feat: enhance tiered billing functionality and UI components
- Introduced new fields for billing mode and expression in the Pricing model. - Implemented dynamic pricing breakdown component to display tiered billing details. - Updated various components to support and render tiered billing information. - Enhanced pricing calculation logic to accommodate dynamic pricing scenarios. - Added tests for new billing expression functionalities and UI components.
This commit is contained in:
CaIon
@@ -1,6 +1,7 @@
|
||||
package service
|
||||
|
||||
import (
|
||||
"encoding/base64"
|
||||
"strings"
|
||||
|
||||
"github.com/QuantumNous/new-api/common"
|
||||
@@ -216,41 +217,17 @@ func GenerateMjOtherInfo(relayInfo *relaycommon.RelayInfo, priceData types.Price
|
||||
return other
|
||||
}
|
||||
|
||||
func GenerateTieredOtherInfo(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, result *billingexpr.TieredResult) map[string]interface{} {
|
||||
other := make(map[string]interface{})
|
||||
other["billing_mode"] = "tiered_expr"
|
||||
|
||||
// InjectTieredBillingInfo overlays tiered billing fields onto an existing
|
||||
// module-specific other map. Call this after GenerateTextOtherInfo /
|
||||
// GenerateClaudeOtherInfo / etc. when the request used tiered_expr billing.
|
||||
func InjectTieredBillingInfo(other map[string]interface{}, relayInfo *relaycommon.RelayInfo, result *billingexpr.TieredResult) {
|
||||
snap := relayInfo.TieredBillingSnapshot
|
||||
if snap != nil {
|
||||
other["group_ratio"] = snap.GroupRatio
|
||||
other["expr_hash"] = snap.ExprHash
|
||||
other["estimated_prompt_tokens"] = snap.EstimatedPromptTokens
|
||||
other["estimated_completion_tokens"] = snap.EstimatedCompletionTokens
|
||||
other["estimated_quota_before_group"] = snap.EstimatedQuotaBeforeGroup
|
||||
other["estimated_quota_after_group"] = snap.EstimatedQuotaAfterGroup
|
||||
other["estimated_tier"] = snap.EstimatedTier
|
||||
if snap == nil {
|
||||
return
|
||||
}
|
||||
|
||||
other["billing_mode"] = "tiered_expr"
|
||||
other["expr_b64"] = base64.StdEncoding.EncodeToString([]byte(snap.ExprString))
|
||||
if result != nil {
|
||||
other["actual_quota_before_group"] = result.ActualQuotaBeforeGroup
|
||||
other["actual_quota_after_group"] = result.ActualQuotaAfterGroup
|
||||
other["matched_tier"] = result.MatchedTier
|
||||
other["crossed_tier"] = result.CrossedTier
|
||||
}
|
||||
|
||||
other["frt"] = float64(relayInfo.FirstResponseTime.UnixMilli() - relayInfo.StartTime.UnixMilli())
|
||||
if relayInfo.IsModelMapped {
|
||||
other["is_model_mapped"] = true
|
||||
other["upstream_model_name"] = relayInfo.UpstreamModelName
|
||||
}
|
||||
|
||||
adminInfo := make(map[string]interface{})
|
||||
adminInfo["use_channel"] = ctx.GetStringSlice("use_channel")
|
||||
AppendChannelAffinityAdminInfo(ctx, adminInfo)
|
||||
other["admin_info"] = adminInfo
|
||||
|
||||
appendRequestPath(ctx, relayInfo, other)
|
||||
appendRequestConversionChain(relayInfo, other)
|
||||
appendBillingInfo(relayInfo, other)
|
||||
return other
|
||||
}
|
||||
|
||||
+35
-61
@@ -158,13 +158,13 @@ func PreWssConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, usag
|
||||
func PostWssConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, modelName string,
|
||||
usage *dto.RealtimeUsage, extraContent string) {
|
||||
|
||||
// Tiered billing early return
|
||||
if ok, tieredQuota, tieredResult := TryTieredSettle(relayInfo, billingexpr.TokenParams{
|
||||
var tieredResult *billingexpr.TieredResult
|
||||
tieredOk, tieredQuota, tieredRes := TryTieredSettle(relayInfo, billingexpr.TokenParams{
|
||||
P: float64(usage.InputTokens),
|
||||
C: float64(usage.OutputTokens),
|
||||
}); ok {
|
||||
postConsumeQuotaTieredService(ctx, relayInfo, modelName, usage.InputTokens, usage.OutputTokens, usage.TotalTokens, tieredQuota, tieredResult, extraContent)
|
||||
return
|
||||
})
|
||||
if tieredOk {
|
||||
tieredResult = tieredRes
|
||||
}
|
||||
|
||||
useTimeSeconds := time.Now().Unix() - relayInfo.StartTime.Unix()
|
||||
@@ -200,6 +200,9 @@ func PostWssConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, mod
|
||||
}
|
||||
|
||||
quota := calculateAudioQuota(quotaInfo)
|
||||
if tieredOk {
|
||||
quota = tieredQuota
|
||||
}
|
||||
|
||||
totalTokens := usage.TotalTokens
|
||||
var logContent string
|
||||
@@ -229,6 +232,9 @@ func PostWssConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, mod
|
||||
}
|
||||
other := GenerateWssOtherInfo(ctx, relayInfo, usage, modelRatio, groupRatio,
|
||||
completionRatio.InexactFloat64(), audioRatio.InexactFloat64(), audioCompletionRatio.InexactFloat64(), modelPrice, relayInfo.PriceData.GroupRatioInfo.GroupSpecialRatio)
|
||||
if tieredResult != nil {
|
||||
InjectTieredBillingInfo(other, relayInfo, tieredResult)
|
||||
}
|
||||
model.RecordConsumeLog(ctx, relayInfo.UserId, model.RecordConsumeLogParams{
|
||||
ChannelId: relayInfo.ChannelId,
|
||||
PromptTokens: usage.InputTokens,
|
||||
@@ -250,16 +256,16 @@ func PostClaudeConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
|
||||
ObserveChannelAffinityUsageCacheByRelayFormat(ctx, usage, relayInfo.GetFinalRequestRelayFormat())
|
||||
}
|
||||
|
||||
// Tiered billing early return
|
||||
if ok, tieredQuota, tieredResult := TryTieredSettle(relayInfo, billingexpr.TokenParams{
|
||||
var tieredResult *billingexpr.TieredResult
|
||||
tieredOk, tieredQuota, tieredRes := TryTieredSettle(relayInfo, billingexpr.TokenParams{
|
||||
P: float64(usage.PromptTokens),
|
||||
C: float64(usage.CompletionTokens),
|
||||
CR: float64(usage.PromptTokensDetails.CachedTokens),
|
||||
CC: float64(usage.PromptTokensDetails.CachedCreationTokens - usage.ClaudeCacheCreation1hTokens),
|
||||
CC1h: float64(usage.ClaudeCacheCreation1hTokens),
|
||||
}); ok {
|
||||
postConsumeQuotaTieredService(ctx, relayInfo, relayInfo.OriginModelName, usage.PromptTokens, usage.CompletionTokens, usage.PromptTokens+usage.CompletionTokens, tieredQuota, tieredResult, "")
|
||||
return
|
||||
})
|
||||
if tieredOk {
|
||||
tieredResult = tieredRes
|
||||
}
|
||||
|
||||
useTimeSeconds := time.Now().Unix() - relayInfo.StartTime.Unix()
|
||||
@@ -315,6 +321,9 @@ func PostClaudeConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
|
||||
}
|
||||
|
||||
quota := int(calculateQuota)
|
||||
if tieredOk {
|
||||
quota = tieredQuota
|
||||
}
|
||||
|
||||
totalTokens := promptTokens + completionTokens
|
||||
|
||||
@@ -342,6 +351,9 @@ func PostClaudeConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
|
||||
cacheCreationTokens5m, cacheCreationRatio5m,
|
||||
cacheCreationTokens1h, cacheCreationRatio1h,
|
||||
modelPrice, relayInfo.PriceData.GroupRatioInfo.GroupSpecialRatio)
|
||||
if tieredResult != nil {
|
||||
InjectTieredBillingInfo(other, relayInfo, tieredResult)
|
||||
}
|
||||
model.RecordConsumeLog(ctx, relayInfo.UserId, model.RecordConsumeLogParams{
|
||||
ChannelId: relayInfo.ChannelId,
|
||||
PromptTokens: promptTokens,
|
||||
@@ -382,14 +394,16 @@ func CalcOpenRouterCacheCreateTokens(usage dto.Usage, priceData types.PriceData)
|
||||
|
||||
func PostAudioConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, usage *dto.Usage, extraContent string) {
|
||||
|
||||
// Tiered billing early return
|
||||
if ok, tieredQuota, tieredResult := TryTieredSettle(relayInfo, billingexpr.TokenParams{
|
||||
var tieredResult *billingexpr.TieredResult
|
||||
tieredOk, tieredQuota, tieredRes := TryTieredSettle(relayInfo, billingexpr.TokenParams{
|
||||
P: float64(usage.PromptTokens),
|
||||
C: float64(usage.CompletionTokens),
|
||||
CR: float64(usage.PromptTokensDetails.CachedTokens),
|
||||
}); ok {
|
||||
postConsumeQuotaTieredService(ctx, relayInfo, relayInfo.OriginModelName, usage.PromptTokens, usage.CompletionTokens, usage.TotalTokens, tieredQuota, tieredResult, extraContent)
|
||||
return
|
||||
AI: float64(usage.PromptTokensDetails.AudioTokens),
|
||||
AO: float64(usage.CompletionTokenDetails.AudioTokens),
|
||||
})
|
||||
if tieredOk {
|
||||
tieredResult = tieredRes
|
||||
}
|
||||
|
||||
useTimeSeconds := time.Now().Unix() - relayInfo.StartTime.Unix()
|
||||
@@ -425,6 +439,9 @@ func PostAudioConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, u
|
||||
}
|
||||
|
||||
quota := calculateAudioQuota(quotaInfo)
|
||||
if tieredOk {
|
||||
quota = tieredQuota
|
||||
}
|
||||
|
||||
totalTokens := usage.TotalTokens
|
||||
var logContent string
|
||||
@@ -458,6 +475,9 @@ func PostAudioConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, u
|
||||
}
|
||||
other := GenerateAudioOtherInfo(ctx, relayInfo, usage, modelRatio, groupRatio,
|
||||
completionRatio.InexactFloat64(), audioRatio.InexactFloat64(), audioCompletionRatio.InexactFloat64(), modelPrice, relayInfo.PriceData.GroupRatioInfo.GroupSpecialRatio)
|
||||
if tieredResult != nil {
|
||||
InjectTieredBillingInfo(other, relayInfo, tieredResult)
|
||||
}
|
||||
model.RecordConsumeLog(ctx, relayInfo.UserId, model.RecordConsumeLogParams{
|
||||
ChannelId: relayInfo.ChannelId,
|
||||
PromptTokens: usage.PromptTokens,
|
||||
@@ -640,49 +660,3 @@ func checkAndSendSubscriptionQuotaNotify(relayInfo *relaycommon.RelayInfo) {
|
||||
})
|
||||
}
|
||||
|
||||
func postConsumeQuotaTieredService(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, modelName string,
|
||||
promptTokens, completionTokens, totalTokens, quota int, tieredResult *TieredResultWrapper, extraContent string) {
|
||||
|
||||
useTimeSeconds := time.Now().Unix() - relayInfo.StartTime.Unix()
|
||||
tokenName := ctx.GetString("token_name")
|
||||
groupRatio := relayInfo.PriceData.GroupRatioInfo.GroupRatio
|
||||
|
||||
var logContent string
|
||||
if totalTokens == 0 {
|
||||
quota = 0
|
||||
logContent = "上游没有返回计费信息（可能是上游超时）"
|
||||
logger.LogError(ctx, fmt.Sprintf("tiered billing: total tokens is 0, userId %d, channelId %d, tokenId %d, model %s, pre-consumed %d",
|
||||
relayInfo.UserId, relayInfo.ChannelId, relayInfo.TokenId, modelName, relayInfo.FinalPreConsumedQuota))
|
||||
} else {
|
||||
if groupRatio != 0 && quota == 0 {
|
||||
quota = 1
|
||||
}
|
||||
model.UpdateUserUsedQuotaAndRequestCount(relayInfo.UserId, quota)
|
||||
model.UpdateChannelUsedQuota(relayInfo.ChannelId, quota)
|
||||
}
|
||||
|
||||
if err := SettleBilling(ctx, relayInfo, quota); err != nil {
|
||||
logger.LogError(ctx, "error settling tiered billing: "+err.Error())
|
||||
}
|
||||
|
||||
if extraContent != "" {
|
||||
logContent += extraContent
|
||||
}
|
||||
|
||||
other := GenerateTieredOtherInfo(ctx, relayInfo, tieredResult)
|
||||
|
||||
model.RecordConsumeLog(ctx, relayInfo.UserId, model.RecordConsumeLogParams{
|
||||
ChannelId: relayInfo.ChannelId,
|
||||
PromptTokens: promptTokens,
|
||||
CompletionTokens: completionTokens,
|
||||
ModelName: modelName,
|
||||
TokenName: tokenName,
|
||||
Quota: quota,
|
||||
Content: logContent,
|
||||
TokenId: relayInfo.TokenId,
|
||||
UseTimeSeconds: int(useTimeSeconds),
|
||||
IsStream: relayInfo.IsStream,
|
||||
Group: relayInfo.UsingGroup,
|
||||
Other: other,
|
||||
})
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user