Files
new-api/service/task_billing.go
T
Calcium-Ion 86ac0f7745 refactor: extract protocol conversion layer into standalone relaykit module (#6369)
* test(relayconvert): add golden snapshot matrix and relaykit boundary guard

Phase 0 of the relaykit extraction plan: pin byte-level output of every
registered (from,to) request/response/stream conversion route, and
forbid kit-bound packages from growing host-only imports.

* wip(relayconvert): drop gin.Context from converter signatures; add convmeta draft

Phase 1 in progress: relayconvert now takes context.Context; host media
resolver adapts gin.Context back at the service boundary.

* refactor(relayconvert): decouple converters from RelayInfo, gin, and settings

Phase 1 of the relaykit extraction plan:
- converters now depend on convmeta.Meta (implemented by RelayInfo) instead
  of *relaycommon.RelayInfo; ClaudeConvertInfo and the format guesser move
  to convmeta with aliases left behind
- host settings reach converters via a convmeta.Options snapshot built in
  RelayInfo.ConvOptions; no more model_setting/reasoning global reads inside
  the conversion layer
- effort-suffix helpers move to service/relayconvert/reasoning (old package
  forwards); chat-to-responses upgrade policy moves to service (host routing
  logic, not conversion)
- golden conversion matrix unchanged

* test(relayconvert): tighten boundary — kit packages now free of gin/setting imports

* refactor(dto): drop gin and logger dependencies

Phase 2 (part 1): dto.Request.IsStream now takes *http.Request instead of
*gin.Context (Gemini's impl reads query/path off the std request); dto's
three logger calls become common.SysError. Boundary test allowlist is now
empty — kit-bound packages import no gin/setting/logger/model.

* refactor(kit): extract dependency-free kitutil; dto/types/relayconvert stop importing common

Phase 2 of the relaykit extraction plan:
- new service/relayconvert/kitutil holds the pure helpers the kit needs
  (JSON wrappers, pointer/string/uuid/timestamp utils, MaskSensitiveInfo,
  pluggable LogInfo/LogError hooks, Debug flag)
- dto, types, and all relayconvert packages now use kitutil; their only
  remaining internal deps are dto/types/constant
- common keeps every original symbol (MaskSensitiveInfo delegates to
  kitutil) so host code is untouched; main.go routes kit logging into
  common.SysLog/SysError and mirrors DebugEnabled
- golden conversion matrix unchanged

* refactor(kit): move EndpointType/FinishReason to types; OpenRouter dialect via Options

Kit packages (dto/types/relayconvert/reasonmap) no longer import constant:
- EndpointType and finish-reason values live in types; constant re-exports
- the OpenRouter special-case in claude->openai request conversion reads
  Options.OpenRouterDialect, set by the host from the channel type;
  InitChannelMeta invalidates the cached snapshot on channel switch

* refactor: extract relaykit submodule (dto/types/relayconvert/reasonmap)

Phase 3 of the relaykit extraction plan:
- new go module github.com/QuantumNous/new-api/relaykit containing dto
  (minus task family), types, relayconvert (with convmeta/kitutil/reasoning),
  and reasonmap; host consumes it via require + replace, go.work for dev
- task-family dto (task/suno/midjourney/video) stays in the host dto
  package; dual-consumer host files alias it as taskdto
- relaykit builds and tests standalone (GOWORK=off): no host imports,
  no gin, no DB, no settings
- golden conversion matrix unchanged

* build(docker): copy relaykit/go.mod before go mod download

The local-replace submodule's go.mod must exist inside the build context
for the main module graph to resolve.

* fix: address relaykit extraction regressions

* fix: address relaykit review regressions

* docs: document Meta nil receiver contract

* fix(relaykit): fail OpenAI→Claude conversion without max_tokens; reject negative default_max_tokens

The Claude Messages API requires max_tokens (omitting it is a 400
"Field required"), but with a nil Options.Claude.DefaultMaxTokens hook
the converters silently emitted a request the upstream is guaranteed to
reject. Both OpenAI Chat and Responses → Claude conversions now return
sharedclaude.ErrMissingMaxTokens when no path (client value, default
hook, thinking-adapter floor) supplied one. Unreachable in the host,
which always configures the hook.

Host side, claude.default_max_tokens now rejects negative values at the
option API before persisting — they would wrap into huge unsigned values
during conversion. Zero stays allowed: the current API treats
max_tokens: 0 as cache pre-warming.

* fix: make Gemini safety settings read path race-free
2026-07-27 15:56:21 +08:00

328 lines
11 KiB
Go
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
package service
import (
"context"
"fmt"
"strings"
"github.com/QuantumNous/new-api/common"
"github.com/QuantumNous/new-api/constant"
"github.com/QuantumNous/new-api/logger"
"github.com/QuantumNous/new-api/model"
relaycommon "github.com/QuantumNous/new-api/relay/common"
"github.com/QuantumNous/new-api/relaykit/types"
"github.com/QuantumNous/new-api/setting/ratio_setting"
"github.com/gin-gonic/gin"
)
// LogTaskConsumption 记录任务消费日志和统计信息(仅记录,不涉及实际扣费)。
// 实际扣费已由 BillingSessionPreConsumeBilling + SettleBilling)完成。
func LogTaskConsumption(c *gin.Context, info *relaycommon.RelayInfo) {
tokenName := c.GetString("token_name")
logContent := fmt.Sprintf("操作 %s", info.Action)
// 支持任务仅按次计费
if common.StringsContains(constant.TaskPricePatches, info.OriginModelName) {
logContent = fmt.Sprintf("%s,按次计费", logContent)
} else {
if otherRatios := info.PriceData.OtherRatios(); len(otherRatios) > 0 {
var contents []string
for key, ra := range otherRatios {
if 1.0 != ra {
contents = append(contents, fmt.Sprintf("%s: %.2f", key, ra))
}
}
if len(contents) > 0 {
logContent = fmt.Sprintf("%s, 计算参数:%s", logContent, strings.Join(contents, ", "))
}
}
}
other := make(map[string]interface{})
other["is_task"] = true
other["request_path"] = c.Request.URL.Path
other["model_price"] = info.PriceData.ModelPrice
if info.PriceData.ModelRatio > 0 {
other["model_ratio"] = info.PriceData.ModelRatio
}
other["group_ratio"] = info.PriceData.GroupRatioInfo.GroupRatio
if info.PriceData.GroupRatioInfo.HasSpecialRatio {
other["user_group_ratio"] = info.PriceData.GroupRatioInfo.GroupSpecialRatio
}
if info.IsModelMapped {
other["is_model_mapped"] = true
other["upstream_model_name"] = info.UpstreamModelName
}
attachQuotaSaturation(c, info, other)
model.RecordConsumeLog(c, info.UserId, model.RecordConsumeLogParams{
ChannelId: info.ChannelId,
ModelName: info.OriginModelName,
TokenName: tokenName,
Quota: info.PriceData.Quota,
Content: logContent,
TokenId: info.TokenId,
Group: info.UsingGroup,
Other: other,
})
model.UpdateUserUsedQuotaAndRequestCount(info.UserId, info.PriceData.Quota)
model.UpdateChannelUsedQuota(info.ChannelId, info.PriceData.Quota)
}
// ---------------------------------------------------------------------------
// 异步任务计费辅助函数
// ---------------------------------------------------------------------------
// resolveTokenKey 通过 TokenId 运行时获取令牌 Key(用于 Redis 缓存操作)。
// 如果令牌已被删除或查询失败,返回空字符串。
func resolveTokenKey(ctx context.Context, tokenId int, taskID string) string {
token, err := model.GetTokenById(tokenId)
if err != nil {
logger.LogWarn(ctx, fmt.Sprintf("获取令牌 key 失败 (tokenId=%d, task=%s): %s", tokenId, taskID, err.Error()))
return ""
}
return token.Key
}
// taskIsSubscription 判断任务是否通过订阅计费。
func taskIsSubscription(task *model.Task) bool {
return task.PrivateData.BillingSource == BillingSourceSubscription && task.PrivateData.SubscriptionId > 0
}
// taskAdjustFunding 调整任务的资金来源(钱包或订阅),delta > 0 表示扣费,delta < 0 表示退还。
func taskAdjustFunding(task *model.Task, delta int) error {
if taskIsSubscription(task) {
return model.PostConsumeUserSubscriptionDelta(task.PrivateData.SubscriptionId, int64(delta))
}
if delta > 0 {
return model.DecreaseUserQuota(task.UserId, delta, false)
}
return model.IncreaseUserQuota(task.UserId, -delta, false)
}
// taskAdjustTokenQuota 调整任务的令牌额度,delta > 0 表示扣费,delta < 0 表示退还。
// 需要通过 resolveTokenKey 运行时获取 key(不从 PrivateData 中读取)。
func taskAdjustTokenQuota(ctx context.Context, task *model.Task, delta int) {
if task.PrivateData.TokenId <= 0 || delta == 0 {
return
}
tokenKey := resolveTokenKey(ctx, task.PrivateData.TokenId, task.TaskID)
if tokenKey == "" {
return
}
var err error
if delta > 0 {
err = model.DecreaseTokenQuota(task.PrivateData.TokenId, tokenKey, delta)
} else {
err = model.IncreaseTokenQuota(task.PrivateData.TokenId, tokenKey, -delta)
}
if err != nil {
logger.LogWarn(ctx, fmt.Sprintf("调整令牌额度失败 (delta=%d, task=%s): %s", delta, task.TaskID, err.Error()))
}
}
// taskBillingOther 从 task 的 BillingContext 构建日志 Other 字段。
func taskBillingOther(task *model.Task) map[string]interface{} {
other := make(map[string]interface{})
if bc := task.PrivateData.BillingContext; bc != nil {
other["model_price"] = bc.ModelPrice
if bc.ModelRatio > 0 {
other["model_ratio"] = bc.ModelRatio
}
other["group_ratio"] = bc.GroupRatio
if priceData := taskBillingContextPriceData(bc); priceData != nil {
for k, v := range priceData.OtherRatios() {
other[k] = v
}
}
}
props := task.Properties
if props.UpstreamModelName != "" && props.UpstreamModelName != props.OriginModelName {
other["is_model_mapped"] = true
other["upstream_model_name"] = props.UpstreamModelName
}
return other
}
func taskBillingContextPriceData(bc *model.TaskBillingContext) *types.PriceData {
if bc == nil || len(bc.OtherRatios) == 0 {
return nil
}
priceData := &types.PriceData{}
if !priceData.ReplaceOtherRatios(bc.OtherRatios) {
return nil
}
return priceData
}
// taskModelName 从 BillingContext 或 Properties 中获取模型名称。
func taskModelName(task *model.Task) string {
if bc := task.PrivateData.BillingContext; bc != nil && bc.OriginModelName != "" {
return bc.OriginModelName
}
return task.Properties.OriginModelName
}
// RefundTaskQuota 统一的任务失败退款逻辑。
// 当异步任务失败时,将预扣的 quota 退还给用户(支持钱包和订阅),并退还令牌额度。
// 返回资金来源是否已成功退还;失败时保留 quota,供显式重试或人工对账。
func RefundTaskQuota(ctx context.Context, task *model.Task, reason string) bool {
quota := task.Quota
if quota == 0 {
return true
}
// 1. 退还资金来源(钱包或订阅)
if err := taskAdjustFunding(task, -quota); err != nil {
logger.LogWarn(ctx, fmt.Sprintf("退还资金来源失败 task %s: %s", task.TaskID, err.Error()))
return false
}
// 2. 退还令牌额度
taskAdjustTokenQuota(ctx, task, -quota)
// 3. 记录日志
other := taskBillingOther(task)
other["task_id"] = task.TaskID
other["reason"] = reason
model.RecordTaskBillingLog(model.RecordTaskBillingLogParams{
UserId: task.UserId,
LogType: model.LogTypeRefund,
Content: "",
ChannelId: task.ChannelId,
ModelName: taskModelName(task),
Quota: quota,
TokenId: task.PrivateData.TokenId,
Group: task.Group,
Other: other,
})
// 4. 资金退款完成后再清除持久化标记。
// 回写失败必须显式告警,避免漏掉潜在的重复退款风险。
task.Quota = 0
if err := task.UpdateQuota(); err != nil {
logger.LogError(ctx, fmt.Sprintf("退款成功但清除 task quota 失败 task %s: %s", task.TaskID, err.Error()))
}
return true
}
// RecalculateTaskQuota 通用的异步差额结算。
// actualQuota 是任务完成后的实际应扣额度,与预扣额度 (task.Quota) 做差额结算。
// reason 用于日志记录(例如 "token重算" 或 "adaptor调整")。
// clamps 可选:若计算 actualQuota 时发生额度饱和,将其记入日志 admin_info(仅管理员可见)。
func RecalculateTaskQuota(ctx context.Context, task *model.Task, actualQuota int, reason string, clamps ...*common.QuotaClamp) {
if actualQuota <= 0 {
return
}
preConsumedQuota := task.Quota
quotaDelta := actualQuota - preConsumedQuota
if quotaDelta == 0 {
logger.LogInfo(ctx, fmt.Sprintf("任务 %s 预扣费准确(%s%s",
task.TaskID, logger.LogQuota(actualQuota), reason))
return
}
logger.LogInfo(ctx, fmt.Sprintf("任务 %s 差额结算:delta=%s(实际:%s,预扣:%s%s",
task.TaskID,
logger.LogQuota(quotaDelta),
logger.LogQuota(actualQuota),
logger.LogQuota(preConsumedQuota),
reason,
))
// 调整资金来源
if err := taskAdjustFunding(task, quotaDelta); err != nil {
logger.LogError(ctx, fmt.Sprintf("差额结算资金调整失败 task %s: %s", task.TaskID, err.Error()))
return
}
// 调整令牌额度
taskAdjustTokenQuota(ctx, task, quotaDelta)
task.Quota = actualQuota
if err := task.UpdateQuota(); err != nil {
logger.LogError(ctx, fmt.Sprintf("差额结算回写 quota 失败 task %s: %s", task.TaskID, err.Error()))
}
var logType int
var logQuota int
if quotaDelta > 0 {
logType = model.LogTypeConsume
logQuota = quotaDelta
model.UpdateUserUsedQuotaAndRequestCount(task.UserId, quotaDelta)
model.UpdateChannelUsedQuota(task.ChannelId, quotaDelta)
} else {
logType = model.LogTypeRefund
logQuota = -quotaDelta
}
other := taskBillingOther(task)
other["task_id"] = task.TaskID
other["pre_consumed_quota"] = preConsumedQuota
other["actual_quota"] = actualQuota
for _, clamp := range clamps {
attachQuotaSaturationToOther(other, clamp)
}
model.RecordTaskBillingLog(model.RecordTaskBillingLogParams{
UserId: task.UserId,
LogType: logType,
Content: reason,
ChannelId: task.ChannelId,
ModelName: taskModelName(task),
Quota: logQuota,
TokenId: task.PrivateData.TokenId,
Group: task.Group,
Other: other,
NodeName: task.PrivateData.NodeName,
})
}
// RecalculateTaskQuotaByTokens 根据实际 token 消耗重新计费(异步差额结算)。
// 当任务成功且返回了 totalTokens 时,根据模型倍率和分组倍率重新计算实际扣费额度,
// 与预扣费的差额进行补扣或退还。支持钱包和订阅计费来源。
func RecalculateTaskQuotaByTokens(ctx context.Context, task *model.Task, totalTokens int) {
if totalTokens <= 0 {
return
}
modelName := taskModelName(task)
// 获取模型价格和倍率
modelRatio, hasRatioSetting, _ := ratio_setting.GetModelRatio(modelName)
// 只有配置了倍率(非固定价格)时才按 token 重新计费
if !hasRatioSetting || modelRatio <= 0 {
return
}
// 获取用户和组的倍率信息
group := task.Group
if group == "" {
user, err := model.GetUserById(task.UserId, false)
if err == nil {
group = user.Group
}
}
if group == "" {
return
}
groupRatio := ratio_setting.GetGroupRatio(group)
userGroupRatio, hasUserGroupRatio := ratio_setting.GetGroupGroupRatio(group, group)
var finalGroupRatio float64
if hasUserGroupRatio {
finalGroupRatio = userGroupRatio
} else {
finalGroupRatio = groupRatio
}
// 计算 OtherRatios 乘积(视频折扣、时长等)
otherMultiplier := 1.0
if priceData := taskBillingContextPriceData(task.PrivateData.BillingContext); priceData != nil {
otherMultiplier = priceData.OtherRatioMultiplier()
}
// 计算实际应扣费额度: totalTokens * modelRatio * groupRatio * otherMultiplier(饱和转换,防止溢出成负数)
actualQuota, clamp := common.QuotaFromFloatChecked(float64(totalTokens) * modelRatio * finalGroupRatio * otherMultiplier)
reason := fmt.Sprintf("token重算:tokens=%d, modelRatio=%.2f, groupRatio=%.2f, otherMultiplier=%.4f", totalTokens, modelRatio, finalGroupRatio, otherMultiplier)
RecalculateTaskQuota(ctx, task, actualQuota, reason, clamp)
}