mirror of
https://github.com/Wei-Shaw/sub2api.git
synced 2026-11-20 15:06:36 +00:00
merge: integrate upstream v0.1.94 into release/custom-0.1.94
Merge upstream v0.1.94 onto our custom branch (based on release/custom-0.1.93). Key additions beyond upstream: - Startup stale concurrency slot cleanup on process restart - iframe language passthrough for embedded payment/custom pages - embedded-url.spec.ts frontend tests - Admin payment integration API docs update Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
commit
028442d00b
125 files changed
+6345
-603
No files matched your search
@@ -733,6 +733,42 @@ func (h *AccountHandler) Test(c *gin.Context) {
|
||||
// Error already sent via SSE, just log
|
||||
return
|
||||
}
|
||||
|
||||
if h.rateLimitService != nil {
|
||||
if _, err := h.rateLimitService.RecoverAccountAfterSuccessfulTest(c.Request.Context(), accountID); err != nil {
|
||||
_ = c.Error(err)
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// RecoverState handles unified recovery of recoverable account runtime state.
|
||||
// POST /api/v1/admin/accounts/:id/recover-state
|
||||
func (h *AccountHandler) RecoverState(c *gin.Context) {
|
||||
accountID, err := strconv.ParseInt(c.Param("id"), 10, 64)
|
||||
if err != nil {
|
||||
response.BadRequest(c, "Invalid account ID")
|
||||
return
|
||||
}
|
||||
|
||||
if h.rateLimitService == nil {
|
||||
response.Error(c, http.StatusServiceUnavailable, "Rate limit service unavailable")
|
||||
return
|
||||
}
|
||||
|
||||
if _, err := h.rateLimitService.RecoverAccountState(c.Request.Context(), accountID, service.AccountRecoveryOptions{
|
||||
InvalidateToken: true,
|
||||
}); err != nil {
|
||||
response.ErrorFrom(c, err)
|
||||
return
|
||||
}
|
||||
|
||||
account, err := h.adminService.GetAccount(c.Request.Context(), accountID)
|
||||
if err != nil {
|
||||
response.ErrorFrom(c, err)
|
||||
return
|
||||
}
|
||||
|
||||
response.Success(c, h.buildAccountResponseWithRuntime(c.Request.Context(), account))
|
||||
}
|
||||
|
||||
// SyncFromCRS handles syncing accounts from claude-relay-service (CRS)
|
||||
|
||||
@@ -25,6 +25,7 @@ type createScheduledTestPlanRequest struct {
|
||||
CronExpression string `json:"cron_expression" binding:"required"`
|
||||
Enabled *bool `json:"enabled"`
|
||||
MaxResults int `json:"max_results"`
|
||||
AutoRecover *bool `json:"auto_recover"`
|
||||
}
|
||||
|
||||
type updateScheduledTestPlanRequest struct {
|
||||
@@ -32,6 +33,7 @@ type updateScheduledTestPlanRequest struct {
|
||||
CronExpression string `json:"cron_expression"`
|
||||
Enabled *bool `json:"enabled"`
|
||||
MaxResults int `json:"max_results"`
|
||||
AutoRecover *bool `json:"auto_recover"`
|
||||
}
|
||||
|
||||
// ListByAccount GET /admin/accounts/:id/scheduled-test-plans
|
||||
@@ -68,6 +70,9 @@ func (h *ScheduledTestHandler) Create(c *gin.Context) {
|
||||
if req.Enabled != nil {
|
||||
plan.Enabled = *req.Enabled
|
||||
}
|
||||
if req.AutoRecover != nil {
|
||||
plan.AutoRecover = *req.AutoRecover
|
||||
}
|
||||
|
||||
created, err := h.scheduledTestSvc.CreatePlan(c.Request.Context(), plan)
|
||||
if err != nil {
|
||||
@@ -109,6 +114,9 @@ func (h *ScheduledTestHandler) Update(c *gin.Context) {
|
||||
if req.MaxResults > 0 {
|
||||
existing.MaxResults = req.MaxResults
|
||||
}
|
||||
if req.AutoRecover != nil {
|
||||
existing.AutoRecover = *req.AutoRecover
|
||||
}
|
||||
|
||||
updated, err := h.scheduledTestSvc.UpdatePlan(c.Request.Context(), existing)
|
||||
if err != nil {
|
||||
|
||||
@@ -1414,6 +1414,63 @@ func (h *SettingHandler) TestSoraS3Connection(c *gin.Context) {
|
||||
response.Success(c, gin.H{"message": "S3 连接成功"})
|
||||
}
|
||||
|
||||
// GetRectifierSettings 获取请求整流器配置
|
||||
// GET /api/v1/admin/settings/rectifier
|
||||
func (h *SettingHandler) GetRectifierSettings(c *gin.Context) {
|
||||
settings, err := h.settingService.GetRectifierSettings(c.Request.Context())
|
||||
if err != nil {
|
||||
response.ErrorFrom(c, err)
|
||||
return
|
||||
}
|
||||
|
||||
response.Success(c, dto.RectifierSettings{
|
||||
Enabled: settings.Enabled,
|
||||
ThinkingSignatureEnabled: settings.ThinkingSignatureEnabled,
|
||||
ThinkingBudgetEnabled: settings.ThinkingBudgetEnabled,
|
||||
})
|
||||
}
|
||||
|
||||
// UpdateRectifierSettingsRequest 更新整流器配置请求
|
||||
type UpdateRectifierSettingsRequest struct {
|
||||
Enabled bool `json:"enabled"`
|
||||
ThinkingSignatureEnabled bool `json:"thinking_signature_enabled"`
|
||||
ThinkingBudgetEnabled bool `json:"thinking_budget_enabled"`
|
||||
}
|
||||
|
||||
// UpdateRectifierSettings 更新请求整流器配置
|
||||
// PUT /api/v1/admin/settings/rectifier
|
||||
func (h *SettingHandler) UpdateRectifierSettings(c *gin.Context) {
|
||||
var req UpdateRectifierSettingsRequest
|
||||
if err := c.ShouldBindJSON(&req); err != nil {
|
||||
response.BadRequest(c, "Invalid request: "+err.Error())
|
||||
return
|
||||
}
|
||||
|
||||
settings := &service.RectifierSettings{
|
||||
Enabled: req.Enabled,
|
||||
ThinkingSignatureEnabled: req.ThinkingSignatureEnabled,
|
||||
ThinkingBudgetEnabled: req.ThinkingBudgetEnabled,
|
||||
}
|
||||
|
||||
if err := h.settingService.SetRectifierSettings(c.Request.Context(), settings); err != nil {
|
||||
response.BadRequest(c, err.Error())
|
||||
return
|
||||
}
|
||||
|
||||
// 重新获取设置返回
|
||||
updatedSettings, err := h.settingService.GetRectifierSettings(c.Request.Context())
|
||||
if err != nil {
|
||||
response.ErrorFrom(c, err)
|
||||
return
|
||||
}
|
||||
|
||||
response.Success(c, dto.RectifierSettings{
|
||||
Enabled: updatedSettings.Enabled,
|
||||
ThinkingSignatureEnabled: updatedSettings.ThinkingSignatureEnabled,
|
||||
ThinkingBudgetEnabled: updatedSettings.ThinkingBudgetEnabled,
|
||||
})
|
||||
}
|
||||
|
||||
// UpdateStreamTimeoutSettingsRequest 更新流超时配置请求
|
||||
type UpdateStreamTimeoutSettingsRequest struct {
|
||||
Enabled bool `json:"enabled"`
|
||||
|
||||
@@ -71,7 +71,7 @@ func APIKeyFromService(k *service.APIKey) *APIKey {
|
||||
if k == nil {
|
||||
return nil
|
||||
}
|
||||
return &APIKey{
|
||||
out := &APIKey{
|
||||
ID: k.ID,
|
||||
UserID: k.UserID,
|
||||
Key: k.Key,
|
||||
@@ -98,6 +98,19 @@ func APIKeyFromService(k *service.APIKey) *APIKey {
|
||||
User: UserFromServiceShallow(k.User),
|
||||
Group: GroupFromServiceShallow(k.Group),
|
||||
}
|
||||
if k.Window5hStart != nil && !service.IsWindowExpired(k.Window5hStart, service.RateLimitWindow5h) {
|
||||
t := k.Window5hStart.Add(service.RateLimitWindow5h)
|
||||
out.Reset5hAt = &t
|
||||
}
|
||||
if k.Window1dStart != nil && !service.IsWindowExpired(k.Window1dStart, service.RateLimitWindow1d) {
|
||||
t := k.Window1dStart.Add(service.RateLimitWindow1d)
|
||||
out.Reset1dAt = &t
|
||||
}
|
||||
if k.Window7dStart != nil && !service.IsWindowExpired(k.Window7dStart, service.RateLimitWindow7d) {
|
||||
t := k.Window7dStart.Add(service.RateLimitWindow7d)
|
||||
out.Reset7dAt = &t
|
||||
}
|
||||
return out
|
||||
}
|
||||
|
||||
func GroupFromServiceShallow(g *service.Group) *Group {
|
||||
@@ -490,6 +503,7 @@ func usageLogFromServiceUser(l *service.UsageLog) UsageLog {
|
||||
AccountID: l.AccountID,
|
||||
RequestID: l.RequestID,
|
||||
Model: l.Model,
|
||||
ServiceTier: l.ServiceTier,
|
||||
ReasoningEffort: l.ReasoningEffort,
|
||||
GroupID: l.GroupID,
|
||||
SubscriptionID: l.SubscriptionID,
|
||||
|
||||
@@ -71,3 +71,29 @@ func TestRequestTypeStringPtrNil(t *testing.T) {
|
||||
t.Parallel()
|
||||
require.Nil(t, requestTypeStringPtr(nil))
|
||||
}
|
||||
|
||||
func TestUsageLogFromService_IncludesServiceTierForUserAndAdmin(t *testing.T) {
|
||||
t.Parallel()
|
||||
|
||||
serviceTier := "priority"
|
||||
log := &service.UsageLog{
|
||||
RequestID: "req_3",
|
||||
Model: "gpt-5.4",
|
||||
ServiceTier: &serviceTier,
|
||||
AccountRateMultiplier: f64Ptr(1.5),
|
||||
}
|
||||
|
||||
userDTO := UsageLogFromService(log)
|
||||
adminDTO := UsageLogFromServiceAdmin(log)
|
||||
|
||||
require.NotNil(t, userDTO.ServiceTier)
|
||||
require.Equal(t, serviceTier, *userDTO.ServiceTier)
|
||||
require.NotNil(t, adminDTO.ServiceTier)
|
||||
require.Equal(t, serviceTier, *adminDTO.ServiceTier)
|
||||
require.NotNil(t, adminDTO.AccountRateMultiplier)
|
||||
require.InDelta(t, 1.5, *adminDTO.AccountRateMultiplier, 1e-12)
|
||||
}
|
||||
|
||||
func f64Ptr(value float64) *float64 {
|
||||
return &value
|
||||
}
|
||||
@@ -171,6 +171,13 @@ type StreamTimeoutSettings struct {
|
||||
ThresholdWindowMinutes int `json:"threshold_window_minutes"`
|
||||
}
|
||||
|
||||
// RectifierSettings 请求整流器配置 DTO
|
||||
type RectifierSettings struct {
|
||||
Enabled bool `json:"enabled"`
|
||||
ThinkingSignatureEnabled bool `json:"thinking_signature_enabled"`
|
||||
ThinkingBudgetEnabled bool `json:"thinking_budget_enabled"`
|
||||
}
|
||||
|
||||
// ParseCustomMenuItems parses a JSON string into a slice of CustomMenuItem.
|
||||
// Returns empty slice on empty/invalid input.
|
||||
func ParseCustomMenuItems(raw string) []CustomMenuItem {
|
||||
|
||||
@@ -57,6 +57,9 @@ type APIKey struct {
|
||||
Window5hStart *time.Time `json:"window_5h_start"`
|
||||
Window1dStart *time.Time `json:"window_1d_start"`
|
||||
Window7dStart *time.Time `json:"window_7d_start"`
|
||||
Reset5hAt *time.Time `json:"reset_5h_at,omitempty"`
|
||||
Reset1dAt *time.Time `json:"reset_1d_at,omitempty"`
|
||||
Reset7dAt *time.Time `json:"reset_7d_at,omitempty"`
|
||||
|
||||
User *User `json:"user,omitempty"`
|
||||
Group *Group `json:"group,omitempty"`
|
||||
@@ -329,6 +332,8 @@ type UsageLog struct {
|
||||
AccountID int64 `json:"account_id"`
|
||||
RequestID string `json:"request_id"`
|
||||
Model string `json:"model"`
|
||||
// ServiceTier records the OpenAI service tier used for billing, e.g. "priority" / "flex".
|
||||
ServiceTier *string `json:"service_tier,omitempty"`
|
||||
// ReasoningEffort is the request's reasoning effort level (OpenAI Responses API).
|
||||
// nil means not provided / not applicable.
|
||||
ReasoningEffort *string `json:"reasoning_effort,omitempty"`
|
||||
|
||||
@@ -30,7 +30,7 @@ const (
|
||||
|
||||
const (
|
||||
// maxSameAccountRetries 同账号重试次数上限(针对 RetryableOnSameAccount 错误)
|
||||
maxSameAccountRetries = 2
|
||||
maxSameAccountRetries = 3
|
||||
// sameAccountRetryDelay 同账号重试间隔
|
||||
sameAccountRetryDelay = 500 * time.Millisecond
|
||||
// singleAccountBackoffDelay 单账号分组 503 退避重试固定延时。
|
||||
|
||||
@@ -291,35 +291,31 @@ func TestHandleFailoverError_SameAccountRetry(t *testing.T) {
|
||||
require.Less(t, elapsed, 2*time.Second)
|
||||
})
|
||||
|
||||
t.Run("第二次重试仍返回FailoverContinue", func(t *testing.T) {
|
||||
t.Run("达到最大重试次数前均返回FailoverContinue", func(t *testing.T) {
|
||||
mock := &mockTempUnscheduler{}
|
||||
fs := NewFailoverState(3, false)
|
||||
err := newTestFailoverErr(400, true, false)
|
||||
|
||||
// 第一次
|
||||
action := fs.HandleFailoverError(context.Background(), mock, 100, "openai", err)
|
||||
require.Equal(t, FailoverContinue, action)
|
||||
require.Equal(t, 1, fs.SameAccountRetryCount[100])
|
||||
for i := 1; i <= maxSameAccountRetries; i++ {
|
||||
action := fs.HandleFailoverError(context.Background(), mock, 100, "openai", err)
|
||||
require.Equal(t, FailoverContinue, action)
|
||||
require.Equal(t, i, fs.SameAccountRetryCount[100])
|
||||
}
|
||||
|
||||
// 第二次
|
||||
action = fs.HandleFailoverError(context.Background(), mock, 100, "openai", err)
|
||||
require.Equal(t, FailoverContinue, action)
|
||||
require.Equal(t, 2, fs.SameAccountRetryCount[100])
|
||||
|
||||
require.Empty(t, mock.calls, "两次重试期间均不应调用 TempUnschedule")
|
||||
require.Empty(t, mock.calls, "达到最大重试次数前均不应调用 TempUnschedule")
|
||||
})
|
||||
|
||||
t.Run("第三次重试耗尽_触发TempUnschedule并切换", func(t *testing.T) {
|
||||
t.Run("超过最大重试次数后触发TempUnschedule并切换", func(t *testing.T) {
|
||||
mock := &mockTempUnscheduler{}
|
||||
fs := NewFailoverState(3, false)
|
||||
err := newTestFailoverErr(400, true, false)
|
||||
|
||||
// 第一次、第二次重试
|
||||
fs.HandleFailoverError(context.Background(), mock, 100, "openai", err)
|
||||
fs.HandleFailoverError(context.Background(), mock, 100, "openai", err)
|
||||
require.Equal(t, 2, fs.SameAccountRetryCount[100])
|
||||
for i := 0; i < maxSameAccountRetries; i++ {
|
||||
fs.HandleFailoverError(context.Background(), mock, 100, "openai", err)
|
||||
}
|
||||
require.Equal(t, maxSameAccountRetries, fs.SameAccountRetryCount[100])
|
||||
|
||||
// 第三次:重试已达到 maxSameAccountRetries(2),应切换账号
|
||||
// 第 maxSameAccountRetries+1 次:重试耗尽,应切换账号
|
||||
action := fs.HandleFailoverError(context.Background(), mock, 100, "openai", err)
|
||||
require.Equal(t, FailoverContinue, action)
|
||||
require.Equal(t, 1, fs.SwitchCount)
|
||||
@@ -354,13 +350,14 @@ func TestHandleFailoverError_SameAccountRetry(t *testing.T) {
|
||||
err := newTestFailoverErr(400, true, false)
|
||||
|
||||
// 耗尽账号 100 的重试
|
||||
fs.HandleFailoverError(context.Background(), mock, 100, "openai", err)
|
||||
fs.HandleFailoverError(context.Background(), mock, 100, "openai", err)
|
||||
// 第三次: 重试耗尽 → 切换
|
||||
for i := 0; i < maxSameAccountRetries; i++ {
|
||||
fs.HandleFailoverError(context.Background(), mock, 100, "openai", err)
|
||||
}
|
||||
// 第 maxSameAccountRetries+1 次: 重试耗尽 → 切换
|
||||
action := fs.HandleFailoverError(context.Background(), mock, 100, "openai", err)
|
||||
require.Equal(t, FailoverContinue, action)
|
||||
|
||||
// 再次遇到账号 100,计数仍为 2,条件不满足 → 直接切换
|
||||
// 再次遇到账号 100,计数仍为 maxSameAccountRetries,条件不满足 → 直接切换
|
||||
action = fs.HandleFailoverError(context.Background(), mock, 100, "openai", err)
|
||||
require.Equal(t, FailoverContinue, action)
|
||||
require.Len(t, mock.calls, 2, "第二次耗尽也应调用 TempUnschedule")
|
||||
@@ -386,9 +383,10 @@ func TestHandleFailoverError_TempUnschedule(t *testing.T) {
|
||||
fs := NewFailoverState(3, false)
|
||||
err := newTestFailoverErr(502, true, false)
|
||||
|
||||
// 耗尽重试
|
||||
fs.HandleFailoverError(context.Background(), mock, 42, "openai", err)
|
||||
fs.HandleFailoverError(context.Background(), mock, 42, "openai", err)
|
||||
for i := 0; i < maxSameAccountRetries; i++ {
|
||||
fs.HandleFailoverError(context.Background(), mock, 42, "openai", err)
|
||||
}
|
||||
// 再次触发时才会执行 TempUnschedule + 切换
|
||||
fs.HandleFailoverError(context.Background(), mock, 42, "openai", err)
|
||||
|
||||
require.Len(t, mock.calls, 1)
|
||||
@@ -521,17 +519,16 @@ func TestHandleFailoverError_IntegrationScenario(t *testing.T) {
|
||||
mock := &mockTempUnscheduler{}
|
||||
fs := NewFailoverState(3, true) // hasBoundSession=true
|
||||
|
||||
// 1. 账号 100 遇到可重试错误,同账号重试 2 次
|
||||
// 1. 账号 100 遇到可重试错误,同账号重试 maxSameAccountRetries 次
|
||||
retryErr := newTestFailoverErr(400, true, false)
|
||||
action := fs.HandleFailoverError(context.Background(), mock, 100, "openai", retryErr)
|
||||
require.Equal(t, FailoverContinue, action)
|
||||
for i := 0; i < maxSameAccountRetries; i++ {
|
||||
action := fs.HandleFailoverError(context.Background(), mock, 100, "openai", retryErr)
|
||||
require.Equal(t, FailoverContinue, action)
|
||||
}
|
||||
require.True(t, fs.ForceCacheBilling, "hasBoundSession=true 应设置 ForceCacheBilling")
|
||||
|
||||
action = fs.HandleFailoverError(context.Background(), mock, 100, "openai", retryErr)
|
||||
require.Equal(t, FailoverContinue, action)
|
||||
|
||||
// 2. 账号 100 重试耗尽 → TempUnschedule + 切换
|
||||
action = fs.HandleFailoverError(context.Background(), mock, 100, "openai", retryErr)
|
||||
// 2. 账号 100 超过重试上限 → TempUnschedule + 切换
|
||||
action := fs.HandleFailoverError(context.Background(), mock, 100, "openai", retryErr)
|
||||
require.Equal(t, FailoverContinue, action)
|
||||
require.Equal(t, 1, fs.SwitchCount)
|
||||
require.Len(t, mock.calls, 1)
|
||||
|
||||
@@ -975,33 +975,45 @@ func (h *GatewayHandler) usageQuotaLimited(c *gin.Context, ctx context.Context,
|
||||
var rateLimits []gin.H
|
||||
if apiKey.RateLimit5h > 0 {
|
||||
used := rateLimitData.EffectiveUsage5h()
|
||||
rateLimits = append(rateLimits, gin.H{
|
||||
entry := gin.H{
|
||||
"window": "5h",
|
||||
"limit": apiKey.RateLimit5h,
|
||||
"used": used,
|
||||
"remaining": max(0, apiKey.RateLimit5h-used),
|
||||
"window_start": rateLimitData.Window5hStart,
|
||||
})
|
||||
}
|
||||
if rateLimitData.Window5hStart != nil && !service.IsWindowExpired(rateLimitData.Window5hStart, service.RateLimitWindow5h) {
|
||||
entry["reset_at"] = rateLimitData.Window5hStart.Add(service.RateLimitWindow5h)
|
||||
}
|
||||
rateLimits = append(rateLimits, entry)
|
||||
}
|
||||
if apiKey.RateLimit1d > 0 {
|
||||
used := rateLimitData.EffectiveUsage1d()
|
||||
rateLimits = append(rateLimits, gin.H{
|
||||
entry := gin.H{
|
||||
"window": "1d",
|
||||
"limit": apiKey.RateLimit1d,
|
||||
"used": used,
|
||||
"remaining": max(0, apiKey.RateLimit1d-used),
|
||||
"window_start": rateLimitData.Window1dStart,
|
||||
})
|
||||
}
|
||||
if rateLimitData.Window1dStart != nil && !service.IsWindowExpired(rateLimitData.Window1dStart, service.RateLimitWindow1d) {
|
||||
entry["reset_at"] = rateLimitData.Window1dStart.Add(service.RateLimitWindow1d)
|
||||
}
|
||||
rateLimits = append(rateLimits, entry)
|
||||
}
|
||||
if apiKey.RateLimit7d > 0 {
|
||||
used := rateLimitData.EffectiveUsage7d()
|
||||
rateLimits = append(rateLimits, gin.H{
|
||||
entry := gin.H{
|
||||
"window": "7d",
|
||||
"limit": apiKey.RateLimit7d,
|
||||
"used": used,
|
||||
"remaining": max(0, apiKey.RateLimit7d-used),
|
||||
"window_start": rateLimitData.Window7dStart,
|
||||
})
|
||||
}
|
||||
if rateLimitData.Window7dStart != nil && !service.IsWindowExpired(rateLimitData.Window7dStart, service.RateLimitWindow7d) {
|
||||
entry["reset_at"] = rateLimitData.Window7dStart.Add(service.RateLimitWindow7d)
|
||||
}
|
||||
rateLimits = append(rateLimits, entry)
|
||||
}
|
||||
if len(rateLimits) > 0 {
|
||||
resp["rate_limits"] = rateLimits
|
||||
|
||||
@@ -127,6 +127,7 @@ func (f *fakeConcurrencyCache) GetAccountConcurrencyBatch(_ context.Context, acc
|
||||
return result, nil
|
||||
}
|
||||
func (f *fakeConcurrencyCache) CleanupExpiredAccountSlots(context.Context, int64) error { return nil }
|
||||
func (f *fakeConcurrencyCache) CleanupStaleProcessSlots(context.Context, string) error { return nil }
|
||||
|
||||
func newTestGatewayHandler(t *testing.T, group *service.Group, accounts []*service.Account) (*GatewayHandler, func()) {
|
||||
t.Helper()
|
||||
@@ -155,6 +156,7 @@ func newTestGatewayHandler(t *testing.T, group *service.Group, accounts []*servi
|
||||
nil, // sessionLimitCache
|
||||
nil, // rpmCache
|
||||
nil, // digestStore
|
||||
nil, // settingService
|
||||
)
|
||||
|
||||
// RunModeSimple:跳过计费检查,避免引入 repo/cache 依赖。
|
||||
|
||||
@@ -89,6 +89,10 @@ func (m *concurrencyCacheMock) CleanupExpiredAccountSlots(ctx context.Context, a
|
||||
return nil
|
||||
}
|
||||
|
||||
func (m *concurrencyCacheMock) CleanupStaleProcessSlots(ctx context.Context, activeRequestPrefix string) error {
|
||||
return nil
|
||||
}
|
||||
|
||||
func TestConcurrencyHelper_TryAcquireUserSlot(t *testing.T) {
|
||||
cache := &concurrencyCacheMock{
|
||||
acquireUserSlotFn: func(ctx context.Context, userID int64, maxConcurrency int, requestID string) (bool, error) {
|
||||
|
||||
@@ -120,6 +120,10 @@ func (s *helperConcurrencyCacheStub) CleanupExpiredAccountSlots(ctx context.Cont
|
||||
return nil
|
||||
}
|
||||
|
||||
func (s *helperConcurrencyCacheStub) CleanupStaleProcessSlots(ctx context.Context, activeRequestPrefix string) error {
|
||||
return nil
|
||||
}
|
||||
|
||||
func newHelperTestContext(method, path string) (*gin.Context, *httptest.ResponseRecorder) {
|
||||
gin.SetMode(gin.TestMode)
|
||||
rec := httptest.NewRecorder()
|
||||
|
||||
@@ -20,6 +20,7 @@ import (
|
||||
|
||||
coderws "github.com/coder/websocket"
|
||||
"github.com/gin-gonic/gin"
|
||||
"github.com/google/uuid"
|
||||
"github.com/tidwall/gjson"
|
||||
"go.uber.org/zap"
|
||||
)
|
||||
@@ -212,6 +213,7 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
|
||||
maxAccountSwitches := h.maxAccountSwitches
|
||||
switchCount := 0
|
||||
failedAccountIDs := make(map[int64]struct{})
|
||||
sameAccountRetryCount := make(map[int64]int)
|
||||
var lastFailoverErr *service.UpstreamFailoverError
|
||||
|
||||
for {
|
||||
@@ -259,6 +261,7 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
|
||||
zap.Float64("load_skew", scheduleDecision.LoadSkew),
|
||||
)
|
||||
account := selection.Account
|
||||
sessionHash = ensureOpenAIPoolModeSessionHash(sessionHash, account)
|
||||
reqLog.Debug("openai.account_selected", zap.Int64("account_id", account.ID), zap.String("account_name", account.Name))
|
||||
setOpsSelectedAccount(c, account.ID, account.Platform)
|
||||
|
||||
@@ -288,6 +291,25 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
|
||||
var failoverErr *service.UpstreamFailoverError
|
||||
if errors.As(err, &failoverErr) {
|
||||
h.gatewayService.ReportOpenAIAccountScheduleResult(account.ID, false, nil)
|
||||
// 池模式:同账号重试
|
||||
if failoverErr.RetryableOnSameAccount {
|
||||
retryLimit := account.GetPoolModeRetryCount()
|
||||
if sameAccountRetryCount[account.ID] < retryLimit {
|
||||
sameAccountRetryCount[account.ID]++
|
||||
reqLog.Warn("openai.pool_mode_same_account_retry",
|
||||
zap.Int64("account_id", account.ID),
|
||||
zap.Int("upstream_status", failoverErr.StatusCode),
|
||||
zap.Int("retry_limit", retryLimit),
|
||||
zap.Int("retry_count", sameAccountRetryCount[account.ID]),
|
||||
)
|
||||
select {
|
||||
case <-c.Request.Context().Done():
|
||||
return
|
||||
case <-time.After(sameAccountRetryDelay):
|
||||
}
|
||||
continue
|
||||
}
|
||||
}
|
||||
h.gatewayService.RecordOpenAIAccountSwitch()
|
||||
failedAccountIDs[account.ID] = struct{}{}
|
||||
lastFailoverErr = failoverErr
|
||||
@@ -538,9 +560,25 @@ func (h *OpenAIGatewayHandler) Messages(c *gin.Context) {
|
||||
sessionHash := h.gatewayService.GenerateSessionHash(c, body)
|
||||
promptCacheKey := h.gatewayService.ExtractSessionID(c, body)
|
||||
|
||||
// Anthropic 格式的请求在 metadata.user_id 中携带 session 标识,
|
||||
// 而非 OpenAI 的 session_id/conversation_id headers。
|
||||
// 从中派生 sessionHash(sticky session)和 promptCacheKey(upstream cache)。
|
||||
if sessionHash == "" || promptCacheKey == "" {
|
||||
if userID := strings.TrimSpace(gjson.GetBytes(body, "metadata.user_id").String()); userID != "" {
|
||||
seed := reqModel + "-" + userID
|
||||
if promptCacheKey == "" {
|
||||
promptCacheKey = service.GenerateSessionUUID(seed)
|
||||
}
|
||||
if sessionHash == "" {
|
||||
sessionHash = service.DeriveSessionHashFromSeed(seed)
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
maxAccountSwitches := h.maxAccountSwitches
|
||||
switchCount := 0
|
||||
failedAccountIDs := make(map[int64]struct{})
|
||||
sameAccountRetryCount := make(map[int64]int)
|
||||
var lastFailoverErr *service.UpstreamFailoverError
|
||||
|
||||
for {
|
||||
@@ -602,6 +640,7 @@ func (h *OpenAIGatewayHandler) Messages(c *gin.Context) {
|
||||
return
|
||||
}
|
||||
account := selection.Account
|
||||
sessionHash = ensureOpenAIPoolModeSessionHash(sessionHash, account)
|
||||
reqLog.Debug("openai_messages.account_selected", zap.Int64("account_id", account.ID), zap.String("account_name", account.Name))
|
||||
_ = scheduleDecision
|
||||
setOpsSelectedAccount(c, account.ID, account.Platform)
|
||||
@@ -641,6 +680,25 @@ func (h *OpenAIGatewayHandler) Messages(c *gin.Context) {
|
||||
var failoverErr *service.UpstreamFailoverError
|
||||
if errors.As(err, &failoverErr) {
|
||||
h.gatewayService.ReportOpenAIAccountScheduleResult(account.ID, false, nil)
|
||||
// 池模式:同账号重试
|
||||
if failoverErr.RetryableOnSameAccount {
|
||||
retryLimit := account.GetPoolModeRetryCount()
|
||||
if sameAccountRetryCount[account.ID] < retryLimit {
|
||||
sameAccountRetryCount[account.ID]++
|
||||
reqLog.Warn("openai_messages.pool_mode_same_account_retry",
|
||||
zap.Int64("account_id", account.ID),
|
||||
zap.Int("upstream_status", failoverErr.StatusCode),
|
||||
zap.Int("retry_limit", retryLimit),
|
||||
zap.Int("retry_count", sameAccountRetryCount[account.ID]),
|
||||
)
|
||||
select {
|
||||
case <-c.Request.Context().Done():
|
||||
return
|
||||
case <-time.After(sameAccountRetryDelay):
|
||||
}
|
||||
continue
|
||||
}
|
||||
}
|
||||
h.gatewayService.RecordOpenAIAccountSwitch()
|
||||
failedAccountIDs[account.ID] = struct{}{}
|
||||
lastFailoverErr = failoverErr
|
||||
@@ -1456,6 +1514,14 @@ func setOpenAIClientTransportWS(c *gin.Context) {
|
||||
service.SetOpenAIClientTransport(c, service.OpenAIClientTransportWS)
|
||||
}
|
||||
|
||||
func ensureOpenAIPoolModeSessionHash(sessionHash string, account *service.Account) string {
|
||||
if sessionHash != "" || account == nil || !account.IsPoolMode() {
|
||||
return sessionHash
|
||||
}
|
||||
// 为当前请求生成一次性粘性会话键,确保同账号重试不会重新负载均衡到其他账号。
|
||||
return "openai-pool-retry-" + uuid.NewString()
|
||||
}
|
||||
|
||||
func openAIWSIngressFallbackSessionSeed(userID, apiKeyID int64, groupID *int64) string {
|
||||
gid := int64(0)
|
||||
if groupID != nil {
|
||||
|
||||
@@ -2210,7 +2210,7 @@ func (s *stubSoraClientForHandler) GetVideoTask(_ context.Context, _ *service.Ac
|
||||
func newMinimalGatewayService(accountRepo service.AccountRepository) *service.GatewayService {
|
||||
return service.NewGatewayService(
|
||||
accountRepo, nil, nil, nil, nil, nil, nil, nil,
|
||||
nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil,
|
||||
nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil, nil,
|
||||
)
|
||||
}
|
||||
|
||||
|
||||
@@ -445,6 +445,7 @@ func TestSoraGatewayHandler_ChatCompletions(t *testing.T) {
|
||||
testutil.StubSessionLimitCache{},
|
||||
nil, // rpmCache
|
||||
nil, // digestStore
|
||||
nil, // settingService
|
||||
)
|
||||
|
||||
soraClient := &stubSoraClient{imageURLs: []string{"https://example.com/a.png"}}
|
||||
|
||||
Reference in new issue
Block a user