feat: 按上游计费倍率调度 OpenAI 账号

This commit is contained in:
Tian Lee
2026-07-16 00:40:46 +08:00
parent 0765d10c1d
commit 90ee85f3ef
34 changed files with 2299 additions and 252 deletions
+36 -10
View File
@@ -205,6 +205,8 @@ func (s *SettingService) InitializeDefaultSettings(ctx context.Context) error {
// 分组隔离(默认不允许未分组 Key 调度)
SettingKeyAllowUngroupedKeyScheduling: "false",
SettingKeyOpenAILowUpstreamRatePriorityEnabled: "false",
SettingKeyOpenAIOAuthSchedulingRateMultiplier: "1",
SettingKeyEnableAnthropicCacheTTL1hInjection: "false",
SettingKeyRewriteMessageCacheControl: strconv.FormatBool(s.defaultRewriteMessageCacheControl()),
SettingKeyEnableClientDatelineNormalization: "true",
@@ -225,6 +227,7 @@ func (s *SettingService) InitializeDefaultSettings(ctx context.Context) error {
SettingKeyOpenAIAdvancedSchedulerWeightTTFT: "",
SettingKeyOpenAIAdvancedSchedulerWeightReset: "",
SettingKeyOpenAIAdvancedSchedulerWeightQuotaHeadroom: "",
SettingKeyOpenAIAdvancedSchedulerWeightUpstreamCost: "",
SettingKeyOpenAIAdvancedSchedulerWeightPreviousResponse: "",
SettingKeyOpenAIAdvancedSchedulerWeightSessionSticky: "",
@@ -789,6 +792,8 @@ func (s *SettingService) parseSettings(settings map[string]string) *SystemSettin
result.PaymentVisibleMethodWxpaySource = NormalizeVisibleMethodSource("wxpay", settings[SettingPaymentVisibleMethodWxpaySource])
result.PaymentVisibleMethodAlipayEnabled = settings[SettingPaymentVisibleMethodAlipayEnabled] == "true"
result.PaymentVisibleMethodWxpayEnabled = settings[SettingPaymentVisibleMethodWxpayEnabled] == "true"
result.OpenAILowUpstreamRatePriorityEnabled = settings[SettingKeyOpenAILowUpstreamRatePriorityEnabled] == "true"
result.OpenAIOAuthSchedulingRateMultiplier = parseOpenAIOAuthSchedulingRateMultiplier(settings[SettingKeyOpenAIOAuthSchedulingRateMultiplier])
result.OpenAIAdvancedSchedulerEnabled = settings[openAIAdvancedSchedulerSettingKey] == "true"
result.OpenAIAdvancedSchedulerStickyWeightedEnabled = settings[SettingKeyOpenAIAdvancedSchedulerStickyWeightedEnabled] == "true"
result.OpenAIAdvancedSchedulerSubscriptionPriorityEnabled = settings[SettingKeyOpenAIAdvancedSchedulerSubscriptionPriorityEnabled] == "true"
@@ -800,6 +805,7 @@ func (s *SettingService) parseSettings(settings map[string]string) *SystemSettin
result.OpenAIAdvancedSchedulerWeightTTFT = strings.TrimSpace(settings[SettingKeyOpenAIAdvancedSchedulerWeightTTFT])
result.OpenAIAdvancedSchedulerWeightReset = strings.TrimSpace(settings[SettingKeyOpenAIAdvancedSchedulerWeightReset])
result.OpenAIAdvancedSchedulerWeightQuotaHeadroom = strings.TrimSpace(settings[SettingKeyOpenAIAdvancedSchedulerWeightQuotaHeadroom])
result.OpenAIAdvancedSchedulerWeightUpstreamCost = strings.TrimSpace(settings[SettingKeyOpenAIAdvancedSchedulerWeightUpstreamCost])
result.OpenAIAdvancedSchedulerWeightPreviousResponse = strings.TrimSpace(settings[SettingKeyOpenAIAdvancedSchedulerWeightPreviousResponse])
result.OpenAIAdvancedSchedulerWeightSessionSticky = strings.TrimSpace(settings[SettingKeyOpenAIAdvancedSchedulerWeightSessionSticky])
result.OpenAIAdvancedSchedulerEffectiveLBTopK = s.openAIAdvancedSchedulerEffectiveLBTopK()
@@ -811,6 +817,7 @@ func (s *SettingService) parseSettings(settings map[string]string) *SystemSettin
result.OpenAIAdvancedSchedulerEffectiveWeightTTFT = formatOpenAIAdvancedSchedulerFloat(effectiveWeights.TTFT)
result.OpenAIAdvancedSchedulerEffectiveWeightReset = formatOpenAIAdvancedSchedulerFloat(effectiveWeights.Reset)
result.OpenAIAdvancedSchedulerEffectiveWeightQuotaHeadroom = formatOpenAIAdvancedSchedulerFloat(effectiveWeights.QuotaHeadroom)
result.OpenAIAdvancedSchedulerEffectiveWeightUpstreamCost = formatOpenAIAdvancedSchedulerFloat(effectiveWeights.UpstreamCost)
result.OpenAIAdvancedSchedulerEffectiveWeightPreviousResponse = formatOpenAIAdvancedSchedulerFloat(effectiveWeights.PreviousResponse)
result.OpenAIAdvancedSchedulerEffectiveWeightSessionSticky = formatOpenAIAdvancedSchedulerFloat(effectiveWeights.SessionSticky)
@@ -901,6 +908,7 @@ func (s *SettingService) openAIAdvancedSchedulerEffectiveWeights() config.Gatewa
TTFT: 0.5,
Reset: 0.0,
QuotaHeadroom: 0.0,
UpstreamCost: 0.0,
PreviousResponse: 5.0,
SessionSticky: 3.0,
}
@@ -909,8 +917,7 @@ func (s *SettingService) openAIAdvancedSchedulerEffectiveWeights() config.Gatewa
}
weights := s.cfg.Gateway.OpenAIWS.SchedulerScoreWeights
baseSum := weights.Priority + weights.Load + weights.Queue + weights.ErrorRate + weights.TTFT + weights.QuotaHeadroom
if baseSum <= 0 {
if !weights.IsValid() {
return defaults
}
return weights
@@ -921,6 +928,10 @@ func formatOpenAIAdvancedSchedulerFloat(value float64) string {
}
func (s *SettingService) normalizeOpenAIAdvancedSchedulerOverrides(settings *SystemSettings) error {
if rate := settings.OpenAIOAuthSchedulingRateMultiplier; rate < 0 || math.IsNaN(rate) || math.IsInf(rate, 0) {
return infraerrors.BadRequest("INVALID_OPENAI_OAUTH_SCHEDULING_RATE_MULTIPLIER", "OpenAI OAuth scheduling rate multiplier must be a finite non-negative number")
}
lbTopK, err := normalizeOptionalPositiveIntString(settings.OpenAIAdvancedSchedulerLBTopK)
if err != nil {
return infraerrors.BadRequest("INVALID_OPENAI_ADVANCED_SCHEDULER_LB_TOP_K", "openai advanced scheduler TopK must be a positive integer or empty")
@@ -935,6 +946,7 @@ func (s *SettingService) normalizeOpenAIAdvancedSchedulerOverrides(settings *Sys
&settings.OpenAIAdvancedSchedulerWeightTTFT,
&settings.OpenAIAdvancedSchedulerWeightReset,
&settings.OpenAIAdvancedSchedulerWeightQuotaHeadroom,
&settings.OpenAIAdvancedSchedulerWeightUpstreamCost,
&settings.OpenAIAdvancedSchedulerWeightPreviousResponse,
&settings.OpenAIAdvancedSchedulerWeightSessionSticky,
}
@@ -950,18 +962,32 @@ func (s *SettingService) normalizeOpenAIAdvancedSchedulerOverrides(settings *Sys
// 覆盖值(空则回退到生效的配置值)叠加后的基础权重和不允许为 0,
// 否则调度会静默退化为 TopK 内均匀随机。
effective := s.openAIAdvancedSchedulerEffectiveWeights()
baseSum := resolveOpenAIAdvancedSchedulerWeight(settings.OpenAIAdvancedSchedulerWeightPriority, effective.Priority) +
resolveOpenAIAdvancedSchedulerWeight(settings.OpenAIAdvancedSchedulerWeightLoad, effective.Load) +
resolveOpenAIAdvancedSchedulerWeight(settings.OpenAIAdvancedSchedulerWeightQueue, effective.Queue) +
resolveOpenAIAdvancedSchedulerWeight(settings.OpenAIAdvancedSchedulerWeightErrorRate, effective.ErrorRate) +
resolveOpenAIAdvancedSchedulerWeight(settings.OpenAIAdvancedSchedulerWeightTTFT, effective.TTFT) +
resolveOpenAIAdvancedSchedulerWeight(settings.OpenAIAdvancedSchedulerWeightQuotaHeadroom, effective.QuotaHeadroom)
if baseSum <= 0 {
return infraerrors.BadRequest("INVALID_OPENAI_ADVANCED_SCHEDULER_WEIGHT", "openai advanced scheduler base weights must not all be zero")
resolved := config.GatewayOpenAIWSSchedulerScoreWeights{
Priority: resolveOpenAIAdvancedSchedulerWeight(settings.OpenAIAdvancedSchedulerWeightPriority, effective.Priority),
Load: resolveOpenAIAdvancedSchedulerWeight(settings.OpenAIAdvancedSchedulerWeightLoad, effective.Load),
Queue: resolveOpenAIAdvancedSchedulerWeight(settings.OpenAIAdvancedSchedulerWeightQueue, effective.Queue),
ErrorRate: resolveOpenAIAdvancedSchedulerWeight(settings.OpenAIAdvancedSchedulerWeightErrorRate, effective.ErrorRate),
TTFT: resolveOpenAIAdvancedSchedulerWeight(settings.OpenAIAdvancedSchedulerWeightTTFT, effective.TTFT),
Reset: resolveOpenAIAdvancedSchedulerWeight(settings.OpenAIAdvancedSchedulerWeightReset, effective.Reset),
QuotaHeadroom: resolveOpenAIAdvancedSchedulerWeight(settings.OpenAIAdvancedSchedulerWeightQuotaHeadroom, effective.QuotaHeadroom),
UpstreamCost: resolveOpenAIAdvancedSchedulerWeight(settings.OpenAIAdvancedSchedulerWeightUpstreamCost, effective.UpstreamCost),
PreviousResponse: resolveOpenAIAdvancedSchedulerWeight(settings.OpenAIAdvancedSchedulerWeightPreviousResponse, effective.PreviousResponse),
SessionSticky: resolveOpenAIAdvancedSchedulerWeight(settings.OpenAIAdvancedSchedulerWeightSessionSticky, effective.SessionSticky),
}
if !resolved.IsValid() {
return infraerrors.BadRequest("INVALID_OPENAI_ADVANCED_SCHEDULER_WEIGHT", "openai advanced scheduler weights must have finite non-zero base and total sums")
}
return nil
}
func parseOpenAIOAuthSchedulingRateMultiplier(raw string) float64 {
value, err := strconv.ParseFloat(strings.TrimSpace(raw), 64)
if err != nil || value < 0 || math.IsNaN(value) || math.IsInf(value, 0) {
return defaultOpenAIOAuthSchedulingRateMultiplier
}
return value
}
// resolveOpenAIAdvancedSchedulerWeight 返回覆盖值(已归一化的非空字符串),空则回退默认值。
func resolveOpenAIAdvancedSchedulerWeight(normalized string, fallback float64) float64 {
if normalized == "" {