revert: completely remove all Sora functionality

This commit is contained in:
erio
2026-04-05 17:11:01 +08:00
parent dbb248df52
commit 62e80c602d
136 changed files with 256 additions and 24221 deletions

View File

@@ -60,13 +60,6 @@ const (
claudeMimicDebugInfoKey = "claude_mimic_debug_info"
)
// MediaType 媒体类型常量
const (
MediaTypeImage = "image"
MediaTypeVideo = "video"
MediaTypePrompt = "prompt"
)
// ForceCacheBillingContextKey 强制缓存计费上下文键
// 用于粘性会话切换时,将 input_tokens 转为 cache_read_input_tokens 计费
type forceCacheBillingKeyType struct{}
@@ -510,10 +503,6 @@ type ForwardResult struct {
// 图片生成计费字段(图片生成模型使用)
ImageCount int // 生成的图片数量
ImageSize string // 图片尺寸 "1K", "2K", "4K"
// Sora 媒体字段
MediaType string // image / video / prompt
MediaURL string // 生成后的媒体地址(可选)
}
// UpstreamFailoverError indicates an upstream error that should trigger account failover.
@@ -1341,6 +1330,11 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
ctx = s.withWindowCostPrefetch(ctx, accounts)
ctx = s.withRPMPrefetch(ctx, accounts)
// 提前构建 accountByID供 Layer 1 和 Layer 1.5 使用)
accountByID := make(map[int64]*Account, len(accounts))
for i := range accounts {
accountByID[accounts[i].ID] = &accounts[i]
}
isExcluded := func(accountID int64) bool {
if excludedIDs == nil {
return false
@@ -1349,12 +1343,6 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
return excluded
}
// 提前构建 accountByID供 Layer 1 和 Layer 1.5 使用)
accountByID := make(map[int64]*Account, len(accounts))
for i := range accounts {
accountByID[accounts[i].ID] = &accounts[i]
}
// 获取模型路由配置(仅 anthropic 平台)
var routingAccountIDs []int64
if group != nil && requestedModel != "" && group.Platform == PlatformAnthropic {
@@ -1442,24 +1430,19 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
if containsInt64(routingAccountIDs, stickyAccountID) && !isExcluded(stickyAccountID) {
// 粘性账号在路由列表中,优先使用
if stickyAccount, ok := accountByID[stickyAccountID]; ok {
var stickyCacheMissReason string
gatePass := s.isAccountSchedulableForSelection(stickyAccount) &&
if s.isAccountSchedulableForSelection(stickyAccount) &&
s.isAccountAllowedForPlatform(stickyAccount, platform, useMixed) &&
(requestedModel == "" || s.isModelSupportedByAccountWithContext(ctx, stickyAccount, requestedModel)) &&
s.isAccountSchedulableForModelSelection(ctx, stickyAccount, requestedModel) &&
s.isAccountSchedulableForQuota(stickyAccount) &&
s.isAccountSchedulableForWindowCost(ctx, stickyAccount, true)
s.isAccountSchedulableForWindowCost(ctx, stickyAccount, true) &&
rpmPass := gatePass && s.isAccountSchedulableForRPM(ctx, stickyAccount, true)
if rpmPass { // 粘性会话窗口费用+RPM 检查
s.isAccountSchedulableForRPM(ctx, stickyAccount, true) { // 粘性会话窗口费用+RPM 检查
result, err := s.tryAcquireAccountSlot(ctx, stickyAccountID, stickyAccount.Concurrency)
if err == nil && result.Acquired {
// 会话数量限制检查
if !s.checkAndRegisterSession(ctx, stickyAccount, sessionHash) {
result.ReleaseFunc() // 释放槽位
stickyCacheMissReason = "session_limit"
// 继续到负载感知选择
} else {
if s.debugModelRoutingEnabled() {
@@ -1473,49 +1456,27 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
}
}
if stickyCacheMissReason == "" {
waitingCount, _ := s.concurrencyService.GetAccountWaitingCount(ctx, stickyAccountID)
if waitingCount < cfg.StickySessionMaxWaiting {
// 会话数量限制检查(等待计划也需要占用会话配额)
if !s.checkAndRegisterSession(ctx, stickyAccount, sessionHash) {
stickyCacheMissReason = "session_limit"
// 会话限制已满,继续到负载感知选择
} else {
return &AccountSelectionResult{
Account: stickyAccount,
WaitPlan: &AccountWaitPlan{
AccountID: stickyAccountID,
MaxConcurrency: stickyAccount.Concurrency,
Timeout: cfg.StickySessionWaitTimeout,
MaxWaiting: cfg.StickySessionMaxWaiting,
},
}, nil
}
waitingCount, _ := s.concurrencyService.GetAccountWaitingCount(ctx, stickyAccountID)
if waitingCount < cfg.StickySessionMaxWaiting {
// 会话数量限制检查(等待计划也需要占用会话配额)
if !s.checkAndRegisterSession(ctx, stickyAccount, sessionHash) {
// 会话限制已满,继续到负载感知选择
} else {
stickyCacheMissReason = "wait_queue_full"
return &AccountSelectionResult{
Account: stickyAccount,
WaitPlan: &AccountWaitPlan{
AccountID: stickyAccountID,
MaxConcurrency: stickyAccount.Concurrency,
Timeout: cfg.StickySessionWaitTimeout,
MaxWaiting: cfg.StickySessionMaxWaiting,
},
}, nil
}
}
// 粘性账号槽位满且等待队列已满,继续使用负载感知选择
} else if !gatePass {
stickyCacheMissReason = "gate_check"
} else {
stickyCacheMissReason = "rpm_red"
}
// 记录粘性缓存未命中的结构化日志
if stickyCacheMissReason != "" {
baseRPM := stickyAccount.GetBaseRPM()
var currentRPM int
if count, ok := rpmFromPrefetchContext(ctx, stickyAccount.ID); ok {
currentRPM = count
}
logger.LegacyPrintf("service.gateway", "[StickyCacheMiss] reason=%s account_id=%d session=%s current_rpm=%d base_rpm=%d",
stickyCacheMissReason, stickyAccountID, shortSessionHash(sessionHash), currentRPM, baseRPM)
}
} else {
_ = s.cache.DeleteSessionAccountID(ctx, derefGroupID(groupID), sessionHash)
logger.LegacyPrintf("service.gateway", "[StickyCacheMiss] reason=account_cleared account_id=%d session=%s current_rpm=0 base_rpm=0",
stickyAccountID, shortSessionHash(sessionHash))
}
}
}
@@ -1621,7 +1582,6 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
account, ok := accountByID[accountID]
if ok {
// 检查账户是否需要清理粘性会话绑定
// Check if the account needs sticky session cleanup
clearSticky := shouldClearStickySession(account, requestedModel)
if clearSticky {
_ = s.cache.DeleteSessionAccountID(ctx, derefGroupID(groupID), sessionHash)
@@ -1637,7 +1597,6 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
result, err := s.tryAcquireAccountSlot(ctx, accountID, account.Concurrency)
if err == nil && result.Acquired {
// 会话数量限制检查
// Session count limit check
if !s.checkAndRegisterSession(ctx, account, sessionHash) {
result.ReleaseFunc() // 释放槽位,继续到 Layer 2
} else {
@@ -1652,10 +1611,8 @@ func (s *GatewayService) SelectAccountWithLoadAwareness(ctx context.Context, gro
waitingCount, _ := s.concurrencyService.GetAccountWaitingCount(ctx, accountID)
if waitingCount < cfg.StickySessionMaxWaiting {
// 会话数量限制检查(等待计划也需要占用会话配额)
// Session count limit check (wait plan also requires session quota)
if !s.checkAndRegisterSession(ctx, account, sessionHash) {
// 会话限制已满,继续到 Layer 2
// Session limit full, continue to Layer 2
} else {
return &AccountSelectionResult{
Account: account,
@@ -1971,9 +1928,6 @@ func (s *GatewayService) resolvePlatform(ctx context.Context, groupID *int64, gr
}
func (s *GatewayService) listSchedulableAccounts(ctx context.Context, groupID *int64, platform string, hasForcePlatform bool) ([]Account, bool, error) {
if platform == PlatformSora {
return s.listSoraSchedulableAccounts(ctx, groupID)
}
if s.schedulerSnapshot != nil {
accounts, useMixed, err := s.schedulerSnapshot.ListSchedulableAccounts(ctx, groupID, platform, hasForcePlatform)
if err == nil {
@@ -2070,53 +2024,6 @@ func (s *GatewayService) listSchedulableAccounts(ctx context.Context, groupID *i
return accounts, useMixed, nil
}
func (s *GatewayService) listSoraSchedulableAccounts(ctx context.Context, groupID *int64) ([]Account, bool, error) {
const useMixed = false
var accounts []Account
var err error
if s.cfg != nil && s.cfg.RunMode == config.RunModeSimple {
accounts, err = s.accountRepo.ListByPlatform(ctx, PlatformSora)
} else if groupID != nil {
accounts, err = s.accountRepo.ListByGroup(ctx, *groupID)
} else {
accounts, err = s.accountRepo.ListByPlatform(ctx, PlatformSora)
}
if err != nil {
slog.Debug("account_scheduling_list_failed",
"group_id", derefGroupID(groupID),
"platform", PlatformSora,
"error", err)
return nil, useMixed, err
}
filtered := make([]Account, 0, len(accounts))
for _, acc := range accounts {
if acc.Platform != PlatformSora {
continue
}
if !s.isSoraAccountSchedulable(&acc) {
continue
}
filtered = append(filtered, acc)
}
slog.Debug("account_scheduling_list_sora",
"group_id", derefGroupID(groupID),
"platform", PlatformSora,
"raw_count", len(accounts),
"filtered_count", len(filtered))
for _, acc := range filtered {
slog.Debug("account_scheduling_account_detail",
"account_id", acc.ID,
"name", acc.Name,
"platform", acc.Platform,
"type", acc.Type,
"status", acc.Status,
"tls_fingerprint", acc.IsTLSFingerprintEnabled())
}
return filtered, useMixed, nil
}
// IsSingleAntigravityAccountGroup 检查指定分组是否只有一个 antigravity 平台的可调度账号。
// 用于 Handler 层在首次请求时提前设置 SingleAccountRetry context
// 避免单账号分组收到 503 时错误地设置模型限流标记导致后续请求连续快速失败。
@@ -2141,33 +2048,10 @@ func (s *GatewayService) isAccountAllowedForPlatform(account *Account, platform
return account.Platform == platform
}
func (s *GatewayService) isSoraAccountSchedulable(account *Account) bool {
return s.soraUnschedulableReason(account) == ""
}
func (s *GatewayService) soraUnschedulableReason(account *Account) string {
if account == nil {
return "account_nil"
}
if account.Status != StatusActive {
return fmt.Sprintf("status=%s", account.Status)
}
if !account.Schedulable {
return "schedulable=false"
}
if account.TempUnschedulableUntil != nil && time.Now().Before(*account.TempUnschedulableUntil) {
return fmt.Sprintf("temp_unschedulable_until=%s", account.TempUnschedulableUntil.UTC().Format(time.RFC3339))
}
return ""
}
func (s *GatewayService) isAccountSchedulableForSelection(account *Account) bool {
if account == nil {
return false
}
if account.Platform == PlatformSora {
return s.isSoraAccountSchedulable(account)
}
return account.IsSchedulable()
}
@@ -2175,12 +2059,6 @@ func (s *GatewayService) isAccountSchedulableForModelSelection(ctx context.Conte
if account == nil {
return false
}
if account.Platform == PlatformSora {
if !s.isSoraAccountSchedulable(account) {
return false
}
return account.GetRateLimitRemainingTimeWithContext(ctx, requestedModel) <= 0
}
return account.IsSchedulableForModelWithContext(ctx, requestedModel)
}
@@ -2795,12 +2673,6 @@ func (s *GatewayService) selectAccountForModelWithPlatform(ctx context.Context,
preferOAuth := platform == PlatformGemini
routingAccountIDs := s.routingAccountIDsForRequest(ctx, groupID, requestedModel, platform)
// require_privacy_set: 获取分组信息
var schedGroup *Group
if groupID != nil && s.groupRepo != nil {
schedGroup, _ = s.groupRepo.GetByID(ctx, *groupID)
}
var accounts []Account
accountsLoaded := false
@@ -2824,7 +2696,7 @@ func (s *GatewayService) selectAccountForModelWithPlatform(ctx context.Context,
if clearSticky {
_ = s.cache.DeleteSessionAccountID(ctx, derefGroupID(groupID), sessionHash)
}
if !clearSticky && s.isAccountInGroup(account, groupID) && account.Platform == platform && (requestedModel == "" || s.isModelSupportedByAccountWithContext(ctx, account, requestedModel)) && s.isAccountSchedulableForModelSelection(ctx, account, requestedModel) && s.isAccountSchedulableForQuota(account) && s.isAccountSchedulableForWindowCost(ctx, account, true) && s.isAccountSchedulableForRPM(ctx, account, true) {
if !clearSticky && s.isAccountInGroup(account, groupID) && account.Platform == platform && (requestedModel == "" || s.isModelSupportedByAccountWithContext(ctx, account, requestedModel)) && s.isAccountSchedulableForModelSelection(ctx, account, requestedModel) && s.isAccountSchedulableForQuota(account) && s.isAccountSchedulableForWindowCost(ctx, account, true) && s.isAccountSchedulableForRPM(ctx, account, true) && !s.isStickyAccountUpstreamRestricted(ctx, groupID, account, requestedModel) {
if s.debugModelRoutingEnabled() {
logger.LegacyPrintf("service.gateway", "[ModelRoutingDebug] legacy routed sticky hit: group_id=%v model=%s session=%s account=%d", derefGroupID(groupID), requestedModel, shortSessionHash(sessionHash), accountID)
}
@@ -2872,12 +2744,6 @@ func (s *GatewayService) selectAccountForModelWithPlatform(ctx context.Context,
if !s.isAccountSchedulableForSelection(acc) {
continue
}
// require_privacy_set: 跳过 privacy 未设置的账号并标记异常
if schedGroup != nil && schedGroup.RequirePrivacySet && !acc.IsPrivacySet() {
_ = s.accountRepo.SetError(ctx, acc.ID,
fmt.Sprintf("Privacy not set, required by group [%s]", schedGroup.Name))
continue
}
if requestedModel != "" && !s.isModelSupportedByAccountWithContext(ctx, acc, requestedModel) {
continue
}
@@ -2983,12 +2849,6 @@ func (s *GatewayService) selectAccountForModelWithPlatform(ctx context.Context,
if !s.isAccountSchedulableForSelection(acc) {
continue
}
// require_privacy_set: 跳过 privacy 未设置的账号并标记异常
if schedGroup != nil && schedGroup.RequirePrivacySet && !acc.IsPrivacySet() {
_ = s.accountRepo.SetError(ctx, acc.ID,
fmt.Sprintf("Privacy not set, required by group [%s]", schedGroup.Name))
continue
}
if requestedModel != "" && !s.isModelSupportedByAccountWithContext(ctx, acc, requestedModel) {
continue
}
@@ -3055,12 +2915,6 @@ func (s *GatewayService) selectAccountWithMixedScheduling(ctx context.Context, g
preferOAuth := nativePlatform == PlatformGemini
routingAccountIDs := s.routingAccountIDsForRequest(ctx, groupID, requestedModel, nativePlatform)
// require_privacy_set: 获取分组信息
var schedGroup *Group
if groupID != nil && s.groupRepo != nil {
schedGroup, _ = s.groupRepo.GetByID(ctx, *groupID)
}
var accounts []Account
accountsLoaded := false
@@ -3128,12 +2982,6 @@ func (s *GatewayService) selectAccountWithMixedScheduling(ctx context.Context, g
if !s.isAccountSchedulableForSelection(acc) {
continue
}
// require_privacy_set: 跳过 privacy 未设置的账号并标记异常
if schedGroup != nil && schedGroup.RequirePrivacySet && !acc.IsPrivacySet() {
_ = s.accountRepo.SetError(ctx, acc.ID,
fmt.Sprintf("Privacy not set, required by group [%s]", schedGroup.Name))
continue
}
// 过滤原生平台直接通过antigravity 需要启用混合调度
if acc.Platform == PlatformAntigravity && !acc.IsMixedSchedulingEnabled() {
continue
@@ -3203,7 +3051,7 @@ func (s *GatewayService) selectAccountWithMixedScheduling(ctx context.Context, g
if clearSticky {
_ = s.cache.DeleteSessionAccountID(ctx, derefGroupID(groupID), sessionHash)
}
if !clearSticky && s.isAccountInGroup(account, groupID) && (requestedModel == "" || s.isModelSupportedByAccountWithContext(ctx, account, requestedModel)) && s.isAccountSchedulableForModelSelection(ctx, account, requestedModel) && s.isAccountSchedulableForQuota(account) && s.isAccountSchedulableForWindowCost(ctx, account, true) && s.isAccountSchedulableForRPM(ctx, account, true) {
if !clearSticky && s.isAccountInGroup(account, groupID) && (requestedModel == "" || s.isModelSupportedByAccountWithContext(ctx, account, requestedModel)) && s.isAccountSchedulableForModelSelection(ctx, account, requestedModel) && s.isAccountSchedulableForQuota(account) && s.isAccountSchedulableForWindowCost(ctx, account, true) && s.isAccountSchedulableForRPM(ctx, account, true) && !s.isStickyAccountUpstreamRestricted(ctx, groupID, account, requestedModel) {
if account.Platform == nativePlatform || (account.Platform == PlatformAntigravity && account.IsMixedSchedulingEnabled()) {
return account, nil
}
@@ -3227,7 +3075,6 @@ func (s *GatewayService) selectAccountWithMixedScheduling(ctx context.Context, g
ctx = s.withRPMPrefetch(ctx, accounts)
// 3. 按优先级+最久未用选择(考虑模型支持和混合调度)
// needsUpstreamCheck 仅在主选择循环中使用;粘性会话命中时跳过此检查。
needsUpstreamCheck := s.needsUpstreamChannelRestrictionCheck(ctx, groupID)
var selected *Account
for i := range accounts {
@@ -3240,12 +3087,6 @@ func (s *GatewayService) selectAccountWithMixedScheduling(ctx context.Context, g
if !s.isAccountSchedulableForSelection(acc) {
continue
}
// require_privacy_set: 跳过 privacy 未设置的账号并标记异常
if schedGroup != nil && schedGroup.RequirePrivacySet && !acc.IsPrivacySet() {
_ = s.accountRepo.SetError(ctx, acc.ID,
fmt.Sprintf("Privacy not set, required by group [%s]", schedGroup.Name))
continue
}
// 过滤原生平台直接通过antigravity 需要启用混合调度
if acc.Platform == PlatformAntigravity && !acc.IsMixedSchedulingEnabled() {
continue
@@ -3357,9 +3198,6 @@ func (s *GatewayService) logDetailedSelectionFailure(
stats.SampleMappingIDs,
stats.SampleRateLimitIDs,
)
if platform == PlatformSora {
s.logSoraSelectionFailureDetails(ctx, groupID, sessionHash, requestedModel, accounts, excludedIDs, allowMixedScheduling)
}
return stats
}
@@ -3417,9 +3255,6 @@ func (s *GatewayService) diagnoseSelectionFailure(
}
if !s.isAccountSchedulableForSelection(acc) {
detail := "generic_unschedulable"
if acc.Platform == PlatformSora {
detail = s.soraUnschedulableReason(acc)
}
return selectionFailureDiagnosis{Category: "unschedulable", Detail: detail}
}
if isPlatformFilteredForSelection(acc, platform, allowMixedScheduling) {
@@ -3444,57 +3279,7 @@ func (s *GatewayService) diagnoseSelectionFailure(
return selectionFailureDiagnosis{Category: "eligible"}
}
func (s *GatewayService) logSoraSelectionFailureDetails(
ctx context.Context,
groupID *int64,
sessionHash string,
requestedModel string,
accounts []Account,
excludedIDs map[int64]struct{},
allowMixedScheduling bool,
) {
const maxLines = 30
logged := 0
for i := range accounts {
if logged >= maxLines {
break
}
acc := &accounts[i]
diagnosis := s.diagnoseSelectionFailure(ctx, acc, requestedModel, PlatformSora, excludedIDs, allowMixedScheduling)
if diagnosis.Category == "eligible" {
continue
}
detail := diagnosis.Detail
if detail == "" {
detail = "-"
}
logger.LegacyPrintf(
"service.gateway",
"[SelectAccountDetailed:Sora] group_id=%v model=%s session=%s account_id=%d account_platform=%s category=%s detail=%s",
derefGroupID(groupID),
requestedModel,
shortSessionHash(sessionHash),
acc.ID,
acc.Platform,
diagnosis.Category,
detail,
)
logged++
}
if len(accounts) > maxLines {
logger.LegacyPrintf(
"service.gateway",
"[SelectAccountDetailed:Sora] group_id=%v model=%s session=%s truncated=true total=%d logged=%d",
derefGroupID(groupID),
requestedModel,
shortSessionHash(sessionHash),
len(accounts),
logged,
)
}
}
// GetAccessToken 获取账号凭证
func isPlatformFilteredForSelection(acc *Account, platform string, allowMixedScheduling bool) bool {
if acc == nil {
return true
@@ -3573,13 +3358,14 @@ func (s *GatewayService) isModelSupportedByAccount(account *Account, requestedMo
}
return mapAntigravityModel(account, requestedModel) != ""
}
if account.Platform == PlatformSora {
return s.isSoraModelSupportedByAccount(account, requestedModel)
}
if account.IsBedrock() {
_, ok := ResolveBedrockModelID(account, requestedModel)
return ok
}
// OpenAI 透传模式:仅替换认证,允许所有模型
if account.Platform == PlatformOpenAI && account.IsOpenAIPassthroughEnabled() {
return true
}
// OAuth/SetupToken 账号使用 Anthropic 标准映射短ID → 长ID
if account.Platform == PlatformAnthropic && account.Type != AccountTypeAPIKey {
requestedModel = claude.NormalizeModelID(requestedModel)
@@ -3588,143 +3374,6 @@ func (s *GatewayService) isModelSupportedByAccount(account *Account, requestedMo
return account.IsModelSupported(requestedModel)
}
func (s *GatewayService) isSoraModelSupportedByAccount(account *Account, requestedModel string) bool {
if account == nil {
return false
}
if strings.TrimSpace(requestedModel) == "" {
return true
}
// 先走原始精确/通配符匹配。
mapping := account.GetModelMapping()
if len(mapping) == 0 || account.IsModelSupported(requestedModel) {
return true
}
aliases := buildSoraModelAliases(requestedModel)
if len(aliases) == 0 {
return false
}
hasSoraSelector := false
for pattern := range mapping {
if !isSoraModelSelector(pattern) {
continue
}
hasSoraSelector = true
if matchPatternAnyAlias(pattern, aliases) {
return true
}
}
// 兼容旧账号mapping 存在但未配置任何 Sora 选择器(例如只含 gpt-*
// 此时不应误拦截 Sora 模型请求。
if !hasSoraSelector {
return true
}
return false
}
func matchPatternAnyAlias(pattern string, aliases []string) bool {
normalizedPattern := strings.ToLower(strings.TrimSpace(pattern))
if normalizedPattern == "" {
return false
}
for _, alias := range aliases {
if matchWildcard(normalizedPattern, alias) {
return true
}
}
return false
}
func isSoraModelSelector(pattern string) bool {
p := strings.ToLower(strings.TrimSpace(pattern))
if p == "" {
return false
}
switch {
case strings.HasPrefix(p, "sora"),
strings.HasPrefix(p, "gpt-image"),
strings.HasPrefix(p, "prompt-enhance"),
strings.HasPrefix(p, "sy_"):
return true
}
return p == "video" || p == "image"
}
func buildSoraModelAliases(requestedModel string) []string {
modelID := strings.ToLower(strings.TrimSpace(requestedModel))
if modelID == "" {
return nil
}
aliases := make([]string, 0, 8)
addAlias := func(value string) {
v := strings.ToLower(strings.TrimSpace(value))
if v == "" {
return
}
for _, existing := range aliases {
if existing == v {
return
}
}
aliases = append(aliases, v)
}
addAlias(modelID)
cfg, ok := GetSoraModelConfig(modelID)
if ok {
addAlias(cfg.Model)
switch cfg.Type {
case "video":
addAlias("video")
addAlias("sora")
addAlias(soraVideoFamilyAlias(modelID))
case "image":
addAlias("image")
addAlias("gpt-image")
case "prompt_enhance":
addAlias("prompt-enhance")
}
return aliases
}
switch {
case strings.HasPrefix(modelID, "sora"):
addAlias("video")
addAlias("sora")
addAlias(soraVideoFamilyAlias(modelID))
case strings.HasPrefix(modelID, "gpt-image"):
addAlias("image")
addAlias("gpt-image")
case strings.HasPrefix(modelID, "prompt-enhance"):
addAlias("prompt-enhance")
default:
return nil
}
return aliases
}
func soraVideoFamilyAlias(modelID string) string {
switch {
case strings.HasPrefix(modelID, "sora2pro-hd"):
return "sora2pro-hd"
case strings.HasPrefix(modelID, "sora2pro"):
return "sora2pro"
case strings.HasPrefix(modelID, "sora2"):
return "sora2"
default:
return ""
}
}
// GetAccessToken 获取账号凭证
func (s *GatewayService) GetAccessToken(ctx context.Context, account *Account) (string, string, error) {
switch account.Type {
@@ -7434,6 +7083,7 @@ func (s *GatewayService) getUserGroupRateMultiplier(ctx context.Context, userID,
// RecordUsageInput 记录使用量的输入参数
type RecordUsageInput struct {
Result *ForwardResult
ParsedRequest *ParsedRequest
APIKey *APIKey
User *User
Account *Account
@@ -7745,12 +7395,10 @@ func writeUsageLogBestEffort(ctx context.Context, repo UsageLogRepository, usage
// recordUsageOpts 内部选项,参数化 RecordUsage 与 RecordUsageWithLongContext 的差异点。
type recordUsageOpts struct {
// Claude Max 策略所需的 ParsedRequest可选仅 Claude 路径传入)
// ParsedRequest可选仅 Claude 路径传入)
ParsedRequest *ParsedRequest
// EnableClaudePath 启用 Claude 路径特有逻辑:
// - Claude Max 缓存计费策略
// - Sora 媒体类型分支image/video/prompt
// - MediaType 字段写入使用日志
EnableClaudePath bool
@@ -7776,6 +7424,7 @@ func (s *GatewayService) RecordUsage(ctx context.Context, input *RecordUsageInpu
APIKeyService: input.APIKeyService,
ChannelUsageFields: input.ChannelUsageFields,
}, &recordUsageOpts{
ParsedRequest: input.ParsedRequest,
EnableClaudePath: true,
})
}
@@ -7841,8 +7490,6 @@ type recordUsageCoreInput struct {
// recordUsageCore 是 RecordUsage 和 RecordUsageWithLongContext 的统一实现。
// opts 中的字段控制两者之间的差异行为:
// - ParsedRequest != nil → 启用 Claude Max 缓存计费策略
// - EnableSoraMedia → 启用 Sora MediaType 分支image/video/prompt
// - LongContextThreshold > 0 → Token 计费回退走 CalculateCostWithLongContext
func (s *GatewayService) recordUsageCore(ctx context.Context, input *recordUsageCoreInput, opts *recordUsageOpts) error {
result := input.Result
@@ -7944,16 +7591,6 @@ func (s *GatewayService) calculateRecordUsageCost(
multiplier float64,
opts *recordUsageOpts,
) *CostBreakdown {
// Sora 媒体类型分支(仅 Claude 路径启用)
if opts.EnableClaudePath {
if result.MediaType == MediaTypeImage || result.MediaType == MediaTypeVideo {
return s.calculateSoraMediaCost(result, apiKey, billingModel, multiplier)
}
if result.MediaType == MediaTypePrompt {
return &CostBreakdown{}
}
}
// 图片生成计费
if result.ImageCount > 0 {
return s.calculateImageCost(ctx, result, apiKey, billingModel, multiplier)
@@ -7963,28 +7600,6 @@ func (s *GatewayService) calculateRecordUsageCost(
return s.calculateTokenCost(ctx, result, apiKey, billingModel, multiplier, opts)
}
// calculateSoraMediaCost 计算 Sora 图片/视频的费用。
func (s *GatewayService) calculateSoraMediaCost(
result *ForwardResult,
apiKey *APIKey,
billingModel string,
multiplier float64,
) *CostBreakdown {
var soraConfig *SoraPriceConfig
if apiKey.Group != nil {
soraConfig = &SoraPriceConfig{
ImagePrice360: apiKey.Group.SoraImagePrice360,
ImagePrice540: apiKey.Group.SoraImagePrice540,
VideoPricePerRequest: apiKey.Group.SoraVideoPricePerRequest,
VideoPricePerRequestHD: apiKey.Group.SoraVideoPricePerRequestHD,
}
}
if result.MediaType == MediaTypeImage {
return s.billingService.CalculateSoraImageCost(result.ImageSize, result.ImageCount, soraConfig, multiplier)
}
return s.billingService.CalculateSoraVideoCost(billingModel, soraConfig, multiplier)
}
// resolveChannelPricing 检查指定模型是否存在渠道级别定价。
// 返回非 nil 的 ResolvedPricing 表示有渠道定价nil 表示走默认定价路径。
func (s *GatewayService) resolveChannelPricing(ctx context.Context, billingModel string, apiKey *APIKey) *ResolvedPricing {
@@ -8163,13 +7778,7 @@ func (s *GatewayService) buildRecordUsageLog(
}
// resolveBillingMode 根据计费结果和请求类型确定计费模式。
// Sora 媒体类型自身已确定计费模式(由上游处理),返回 nil 跳过。
func resolveBillingMode(opts *recordUsageOpts, result *ForwardResult, cost *CostBreakdown) *string {
isSoraMedia := opts.EnableClaudePath &&
(result.MediaType == MediaTypeImage || result.MediaType == MediaTypeVideo || result.MediaType == MediaTypePrompt)
if isSoraMedia {
return nil
}
var mode string
switch {
case cost != nil && cost.BillingMode != "":
@@ -8183,9 +7792,6 @@ func resolveBillingMode(opts *recordUsageOpts, result *ForwardResult, cost *Cost
}
func resolveMediaType(opts *recordUsageOpts, result *ForwardResult) *string {
if opts.EnableClaudePath && strings.TrimSpace(result.MediaType) != "" {
return &result.MediaType
}
return nil
}
@@ -8293,6 +7899,19 @@ func (s *GatewayService) needsUpstreamChannelRestrictionCheck(ctx context.Contex
return ch.BillingModelSource == BillingModelSourceUpstream
}
// isStickyAccountUpstreamRestricted 检查粘性会话命中的账号是否受 upstream 渠道限制。
// 合并 needsUpstreamChannelRestrictionCheck + isUpstreamModelRestrictedByChannel 两步调用,
// 供 sticky session 条件链使用,避免内联多个函数调用导致行过长。
func (s *GatewayService) isStickyAccountUpstreamRestricted(ctx context.Context, groupID *int64, account *Account, requestedModel string) bool {
if groupID == nil {
return false
}
if !s.needsUpstreamChannelRestrictionCheck(ctx, groupID) {
return false
}
return s.isUpstreamModelRestrictedByChannel(ctx, *groupID, account, requestedModel)
}
// ForwardCountTokens 转发 count_tokens 请求到上游 API
// 特点:不记录使用量、仅支持非流式响应
func (s *GatewayService) ForwardCountTokens(ctx context.Context, c *gin.Context, account *Account, parsed *ParsedRequest) error {