mirror of
https://github.com/Wei-Shaw/sub2api.git
synced 2026-10-24 23:43:46 +08:00
shaw
80a229bce5
修复 PR #3768 批量图像 MVP 合并后审计报告中的全部问题： 结算与计费（高危）： - 所有 SETTLEMENT_* 失败（超冻结/计数非法/manifest 冲突/定价缺失/扣费失败） 统一计入 retry_count 并在耗尽时释放冻结转 failed，消灭 settling 无限 requeue 导致的冻结余额永久锁死 - 耗尽出口的释放指纹统一为 RequestHash，与 processor/Cancel/recovery 一致； release 遇同 request id 指纹冲突视为幂等成功，治愈历史毒消息 - 管理端校验 hold_multiplier >= discount_multiplier，定价快照对存量脏数据钳制 - 释放前校验 per-job hold claim（dedup+归档表），杜绝幻影释放 索引对账（高危）： - provider 输出与提交 custom_id 集对账：未知条目丢弃并记事件， 漏项补 PROVIDER_RESULT_MISSING 失败行，保证 success+fail == item_count 提交与恢复（高危）： - 提交前转 uploading 并在 provider.Submit 期间心跳刷新 updated_at； 恢复扫描改为原子复核（FailStaleUnsubmittedBatchImageJob）， 消灭慢提交被误杀退款而上游任务照常计费的孤儿场景 - 上游任务创建成功但本地状态推进失败时，尽力取消上游并清理输入 - recovery 释放失败时入队交由 worker releaseTerminalHold 兜底重试 队列与并发（中危）： - Enqueue（SetNX+LPush）与 Reserve（BRPop+ZAdd）均改为 Lua 原子脚本， 消灭崩溃窗口导致 job 脱离队列、被 7 天 inflight 键锁死 - 锁冲突按 LockConflictDelay 重新入队（原直接丢弃需等 10 分钟 stale 恢复） - 处理期间心跳：active zset 续期（ZAddXX 防幽灵成员）+ 锁 TTL 续期 - ReplaceBatchImageItemsForJob 增加 indexing 状态守卫，防掉队 worker 重写账目 存量回归（中危）： - image-only 定价条目（仅图片价无 token 价）恢复 token 计费 fail-closed， 不再按 $0 计费；图片计费路径不受影响 - 鉴权余额门槛恢复 balance <= 0 语义，MinimumBalanceReserve 不再作硬 403 加固： - ZIP max_items 钳制到管理员上限；Submit 补齐 Platform==Gemini 校验； gemini downloadUri 跟随前做 host 白名单校验 - 批量客户端改用共享 httpclient（拨号/TLS/响应头超时有界） - 审计点名的忽略错误（MarkDownloaded/SettlementFailed/AppendEvent 等）改为记日志
384 lines
13 KiB
Go
384 lines
13 KiB
Go
package service
|
|
|
|
import (
|
|
"encoding/json"
|
|
"os"
|
|
"path/filepath"
|
|
"testing"
|
|
|
|
"github.com/Wei-Shaw/sub2api/internal/config"
|
|
"github.com/stretchr/testify/require"
|
|
)
|
|
|
|
func TestParsePricingData_ParsesPriorityAndServiceTierFields(t *testing.T) {
|
|
svc := &PricingService{}
|
|
body := []byte(`{
|
|
"gpt-5.4": {
|
|
"input_cost_per_token": 0.0000025,
|
|
"input_cost_per_token_priority": 0.000005,
|
|
"output_cost_per_token": 0.000015,
|
|
"output_cost_per_token_priority": 0.00003,
|
|
"cache_creation_input_token_cost": 0.0000025,
|
|
"cache_read_input_token_cost": 0.00000025,
|
|
"cache_read_input_token_cost_priority": 0.0000005,
|
|
"supports_service_tier": true,
|
|
"supports_prompt_caching": true,
|
|
"litellm_provider": "openai",
|
|
"mode": "chat"
|
|
}
|
|
}`)
|
|
|
|
data, err := svc.parsePricingData(body)
|
|
require.NoError(t, err)
|
|
pricing := data["gpt-5.4"]
|
|
require.NotNil(t, pricing)
|
|
require.InDelta(t, 5e-6, pricing.InputCostPerTokenPriority, 1e-12)
|
|
require.InDelta(t, 3e-5, pricing.OutputCostPerTokenPriority, 1e-12)
|
|
require.InDelta(t, 5e-7, pricing.CacheReadInputTokenCostPriority, 1e-12)
|
|
require.True(t, pricing.SupportsServiceTier)
|
|
}
|
|
|
|
func TestParsePricingData_KeepsImageOnlyPricing(t *testing.T) {
|
|
svc := &PricingService{}
|
|
body := []byte(`{
|
|
"image-only-model": {
|
|
"output_cost_per_image": 0.034,
|
|
"litellm_provider": "vertex_ai-language-models",
|
|
"mode": "image_generation"
|
|
}
|
|
}`)
|
|
|
|
data, err := svc.parsePricingData(body)
|
|
require.NoError(t, err)
|
|
pricing := data["image-only-model"]
|
|
require.NotNil(t, pricing)
|
|
require.InDelta(t, 0.034, pricing.OutputCostPerImage, 1e-12)
|
|
require.Equal(t, "image_generation", pricing.Mode)
|
|
// 仅有图片价的条目必须标记 token 价缺失，供 token 计费路径 fail-closed。
|
|
require.True(t, pricing.TokenPricingAbsent)
|
|
}
|
|
|
|
func TestBillingService_GetModelPricing_FailsClosedForImageOnlyEntries(t *testing.T) {
|
|
pricingSvc := &PricingService{}
|
|
data, err := pricingSvc.parsePricingData([]byte(`{
|
|
"imagen-9.0-generate": {
|
|
"output_cost_per_image": 0.04,
|
|
"litellm_provider": "vertex_ai-image-models",
|
|
"mode": "image_generation"
|
|
},
|
|
"gemini-image-with-token-price": {
|
|
"input_cost_per_token": 0.0,
|
|
"output_cost_per_token": 0.0,
|
|
"output_cost_per_image": 0.034,
|
|
"litellm_provider": "vertex_ai-language-models",
|
|
"mode": "image_generation"
|
|
}
|
|
}`))
|
|
require.NoError(t, err)
|
|
pricingSvc.pricingData = data
|
|
billingSvc := NewBillingService(&config.Config{}, pricingSvc)
|
|
|
|
// image-only 条目不得进入 token 计费（否则 token 流量按 $0 计费），
|
|
// 必须落到 fallback / ErrModelPricingUnavailable 的 fail-closed 路径。
|
|
_, err = billingSvc.GetModelPricing("imagen-9.0-generate")
|
|
require.ErrorIs(t, err, ErrModelPricingUnavailable)
|
|
|
|
// 显式 0 token 价的免费条目保持历史行为：正常返回。
|
|
pricing, err := billingSvc.GetModelPricing("gemini-image-with-token-price")
|
|
require.NoError(t, err)
|
|
require.Zero(t, pricing.InputPricePerToken)
|
|
|
|
// 图片计费路径不受影响：仍能读到 image-only 条目的图片单价。
|
|
raw := pricingSvc.GetModelPricing("imagen-9.0-generate")
|
|
require.NotNil(t, raw)
|
|
require.InDelta(t, 0.04, raw.OutputCostPerImage, 1e-12)
|
|
}
|
|
|
|
func TestPricingService_MergesFallbackOnlyModels(t *testing.T) {
|
|
dir := t.TempDir()
|
|
fallbackFile := filepath.Join(dir, "fallback.json")
|
|
require.NoError(t, os.WriteFile(fallbackFile, []byte(`{
|
|
"remote-model": {
|
|
"input_cost_per_token": 0.000001,
|
|
"litellm_provider": "test",
|
|
"mode": "chat"
|
|
},
|
|
"gemini-3.1-flash-lite-image": {
|
|
"output_cost_per_image": 0.034,
|
|
"litellm_provider": "vertex_ai-language-models",
|
|
"mode": "image_generation"
|
|
}
|
|
}`), 0644))
|
|
|
|
svc := &PricingService{cfg: &config.Config{}}
|
|
svc.cfg.Pricing.FallbackFile = fallbackFile
|
|
remoteData, err := svc.parsePricingData([]byte(`{
|
|
"remote-model": {
|
|
"input_cost_per_token": 0.000002,
|
|
"litellm_provider": "test",
|
|
"mode": "chat"
|
|
}
|
|
}`))
|
|
require.NoError(t, err)
|
|
|
|
merged := svc.mergeFallbackPricingData(remoteData)
|
|
require.InDelta(t, 0.000002, merged["remote-model"].InputCostPerToken, 1e-12)
|
|
require.NotNil(t, merged["gemini-3.1-flash-lite-image"])
|
|
require.InDelta(t, 0.034, merged["gemini-3.1-flash-lite-image"].OutputCostPerImage, 1e-12)
|
|
}
|
|
|
|
func TestGetModelPricing_Gpt53CodexSparkUsesGpt51CodexPricing(t *testing.T) {
|
|
sparkPricing := &LiteLLMModelPricing{InputCostPerToken: 1}
|
|
gpt53Pricing := &LiteLLMModelPricing{InputCostPerToken: 9}
|
|
|
|
svc := &PricingService{
|
|
pricingData: map[string]*LiteLLMModelPricing{
|
|
"gpt-5.1-codex": sparkPricing,
|
|
"gpt-5.3": gpt53Pricing,
|
|
},
|
|
}
|
|
|
|
got := svc.GetModelPricing("gpt-5.3-codex-spark")
|
|
require.Same(t, sparkPricing, got)
|
|
}
|
|
|
|
func TestGetModelPricing_Gpt53CodexFallbackStillUsesGpt52Codex(t *testing.T) {
|
|
gpt52CodexPricing := &LiteLLMModelPricing{InputCostPerToken: 2}
|
|
|
|
svc := &PricingService{
|
|
pricingData: map[string]*LiteLLMModelPricing{
|
|
"gpt-5.2-codex": gpt52CodexPricing,
|
|
},
|
|
}
|
|
|
|
got := svc.GetModelPricing("gpt-5.3-codex")
|
|
require.Same(t, gpt52CodexPricing, got)
|
|
}
|
|
|
|
func TestGetModelPricing_OpenAIFallbackMatchedLoggedAsInfo(t *testing.T) {
|
|
logSink, restore := captureStructuredLog(t)
|
|
defer restore()
|
|
|
|
gpt52CodexPricing := &LiteLLMModelPricing{InputCostPerToken: 2}
|
|
svc := &PricingService{
|
|
pricingData: map[string]*LiteLLMModelPricing{
|
|
"gpt-5.2-codex": gpt52CodexPricing,
|
|
},
|
|
}
|
|
|
|
got := svc.GetModelPricing("gpt-5.3-codex")
|
|
require.Same(t, gpt52CodexPricing, got)
|
|
|
|
require.True(t, logSink.ContainsMessageAtLevel("[Pricing] OpenAI fallback matched gpt-5.3-codex -> gpt-5.2-codex", "info"))
|
|
require.False(t, logSink.ContainsMessageAtLevel("[Pricing] OpenAI fallback matched gpt-5.3-codex -> gpt-5.2-codex", "warn"))
|
|
}
|
|
|
|
func TestGetModelPricing_Gpt54UsesStaticFallbackWhenRemoteMissing(t *testing.T) {
|
|
svc := &PricingService{
|
|
pricingData: map[string]*LiteLLMModelPricing{
|
|
"gpt-5.1-codex": &LiteLLMModelPricing{InputCostPerToken: 1.25e-6},
|
|
},
|
|
}
|
|
|
|
got := svc.GetModelPricing("gpt-5.4")
|
|
require.NotNil(t, got)
|
|
require.InDelta(t, 2.5e-6, got.InputCostPerToken, 1e-12)
|
|
require.InDelta(t, 1.5e-5, got.OutputCostPerToken, 1e-12)
|
|
require.InDelta(t, 2.5e-7, got.CacheReadInputTokenCost, 1e-12)
|
|
require.Equal(t, 272000, got.LongContextInputTokenThreshold)
|
|
require.InDelta(t, 2.0, got.LongContextInputCostMultiplier, 1e-12)
|
|
require.InDelta(t, 1.5, got.LongContextOutputCostMultiplier, 1e-12)
|
|
}
|
|
|
|
func TestGetModelPricing_OpenAICompactAliasUsesStaticFallback(t *testing.T) {
|
|
svc := &PricingService{
|
|
pricingData: map[string]*LiteLLMModelPricing{
|
|
"gpt-5.1-codex": {InputCostPerToken: 1.25e-6},
|
|
},
|
|
}
|
|
|
|
got := svc.GetModelPricing("openai/gpt5.5")
|
|
require.NotNil(t, got)
|
|
require.InDelta(t, 2.5e-6, got.InputCostPerToken, 1e-12)
|
|
require.InDelta(t, 1.5e-5, got.OutputCostPerToken, 1e-12)
|
|
}
|
|
|
|
func TestDefaultPricingIncludesCodexAutoReview(t *testing.T) {
|
|
data, err := os.ReadFile(filepath.Join("..", "..", "resources", "model-pricing", "model_prices_and_context_window.json"))
|
|
require.NoError(t, err)
|
|
|
|
svc := &PricingService{}
|
|
pricingData, err := svc.parsePricingData(data)
|
|
require.NoError(t, err)
|
|
svc.pricingData = pricingData
|
|
|
|
got := svc.GetModelPricing("codex-auto-review")
|
|
require.NotNil(t, got)
|
|
require.InDelta(t, 5e-6, got.InputCostPerToken, 1e-12)
|
|
require.InDelta(t, 3e-5, got.OutputCostPerToken, 1e-12)
|
|
require.InDelta(t, 5e-7, got.CacheReadInputTokenCost, 1e-12)
|
|
}
|
|
|
|
func TestGetModelPricing_Gpt54MiniUsesDedicatedStaticFallbackWhenRemoteMissing(t *testing.T) {
|
|
svc := &PricingService{
|
|
pricingData: map[string]*LiteLLMModelPricing{
|
|
"gpt-5.1-codex": {InputCostPerToken: 1.25e-6},
|
|
},
|
|
}
|
|
|
|
got := svc.GetModelPricing("gpt-5.4-mini")
|
|
require.NotNil(t, got)
|
|
require.InDelta(t, 7.5e-7, got.InputCostPerToken, 1e-12)
|
|
require.InDelta(t, 4.5e-6, got.OutputCostPerToken, 1e-12)
|
|
require.InDelta(t, 7.5e-8, got.CacheReadInputTokenCost, 1e-12)
|
|
require.Zero(t, got.LongContextInputTokenThreshold)
|
|
}
|
|
|
|
func TestGetModelPricing_Gpt54NanoUsesDedicatedStaticFallbackWhenRemoteMissing(t *testing.T) {
|
|
svc := &PricingService{
|
|
pricingData: map[string]*LiteLLMModelPricing{
|
|
"gpt-5.1-codex": {InputCostPerToken: 1.25e-6},
|
|
},
|
|
}
|
|
|
|
got := svc.GetModelPricing("gpt-5.4-nano")
|
|
require.NotNil(t, got)
|
|
require.InDelta(t, 2e-7, got.InputCostPerToken, 1e-12)
|
|
require.InDelta(t, 1.25e-6, got.OutputCostPerToken, 1e-12)
|
|
require.InDelta(t, 2e-8, got.CacheReadInputTokenCost, 1e-12)
|
|
require.Zero(t, got.LongContextInputTokenThreshold)
|
|
}
|
|
|
|
func TestGetModelPricing_ImageModelDoesNotFallbackToTextModel(t *testing.T) {
|
|
imagePricing := &LiteLLMModelPricing{InputCostPerToken: 3}
|
|
textPricing := &LiteLLMModelPricing{InputCostPerToken: 9}
|
|
|
|
svc := &PricingService{
|
|
pricingData: map[string]*LiteLLMModelPricing{
|
|
"gpt-image-2": imagePricing,
|
|
"gpt-5.4": textPricing,
|
|
},
|
|
}
|
|
|
|
got := svc.GetModelPricing("gpt-image-3")
|
|
require.Same(t, imagePricing, got)
|
|
}
|
|
|
|
func TestParsePricingData_PreservesPriorityAndServiceTierFields(t *testing.T) {
|
|
raw := map[string]any{
|
|
"gpt-5.4": map[string]any{
|
|
"input_cost_per_token": 2.5e-6,
|
|
"input_cost_per_token_priority": 5e-6,
|
|
"output_cost_per_token": 15e-6,
|
|
"output_cost_per_token_priority": 30e-6,
|
|
"cache_read_input_token_cost": 0.25e-6,
|
|
"cache_read_input_token_cost_priority": 0.5e-6,
|
|
"supports_service_tier": true,
|
|
"supports_prompt_caching": true,
|
|
"litellm_provider": "openai",
|
|
"mode": "chat",
|
|
},
|
|
}
|
|
body, err := json.Marshal(raw)
|
|
require.NoError(t, err)
|
|
|
|
svc := &PricingService{}
|
|
pricingMap, err := svc.parsePricingData(body)
|
|
require.NoError(t, err)
|
|
|
|
pricing := pricingMap["gpt-5.4"]
|
|
require.NotNil(t, pricing)
|
|
require.InDelta(t, 2.5e-6, pricing.InputCostPerToken, 1e-12)
|
|
require.InDelta(t, 5e-6, pricing.InputCostPerTokenPriority, 1e-12)
|
|
require.InDelta(t, 15e-6, pricing.OutputCostPerToken, 1e-12)
|
|
require.InDelta(t, 30e-6, pricing.OutputCostPerTokenPriority, 1e-12)
|
|
require.InDelta(t, 0.25e-6, pricing.CacheReadInputTokenCost, 1e-12)
|
|
require.InDelta(t, 0.5e-6, pricing.CacheReadInputTokenCostPriority, 1e-12)
|
|
require.True(t, pricing.SupportsServiceTier)
|
|
}
|
|
|
|
func TestParsePricingData_PreservesServiceTierPriorityFields(t *testing.T) {
|
|
svc := &PricingService{}
|
|
pricingData, err := svc.parsePricingData([]byte(`{
|
|
"gpt-5.4": {
|
|
"input_cost_per_token": 0.0000025,
|
|
"input_cost_per_token_priority": 0.000005,
|
|
"output_cost_per_token": 0.000015,
|
|
"output_cost_per_token_priority": 0.00003,
|
|
"cache_read_input_token_cost": 0.00000025,
|
|
"cache_read_input_token_cost_priority": 0.0000005,
|
|
"supports_service_tier": true,
|
|
"litellm_provider": "openai",
|
|
"mode": "chat"
|
|
}
|
|
}`))
|
|
require.NoError(t, err)
|
|
|
|
pricing := pricingData["gpt-5.4"]
|
|
require.NotNil(t, pricing)
|
|
require.InDelta(t, 0.0000025, pricing.InputCostPerToken, 1e-12)
|
|
require.InDelta(t, 0.000005, pricing.InputCostPerTokenPriority, 1e-12)
|
|
require.InDelta(t, 0.000015, pricing.OutputCostPerToken, 1e-12)
|
|
require.InDelta(t, 0.00003, pricing.OutputCostPerTokenPriority, 1e-12)
|
|
require.InDelta(t, 0.00000025, pricing.CacheReadInputTokenCost, 1e-12)
|
|
require.InDelta(t, 0.0000005, pricing.CacheReadInputTokenCostPriority, 1e-12)
|
|
require.True(t, pricing.SupportsServiceTier)
|
|
}
|
|
|
|
// ---------------------------------------------------------------------------
|
|
// ListModelNamesByProvider
|
|
// ---------------------------------------------------------------------------
|
|
|
|
func TestListModelNamesByProvider_ReturnsMatchingModels(t *testing.T) {
|
|
svc := &PricingService{
|
|
pricingData: map[string]*LiteLLMModelPricing{
|
|
"claude-opus-4-5-20251101": {LiteLLMProvider: "anthropic", InputCostPerToken: 1.5e-5},
|
|
"claude-sonnet-4-5": {LiteLLMProvider: "anthropic", InputCostPerToken: 3e-6},
|
|
"gpt-4o": {LiteLLMProvider: "openai", InputCostPerToken: 5e-6},
|
|
"gemini-2.5-pro": {LiteLLMProvider: "google", InputCostPerToken: 1.25e-6},
|
|
},
|
|
}
|
|
|
|
got := svc.ListModelNamesByProvider("anthropic")
|
|
require.ElementsMatch(t, []string{"claude-opus-4-5-20251101", "claude-sonnet-4-5"}, got)
|
|
// Must be sorted
|
|
require.Equal(t, "claude-opus-4-5-20251101", got[0])
|
|
require.Equal(t, "claude-sonnet-4-5", got[1])
|
|
}
|
|
|
|
func TestListModelNamesByProvider_CaseInsensitive(t *testing.T) {
|
|
svc := &PricingService{
|
|
pricingData: map[string]*LiteLLMModelPricing{
|
|
"gpt-4o": {LiteLLMProvider: "OpenAI", InputCostPerToken: 5e-6},
|
|
},
|
|
}
|
|
|
|
got := svc.ListModelNamesByProvider("openai")
|
|
require.Equal(t, []string{"gpt-4o"}, got)
|
|
|
|
got2 := svc.ListModelNamesByProvider("OPENAI")
|
|
require.Equal(t, []string{"gpt-4o"}, got2)
|
|
}
|
|
|
|
func TestListModelNamesByProvider_NoMatch(t *testing.T) {
|
|
svc := &PricingService{
|
|
pricingData: map[string]*LiteLLMModelPricing{
|
|
"gpt-4o": {LiteLLMProvider: "openai", InputCostPerToken: 5e-6},
|
|
},
|
|
}
|
|
|
|
got := svc.ListModelNamesByProvider("anthropic")
|
|
require.NotNil(t, got)
|
|
require.Empty(t, got)
|
|
}
|
|
|
|
func TestListModelNamesByProvider_EmptyCatalog(t *testing.T) {
|
|
svc := &PricingService{
|
|
pricingData: map[string]*LiteLLMModelPricing{},
|
|
}
|
|
|
|
got := svc.ListModelNamesByProvider("openai")
|
|
require.NotNil(t, got)
|
|
require.Empty(t, got)
|
|
}
|