Files
new-api/model/pricing.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

434 lines
14 KiB
Go

package model
import (
"fmt"
"strings"
"sync"
"time"
"github.com/QuantumNous/new-api/common"
"github.com/QuantumNous/new-api/constant"
"github.com/QuantumNous/new-api/relaykit/dto"
"github.com/QuantumNous/new-api/relaykit/types"
"github.com/QuantumNous/new-api/setting/billing_setting"
"github.com/QuantumNous/new-api/setting/ratio_setting"
)
type Pricing struct {
ModelName string `json:"model_name"`
Description string `json:"description,omitempty"`
Icon string `json:"icon,omitempty"`
Tags string `json:"tags,omitempty"`
VendorID int `json:"vendor_id,omitempty"`
QuotaType int `json:"quota_type"`
ModelRatio float64 `json:"model_ratio"`
ModelPrice float64 `json:"model_price"`
OwnerBy string `json:"owner_by"`
CompletionRatio float64 `json:"completion_ratio"`
CacheRatio *float64 `json:"cache_ratio,omitempty"`
CreateCacheRatio *float64 `json:"create_cache_ratio,omitempty"`
ImageRatio *float64 `json:"image_ratio,omitempty"`
AudioRatio *float64 `json:"audio_ratio,omitempty"`
AudioCompletionRatio *float64 `json:"audio_completion_ratio,omitempty"`
EnableGroup []string `json:"enable_groups"`
SupportedEndpointTypes []constant.EndpointType `json:"supported_endpoint_types"`
BillingMode string `json:"billing_mode,omitempty"`
BillingExpr string `json:"billing_expr,omitempty"`
PricingVersion string `json:"pricing_version,omitempty"`
}
type PricingVendor struct {
ID int `json:"id"`
Name string `json:"name"`
Description string `json:"description,omitempty"`
Icon string `json:"icon,omitempty"`
}
var (
pricingMap []Pricing
vendorsList []PricingVendor
supportedEndpointMap map[string]common.EndpointInfo
lastGetPricingTime time.Time
updatePricingLock sync.Mutex
// 缓存映射:模型名 -> 启用分组 / 计费类型
modelEnableGroups = make(map[string][]string)
modelQuotaTypeMap = make(map[string]int)
modelEnableGroupsLock = sync.RWMutex{}
)
var (
modelSupportEndpointTypes = make(map[string][]constant.EndpointType)
modelSupportEndpointsLock = sync.RWMutex{}
)
func GetPricing() []Pricing {
if time.Since(lastGetPricingTime) > time.Minute*1 || len(pricingMap) == 0 {
updatePricingLock.Lock()
defer updatePricingLock.Unlock()
// Double check after acquiring the lock
if time.Since(lastGetPricingTime) > time.Minute*1 || len(pricingMap) == 0 {
modelSupportEndpointsLock.Lock()
defer modelSupportEndpointsLock.Unlock()
updatePricing()
}
}
return pricingMap
}
func InvalidatePricingCache() {
updatePricingLock.Lock()
defer updatePricingLock.Unlock()
pricingMap = nil
vendorsList = nil
lastGetPricingTime = time.Time{}
}
// GetVendors 返回当前定价接口使用到的供应商信息
func GetVendors() []PricingVendor {
if time.Since(lastGetPricingTime) > time.Minute*1 || len(pricingMap) == 0 {
// 保证先刷新一次
GetPricing()
}
return vendorsList
}
func GetModelSupportEndpointTypes(model string) []constant.EndpointType {
if model == "" {
return make([]constant.EndpointType, 0)
}
modelSupportEndpointsLock.RLock()
defer modelSupportEndpointsLock.RUnlock()
if endpoints, ok := modelSupportEndpointTypes[model]; ok {
return endpoints
}
return make([]constant.EndpointType, 0)
}
func getPricingEndpointTypesForAbility(ability AbilityWithChannel, advancedCustomConfigs map[int]*dto.AdvancedCustomConfig) []constant.EndpointType {
if ability.ChannelType != constant.ChannelTypeAdvancedCustom {
return common.GetEndpointTypesByChannelType(ability.ChannelType, ability.Model)
}
if config := advancedCustomConfigs[ability.ChannelId]; config != nil {
return config.SupportedEndpointTypesForModel(ability.Model)
}
return common.GetEndpointTypesByChannelType(ability.ChannelType, ability.Model)
}
// loadPricingAdvancedCustomConfigs runs inside updatePricing while
// updatePricingLock is held, and nests channelSyncLock.RLock. This defines the
// global lock order updatePricingLock -> channelSyncLock: any code path holding
// channelSyncLock must release it before touching the pricing cache (see
// InitChannelCache / CacheUpdateChannel), otherwise it deadlocks.
// The returned configs are pointers shared with the channel cache; they are
// replaced wholesale on update and never mutated in place, so reading them after
// RUnlock is safe.
func loadPricingAdvancedCustomConfigs(enableAbilities []AbilityWithChannel) map[int]*dto.AdvancedCustomConfig {
channelIDs := make([]int, 0)
seen := make(map[int]struct{})
for _, ability := range enableAbilities {
if ability.ChannelType != constant.ChannelTypeAdvancedCustom {
continue
}
if _, exists := seen[ability.ChannelId]; exists {
continue
}
seen[ability.ChannelId] = struct{}{}
channelIDs = append(channelIDs, ability.ChannelId)
}
if len(channelIDs) == 0 {
return nil
}
configs := make(map[int]*dto.AdvancedCustomConfig, len(channelIDs))
if common.MemoryCacheEnabled {
channelSyncLock.RLock()
defer channelSyncLock.RUnlock()
for _, channelID := range channelIDs {
if config := channel2advancedCustomConfig[channelID]; config != nil {
configs[channelID] = config
}
}
return configs
}
for _, channelID := range channelIDs {
channel, err := CacheGetChannel(channelID)
if err != nil {
common.SysLog(fmt.Sprintf("load advanced custom channel settings error: channel_id=%d, error=%v", channelID, err))
continue
}
if channel.Type != constant.ChannelTypeAdvancedCustom {
continue
}
if config := channel.GetOtherSettings().AdvancedCustom; config != nil {
configs[channelID] = config
}
}
return configs
}
func appendPricingEndpoint(endpoints []string, endpoint string) []string {
if endpoint == "" || common.StringsContains(endpoints, endpoint) {
return endpoints
}
return append(endpoints, endpoint)
}
func updatePricing() {
//modelRatios := common.GetModelRatios()
enableAbilities, err := GetAllEnableAbilityWithChannels()
if err != nil {
common.SysLog(fmt.Sprintf("GetAllEnableAbilityWithChannels error: %v", err))
return
}
// 预加载模型元数据与供应商一次,避免循环查询
var allMeta []Model
_ = DB.Find(&allMeta).Error
metaMap := make(map[string]*Model)
prefixList := make([]*Model, 0)
suffixList := make([]*Model, 0)
containsList := make([]*Model, 0)
for i := range allMeta {
m := &allMeta[i]
if m.NameRule == NameRuleExact {
metaMap[m.ModelName] = m
} else {
switch m.NameRule {
case NameRulePrefix:
prefixList = append(prefixList, m)
case NameRuleSuffix:
suffixList = append(suffixList, m)
case NameRuleContains:
containsList = append(containsList, m)
}
}
}
// 将非精确规则模型匹配到 metaMap
for _, m := range prefixList {
for _, pricingModel := range enableAbilities {
if strings.HasPrefix(pricingModel.Model, m.ModelName) {
if _, exists := metaMap[pricingModel.Model]; !exists {
metaMap[pricingModel.Model] = m
}
}
}
}
for _, m := range suffixList {
for _, pricingModel := range enableAbilities {
if strings.HasSuffix(pricingModel.Model, m.ModelName) {
if _, exists := metaMap[pricingModel.Model]; !exists {
metaMap[pricingModel.Model] = m
}
}
}
}
for _, m := range containsList {
for _, pricingModel := range enableAbilities {
if strings.Contains(pricingModel.Model, m.ModelName) {
if _, exists := metaMap[pricingModel.Model]; !exists {
metaMap[pricingModel.Model] = m
}
}
}
}
// 预加载供应商
var vendors []Vendor
_ = DB.Find(&vendors).Error
vendorMap := make(map[int]*Vendor)
for i := range vendors {
vendorMap[vendors[i].Id] = &vendors[i]
}
// 初始化默认供应商映射
initDefaultVendorMapping(metaMap, vendorMap, enableAbilities)
// 构建对前端友好的供应商列表
vendorsList = make([]PricingVendor, 0, len(vendorMap))
for _, v := range vendorMap {
vendorsList = append(vendorsList, PricingVendor{
ID: v.Id,
Name: v.Name,
Description: v.Description,
Icon: v.Icon,
})
}
modelGroupsMap := make(map[string]*types.Set[string])
for _, ability := range enableAbilities {
groups, ok := modelGroupsMap[ability.Model]
if !ok {
groups = types.NewSet[string]()
modelGroupsMap[ability.Model] = groups
}
groups.Add(ability.Group)
}
//这里使用切片而不是Set,因为一个模型可能支持多个端点类型,并且第一个端点是优先使用端点
modelSupportEndpointsStr := make(map[string][]string)
advancedCustomConfigs := loadPricingAdvancedCustomConfigs(enableAbilities)
// 先根据已有能力填充原生端点
for _, ability := range enableAbilities {
endpoints := modelSupportEndpointsStr[ability.Model]
channelTypes := getPricingEndpointTypesForAbility(ability, advancedCustomConfigs)
for _, channelType := range channelTypes {
if !common.StringsContains(endpoints, string(channelType)) {
endpoints = append(endpoints, string(channelType))
}
}
modelSupportEndpointsStr[ability.Model] = endpoints
}
// 再补充模型自定义端点:若配置有效则追加到已有推断,不再裁剪渠道真实能力
for modelName, meta := range metaMap {
if strings.TrimSpace(meta.Endpoints) == "" {
continue
}
var raw map[string]interface{}
if err := common.Unmarshal([]byte(meta.Endpoints), &raw); err == nil {
endpoints := modelSupportEndpointsStr[modelName]
for k, v := range raw {
switch v.(type) {
case string, map[string]interface{}:
endpoints = appendPricingEndpoint(endpoints, k)
}
}
if len(endpoints) > 0 {
modelSupportEndpointsStr[modelName] = endpoints
}
}
}
modelSupportEndpointTypes = make(map[string][]constant.EndpointType)
for model, endpoints := range modelSupportEndpointsStr {
supportedEndpoints := make([]constant.EndpointType, 0)
for _, endpointStr := range endpoints {
endpointType := constant.EndpointType(endpointStr)
supportedEndpoints = append(supportedEndpoints, endpointType)
}
modelSupportEndpointTypes[model] = supportedEndpoints
}
// 构建全局 supportedEndpointMap(默认 + 自定义覆盖)
supportedEndpointMap = make(map[string]common.EndpointInfo)
// 1. 默认端点
for _, endpoints := range modelSupportEndpointTypes {
for _, et := range endpoints {
if info, ok := common.GetDefaultEndpointInfo(et); ok {
if _, exists := supportedEndpointMap[string(et)]; !exists {
supportedEndpointMap[string(et)] = info
}
}
}
}
// 2. 自定义端点(models 表)覆盖默认
for _, meta := range metaMap {
if strings.TrimSpace(meta.Endpoints) == "" {
continue
}
var raw map[string]interface{}
if err := common.Unmarshal([]byte(meta.Endpoints), &raw); err == nil {
for k, v := range raw {
switch val := v.(type) {
case string:
supportedEndpointMap[k] = common.EndpointInfo{Path: val, Method: "POST"}
case map[string]interface{}:
ep := common.EndpointInfo{Method: "POST"}
if p, ok := val["path"].(string); ok {
ep.Path = p
}
if m, ok := val["method"].(string); ok {
ep.Method = strings.ToUpper(m)
}
supportedEndpointMap[k] = ep
default:
// ignore unsupported types
}
}
}
}
pricingMap = make([]Pricing, 0)
for model, groups := range modelGroupsMap {
pricing := Pricing{
ModelName: model,
EnableGroup: groups.Items(),
SupportedEndpointTypes: modelSupportEndpointTypes[model],
}
// 补充模型元数据(描述、标签、供应商、状态)
if meta, ok := metaMap[model]; ok {
// 若模型被禁用(status!=1),则直接跳过,不返回给前端
if meta.Status != 1 {
continue
}
pricing.Description = meta.Description
pricing.Icon = meta.Icon
pricing.Tags = meta.Tags
pricing.VendorID = meta.VendorID
}
modelPrice, findPrice := ratio_setting.GetModelPrice(model, false)
if findPrice {
pricing.ModelPrice = modelPrice
pricing.QuotaType = 1
} else {
modelRatio, _, _ := ratio_setting.GetModelRatio(model)
pricing.ModelRatio = modelRatio
pricing.CompletionRatio = ratio_setting.GetCompletionRatio(model)
pricing.QuotaType = 0
}
if cacheRatio, ok := ratio_setting.GetCacheRatio(model); ok {
pricing.CacheRatio = &cacheRatio
}
if createCacheRatio, ok := ratio_setting.GetCreateCacheRatio(model); ok {
pricing.CreateCacheRatio = &createCacheRatio
}
if imageRatio, ok := ratio_setting.GetImageRatio(model); ok {
pricing.ImageRatio = &imageRatio
}
if ratio_setting.ContainsAudioRatio(model) {
audioRatio := ratio_setting.GetAudioRatio(model)
pricing.AudioRatio = &audioRatio
}
if ratio_setting.ContainsAudioCompletionRatio(model) {
audioCompletionRatio := ratio_setting.GetAudioCompletionRatio(model)
pricing.AudioCompletionRatio = &audioCompletionRatio
}
if billingMode := billing_setting.GetBillingMode(model); billingMode == "tiered_expr" {
if expr, ok := billing_setting.GetBillingExpr(model); ok && strings.TrimSpace(expr) != "" {
pricing.BillingMode = billingMode
pricing.BillingExpr = expr
}
}
pricingMap = append(pricingMap, pricing)
}
// 防止大更新后数据不通用
if len(pricingMap) > 0 {
pricingMap[0].PricingVersion = "5a90f2b86c08bd983a9a2e6d66c255f4eaef9c4bc934386d2b6ae84ef0ff1f1f"
}
// 刷新缓存映射,供高并发快速查询
modelEnableGroupsLock.Lock()
modelEnableGroups = make(map[string][]string)
modelQuotaTypeMap = make(map[string]int)
for _, p := range pricingMap {
modelEnableGroups[p.ModelName] = p.EnableGroup
modelQuotaTypeMap[p.ModelName] = p.QuotaType
}
modelEnableGroupsLock.Unlock()
lastGetPricingTime = time.Now()
}
// GetSupportedEndpointMap 返回全局端点到路径的映射
func GetSupportedEndpointMap() map[string]common.EndpointInfo {
return supportedEndpointMap
}