diff --git a/backend/internal/handler/failover_loop.go b/backend/internal/handler/failover_loop.go index b2583301..6d8ddc72 100644 --- a/backend/internal/handler/failover_loop.go +++ b/backend/internal/handler/failover_loop.go @@ -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。 diff --git a/backend/internal/handler/failover_loop_test.go b/backend/internal/handler/failover_loop_test.go index 5a41b2dd..2c65ebc2 100644 --- a/backend/internal/handler/failover_loop_test.go +++ b/backend/internal/handler/failover_loop_test.go @@ -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) diff --git a/backend/internal/handler/openai_gateway_handler.go b/backend/internal/handler/openai_gateway_handler.go index ffd8b166..2328ffda 100644 --- a/backend/internal/handler/openai_gateway_handler.go +++ b/backend/internal/handler/openai_gateway_handler.go @@ -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 @@ -541,6 +563,7 @@ func (h *OpenAIGatewayHandler) Messages(c *gin.Context) { maxAccountSwitches := h.maxAccountSwitches switchCount := 0 failedAccountIDs := make(map[int64]struct{}) + sameAccountRetryCount := make(map[int64]int) var lastFailoverErr *service.UpstreamFailoverError for { @@ -602,6 +625,7 @@ 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 +665,25 @@ 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 +1499,14 @@ 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 { diff --git a/backend/internal/service/account.go b/backend/internal/service/account.go index fdef2f6c..9d4f73d4 100644 --- a/backend/internal/service/account.go +++ b/backend/internal/service/account.go @@ -647,6 +647,75 @@ func (a *Account) IsCustomErrorCodesEnabled() bool { return false } +// IsPoolMode 检查 API Key 账号是否启用池模式。 +// 池模式下，上游错误不标记本地账号状态，而是在同一账号上重试。 +func (a *Account) IsPoolMode() bool { + if a.Type != AccountTypeAPIKey || a.Credentials == nil { + return false + } + if v, ok := a.Credentials["pool_mode"]; ok { + if enabled, ok := v.(bool); ok { + return enabled + } + } + return false +} + +const ( + defaultPoolModeRetryCount = 3 + maxPoolModeRetryCount = 10 +) + +// GetPoolModeRetryCount 返回池模式同账号重试次数。 +// 未配置或配置非法时回退为默认值 3；小于 0 按 0 处理；过大则截断到 10。 +func (a *Account) GetPoolModeRetryCount() int { + if a == nil || !a.IsPoolMode() || a.Credentials == nil { + return defaultPoolModeRetryCount + } + raw, ok := a.Credentials["pool_mode_retry_count"] + if !ok || raw == nil { + return defaultPoolModeRetryCount + } + count := parsePoolModeRetryCount(raw) + if count < 0 { + return 0 + } + if count > maxPoolModeRetryCount { + return maxPoolModeRetryCount + } + return count +} + +func parsePoolModeRetryCount(value any) int { + switch v := value.(type) { + case int: + return v + case int64: + return int(v) + case float64: + return int(v) + case json.Number: + if i, err := v.Int64(); err == nil { + return int(i) + } + case string: + if i, err := strconv.Atoi(strings.TrimSpace(v)); err == nil { + return i + } + } + return defaultPoolModeRetryCount +} + +// isPoolModeRetryableStatus 池模式下应触发同账号重试的状态码 +func isPoolModeRetryableStatus(statusCode int) bool { + switch statusCode { + case 401, 403, 429: + return true + default: + return false + } +} + func (a *Account) GetCustomErrorCodes() []int { if a.Credentials == nil { return nil diff --git a/backend/internal/service/account_pool_mode_test.go b/backend/internal/service/account_pool_mode_test.go new file mode 100644 index 00000000..98429bb1 --- /dev/null +++ b/backend/internal/service/account_pool_mode_test.go @@ -0,0 +1,117 @@ +//go:build unit + +package service + +import ( + "encoding/json" + "testing" + + "github.com/stretchr/testify/require" +) + +func TestGetPoolModeRetryCount(t *testing.T) { + tests := []struct { + name string + account *Account + expected int + }{ + { + name: "default_when_not_pool_mode", + account: &Account{ + Type: AccountTypeAPIKey, + Platform: PlatformOpenAI, + Credentials: map[string]any{}, + }, + expected: defaultPoolModeRetryCount, + }, + { + name: "default_when_missing_retry_count", + account: &Account{ + Type: AccountTypeAPIKey, + Platform: PlatformOpenAI, + Credentials: map[string]any{ + "pool_mode": true, + }, + }, + expected: defaultPoolModeRetryCount, + }, + { + name: "supports_float64_from_json_credentials", + account: &Account{ + Type: AccountTypeAPIKey, + Platform: PlatformOpenAI, + Credentials: map[string]any{ + "pool_mode": true, + "pool_mode_retry_count": float64(5), + }, + }, + expected: 5, + }, + { + name: "supports_json_number", + account: &Account{ + Type: AccountTypeAPIKey, + Platform: PlatformOpenAI, + Credentials: map[string]any{ + "pool_mode": true, + "pool_mode_retry_count": json.Number("4"), + }, + }, + expected: 4, + }, + { + name: "supports_string_value", + account: &Account{ + Type: AccountTypeAPIKey, + Platform: PlatformOpenAI, + Credentials: map[string]any{ + "pool_mode": true, + "pool_mode_retry_count": "2", + }, + }, + expected: 2, + }, + { + name: "negative_value_is_clamped_to_zero", + account: &Account{ + Type: AccountTypeAPIKey, + Platform: PlatformOpenAI, + Credentials: map[string]any{ + "pool_mode": true, + "pool_mode_retry_count": -1, + }, + }, + expected: 0, + }, + { + name: "oversized_value_is_clamped_to_max", + account: &Account{ + Type: AccountTypeAPIKey, + Platform: PlatformOpenAI, + Credentials: map[string]any{ + "pool_mode": true, + "pool_mode_retry_count": 99, + }, + }, + expected: maxPoolModeRetryCount, + }, + { + name: "invalid_value_falls_back_to_default", + account: &Account{ + Type: AccountTypeAPIKey, + Platform: PlatformOpenAI, + Credentials: map[string]any{ + "pool_mode": true, + "pool_mode_retry_count": "oops", + }, + }, + expected: defaultPoolModeRetryCount, + }, + } + + for _, tt := range tests { + t.Run(tt.name, func(t *testing.T) { + require.Equal(t, tt.expected, tt.account.GetPoolModeRetryCount()) + }) + } +} diff --git a/backend/internal/service/error_policy_test.go b/backend/internal/service/error_policy_test.go index 59375cf5..dd9850bd 100644 --- a/backend/internal/service/error_policy_test.go +++ b/backend/internal/service/error_policy_test.go @@ -177,6 +177,36 @@ func TestCheckErrorPolicy(t *testing.T) { body: []byte(`overloaded`), expected: ErrorPolicyMatched, // custom codes take precedence }, + { + name: "pool_mode_custom_error_codes_hit_returns_matched", + account: &Account{ + ID: 7, + Type: AccountTypeAPIKey, + Platform: PlatformOpenAI, + Credentials: map[string]any{ + "pool_mode": true, + "custom_error_codes_enabled": true, + "custom_error_codes": []any{float64(401), float64(403)}, + }, + }, + statusCode: 401, + body: []byte(`unauthorized`), + expected: ErrorPolicyMatched, + }, + { + name: "pool_mode_without_custom_error_codes_returns_skipped", + account: &Account{ + ID: 8, + Type: AccountTypeAPIKey, + Platform: PlatformOpenAI, + Credentials: map[string]any{ + "pool_mode": true, + }, + }, + statusCode: 401, + body: []byte(`unauthorized`), + expected: ErrorPolicySkipped, + }, } for _, tt := range tests { @@ -190,6 +220,48 @@ func TestCheckErrorPolicy(t *testing.T) { } } +func TestHandleUpstreamError_PoolModeCustomErrorCodesOverride(t *testing.T) { + t.Run("pool_mode_without_custom_error_codes_still_skips", func(t *testing.T) { + repo := &errorPolicyRepoStub{} + svc := NewRateLimitService(repo, nil, &config.Config{}, nil, nil) + account := &Account{ + ID: 30, + Type: AccountTypeAPIKey, + Platform: PlatformOpenAI, + Credentials: map[string]any{ + "pool_mode": true, + }, + } + + shouldDisable := svc.HandleUpstreamError(context.Background(), account, 401, http.Header{}, []byte("unauthorized")) + + require.False(t, shouldDisable) + require.Equal(t, 0, repo.setErrCalls) + require.Equal(t, 0, repo.tempCalls) + }) + + t.Run("pool_mode_with_custom_error_codes_uses_local_error_policy", func(t *testing.T) { + repo := &errorPolicyRepoStub{} + svc := NewRateLimitService(repo, nil, &config.Config{}, nil, nil) + account := &Account{ + ID: 31, + Type: AccountTypeAPIKey, + Platform: PlatformOpenAI, + Credentials: map[string]any{ + "pool_mode": true, + "custom_error_codes_enabled": true, + "custom_error_codes": []any{float64(401)}, + }, + } + + shouldDisable := svc.HandleUpstreamError(context.Background(), account, 401, http.Header{}, []byte("unauthorized")) + + require.True(t, shouldDisable) + require.Equal(t, 1, repo.setErrCalls) + require.Equal(t, 0, repo.tempCalls) + }) +} + // --------------------------------------------------------------------------- // TestApplyErrorPolicy — 4 table-driven cases for the wrapper method // --------------------------------------------------------------------------- diff --git a/backend/internal/service/gateway_service.go b/backend/internal/service/gateway_service.go index 06b182cf..c6bec806 100644 --- a/backend/internal/service/gateway_service.go +++ b/backend/internal/service/gateway_service.go @@ -4319,7 +4319,11 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A return "" }(), }) - return nil, &UpstreamFailoverError{StatusCode: resp.StatusCode, ResponseBody: respBody} + return nil, &UpstreamFailoverError{ + StatusCode: resp.StatusCode, + ResponseBody: respBody, + RetryableOnSameAccount: account.IsPoolMode() && isPoolModeRetryableStatus(resp.StatusCode), + } } return s.handleRetryExhaustedError(ctx, resp, c, account) } @@ -4349,7 +4353,11 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A return "" }(), }) - return nil, &UpstreamFailoverError{StatusCode: resp.StatusCode, ResponseBody: respBody} + return nil, &UpstreamFailoverError{ + StatusCode: resp.StatusCode, + ResponseBody: respBody, + RetryableOnSameAccount: account.IsPoolMode() && isPoolModeRetryableStatus(resp.StatusCode), + } } if resp.StatusCode >= 400 { // 可选：对部分 400 触发 failover（默认关闭以保持语义） @@ -4584,7 +4592,11 @@ func (s *GatewayService) forwardAnthropicAPIKeyPassthrough( return "" }(), }) - return nil, &UpstreamFailoverError{StatusCode: resp.StatusCode, ResponseBody: respBody} + return nil, &UpstreamFailoverError{ + StatusCode: resp.StatusCode, + ResponseBody: respBody, + RetryableOnSameAccount: account.IsPoolMode() && isPoolModeRetryableStatus(resp.StatusCode), + } } return s.handleRetryExhaustedError(ctx, resp, c, account) } @@ -4614,7 +4626,11 @@ func (s *GatewayService) forwardAnthropicAPIKeyPassthrough( return "" }(), }) - return nil, &UpstreamFailoverError{StatusCode: resp.StatusCode, ResponseBody: respBody} + return nil, &UpstreamFailoverError{ + StatusCode: resp.StatusCode, + ResponseBody: respBody, + RetryableOnSameAccount: account.IsPoolMode() && isPoolModeRetryableStatus(resp.StatusCode), + } } if resp.StatusCode >= 400 { diff --git a/backend/internal/service/openai_gateway_service.go b/backend/internal/service/openai_gateway_service.go index 709ee808..63cdb02d 100644 --- a/backend/internal/service/openai_gateway_service.go +++ b/backend/internal/service/openai_gateway_service.go @@ -2040,7 +2040,11 @@ func (s *OpenAIGatewayService) Forward(ctx context.Context, c *gin.Context, acco }) s.handleFailoverSideEffects(ctx, resp, account) - return nil, &UpstreamFailoverError{StatusCode: resp.StatusCode, ResponseBody: respBody} + return nil, &UpstreamFailoverError{ + StatusCode: resp.StatusCode, + ResponseBody: respBody, + RetryableOnSameAccount: account.IsPoolMode() && isPoolModeRetryableStatus(resp.StatusCode), + } } return s.handleErrorResponse(ctx, resp, c, account, body) } @@ -2853,7 +2857,11 @@ func (s *OpenAIGatewayService) handleErrorResponse( Detail: upstreamDetail, }) if shouldDisable { - return nil, &UpstreamFailoverError{StatusCode: resp.StatusCode, ResponseBody: body} + return nil, &UpstreamFailoverError{ + StatusCode: resp.StatusCode, + ResponseBody: body, + RetryableOnSameAccount: account.IsPoolMode() && isPoolModeRetryableStatus(resp.StatusCode), + } } // Return appropriate error response diff --git a/backend/internal/service/ratelimit_service.go b/backend/internal/service/ratelimit_service.go index 64ab7d95..5df2d639 100644 --- a/backend/internal/service/ratelimit_service.go +++ b/backend/internal/service/ratelimit_service.go @@ -98,6 +98,9 @@ func (s *RateLimitService) CheckErrorPolicy(ctx context.Context, account *Accoun slog.Info("account_error_code_skipped", "account_id", account.ID, "status_code", statusCode) return ErrorPolicySkipped } + if account.IsPoolMode() { + return ErrorPolicySkipped + } if s.tryTempUnschedulable(ctx, account, statusCode, responseBody) { return ErrorPolicyTempUnscheduled } @@ -107,9 +110,16 @@ func (s *RateLimitService) CheckErrorPolicy(ctx context.Context, account *Accoun // HandleUpstreamError 处理上游错误响应，标记账号状态 // 返回是否应该停止该账号的调度 func (s *RateLimitService) HandleUpstreamError(ctx context.Context, account *Account, statusCode int, headers http.Header, responseBody []byte) (shouldDisable bool) { + customErrorCodesEnabled := account.IsCustomErrorCodesEnabled() + + // 池模式默认不标记本地账号状态；仅当用户显式配置自定义错误码时按本地策略处理。 + if account.IsPoolMode() && !customErrorCodesEnabled { + slog.Info("pool_mode_error_skipped", "account_id", account.ID, "status_code", statusCode) + return false + } + // apikey 类型账号：检查自定义错误码配置 // 如果启用且错误码不在列表中，则不处理（不停止调度、不标记限流/过载） - customErrorCodesEnabled := account.IsCustomErrorCodesEnabled() if !account.ShouldHandleErrorCode(statusCode) { slog.Info("account_error_code_skipped", "account_id", account.ID, "status_code", statusCode) return false diff --git a/frontend/src/components/account/CreateAccountModal.vue b/frontend/src/components/account/CreateAccountModal.vue index 835ec853..8423c1b9 100644 --- a/frontend/src/components/account/CreateAccountModal.vue +++ b/frontend/src/components/account/CreateAccountModal.vue @@ -1127,6 +1127,58 @@ + +
+
+
+ +

+ {{ t('admin.accounts.poolModeHint') }} +

+
+ +
+
+

+ + {{ t('admin.accounts.poolModeInfo') }} +

+
+
+ + +

+ {{ + t('admin.accounts.poolModeRetryCountHint', { + default: DEFAULT_POOL_MODE_RETRY_COUNT, + max: MAX_POOL_MODE_RETRY_COUNT + }) + }} +

+
+
+
@@ -2629,6 +2681,10 @@ const editQuotaWeeklyLimit = ref(null) const modelMappings = ref([]) const modelRestrictionMode = ref<'whitelist' | 'mapping'>('whitelist') const allowedModels = ref([]) +const DEFAULT_POOL_MODE_RETRY_COUNT = 3 +const MAX_POOL_MODE_RETRY_COUNT = 10 +const poolModeEnabled = ref(false) +const poolModeRetryCount = ref(DEFAULT_POOL_MODE_RETRY_COUNT) const customErrorCodesEnabled = ref(false) const selectedErrorCodes = ref([]) const customErrorCodeInput = ref(null) @@ -3300,6 +3356,8 @@ const resetForm = () => { fetchAntigravityDefaultMappings().then(mappings => { antigravityModelMappings.value = [...mappings] }) + poolModeEnabled.value = false + poolModeRetryCount.value = DEFAULT_POOL_MODE_RETRY_COUNT customErrorCodesEnabled.value = false selectedErrorCodes.value = [] customErrorCodeInput.value = null @@ -3452,6 +3510,20 @@ const handleMixedChannelCancel = () => { clearMixedChannelDialog() } +const normalizePoolModeRetryCount = (value: number) => { + if (!Number.isFinite(value)) { + return DEFAULT_POOL_MODE_RETRY_COUNT + } + const normalized = Math.trunc(value) + if (normalized < 0) { + return 0 + } + if (normalized > MAX_POOL_MODE_RETRY_COUNT) { + return MAX_POOL_MODE_RETRY_COUNT + } + return normalized +} + const handleSubmit = async () => { // For OAuth-based type, handle OAuth flow (goes to step 2) if (isOAuthFlow.value) { @@ -3551,6 +3623,12 @@ const handleSubmit = async () => { } } + // Add pool mode if enabled + if (poolModeEnabled.value) { + credentials.pool_mode = true + credentials.pool_mode_retry_count = normalizePoolModeRetryCount(poolModeRetryCount.value) + } + // Add custom error codes if enabled if (customErrorCodesEnabled.value) { credentials.custom_error_codes_enabled = true diff --git a/frontend/src/components/account/EditAccountModal.vue b/frontend/src/components/account/EditAccountModal.vue index 200f3c3c..1f2e988c 100644 --- a/frontend/src/components/account/EditAccountModal.vue +++ b/frontend/src/components/account/EditAccountModal.vue @@ -251,6 +251,58 @@
+ +
+
+
+ +

+ {{ t('admin.accounts.poolModeHint') }} +

+
+ +
+
+

+ + {{ t('admin.accounts.poolModeInfo') }} +

+
+
+ + +

+ {{ + t('admin.accounts.poolModeRetryCountHint', { + default: DEFAULT_POOL_MODE_RETRY_COUNT, + max: MAX_POOL_MODE_RETRY_COUNT + }) + }} +

+
+
+
@@ -1498,6 +1550,10 @@ const editApiKey = ref('') const modelMappings = ref([]) const modelRestrictionMode = ref<'whitelist' | 'mapping'>('whitelist') const allowedModels = ref([]) +const DEFAULT_POOL_MODE_RETRY_COUNT = 3 +const MAX_POOL_MODE_RETRY_COUNT = 10 +const poolModeEnabled = ref(false) +const poolModeRetryCount = ref(DEFAULT_POOL_MODE_RETRY_COUNT) const customErrorCodesEnabled = ref(false) const selectedErrorCodes = ref([]) const customErrorCodeInput = ref(null) @@ -1658,6 +1714,20 @@ const expiresAtInput = computed({ }) // Watchers +const normalizePoolModeRetryCount = (value: number) => { + if (!Number.isFinite(value)) { + return DEFAULT_POOL_MODE_RETRY_COUNT + } + const normalized = Math.trunc(value) + if (normalized < 0) { + return 0 + } + if (normalized > MAX_POOL_MODE_RETRY_COUNT) { + return MAX_POOL_MODE_RETRY_COUNT + } + return normalized +} + watch( () => props.account, (newAccount) => { @@ -1805,6 +1875,12 @@ watch( allowedModels.value = [] } + // Load pool mode + poolModeEnabled.value = credentials.pool_mode === true + poolModeRetryCount.value = normalizePoolModeRetryCount( + Number(credentials.pool_mode_retry_count ?? DEFAULT_POOL_MODE_RETRY_COUNT) + ) + // Load custom error codes customErrorCodesEnabled.value = credentials.custom_error_codes_enabled === true const existingErrorCodes = credentials.custom_error_codes as number[] | undefined @@ -1851,6 +1927,8 @@ watch( modelMappings.value = [] allowedModels.value = [] } + poolModeEnabled.value = false + poolModeRetryCount.value = DEFAULT_POOL_MODE_RETRY_COUNT customErrorCodesEnabled.value = false selectedErrorCodes.value = [] } @@ -2311,6 +2389,15 @@ const handleSubmit = async () => { newCredentials.model_mapping = currentCredentials.model_mapping } + // Add pool mode if enabled + if (poolModeEnabled.value) { + newCredentials.pool_mode = true + newCredentials.pool_mode_retry_count = normalizePoolModeRetryCount(poolModeRetryCount.value) + } else { + delete newCredentials.pool_mode + delete newCredentials.pool_mode_retry_count + } + // Add custom error codes if enabled if (customErrorCodesEnabled.value) { newCredentials.custom_error_codes_enabled = true diff --git a/frontend/src/i18n/locales/en.ts b/frontend/src/i18n/locales/en.ts index b23b46f6..ecd13940 100644 --- a/frontend/src/i18n/locales/en.ts +++ b/frontend/src/i18n/locales/en.ts @@ -1942,6 +1942,13 @@ export default { addModel: 'Add', modelExists: 'Model already exists', modelCount: '{count} models', + poolMode: 'Pool Mode', + poolModeHint: 'Enable when upstream is an account pool; errors won\'t mark local account status', + poolModeInfo: + 'When enabled, upstream 429/403/401 errors will auto-retry without marking the account as rate-limited or errored. Suitable for upstream pointing to another sub2api instance.', + poolModeRetryCount: 'Same-Account Retries', + poolModeRetryCountHint: + 'Only applies in pool mode. Use 0 to disable in-place retry. Default {default}, maximum {max}.', customErrorCodes: 'Custom Error Codes', customErrorCodesHint: 'Only stop scheduling for selected error codes', customErrorCodesWarning: diff --git a/frontend/src/i18n/locales/zh.ts b/frontend/src/i18n/locales/zh.ts index 777a5602..0238595e 100644 --- a/frontend/src/i18n/locales/zh.ts +++ b/frontend/src/i18n/locales/zh.ts @@ -2086,6 +2086,12 @@ export default { addModel: '填入', modelExists: '该模型已存在', modelCount: '{count} 个模型', + poolMode: '池模式', + poolModeHint: '上游为账号池时启用，错误不标记本地账号状态', + poolModeInfo: + '启用后，上游 429/403/401 错误将自动重试而不标记账号限流或错误，适用于上游指向另一个 sub2api 实例的场景。', + poolModeRetryCount: '同账号重试次数', + poolModeRetryCountHint: '仅在池模式下生效。0 表示不原地重试；默认 {default}，最大 {max}。', customErrorCodes: '自定义错误码', customErrorCodesHint: '仅对选中的错误码停止调度', customErrorCodesWarning: '仅选中的错误码会停止调度，其他错误将返回 500。',