feat: sync deployed v1.0.0-rc.38-custom.42

Sync production source including routing, billing, Astra detection and public uptime. Update custom build instructions and license packaging.

relay/common/relay_info.go

@@ -99,6 +99,11 @@ type RelayInfo struct {
 	UsePrice               bool
 	RelayMode              int
 	OriginModelName        string
+	ResponseModel          *ResponseModel
+	ResponseGuard          *ResponseModelGuard
+	AstraDetectionDisabled bool
+	AstraAutoSwitch        bool
+	UpstreamRequests       int
 
 	// BillingModelName is the pricing identity for this request. It is kept
 	// separate from OriginModelName and UpstreamModelName so virtual pricing
@@ -138,7 +143,11 @@ type RelayInfo struct {
 	ForcePreConsume bool
 	// Billing 是计费会话，封装了预扣费/结算/退款的统一生命周期。
 	// 初始免费组可为 nil；若 auto 重试切换到付费组，会在发送前创建。
-	Billing BillingSettler
+	Billing               BillingSettler
+	DurableAttemptBilling bool
+	DurableAttemptID      int64
+	AttemptDetached       bool
+	GuardDeadline         time.Time
 	// BillingSource indicates whether this request is billed from wallet quota or subscription.
 	// "" or "wallet" => wallet; "subscription" => subscription
 	BillingSource string
@@ -176,6 +185,11 @@ type RelayInfo struct {
 	// and again before settlement. Non-nil only when billing mode is "tiered_expr".
 	TieredBillingSnapshot *billingexpr.BillingSnapshot
 	BillingRequestInput   *billingexpr.RequestInput
+	BillingImageCount     *int
+	// ImageRequestCount is the effective quantity sent on the current attempt;
+	// ImageQuotaBeforeGroup is the frozen legacy estimate before request ratios.
+	ImageRequestCount     int
+	ImageQuotaBeforeGroup float64
 
 	Request dto.Request
 
@@ -187,6 +201,10 @@ type RelayInfo struct {
 	FinalRequestRelayFormat types.RelayFormat
 
 	StreamStatus *StreamStatus
+	// PerformanceOutputTokens is captured by settlement and sampled once at
+	// the request boundary, independently of billing success or failure.
+	PerformanceOutputTokens      int64
+	PerformanceBusinessRejection bool
 
 	// convOptions caches the converter settings snapshot (see ConvOptions).
 	convOptions *convmeta.Options
@@ -204,7 +222,36 @@ type RelayInfo struct {
 	*TaskRelayInfo
 }
 
+// UpdateImageCount replaces the billable quantity without changing the frozen
+// request parameters or multiplying the legacy and expression prices together.
+func (info *RelayInfo) UpdateImageCount(count int64) {
+	if info == nil || count <= 0 || count > int64(dto.MaxImageN) {
+		return
+	}
+	if info.PriceData.UsePrice {
+		info.PriceData.AddOtherRatio("n", float64(count))
+	}
+	if info.TieredBillingSnapshot != nil && info.TieredBillingSnapshot.EstimatedImageCount != nil {
+		n := int(count)
+		info.BillingImageCount = &n
+	}
+}
+
+func (info *RelayInfo) RequestedImageCount() int {
+	if info.ImageRequestCount > 0 {
+		return info.ImageRequestCount
+	}
+	if info.TieredBillingSnapshot != nil && info.TieredBillingSnapshot.EstimatedImageCount != nil {
+		return *info.TieredBillingSnapshot.EstimatedImageCount
+	}
+	if count, ok := info.PriceData.OtherRatios()["n"]; ok && count >= 1 && count <= dto.MaxImageN {
+		return int(count)
+	}
+	return 1
+}
+
 func (info *RelayInfo) InitChannelMeta(c *gin.Context) {
+	info.ResponseModel = nil
 	info.FinalRequestRelayFormat = ""
 	info.RequestConversionChain = nil
 	info.InitRequestConversionChain()
@@ -252,6 +299,15 @@ func (info *RelayInfo) InitChannelMeta(c *gin.Context) {
 		channelMeta.ChannelOtherSettings = channelOtherSettings
 	}
 
+	if channelType == constant.ChannelTypeAdvancedCustom &&
+		!channelMeta.ChannelSetting.PassThroughBodyEnabled &&
+		c.Request != nil && c.Request.URL != nil {
+		route, matched := channelMeta.ChannelOtherSettings.AdvancedCustom.MatchPathForModel(c.Request.URL.Path, info.OriginModelName)
+		if matched && route.PassThroughBodyEnabled {
+			channelMeta.ChannelSetting.PassThroughBodyEnabled = true
+		}
+	}
+
 	if streamSupportedChannels[channelMeta.ChannelType] {
 		channelMeta.SupportStreamOptions = true
 	}
@@ -377,6 +433,8 @@ var streamSupportedChannels = map[int]bool{
 	constant.ChannelTypeAdvancedCustom: true,
 	constant.ChannelTypeSub2API:        true,
 	constant.ChannelTypeNewAPI:         true,
+	constant.ChannelTypeVLLM:           true,
+	constant.ChannelTypeSGLang:         true,
 	constant.ChannelTypeTencent:        true,
 }
 
@@ -498,6 +556,9 @@ func reasoningEffortFromRequest(request dto.Request) string {
 		if req != nil && req.GenerationConfig.ThinkingConfig != nil {
 			config := req.GenerationConfig.ThinkingConfig
 			effort = config.ThinkingLevel
+			if canonical, err := kitreasoning.ParseEffort(effort); err == nil {
+				effort = string(canonical)
+			}
 			if effort == "" && config.ThinkingBudget != nil {
 				effort = string(kitreasoning.EffortFromBudget(*config.ThinkingBudget))
 			}
@@ -551,10 +612,12 @@ func genBaseRelayInfo(c *gin.Context, request dto.Request) *RelayInfo {
 
 		OriginModelName: originModelName,
 
-		TokenId:        common.GetContextKeyInt(c, constant.ContextKeyTokenId),
-		TokenKey:       common.GetContextKeyString(c, constant.ContextKeyTokenKey),
-		TokenUnlimited: common.GetContextKeyBool(c, constant.ContextKeyTokenUnlimited),
-		TokenGroup:     tokenGroup,
+		TokenId:                common.GetContextKeyInt(c, constant.ContextKeyTokenId),
+		TokenKey:               common.GetContextKeyString(c, constant.ContextKeyTokenKey),
+		TokenUnlimited:         common.GetContextKeyBool(c, constant.ContextKeyTokenUnlimited),
+		TokenGroup:             tokenGroup,
+		AstraAutoSwitch:        common.GetContextKeyBool(c, constant.ContextKeyTokenAstraAutoSwitch),
+		AstraDetectionDisabled: common.GetContextKeyBool(c, constant.ContextKeyTokenAstraDetectionDisabled),
 
 		isFirstResponse: true,
 		RelayMode:       relayconstant.Path2RelayMode(c.Request.URL.Path),