fix(billing): 修复 OpenAI fast 档位计费并补齐展示

- 打通 service_tier 在 OpenAI HTTP、WS、passthrough 与 usage 记录中的传递 - 修正 priority/flex 计费逻辑，并将 fast 归一化为 priority - 在用户端和管理端补齐服务档位与计费明细展示 - 补齐前后端测试，并修复 WS 限流信号重复持久化导致的全量回归失败 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-04-22 23:54:45 +08:00 · 2026-03-08 23:22:28 +08:00
parent bcb6444f89
commit 87f4ed591e
29 changed files with 1417 additions and 47 deletions
--- a/backend/internal/service/billing_service_test.go
+++ b/backend/internal/service/billing_service_test.go
@@ -522,3 +522,189 @@ func TestCalculateCost_LargeTokenCount(t *testing.T) {
 	require.False(t, math.IsNaN(cost.TotalCost))
 	require.False(t, math.IsInf(cost.TotalCost, 0))
 }
+
+func TestServiceTierCostMultiplier(t *testing.T) {
+	require.InDelta(t, 2.0, serviceTierCostMultiplier("priority"), 1e-12)
+	require.InDelta(t, 2.0, serviceTierCostMultiplier(" Priority "), 1e-12)
+	require.InDelta(t, 0.5, serviceTierCostMultiplier("flex"), 1e-12)
+	require.InDelta(t, 1.0, serviceTierCostMultiplier(""), 1e-12)
+	require.InDelta(t, 1.0, serviceTierCostMultiplier("default"), 1e-12)
+}
+
+func TestCalculateCostWithServiceTier_OpenAIPriorityUsesPriorityPricing(t *testing.T) {
+	svc := newTestBillingService()
+	tokens := UsageTokens{InputTokens: 100, OutputTokens: 50, CacheReadTokens: 20}
+
+	baseCost, err := svc.CalculateCost("gpt-5.1-codex", tokens, 1.0)
+	require.NoError(t, err)
+
+	priorityCost, err := svc.CalculateCostWithServiceTier("gpt-5.1-codex", tokens, 1.0, "priority")
+	require.NoError(t, err)
+
+	require.InDelta(t, baseCost.InputCost*2, priorityCost.InputCost, 1e-10)
+	require.InDelta(t, baseCost.OutputCost*2, priorityCost.OutputCost, 1e-10)
+	require.InDelta(t, baseCost.CacheReadCost*2, priorityCost.CacheReadCost, 1e-10)
+	require.InDelta(t, baseCost.TotalCost*2, priorityCost.TotalCost, 1e-10)
+}
+
+func TestCalculateCostWithServiceTier_FlexAppliesHalfMultiplier(t *testing.T) {
+	svc := newTestBillingService()
+	tokens := UsageTokens{InputTokens: 100, OutputTokens: 50, CacheCreationTokens: 40, CacheReadTokens: 20}
+
+	baseCost, err := svc.CalculateCost("gpt-5.4", tokens, 1.0)
+	require.NoError(t, err)
+
+	flexCost, err := svc.CalculateCostWithServiceTier("gpt-5.4", tokens, 1.0, "flex")
+	require.NoError(t, err)
+
+	require.InDelta(t, baseCost.InputCost*0.5, flexCost.InputCost, 1e-10)
+	require.InDelta(t, baseCost.OutputCost*0.5, flexCost.OutputCost, 1e-10)
+	require.InDelta(t, baseCost.CacheCreationCost*0.5, flexCost.CacheCreationCost, 1e-10)
+	require.InDelta(t, baseCost.CacheReadCost*0.5, flexCost.CacheReadCost, 1e-10)
+	require.InDelta(t, baseCost.TotalCost*0.5, flexCost.TotalCost, 1e-10)
+}
+
+func TestCalculateCostWithServiceTier_PriorityFallsBackToTierMultiplierWithoutExplicitPriorityPrice(t *testing.T) {
+	svc := newTestBillingService()
+	tokens := UsageTokens{InputTokens: 120, OutputTokens: 30, CacheCreationTokens: 12, CacheReadTokens: 8}
+
+	baseCost, err := svc.CalculateCost("claude-sonnet-4", tokens, 1.0)
+	require.NoError(t, err)
+
+	priorityCost, err := svc.CalculateCostWithServiceTier("claude-sonnet-4", tokens, 1.0, "priority")
+	require.NoError(t, err)
+
+	require.InDelta(t, baseCost.InputCost*2, priorityCost.InputCost, 1e-10)
+	require.InDelta(t, baseCost.OutputCost*2, priorityCost.OutputCost, 1e-10)
+	require.InDelta(t, baseCost.CacheCreationCost*2, priorityCost.CacheCreationCost, 1e-10)
+	require.InDelta(t, baseCost.CacheReadCost*2, priorityCost.CacheReadCost, 1e-10)
+	require.InDelta(t, baseCost.TotalCost*2, priorityCost.TotalCost, 1e-10)
+}
+
+func TestBillingServiceGetModelPricing_UsesDynamicPriorityFields(t *testing.T) {
+	pricingSvc := &PricingService{
+		pricingData: map[string]*LiteLLMModelPricing{
+			"gpt-5.4": {
+				InputCostPerToken:               2.5e-6,
+				InputCostPerTokenPriority:       5e-6,
+				OutputCostPerToken:              15e-6,
+				OutputCostPerTokenPriority:      30e-6,
+				CacheCreationInputTokenCost:     2.5e-6,
+				CacheReadInputTokenCost:         0.25e-6,
+				CacheReadInputTokenCostPriority: 0.5e-6,
+				LongContextInputTokenThreshold:  272000,
+				LongContextInputCostMultiplier:  2.0,
+				LongContextOutputCostMultiplier: 1.5,
+			},
+		},
+	}
+	svc := NewBillingService(&config.Config{}, pricingSvc)
+
+	pricing, err := svc.GetModelPricing("gpt-5.4")
+	require.NoError(t, err)
+	require.InDelta(t, 2.5e-6, pricing.InputPricePerToken, 1e-12)
+	require.InDelta(t, 5e-6, pricing.InputPricePerTokenPriority, 1e-12)
+	require.InDelta(t, 15e-6, pricing.OutputPricePerToken, 1e-12)
+	require.InDelta(t, 30e-6, pricing.OutputPricePerTokenPriority, 1e-12)
+	require.InDelta(t, 0.25e-6, pricing.CacheReadPricePerToken, 1e-12)
+	require.InDelta(t, 0.5e-6, pricing.CacheReadPricePerTokenPriority, 1e-12)
+	require.Equal(t, 272000, pricing.LongContextInputThreshold)
+	require.InDelta(t, 2.0, pricing.LongContextInputMultiplier, 1e-12)
+	require.InDelta(t, 1.5, pricing.LongContextOutputMultiplier, 1e-12)
+}
+
+func TestBillingServiceGetModelPricing_OpenAIFallbackGpt52Variants(t *testing.T) {
+	svc := newTestBillingService()
+
+	gpt52, err := svc.GetModelPricing("gpt-5.2")
+	require.NoError(t, err)
+	require.NotNil(t, gpt52)
+	require.InDelta(t, 1.75e-6, gpt52.InputPricePerToken, 1e-12)
+	require.InDelta(t, 3.5e-6, gpt52.InputPricePerTokenPriority, 1e-12)
+
+	gpt52Codex, err := svc.GetModelPricing("gpt-5.2-codex")
+	require.NoError(t, err)
+	require.NotNil(t, gpt52Codex)
+	require.InDelta(t, 1.75e-6, gpt52Codex.InputPricePerToken, 1e-12)
+	require.InDelta(t, 3.5e-6, gpt52Codex.InputPricePerTokenPriority, 1e-12)
+	require.InDelta(t, 28e-6, gpt52Codex.OutputPricePerTokenPriority, 1e-12)
+}
+
+func TestCalculateCostWithServiceTier_PriorityFallsBackToTierMultiplierWhenExplicitPriceMissing(t *testing.T) {
+	svc := NewBillingService(&config.Config{}, &PricingService{
+		pricingData: map[string]*LiteLLMModelPricing{
+			"custom-no-priority": {
+				InputCostPerToken:           1e-6,
+				OutputCostPerToken:          2e-6,
+				CacheCreationInputTokenCost: 0.5e-6,
+				CacheReadInputTokenCost:     0.25e-6,
+			},
+		},
+	})
+	tokens := UsageTokens{InputTokens: 100, OutputTokens: 50, CacheCreationTokens: 40, CacheReadTokens: 20}
+
+	baseCost, err := svc.CalculateCost("custom-no-priority", tokens, 1.0)
+	require.NoError(t, err)
+
+	priorityCost, err := svc.CalculateCostWithServiceTier("custom-no-priority", tokens, 1.0, "priority")
+	require.NoError(t, err)
+
+	require.InDelta(t, baseCost.InputCost*2, priorityCost.InputCost, 1e-10)
+	require.InDelta(t, baseCost.OutputCost*2, priorityCost.OutputCost, 1e-10)
+	require.InDelta(t, baseCost.CacheCreationCost*2, priorityCost.CacheCreationCost, 1e-10)
+	require.InDelta(t, baseCost.CacheReadCost*2, priorityCost.CacheReadCost, 1e-10)
+	require.InDelta(t, baseCost.TotalCost*2, priorityCost.TotalCost, 1e-10)
+}
+
+func TestGetModelPricing_OpenAIGpt52FallbacksExposePriorityPrices(t *testing.T) {
+	svc := newTestBillingService()
+
+	gpt52, err := svc.GetModelPricing("gpt-5.2")
+	require.NoError(t, err)
+	require.InDelta(t, 1.75e-6, gpt52.InputPricePerToken, 1e-12)
+	require.InDelta(t, 3.5e-6, gpt52.InputPricePerTokenPriority, 1e-12)
+	require.InDelta(t, 14e-6, gpt52.OutputPricePerToken, 1e-12)
+	require.InDelta(t, 28e-6, gpt52.OutputPricePerTokenPriority, 1e-12)
+
+	gpt52Codex, err := svc.GetModelPricing("gpt-5.2-codex")
+	require.NoError(t, err)
+	require.InDelta(t, 1.75e-6, gpt52Codex.InputPricePerToken, 1e-12)
+	require.InDelta(t, 3.5e-6, gpt52Codex.InputPricePerTokenPriority, 1e-12)
+	require.InDelta(t, 14e-6, gpt52Codex.OutputPricePerToken, 1e-12)
+	require.InDelta(t, 28e-6, gpt52Codex.OutputPricePerTokenPriority, 1e-12)
+}
+
+func TestGetModelPricing_MapsDynamicPriorityFieldsIntoBillingPricing(t *testing.T) {
+	svc := NewBillingService(&config.Config{}, &PricingService{
+		pricingData: map[string]*LiteLLMModelPricing{
+			"dynamic-tier-model": {
+				InputCostPerToken:                   1e-6,
+				InputCostPerTokenPriority:           2e-6,
+				OutputCostPerToken:                  3e-6,
+				OutputCostPerTokenPriority:          6e-6,
+				CacheCreationInputTokenCost:         4e-6,
+				CacheCreationInputTokenCostAbove1hr: 5e-6,
+				CacheReadInputTokenCost:             7e-7,
+				CacheReadInputTokenCostPriority:     8e-7,
+				LongContextInputTokenThreshold:      999,
+				LongContextInputCostMultiplier:      1.5,
+				LongContextOutputCostMultiplier:     1.25,
+			},
+		},
+	})
+
+	pricing, err := svc.GetModelPricing("dynamic-tier-model")
+	require.NoError(t, err)
+	require.InDelta(t, 1e-6, pricing.InputPricePerToken, 1e-12)
+	require.InDelta(t, 2e-6, pricing.InputPricePerTokenPriority, 1e-12)
+	require.InDelta(t, 3e-6, pricing.OutputPricePerToken, 1e-12)
+	require.InDelta(t, 6e-6, pricing.OutputPricePerTokenPriority, 1e-12)
+	require.InDelta(t, 4e-6, pricing.CacheCreation5mPrice, 1e-12)
+	require.InDelta(t, 5e-6, pricing.CacheCreation1hPrice, 1e-12)
+	require.True(t, pricing.SupportsCacheBreakdown)
+	require.InDelta(t, 7e-7, pricing.CacheReadPricePerToken, 1e-12)
+	require.InDelta(t, 8e-7, pricing.CacheReadPricePerTokenPriority, 1e-12)
+	require.Equal(t, 999, pricing.LongContextInputThreshold)
+	require.InDelta(t, 1.5, pricing.LongContextInputMultiplier, 1e-12)
+	require.InDelta(t, 1.25, pricing.LongContextOutputMultiplier, 1e-12)
+}