<TData>({
             {t('Time')}
           </div>
           <MobileLogTimeStatus
-            createdAt={cells.get('created_at')?.row.original?.created_at}
-            type={cells.get('created_at')?.row.original?.type}
+            createdAt={rowData?.created_at}
+            type={rowData?.type}
           />
         </div>
         <SummaryField

web/default/src/i18n/locales/en.json

@@ -134,6 +134,7 @@
     "Actual Amount": "Actual Amount",
     "Actual Model": "Actual Model",
     "Actual Model:": "Actual Model:",
+    "Adapt `-thinking` suffix requests to Anthropic native thinking behavior while keeping billing predictable.": "Adapt `-thinking` suffix requests to Anthropic native thinking behavior while keeping billing predictable.",
     "Add": "Add",
     "Add \"{{value}}\"": "Add \"{{value}}\"",
     "Add {{title}}": "Add {{title}}",
@@ -644,6 +645,7 @@
     "Channel Affinity": "Channel Affinity",
     "Channel affinity reuses the last successful channel based on keys extracted from the request context or JSON body.": "Channel affinity reuses the last successful channel based on keys extracted from the request context or JSON body.",
     "Channel Affinity: Upstream Cache Hit": "Channel Affinity: Upstream Cache Hit",
+    "Keep affinity when channel is disabled": "Keep affinity when channel is disabled",
     "Channel copied successfully": "Channel copied successfully",
     "Channel created successfully": "Channel created successfully",
     "Channel deleted successfully": "Channel deleted successfully",
@@ -1191,7 +1193,6 @@
     "Disable selected channels": "Disable selected channels",
     "Disable selected models": "Disable selected models",
     "Disable store passthrough": "Disable store passthrough",
-    "Disable thinking processing models": "Disable thinking processing models",
     "Disable this key?": "Disable this key?",
     "Disable threshold (seconds)": "Disable threshold (seconds)",
     "Disable Two-Factor Authentication": "Disable Two-Factor Authentication",
@@ -1994,6 +1995,7 @@
     "If connecting to upstream One API or New API relay projects, use OpenAI type instead unless you know what you are doing": "If connecting to upstream One API or New API relay projects, use OpenAI type instead unless you know what you are doing",
     "If default auto group is enabled, newly created tokens start with auto instead of an empty group.": "If default auto group is enabled, newly created tokens start with auto instead of an empty group.",
     "If the affinity channel fails and retry succeeds on another channel, update affinity to the successful channel.": "If the affinity channel fails and retry succeeds on another channel, update affinity to the successful channel.",
+    "When enabled, keep the affinity entry even if the affinity channel is disabled or no longer usable for the current group/model. Leave it off to delete the entry and select another channel.": "When enabled, keep the affinity entry even if the affinity channel is disabled or no longer usable for the current group/model. Leave it off to delete the entry and select another channel.",
     "If this keeps happening, please report it on GitHub Issues.": "If this keeps happening, please report it on GitHub Issues.",
     "If you provide generative AI services to the public in mainland China, you will fulfill legal obligations including filing, security assessment, content safety, complaint handling, generated content labeling, log retention, and personal information protection.": "If you provide generative AI services to the public in mainland China, you will fulfill legal obligations including filing, security assessment, content safety, complaint handling, generated content labeling, log retention, and personal information protection.",
     "Ignored upstream models": "Ignored upstream models",
@@ -2389,6 +2391,7 @@
     "Models losing the most positions": "Models losing the most positions",
     "Models not in list, may fail to invoke": "Models not in list, may fail to invoke",
     "Models that are being used but not configured in the system": "Models that are being used but not configured in the system",
+    "Models that skip thinking suffix processing": "Models that skip thinking suffix processing",
     "Models updated successfully": "Models updated successfully",
     "Modify existing subscription plan configuration": "Modify existing subscription plan configuration",
     "Module availability": "Module availability",
@@ -3926,7 +3929,7 @@
     "There are both add and remove models pending, but you only selected one type. Confirm submitting only the selected items?": "There are both add and remove models pending, but you only selected one type. Confirm submitting only the selected items?",
     "These models are still in your selection but were not returned by the upstream listing. Entries that are only model_mapping source aliases are omitted. Toggle to adjust before saving.": "These models are still in your selection but were not returned by the upstream listing. Entries that are only model_mapping source aliases are omitted. Toggle to adjust before saving.",
     "These toggles affect whether certain request fields are passed through to the upstream provider.": "These toggles affect whether certain request fields are passed through to the upstream provider.",
-    "Thinking Adapter": "Thinking Adapter",
+    "Thinking Suffix Adapter": "Thinking Suffix Adapter",
     "Thinking to Content": "Thinking to Content",
     "Third-party account bindings (read-only, managed by user in profile settings)": "Third-party account bindings (read-only, managed by user in profile settings)",
     "Third-party Payment Config": "Third-party Payment Config",
@@ -4105,7 +4108,6 @@
     "Transfer Rewards": "Transfer Rewards",
     "Transfer successful": "Transfer successful",
     "Transfer to Balance": "Transfer to Balance",
-    "Translate `-thinking` suffixes into Anthropic native thinking models while keeping pricing predictable.": "Translate `-thinking` suffixes into Anthropic native thinking models while keeping pricing predictable.",
     "Translation": "Translation",
     "Transparent Billing": "Transparent Billing",
     "Trend": "Trend",

web/default/src/i18n/locales/fr.json

@@ -134,6 +134,7 @@
     "Actual Amount": "Montant réel",
     "Actual Model": "Modèle réel",
     "Actual Model:": "Modèle réel :",
+    "Adapt `-thinking` suffix requests to Anthropic native thinking behavior while keeping billing predictable.": "Adapter les requêtes avec le suffixe `-thinking` au comportement de pensée natif d’Anthropic tout en gardant une facturation prévisible.",
     "Add": "Ajouter",
     "Add \"{{value}}\"": "Ajouter \"{{value}}\"",
     "Add {{title}}": "Ajouter {{title}}",
@@ -644,6 +645,7 @@
     "Channel Affinity": "Affinité de canal",
     "Channel affinity reuses the last successful channel based on keys extracted from the request context or JSON body.": "L'affinité de canal réutilise le dernier canal ayant réussi, en se basant sur les clés extraites du contexte de la requête ou du corps JSON.",
     "Channel Affinity: Upstream Cache Hit": "Affinité de canal : hit de cache en amont",
+    "Keep affinity when channel is disabled": "Conserver l'affinité lorsque le canal est désactivé",
     "Channel copied successfully": "Canal copié avec succès",
     "Channel created successfully": "Canal créé avec succès",
     "Channel deleted successfully": "Canal supprimé avec succès",
@@ -1191,7 +1193,6 @@
     "Disable selected channels": "Désactiver les canaux sélectionnés",
     "Disable selected models": "Désactiver les modèles sélectionnés",
     "Disable store passthrough": "Désactiver la transmission du champ store",
-    "Disable thinking processing models": "Désactiver les modèles de traitement de la pensée",
     "Disable this key?": "Désactiver cette clé ?",
     "Disable threshold (seconds)": "Seuil de désactivation (secondes)",
     "Disable Two-Factor Authentication": "Désactiver l'authentification à deux facteurs",
@@ -1994,6 +1995,7 @@
     "If connecting to upstream One API or New API relay projects, use OpenAI type instead unless you know what you are doing": "Si vous vous connectez à des projets de relais One API ou New API en amont, utilisez le type OpenAI à la place sauf si vous savez ce que vous faites",
     "If default auto group is enabled, newly created tokens start with auto instead of an empty group.": "Si le groupe auto par défaut est activé, les nouveaux jetons commencent avec auto au lieu d’un groupe vide.",
     "If the affinity channel fails and retry succeeds on another channel, update affinity to the successful channel.": "Si le canal affinitaire échoue et qu'une nouvelle tentative réussit sur un autre canal, mettre à jour l'affinité vers le canal ayant réussi.",
+    "When enabled, keep the affinity entry even if the affinity channel is disabled or no longer usable for the current group/model. Leave it off to delete the entry and select another channel.": "Lorsque cette option est activée, conserver l'entrée d'affinité même si le canal affinitaire est désactivé ou n'est plus utilisable pour le groupe/modèle actuel. Laissez-la désactivée pour supprimer l'entrée et sélectionner un autre canal.",
     "If this keeps happening, please report it on GitHub Issues.": "Si cela continue, veuillez le signaler sur GitHub Issues.",
     "If you provide generative AI services to the public in mainland China, you will fulfill legal obligations including filing, security assessment, content safety, complaint handling, generated content labeling, log retention, and personal information protection.": "Si vous fournissez des services d’IA générative au public en Chine continentale, vous remplirez les obligations légales applicables, notamment le dépôt, l’évaluation de sécurité, la sécurité du contenu, le traitement des plaintes, l’étiquetage du contenu généré, la conservation des journaux et la protection des informations personnelles.",
     "Ignored upstream models": "Modèles amont ignorés",
@@ -2389,6 +2391,7 @@
     "Models losing the most positions": "Modèles qui perdent le plus de places",
     "Models not in list, may fail to invoke": "Modèles non listés, peut échouer lors de l'invocation",
     "Models that are being used but not configured in the system": "Modèles utilisés mais non configurés dans le système",
+    "Models that skip thinking suffix processing": "Modèles qui ignorent le traitement des suffixes thinking",
     "Models updated successfully": "Modèles mis à jour avec succès",
     "Modify existing subscription plan configuration": "Modifier la configuration du plan d'abonnement existant",
     "Module availability": "Disponibilité du module",
@@ -3926,7 +3929,7 @@
     "There are both add and remove models pending, but you only selected one type. Confirm submitting only the selected items?": "Il y a à la fois des modèles à ajouter et à supprimer, mais vous n'avez sélectionné qu'un seul type. Confirmer l'envoi uniquement des éléments sélectionnés ?",
     "These models are still in your selection but were not returned by the upstream listing. Entries that are only model_mapping source aliases are omitted. Toggle to adjust before saving.": "Ces modèles restent encore sélectionnés mais ne figurent pas dans la liste renvoyée par l'amont ; les noms qui sont uniquement des clés sources de model_mapping sont exclus. Modifiez la sélection avant d'enregistrer.",
     "These toggles affect whether certain request fields are passed through to the upstream provider.": "Ces bascules déterminent si certains champs de demande sont transmis au fournisseur en amont.",
-    "Thinking Adapter": "Adaptateur de réflexion",
+    "Thinking Suffix Adapter": "Adaptateur de suffixe thinking",
     "Thinking to Content": "Réflexion vers Contenu",
     "Third-party account bindings (read-only, managed by user in profile settings)": "Liaisons de comptes tiers (lecture seule, gérées par l'utilisateur dans les paramètres de profil)",
     "Third-party Payment Config": "Configuration de paiement tiers",
@@ -4105,7 +4108,6 @@
     "Transfer Rewards": "Transférer les récompenses",
     "Transfer successful": "Transfert réussi",
     "Transfer to Balance": "Transférer vers le solde",
-    "Translate `-thinking` suffixes into Anthropic native thinking models while keeping pricing predictable.": "Traduire les suffixes `-thinking` en modèles de réflexion natifs Anthropic tout en gardant une tarification prévisible.",
     "Translation": "Traduction",
     "Transparent Billing": "Facturation transparente",
     "Trend": "Tendance",

web/default/src/i18n/locales/ja.json

@@ -134,6 +134,7 @@
     "Actual Amount": "実際の金額",
     "Actual Model": "実際のモデル",
     "Actual Model:": "実際のモデル：",
+    "Adapt `-thinking` suffix requests to Anthropic native thinking behavior while keeping billing predictable.": "`-thinking` サフィックス付きリクエストを Anthropic ネイティブの思考動作に適配し、課金を予測可能に保ちます。",
     "Add": "追加",
     "Add \"{{value}}\"": "\"{{value}}\" を追加",
     "Add {{title}}": "{{title}}を追加",
@@ -644,6 +645,7 @@
     "Channel Affinity": "チャネルアフィニティ",
     "Channel affinity reuses the last successful channel based on keys extracted from the request context or JSON body.": "チャネルアフィニティは、リクエストコンテキストまたは JSON Body から抽出したキーに基づいて、前回成功したチャネルを優先的に再利用します。",
     "Channel Affinity: Upstream Cache Hit": "チャネルアフィニティ：上流キャッシュヒット",
+    "Keep affinity when channel is disabled": "チャネル無効時にアフィニティを保持",
     "Channel copied successfully": "チャンネルが正常にコピーされました",
     "Channel created successfully": "チャンネルが正常に作成されました",
     "Channel deleted successfully": "チャンネルが正常に削除されました",
@@ -1191,7 +1193,6 @@
     "Disable selected channels": "選択したチャネルを無効にする",
     "Disable selected models": "選択したモデルを無効にする",
     "Disable store passthrough": "ストアパススルーを無効にする",
-    "Disable thinking processing models": "思考処理モデルを無効にする",
     "Disable this key?": "このキーを無効にしますか？",
     "Disable threshold (seconds)": "無効化しきい値（秒）",
     "Disable Two-Factor Authentication": "二要素認証を無効にする",
@@ -1994,6 +1995,7 @@
     "If connecting to upstream One API or New API relay projects, use OpenAI type instead unless you know what you are doing": "上流の One API または New API リレープロジェクトに接続する場合、知っている場合を除き OpenAI タイプを使用してください",
     "If default auto group is enabled, newly created tokens start with auto instead of an empty group.": "デフォルト auto グループを有効にすると、新規トークンは空グループではなく auto で開始します。",
     "If the affinity channel fails and retry succeeds on another channel, update affinity to the successful channel.": "アフィニティチャネルが失敗し、別のチャネルでリトライが成功した場合、アフィニティを成功したチャネルに更新します。",
+    "When enabled, keep the affinity entry even if the affinity channel is disabled or no longer usable for the current group/model. Leave it off to delete the entry and select another channel.": "有効にすると、アフィニティチャネルが無効化された、または現在のグループ/モデルで利用できなくなった場合でも、そのアフィニティエントリを保持します。無効のままにすると、エントリを削除して別のチャネルを選択します。",
     "If this keeps happening, please report it on GitHub Issues.": "この問題が続く場合は、GitHub Issues で報告してください。",
     "If you provide generative AI services to the public in mainland China, you will fulfill legal obligations including filing, security assessment, content safety, complaint handling, generated content labeling, log retention, and personal information protection.": "中国本土で一般向けに生成 AI サービスを提供する場合、届出、セキュリティ評価、コンテンツ安全、苦情対応、生成コンテンツのラベル表示、ログ保存、個人情報保護などの法的義務を履行します。",
     "Ignored upstream models": "無視する上流モデル",
@@ -2389,6 +2391,7 @@
     "Models losing the most positions": "順位を最も落としているモデル",
     "Models not in list, may fail to invoke": "リストにないモデル、呼び出しに失敗する可能性があります",
     "Models that are being used but not configured in the system": "使用されているがシステムに設定されていないモデル",
+    "Models that skip thinking suffix processing": "thinking サフィックス処理をスキップするモデル",
     "Models updated successfully": "モデルが正常に更新されました",
     "Modify existing subscription plan configuration": "既存のサブスクリプションプラン設定を変更",
     "Module availability": "モジュールの可用性",
@@ -3926,7 +3929,7 @@
     "There are both add and remove models pending, but you only selected one type. Confirm submitting only the selected items?": "追加と削除の両方のモデルが保留中ですが、一方のタイプのみ選択されています。選択した項目のみ送信してよろしいですか？",
     "These models are still in your selection but were not returned by the upstream listing. Entries that are only model_mapping source aliases are omitted. Toggle to adjust before saving.": "これらはまだ選択中ですが上流のリストにありません。model_mapping にのみソース別名として載る名前は除外されています。保存前に選択を調整してください。",
     "These toggles affect whether certain request fields are passed through to the upstream provider.": "これらの切り替えは、特定の要求フィールドがアップストリームプロバイダーに渡されるかどうかに影響します。",
-    "Thinking Adapter": "思考アダプター",
+    "Thinking Suffix Adapter": "思考サフィックスアダプター",
     "Thinking to Content": "思考からコンテンツへ",
     "Third-party account bindings (read-only, managed by user in profile settings)": "サードパーティアカウントのバインディング（読み取り専用、プロファイル設定でユーザーが管理）",
     "Third-party Payment Config": "サードパーティ決済設定",
@@ -4105,7 +4108,6 @@
     "Transfer Rewards": "報酬の振替",
     "Transfer successful": "転送が成功しました",
     "Transfer to Balance": "残高への振替",
-    "Translate `-thinking` suffixes into Anthropic native thinking models while keeping pricing predictable.": "`-thinking` サフィックスをAnthropicネイティブの思考モデルに変換し、価格設定を予測可能に保ちます。",
     "Translation": "翻訳",
     "Transparent Billing": "透明性のある請求",
     "Trend": "トレンド",

web/default/src/i18n/locales/ru.json

@@ -134,6 +134,7 @@
     "Actual Amount": "Фактическая сумма",
     "Actual Model": "Фактическая модель",
     "Actual Model:": "Фактическая модель:",
+    "Adapt `-thinking` suffix requests to Anthropic native thinking behavior while keeping billing predictable.": "Адаптировать запросы с суффиксом `-thinking` к собственному режиму размышления Anthropic, сохраняя предсказуемость биллинга.",
     "Add": "Добавить",
     "Add \"{{value}}\"": "Добавить \"{{value}}\"",
     "Add {{title}}": "Добавить {{title}}",
@@ -644,6 +645,7 @@
     "Channel Affinity": "Привязка к каналу",
     "Channel affinity reuses the last successful channel based on keys extracted from the request context or JSON body.": "Привязка к каналу повторно использует последний успешный канал на основе ключей, извлечённых из контекста запроса или тела JSON.",
     "Channel Affinity: Upstream Cache Hit": "Привязка к каналу: попадание в кэш upstream",
+    "Keep affinity when channel is disabled": "Сохранять привязку при отключении канала",
     "Channel copied successfully": "Канал успешно скопирован",
     "Channel created successfully": "Канал успешно создан",
     "Channel deleted successfully": "Канал успешно удалён",
@@ -1191,7 +1193,6 @@
     "Disable selected channels": "Отключить выбранные каналы",
     "Disable selected models": "Отключить выбранные модели",
     "Disable store passthrough": "Отключить сквозной переход магазина",
-    "Disable thinking processing models": "Отключить модели с обработкой размышлений",
     "Disable this key?": "Отключить этот ключ?",
     "Disable threshold (seconds)": "Порог отключения (секунды)",
     "Disable Two-Factor Authentication": "Отключить двухфакторную аутентификацию",
@@ -1994,6 +1995,7 @@
     "If connecting to upstream One API or New API relay projects, use OpenAI type instead unless you know what you are doing": "При подключении к upstream One API или проектам-ретрансляторам New API используйте тип OpenAI, если только вы точно знаете, что делаете",
     "If default auto group is enabled, newly created tokens start with auto instead of an empty group.": "Если группа auto включена по умолчанию, новые токены создаются с auto вместо пустой группы.",
     "If the affinity channel fails and retry succeeds on another channel, update affinity to the successful channel.": "Если привязанный канал не работает и повторная попытка удалась через другой канал, привязка обновляется на успешный канал.",
+    "When enabled, keep the affinity entry even if the affinity channel is disabled or no longer usable for the current group/model. Leave it off to delete the entry and select another channel.": "Если включено, запись привязки сохраняется, даже когда привязанный канал отключён или больше не подходит для текущей группы/модели. Оставьте выключенным, чтобы удалять запись и выбирать другой канал.",
     "If this keeps happening, please report it on GitHub Issues.": "Если проблема повторяется, сообщите о ней в GitHub Issues.",
     "If you provide generative AI services to the public in mainland China, you will fulfill legal obligations including filing, security assessment, content safety, complaint handling, generated content labeling, log retention, and personal information protection.": "Если вы предоставляете услуги генеративного ИИ населению материкового Китая, вы будете выполнять юридические обязанности, включая регистрацию, оценку безопасности, безопасность контента, обработку жалоб, маркировку сгенерированного контента, хранение журналов и защиту персональных данных.",
     "Ignored upstream models": "Игнорируемые upstream-модели",
@@ -2389,6 +2391,7 @@
     "Models losing the most positions": "Модели, потерявшие больше всего позиций",
     "Models not in list, may fail to invoke": "Модели не в списке, вызов может не сработать",
     "Models that are being used but not configured in the system": "Модели, которые используются, но не настроены в системе",
+    "Models that skip thinking suffix processing": "Модели, пропускающие обработку суффиксов thinking",
     "Models updated successfully": "Модели успешно обновлены",
     "Modify existing subscription plan configuration": "Изменить конфигурацию существующего плана",
     "Module availability": "Доступность модуля",
@@ -3926,7 +3929,7 @@
     "There are both add and remove models pending, but you only selected one type. Confirm submitting only the selected items?": "Есть модели для добавления и удаления, но вы выбрали только один тип. Подтвердить отправку только выбранных элементов?",
     "These models are still in your selection but were not returned by the upstream listing. Entries that are only model_mapping source aliases are omitted. Toggle to adjust before saving.": "Эти имена всё ещё отмечены в выборе, но не возвращены в списке upstream; ключи только как источники model_mapping исключены. Скорректируйте выбор перед сохранением.",
     "These toggles affect whether certain request fields are passed through to the upstream provider.": "Эти переключатели влияют на то, передаются ли определенные поля запроса вышестоящему поставщику.",
-    "Thinking Adapter": "Адаптер мышления",
+    "Thinking Suffix Adapter": "Адаптер суффикса thinking",
     "Thinking to Content": "Мышление в контент",
     "Third-party account bindings (read-only, managed by user in profile settings)": "Привязки сторонних учетных записей (только для чтения, управляется пользователем в настройках профиля)",
     "Third-party Payment Config": "Настройка стороннего платежа",
@@ -4105,7 +4108,6 @@
     "Transfer Rewards": "Перевести награды",
     "Transfer successful": "Перевод успешен",
     "Transfer to Balance": "Перевести на баланс",
-    "Translate `-thinking` suffixes into Anthropic native thinking models while keeping pricing predictable.": "Преобразовывать суффиксы `-thinking` в собственные модели мышления Anthropic, сохраняя при этом предсказуемость ценообразования.",
     "Translation": "Перевод",
     "Transparent Billing": "Прозрачная тарификация",
     "Trend": "Тренд",

web/default/src/i18n/locales/vi.json

@@ -134,6 +134,7 @@
     "Actual Amount": "Số tiền thực tế",
     "Actual Model": "Mô hình thực tế",
     "Actual Model:": "Mô hình thực tế:",
+    "Adapt `-thinking` suffix requests to Anthropic native thinking behavior while keeping billing predictable.": "Điều chỉnh các yêu cầu có hậu tố `-thinking` sang hành vi suy luận gốc của Anthropic trong khi vẫn giữ tính phí dễ dự đoán.",
     "Add": "Add",
     "Add \"{{value}}\"": "Thêm \"{{value}}\"",
     "Add {{title}}": "Thêm {{title}}",
@@ -644,6 +645,7 @@
     "Channel Affinity": "Ưu tiên kênh",
     "Channel affinity reuses the last successful channel based on keys extracted from the request context or JSON body.": "Ưu tiên kênh sẽ sử dụng lại kênh thành công gần nhất dựa trên các khóa được trích xuất từ ngữ cảnh yêu cầu hoặc JSON body.",
     "Channel Affinity: Upstream Cache Hit": "Ưu tiên kênh: Cache hit từ upstream",
+    "Keep affinity when channel is disabled": "Giữ ưu tiên khi kênh bị tắt",
     "Channel copied successfully": "Sao chép kênh thành công",
     "Channel created successfully": "Tạo kênh thành công",
     "Channel deleted successfully": "Xóa kênh thành công",
@@ -1191,7 +1193,6 @@
     "Disable selected channels": "Vô hiệu hóa các kênh đã chọn",
     "Disable selected models": "Vô hiệu hóa các mô hình đã chọn",
     "Disable store passthrough": "Vô hiệu hóa chuyển tiếp store",
-    "Disable thinking processing models": "Tắt mô hình xử lý suy nghĩ",
     "Disable this key?": "Vô hiệu hóa khóa này?",
     "Disable threshold (seconds)": "Vô hiệu hóa ngưỡng (giây)",
     "Disable Two-Factor Authentication": "Vô hiệu hóa Xác thực hai yếu tố",
@@ -1994,6 +1995,7 @@
     "If connecting to upstream One API or New API relay projects, use OpenAI type instead unless you know what you are doing": "Nếu kết nối với dự án relay One API hoặc New API upstream, hãy sử dụng loại OpenAI thay thế trừ khi bạn biết mình đang làm gì",
     "If default auto group is enabled, newly created tokens start with auto instead of an empty group.": "Nếu bật nhóm auto mặc định, token mới sẽ bắt đầu với auto thay vì nhóm trống.",
     "If the affinity channel fails and retry succeeds on another channel, update affinity to the successful channel.": "Nếu kênh ưu tiên thất bại và thử lại thành công trên kênh khác, cập nhật ưu tiên sang kênh thành công.",
+    "When enabled, keep the affinity entry even if the affinity channel is disabled or no longer usable for the current group/model. Leave it off to delete the entry and select another channel.": "Khi bật, giữ mục ưu tiên ngay cả khi kênh ưu tiên bị tắt hoặc không còn dùng được cho nhóm/mô hình hiện tại. Để tắt để xóa mục đó và chọn kênh khác.",
     "If this keeps happening, please report it on GitHub Issues.": "Nếu sự cố tiếp tục xảy ra, vui lòng báo cáo trên GitHub Issues.",
     "If you provide generative AI services to the public in mainland China, you will fulfill legal obligations including filing, security assessment, content safety, complaint handling, generated content labeling, log retention, and personal information protection.": "Nếu bạn cung cấp dịch vụ AI tạo sinh cho công chúng tại Trung Quốc đại lục, bạn sẽ thực hiện các nghĩa vụ pháp lý bao gồm đăng ký, đánh giá an toàn, an toàn nội dung, xử lý khiếu nại, gắn nhãn nội dung được tạo, lưu giữ nhật ký và bảo vệ thông tin cá nhân.",
     "Ignored upstream models": "Mô hình upstream bị bỏ qua",
@@ -2389,6 +2391,7 @@
     "Models losing the most positions": "Mô hình tụt hạng nhiều nhất",
     "Models not in list, may fail to invoke": "Các mô hình không có trong danh sách, có thể gọi thất bại",
     "Models that are being used but not configured in the system": "Mô hình đang được sử dụng nhưng chưa được cấu hình trong hệ thống",
+    "Models that skip thinking suffix processing": "Mô hình bỏ qua xử lý hậu tố thinking",
     "Models updated successfully": "Mô hình đã được cập nhật thành công",
     "Modify existing subscription plan configuration": "Sửa đổi cấu hình gói đăng ký hiện có",
     "Module availability": "Khả dụng của mô-đun",
@@ -3926,7 +3929,7 @@
     "There are both add and remove models pending, but you only selected one type. Confirm submitting only the selected items?": "Có cả mô hình cần thêm và xóa đang chờ, nhưng bạn chỉ chọn một loại. Xác nhận chỉ gửi các mục đã chọn?",
     "These models are still in your selection but were not returned by the upstream listing. Entries that are only model_mapping source aliases are omitted. Toggle to adjust before saving.": "Các model này vẫn được chọn nhưng không còn xuất hiện trong danh sách upstream; tên chỉ là khóa nguồn trong model_mapping đã được loại. Điều chỉnh trước khi lưu.",
     "These toggles affect whether certain request fields are passed through to the upstream provider.": "Các chuyển đổi này ảnh hưởng đến việc các trường yêu cầu nhất định có được chuyển đến nhà cung cấp dịch vụ đầu vào hay không.",
-    "Thinking Adapter": "Adapter tư duy",
+    "Thinking Suffix Adapter": "Adapter hậu tố thinking",
     "Thinking to Content": "Suy nghĩ thành Nội dung",
     "Third-party account bindings (read-only, managed by user in profile settings)": "Liên kết tài khoản bên thứ ba (chỉ đọc, do người dùng quản lý trong cài đặt hồ sơ)",
     "Third-party Payment Config": "Cấu hình thanh toán bên thứ ba",
@@ -4105,7 +4108,6 @@
     "Transfer Rewards": "Chuyển thưởng",
     "Transfer successful": "Chuyển thành công",
     "Transfer to Balance": "Chuyển vào số dư",
-    "Translate `-thinking` suffixes into Anthropic native thinking models while keeping pricing predictable.": "Dịch các hậu tố `-thinking` sang các mô hình tư duy gốc của Anthropic đồng thời giữ giá cả có thể dự đoán được.",
     "Translation": "Dịch thuật",
     "Transparent Billing": "Thanh toán minh bạch",
     "Trend": "Xu hướng",

web/default/src/i18n/locales/zh.json

@@ -134,6 +134,7 @@
     "Actual Amount": "实付金额",
     "Actual Model": "实际模型",
     "Actual Model:": "实际模型：",
+    "Adapt `-thinking` suffix requests to Anthropic native thinking behavior while keeping billing predictable.": "将带 `-thinking` 后缀的请求适配为 Anthropic 原生思考请求，并保持计费可预测。",
     "Add": "添加",
     "Add \"{{value}}\"": "添加 \"{{value}}\"",
     "Add {{title}}": "添加{{title}}",
@@ -644,6 +645,7 @@
     "Channel Affinity": "渠道亲和性",
     "Channel affinity reuses the last successful channel based on keys extracted from the request context or JSON body.": "渠道亲和性会基于从请求上下文或 JSON Body 提取的 Key，优先复用上一次成功的渠道。",
     "Channel Affinity: Upstream Cache Hit": "渠道亲和性：上游缓存命中",
+    "Keep affinity when channel is disabled": "渠道禁用后保留亲和",
     "Channel copied successfully": "渠道复制成功",
     "Channel created successfully": "渠道创建成功",
     "Channel deleted successfully": "渠道删除成功",
@@ -1191,7 +1193,6 @@
     "Disable selected channels": "禁用选定的渠道",
     "Disable selected models": "禁用选定的模型",
     "Disable store passthrough": "禁止透传 store",
-    "Disable thinking processing models": "禁用思考处理模型",
     "Disable this key?": "禁用此密钥？",
     "Disable threshold (seconds)": "禁用阈值（秒）",
     "Disable Two-Factor Authentication": "禁用双重身份验证",
@@ -1994,6 +1995,7 @@
     "If connecting to upstream One API or New API relay projects, use OpenAI type instead unless you know what you are doing": "如果连接上游 One API 或 New API 中继项目，除非您知道自己在做什么，否则请使用 OpenAI 类型",
     "If default auto group is enabled, newly created tokens start with auto instead of an empty group.": "如果启用默认 auto 分组，新建令牌会默认使用 auto，而不是空分组。",
     "If the affinity channel fails and retry succeeds on another channel, update affinity to the successful channel.": "如果亲和到的渠道失败，重试到其他渠道成功后，将亲和更新到成功的渠道。",
+    "When enabled, keep the affinity entry even if the affinity channel is disabled or no longer usable for the current group/model. Leave it off to delete the entry and select another channel.": "开启后，亲和到的渠道被禁用，或不再适用于当前分组/模型时，仍保留这条亲和；关闭时会删除并重新选择渠道。",
     "If this keeps happening, please report it on GitHub Issues.": "如果问题持续出现，请到 GitHub Issues 反馈。",
     "If you provide generative AI services to the public in mainland China, you will fulfill legal obligations including filing, security assessment, content safety, complaint handling, generated content labeling, log retention, and personal information protection.": "如果你在中国大陆向公众提供生成式人工智能服务，你将履行备案、安全评估、内容安全、投诉处理、生成内容标识、日志留存和个人信息保护等法律义务。",
     "Ignored upstream models": "已忽略上游模型",
@@ -2389,6 +2391,7 @@
     "Models losing the most positions": "名次下滑最多的模型",
     "Models not in list, may fail to invoke": "模型未加入列表，可能无法调用",
     "Models that are being used but not configured in the system": "正在使用但未在系统中配置的模型",
+    "Models that skip thinking suffix processing": "不自动处理思考后缀的模型",
     "Models updated successfully": "模型更新成功",
     "Modify existing subscription plan configuration": "修改现有订阅套餐的配置",
     "Module availability": "模块可用性",
@@ -3926,7 +3929,7 @@
     "There are both add and remove models pending, but you only selected one type. Confirm submitting only the selected items?": "当前有新增和删除两类待处理模型，但您只勾选了其中一类。确认仅提交已勾选的部分吗？",
     "These models are still in your selection but were not returned by the upstream listing. Entries that are only model_mapping source aliases are omitted. Toggle to adjust before saving.": "这些模型仍然在您的勾选列表中，但上游已不再返回该名称；仅作为 model_mapping 来源键而不会出现在 upstream 列表的别名已从本视图排除，请在保存前调整勾选。",
     "These toggles affect whether certain request fields are passed through to the upstream provider.": "这些开关控制某些请求字段是否透传到上游服务。",
-    "Thinking Adapter": "思维适配器",
+    "Thinking Suffix Adapter": "思考后缀适配器",
     "Thinking to Content": "思维到内容",
     "Third-party account bindings (read-only, managed by user in profile settings)": "第三方账户绑定（只读，由用户在个人资料设置中管理）",
     "Third-party Payment Config": "第三方支付配置",
@@ -4105,7 +4108,6 @@
     "Transfer Rewards": "转移奖励",
     "Transfer successful": "转账成功",
     "Transfer to Balance": "转移到余额",
-    "Translate `-thinking` suffixes into Anthropic native thinking models while keeping pricing predictable.": "将 `-thinking` 后缀转换为 Anthropic 原生思维模型，同时保持价格可预测性。",
     "Translation": "翻译",
     "Transparent Billing": "透明计费",
     "Trend": "趋势",