merge: upstream v0.1.106 into release/custom-0.1.106

Resolved conflicts:
- VERSION: set to 0.1.106.1
- wire_gen.go: keep our version with accountUsageService
- antigravity_gateway_service.go: keep accountUsageService field/param
- openai_gateway_service.go: keep billing model override logic
- account.go: keep both affinity methods and new custom base URL methods
- dto/types.go: keep both affinity fields and new custom base URL fields
- CreateAccountModal.vue: keep both affinity and custom base URL logic
- EditAccountModal.vue: keep both affinity and custom base URL state
This commit is contained in:
erio committed 2026-04-03 15:47:18 +08:00
commit fc5e96aa9f
30 files changed
+816 -112

No files matched your search

+22
View File
@@ -1252,6 +1252,23 @@ func (a *Account) IsAffinityEnabled() bool {
return false
}
// IsCustomBaseURLEnabled 检查是否启用自定义 base URL 中继转发
// 仅适用于 Anthropic OAuth/SetupToken 类型账号
func (a *Account) IsCustomBaseURLEnabled() bool {
if !a.IsAnthropicOAuthOrSetupToken() {
return false
}
if a.Extra == nil {
return false
}
if v, ok := a.Extra["custom_base_url_enabled"]; ok {
if enabled, ok := v.(bool); ok {
return enabled
}
}
return false
}
// IsClientAffinityEnabled 向后兼容别名,内部调用 IsAffinityEnabled
func (a *Account) IsClientAffinityEnabled() bool {
return a.IsAffinityEnabled()
@@ -1467,6 +1484,11 @@ func (a *Account) GetMultiDimAffinityZone(userCount, clientCount, perUserCount i
return worst
}
// GetCustomBaseURL 返回自定义中继服务的 base URL
func (a *Account) GetCustomBaseURL() string {
return a.GetExtraString("custom_base_url")
}
// IsCacheTTLOverrideEnabled 检查是否启用缓存 TTL 强制替换
// 仅适用于 Anthropic OAuth/SetupToken 类型账号
// 启用后将所有 cache creation tokens 归入指定的 TTL 类型(5m 或 1h)
+12
View File
@@ -1885,6 +1885,18 @@ func (s *adminServiceImpl) ClearAccountError(ctx context.Context, id int64) (*Ac
if err := s.accountRepo.ClearError(ctx, id); err != nil {
return nil, err
}
if err := s.accountRepo.ClearRateLimit(ctx, id); err != nil {
return nil, err
}
if err := s.accountRepo.ClearAntigravityQuotaScopes(ctx, id); err != nil {
return nil, err
}
if err := s.accountRepo.ClearModelRateLimits(ctx, id); err != nil {
return nil, err
}
if err := s.accountRepo.ClearTempUnschedulable(ctx, id); err != nil {
return nil, err
}
return s.accountRepo.GetByID(ctx, id)
}
@@ -0,0 +1,86 @@
//go:build unit
package service
import (
"context"
"testing"
"time"
"github.com/stretchr/testify/require"
)
type accountRepoStubForClearAccountError struct {
mockAccountRepoForGemini
account *Account
clearErrorCalls int
clearRateLimitCalls int
clearAntigravityCalls int
clearModelRateLimitCalls int
clearTempUnschedCalls int
}
func (r *accountRepoStubForClearAccountError) GetByID(ctx context.Context, id int64) (*Account, error) {
return r.account, nil
}
func (r *accountRepoStubForClearAccountError) ClearError(ctx context.Context, id int64) error {
r.clearErrorCalls++
r.account.Status = StatusActive
r.account.ErrorMessage = ""
return nil
}
func (r *accountRepoStubForClearAccountError) ClearRateLimit(ctx context.Context, id int64) error {
r.clearRateLimitCalls++
r.account.RateLimitedAt = nil
r.account.RateLimitResetAt = nil
return nil
}
func (r *accountRepoStubForClearAccountError) ClearAntigravityQuotaScopes(ctx context.Context, id int64) error {
r.clearAntigravityCalls++
return nil
}
func (r *accountRepoStubForClearAccountError) ClearModelRateLimits(ctx context.Context, id int64) error {
r.clearModelRateLimitCalls++
return nil
}
func (r *accountRepoStubForClearAccountError) ClearTempUnschedulable(ctx context.Context, id int64) error {
r.clearTempUnschedCalls++
r.account.TempUnschedulableUntil = nil
r.account.TempUnschedulableReason = ""
return nil
}
func TestAdminService_ClearAccountError_AlsoClearsRecoverableRuntimeState(t *testing.T) {
until := time.Now().Add(10 * time.Minute)
resetAt := time.Now().Add(5 * time.Minute)
repo := &accountRepoStubForClearAccountError{
account: &Account{
ID: 31,
Platform: PlatformOpenAI,
Type: AccountTypeOAuth,
Status: StatusError,
ErrorMessage: "refresh failed",
RateLimitResetAt: &resetAt,
TempUnschedulableUntil: &until,
TempUnschedulableReason: "missing refresh token",
},
}
svc := &adminServiceImpl{accountRepo: repo}
updated, err := svc.ClearAccountError(context.Background(), 31)
require.NoError(t, err)
require.NotNil(t, updated)
require.Equal(t, 1, repo.clearErrorCalls)
require.Equal(t, 1, repo.clearRateLimitCalls)
require.Equal(t, 1, repo.clearAntigravityCalls)
require.Equal(t, 1, repo.clearModelRateLimitCalls)
require.Equal(t, 1, repo.clearTempUnschedCalls)
require.Nil(t, updated.RateLimitResetAt)
require.Nil(t, updated.TempUnschedulableUntil)
require.Empty(t, updated.TempUnschedulableReason)
}
+62 -4
View File
@@ -12,6 +12,7 @@ import (
"log/slog"
mathrand "math/rand"
"net/http"
"net/url"
"os"
"path/filepath"
"regexp"
@@ -393,6 +394,8 @@ var allowedHeaders = map[string]bool{
"user-agent": true,
"content-type": true,
"accept-encoding": true,
"x-claude-code-session-id": true,
"x-client-request-id": true,
}
// GatewayCache 定义网关服务的缓存操作接口。
@@ -4412,10 +4415,12 @@ func (s *GatewayService) Forward(ctx context.Context, c *gin.Context, account *A
return nil, err
}
// 获取代理URL
// 获取代理URL(自定义 base URL 模式下,proxy 通过 buildCustomRelayURL 作为查询参数传递)
proxyURL := ""
if account.ProxyID != nil && account.Proxy != nil {
proxyURL = account.Proxy.URL()
if !account.IsCustomBaseURLEnabled() || account.GetCustomBaseURL() == "" {
proxyURL = account.Proxy.URL()
}
}
// 解析 TLS 指纹 profile(同一请求生命周期内不变,避免重试循环中重复解析)
@@ -5891,6 +5896,16 @@ func (s *GatewayService) buildUpstreamRequest(ctx context.Context, c *gin.Contex
}
targetURL = validatedURL + "/v1/messages?beta=true"
}
} else if account.IsCustomBaseURLEnabled() {
customURL := account.GetCustomBaseURL()
if customURL == "" {
return nil, fmt.Errorf("custom_base_url is enabled but not configured for account %d", account.ID)
}
validatedURL, err := s.validateUpstreamBaseURL(customURL)
if err != nil {
return nil, err
}
targetURL = s.buildCustomRelayURL(validatedURL, "/v1/messages", account)
}
clientHeaders := http.Header{}
@@ -6006,6 +6021,15 @@ func (s *GatewayService) buildUpstreamRequest(ctx context.Context, c *gin.Contex
}
}
// 同步 X-Claude-Code-Session-Id 头:取 body 中已处理的 metadata.user_id 的 session_id 覆盖
if sessionHeader := getHeaderRaw(req.Header, "X-Claude-Code-Session-Id"); sessionHeader != "" {
if uid := gjson.GetBytes(body, "metadata.user_id").String(); uid != "" {
if parsed := ParseMetadataUserID(uid); parsed != nil {
setHeaderRaw(req.Header, "X-Claude-Code-Session-Id", parsed.SessionID)
}
}
}
// === DEBUG: 打印上游转发请求(headers + body 摘要),与 CLIENT_ORIGINAL 对比 ===
s.debugLogGatewaySnapshot("UPSTREAM_FORWARD", req.Header, body, map[string]string{
"url": req.URL.String(),
@@ -8559,10 +8583,12 @@ func (s *GatewayService) ForwardCountTokens(ctx context.Context, c *gin.Context,
return err
}
// 获取代理URL
// 获取代理URL(自定义 base URL 模式下,proxy 通过 buildCustomRelayURL 作为查询参数传递)
proxyURL := ""
if account.ProxyID != nil && account.Proxy != nil {
proxyURL = account.Proxy.URL()
if !account.IsCustomBaseURLEnabled() || account.GetCustomBaseURL() == "" {
proxyURL = account.Proxy.URL()
}
}
// 发送请求
@@ -8841,6 +8867,16 @@ func (s *GatewayService) buildCountTokensRequest(ctx context.Context, c *gin.Con
}
targetURL = validatedURL + "/v1/messages/count_tokens?beta=true"
}
} else if account.IsCustomBaseURLEnabled() {
customURL := account.GetCustomBaseURL()
if customURL == "" {
return nil, fmt.Errorf("custom_base_url is enabled but not configured for account %d", account.ID)
}
validatedURL, err := s.validateUpstreamBaseURL(customURL)
if err != nil {
return nil, err
}
targetURL = s.buildCustomRelayURL(validatedURL, "/v1/messages/count_tokens", account)
}
clientHeaders := http.Header{}
@@ -8946,6 +8982,15 @@ func (s *GatewayService) buildCountTokensRequest(ctx context.Context, c *gin.Con
}
}
// 同步 X-Claude-Code-Session-Id 头:取 body 中已处理的 metadata.user_id 的 session_id 覆盖
if sessionHeader := getHeaderRaw(req.Header, "X-Claude-Code-Session-Id"); sessionHeader != "" {
if uid := gjson.GetBytes(body, "metadata.user_id").String(); uid != "" {
if parsed := ParseMetadataUserID(uid); parsed != nil {
setHeaderRaw(req.Header, "X-Claude-Code-Session-Id", parsed.SessionID)
}
}
}
if c != nil && tokenType == "oauth" {
c.Set(claudeMimicDebugInfoKey, buildClaudeMimicDebugLine(req, body, account, tokenType, mimicClaudeCode))
}
@@ -8967,6 +9012,19 @@ func (s *GatewayService) countTokensError(c *gin.Context, status int, errType, m
})
}
// buildCustomRelayURL 构建自定义中继转发 URL
// 在 path 后附加 beta=true 和可选的 proxy 查询参数
func (s *GatewayService) buildCustomRelayURL(baseURL, path string, account *Account) string {
u := strings.TrimRight(baseURL, "/") + path + "?beta=true"
if account.ProxyID != nil && account.Proxy != nil {
proxyURL := account.Proxy.URL()
if proxyURL != "" {
u += "&proxy=" + url.QueryEscape(proxyURL)
}
}
return u
}
func (s *GatewayService) validateUpstreamBaseURL(raw string) (string, error) {
if s.cfg != nil && !s.cfg.Security.URLAllowlist.Enabled {
normalized, err := urlvalidator.ValidateURLFormat(raw, s.cfg.Security.URLAllowlist.AllowInsecureHTTP)
+8
View File
@@ -36,6 +36,11 @@ var headerWireCasing = map[string]string{
"sec-fetch-mode": "sec-fetch-mode",
"accept-encoding": "accept-encoding",
"authorization": "authorization",
// Claude Code 2.1.87+ 新增 header
"x-claude-code-session-id": "X-Claude-Code-Session-Id",
"x-client-request-id": "x-client-request-id",
"content-length": "content-length",
}
// headerWireOrder 定义真实 Claude CLI 发送 header 的顺序(基于抓包)。
@@ -55,11 +60,14 @@ var headerWireOrder = []string{
"authorization",
"x-app",
"User-Agent",
"X-Claude-Code-Session-Id",
"content-type",
"anthropic-beta",
"x-client-request-id",
"accept-language",
"sec-fetch-mode",
"accept-encoding",
"content-length",
"x-stainless-helper-method",
}
@@ -0,0 +1,103 @@
package service
import (
"strings"
"github.com/Wei-Shaw/sub2api/internal/pkg/apicompat"
)
func NormalizeOpenAICompatRequestedModel(model string) string {
trimmed := strings.TrimSpace(model)
if trimmed == "" {
return ""
}
normalized, _, ok := splitOpenAICompatReasoningModel(trimmed)
if !ok || normalized == "" {
return trimmed
}
return normalized
}
func applyOpenAICompatModelNormalization(req *apicompat.AnthropicRequest) {
if req == nil {
return
}
originalModel := strings.TrimSpace(req.Model)
if originalModel == "" {
return
}
normalizedModel, derivedEffort, hasReasoningSuffix := splitOpenAICompatReasoningModel(originalModel)
if hasReasoningSuffix && normalizedModel != "" {
req.Model = normalizedModel
}
if req.OutputConfig != nil && strings.TrimSpace(req.OutputConfig.Effort) != "" {
return
}
claudeEffort := openAIReasoningEffortToClaudeOutputEffort(derivedEffort)
if claudeEffort == "" {
return
}
if req.OutputConfig == nil {
req.OutputConfig = &apicompat.AnthropicOutputConfig{}
}
req.OutputConfig.Effort = claudeEffort
}
func splitOpenAICompatReasoningModel(model string) (normalizedModel string, reasoningEffort string, ok bool) {
trimmed := strings.TrimSpace(model)
if trimmed == "" {
return "", "", false
}
modelID := trimmed
if strings.Contains(modelID, "/") {
parts := strings.Split(modelID, "/")
modelID = parts[len(parts)-1]
}
modelID = strings.TrimSpace(modelID)
if !strings.HasPrefix(strings.ToLower(modelID), "gpt-") {
return trimmed, "", false
}
parts := strings.FieldsFunc(strings.ToLower(modelID), func(r rune) bool {
switch r {
case '-', '_', ' ':
return true
default:
return false
}
})
if len(parts) == 0 {
return trimmed, "", false
}
last := strings.NewReplacer("-", "", "_", "", " ", "").Replace(parts[len(parts)-1])
switch last {
case "none", "minimal":
case "low", "medium", "high":
reasoningEffort = last
case "xhigh", "extrahigh":
reasoningEffort = "xhigh"
default:
return trimmed, "", false
}
return normalizeCodexModel(modelID), reasoningEffort, true
}
func openAIReasoningEffortToClaudeOutputEffort(effort string) string {
switch strings.TrimSpace(effort) {
case "low", "medium", "high":
return effort
case "xhigh":
return "max"
default:
return ""
}
}
@@ -0,0 +1,129 @@
package service
import (
"bytes"
"context"
"io"
"net/http"
"net/http/httptest"
"strings"
"testing"
"github.com/Wei-Shaw/sub2api/internal/pkg/apicompat"
"github.com/gin-gonic/gin"
"github.com/stretchr/testify/require"
"github.com/tidwall/gjson"
)
func TestNormalizeOpenAICompatRequestedModel(t *testing.T) {
t.Parallel()
tests := []struct {
name string
input string
want string
}{
{name: "gpt reasoning alias strips xhigh", input: "gpt-5.4-xhigh", want: "gpt-5.4"},
{name: "gpt reasoning alias strips none", input: "gpt-5.4-none", want: "gpt-5.4"},
{name: "codex max model stays intact", input: "gpt-5.1-codex-max", want: "gpt-5.1-codex-max"},
{name: "non openai model unchanged", input: "claude-opus-4-6", want: "claude-opus-4-6"},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
require.Equal(t, tt.want, NormalizeOpenAICompatRequestedModel(tt.input))
})
}
}
func TestApplyOpenAICompatModelNormalization(t *testing.T) {
t.Parallel()
t.Run("derives xhigh from model suffix when output config missing", func(t *testing.T) {
req := &apicompat.AnthropicRequest{Model: "gpt-5.4-xhigh"}
applyOpenAICompatModelNormalization(req)
require.Equal(t, "gpt-5.4", req.Model)
require.NotNil(t, req.OutputConfig)
require.Equal(t, "max", req.OutputConfig.Effort)
})
t.Run("explicit output config wins over model suffix", func(t *testing.T) {
req := &apicompat.AnthropicRequest{
Model: "gpt-5.4-xhigh",
OutputConfig: &apicompat.AnthropicOutputConfig{Effort: "low"},
}
applyOpenAICompatModelNormalization(req)
require.Equal(t, "gpt-5.4", req.Model)
require.NotNil(t, req.OutputConfig)
require.Equal(t, "low", req.OutputConfig.Effort)
})
t.Run("non openai model is untouched", func(t *testing.T) {
req := &apicompat.AnthropicRequest{Model: "claude-opus-4-6"}
applyOpenAICompatModelNormalization(req)
require.Equal(t, "claude-opus-4-6", req.Model)
require.Nil(t, req.OutputConfig)
})
}
func TestForwardAsAnthropic_NormalizesRoutingAndEffortForGpt54XHigh(t *testing.T) {
t.Parallel()
gin.SetMode(gin.TestMode)
rec := httptest.NewRecorder()
c, _ := gin.CreateTestContext(rec)
body := []byte(`{"model":"gpt-5.4-xhigh","max_tokens":16,"messages":[{"role":"user","content":"hello"}],"stream":false}`)
c.Request = httptest.NewRequest(http.MethodPost, "/v1/messages", bytes.NewReader(body))
c.Request.Header.Set("Content-Type", "application/json")
upstreamBody := strings.Join([]string{
`data: {"type":"response.completed","response":{"id":"resp_1","object":"response","model":"gpt-5.4","status":"completed","output":[{"type":"message","id":"msg_1","role":"assistant","status":"completed","content":[{"type":"output_text","text":"ok"}]}],"usage":{"input_tokens":5,"output_tokens":2,"total_tokens":7}}}`,
"",
"data: [DONE]",
"",
}, "\n")
upstream := &httpUpstreamRecorder{resp: &http.Response{
StatusCode: http.StatusOK,
Header: http.Header{"Content-Type": []string{"text/event-stream"}, "x-request-id": []string{"rid_compat"}},
Body: io.NopCloser(strings.NewReader(upstreamBody)),
}}
svc := &OpenAIGatewayService{httpUpstream: upstream}
account := &Account{
ID: 1,
Name: "openai-oauth",
Platform: PlatformOpenAI,
Type: AccountTypeOAuth,
Concurrency: 1,
Credentials: map[string]any{
"access_token": "oauth-token",
"chatgpt_account_id": "chatgpt-acc",
"model_mapping": map[string]any{
"gpt-5.4": "gpt-5.4",
},
},
}
result, err := svc.ForwardAsAnthropic(context.Background(), c, account, body, "", "gpt-5.1")
require.NoError(t, err)
require.NotNil(t, result)
require.Equal(t, "gpt-5.4-xhigh", result.Model)
require.Equal(t, "gpt-5.4", result.UpstreamModel)
require.Equal(t, "gpt-5.4", result.BillingModel)
require.NotNil(t, result.ReasoningEffort)
require.Equal(t, "xhigh", *result.ReasoningEffort)
require.Equal(t, "gpt-5.4", gjson.GetBytes(upstream.lastBody, "model").String())
require.Equal(t, "xhigh", gjson.GetBytes(upstream.lastBody, "reasoning.effort").String())
require.Equal(t, http.StatusOK, rec.Code)
require.Equal(t, "gpt-5.4-xhigh", gjson.GetBytes(rec.Body.Bytes(), "model").String())
require.Equal(t, "ok", gjson.GetBytes(rec.Body.Bytes(), "content.0.text").String())
t.Logf("upstream body: %s", string(upstream.lastBody))
t.Logf("response body: %s", rec.Body.String())
}
@@ -40,6 +40,7 @@ func (s *OpenAIGatewayService) ForwardAsAnthropic(
return nil, fmt.Errorf("parse anthropic request: %w", err)
}
originalModel := anthropicReq.Model
applyOpenAICompatModelNormalization(&anthropicReq)
clientStream := anthropicReq.Stream // client's original stream preference
// 2. Convert Anthropic → Responses
@@ -59,7 +60,7 @@ func (s *OpenAIGatewayService) ForwardAsAnthropic(
}
// 3. Model mapping
mappedModel := resolveOpenAIForwardModel(account, originalModel, defaultMappedModel)
mappedModel := resolveOpenAIForwardModel(account, anthropicReq.Model, defaultMappedModel)
responsesReq.Model = mappedModel
logger.L().Debug("openai messages: model mapping applied",
@@ -897,14 +897,16 @@ func TestOpenAIGatewayServiceRecordUsage_UsesRequestedModelAndUpstreamModelMetad
require.Equal(t, 1, userRepo.deductCalls)
}
func TestOpenAIGatewayServiceRecordUsage_BillsMappedRequestsUsingUpstreamModelFallback(t *testing.T) {
func TestOpenAIGatewayServiceRecordUsage_BillsMappedRequestsUsingRequestedModel(t *testing.T) {
usageRepo := &openAIRecordUsageLogRepoStub{inserted: true}
userRepo := &openAIRecordUsageUserRepoStub{}
subRepo := &openAIRecordUsageSubRepoStub{}
svc := newOpenAIRecordUsageServiceForTest(usageRepo, userRepo, subRepo, nil)
usage := OpenAIUsage{InputTokens: 20, OutputTokens: 10}
expectedCost, err := svc.billingService.CalculateCost("gpt-5.1-codex", UsageTokens{
// Billing should use the requested model ("gpt-5.1"), not the upstream mapped model ("gpt-5.1-codex").
// This ensures pricing is always based on the model the user requested.
expectedCost, err := svc.billingService.CalculateCost("gpt-5.1", UsageTokens{
InputTokens: 20,
OutputTokens: 10,
}, 1.1)
@@ -502,6 +502,25 @@ func (s *OpenAIOAuthService) RefreshAccountToken(ctx context.Context, account *A
refreshToken := account.GetCredential("refresh_token")
if refreshToken == "" {
accessToken := account.GetCredential("access_token")
if accessToken != "" {
tokenInfo := &OpenAITokenInfo{
AccessToken: accessToken,
RefreshToken: "",
IDToken: account.GetCredential("id_token"),
ClientID: account.GetCredential("client_id"),
Email: account.GetCredential("email"),
ChatGPTAccountID: account.GetCredential("chatgpt_account_id"),
ChatGPTUserID: account.GetCredential("chatgpt_user_id"),
OrganizationID: account.GetCredential("organization_id"),
PlanType: account.GetCredential("plan_type"),
}
if expiresAt := account.GetCredentialAsTime("expires_at"); expiresAt != nil {
tokenInfo.ExpiresAt = expiresAt.Unix()
tokenInfo.ExpiresIn = int64(time.Until(*expiresAt).Seconds())
}
return tokenInfo, nil
}
return nil, infraerrors.New(http.StatusBadRequest, "OPENAI_OAUTH_NO_REFRESH_TOKEN", "no refresh token available")
}
@@ -0,0 +1,54 @@
package service
import (
"context"
"errors"
"sync/atomic"
"testing"
"time"
"github.com/Wei-Shaw/sub2api/internal/pkg/openai"
"github.com/stretchr/testify/require"
)
type openaiOAuthClientRefreshStub struct {
refreshCalls int32
}
func (s *openaiOAuthClientRefreshStub) ExchangeCode(ctx context.Context, code, codeVerifier, redirectURI, proxyURL, clientID string) (*openai.TokenResponse, error) {
return nil, errors.New("not implemented")
}
func (s *openaiOAuthClientRefreshStub) RefreshToken(ctx context.Context, refreshToken, proxyURL string) (*openai.TokenResponse, error) {
atomic.AddInt32(&s.refreshCalls, 1)
return nil, errors.New("not implemented")
}
func (s *openaiOAuthClientRefreshStub) RefreshTokenWithClientID(ctx context.Context, refreshToken, proxyURL string, clientID string) (*openai.TokenResponse, error) {
atomic.AddInt32(&s.refreshCalls, 1)
return nil, errors.New("not implemented")
}
func TestOpenAIOAuthService_RefreshAccountToken_NoRefreshTokenUsesExistingAccessToken(t *testing.T) {
client := &openaiOAuthClientRefreshStub{}
svc := NewOpenAIOAuthService(nil, client)
expiresAt := time.Now().Add(30 * time.Minute).UTC().Format(time.RFC3339)
account := &Account{
ID: 77,
Platform: PlatformOpenAI,
Type: AccountTypeOAuth,
Credentials: map[string]any{
"access_token": "existing-access-token",
"expires_at": expiresAt,
"client_id": "client-id-1",
},
}
info, err := svc.RefreshAccountToken(context.Background(), account)
require.NoError(t, err)
require.NotNil(t, info)
require.Equal(t, "existing-access-token", info.AccessToken)
require.Equal(t, "client-id-1", info.ClientID)
require.Zero(t, atomic.LoadInt32(&client.refreshCalls), "existing access token should be reused without calling refresh")
}
+58 -38
View File
@@ -191,10 +191,38 @@ func (s *PricingService) checkAndUpdatePricing() error {
return s.downloadPricingData()
}
// 检查文件是否过期
// 先加载本地文件(确保服务可用),再检查是否需要更新
if err := s.loadPricingData(pricingFile); err != nil {
logger.LegacyPrintf("service.pricing", "[Pricing] Failed to load local file, downloading: %v", err)
return s.downloadPricingData()
}
// 如果配置了哈希URL,通过远程哈希检查是否有更新
if s.cfg.Pricing.HashURL != "" {
remoteHash, err := s.fetchRemoteHash()
if err != nil {
logger.LegacyPrintf("service.pricing", "[Pricing] Failed to fetch remote hash on startup: %v", err)
return nil // 已加载本地文件,哈希获取失败不影响启动
}
s.mu.RLock()
localHash := s.localHash
s.mu.RUnlock()
if localHash == "" || remoteHash != localHash {
logger.LegacyPrintf("service.pricing", "[Pricing] Remote hash differs on startup (local=%s remote=%s), downloading...",
localHash[:min(8, len(localHash))], remoteHash[:min(8, len(remoteHash))])
if err := s.downloadPricingData(); err != nil {
logger.LegacyPrintf("service.pricing", "[Pricing] Download failed, using existing file: %v", err)
}
}
return nil
}
// 没有哈希URL时,基于文件年龄检查
info, err := os.Stat(pricingFile)
if err != nil {
return s.downloadPricingData()
return nil // 已加载本地文件
}
fileAge := time.Since(info.ModTime())
@@ -207,21 +235,11 @@ func (s *PricingService) checkAndUpdatePricing() error {
}
}
// 加载本地文件
return s.loadPricingData(pricingFile)
return nil
}
// syncWithRemote 与远程同步(基于哈希校验)
func (s *PricingService) syncWithRemote() error {
pricingFile := s.getPricingFilePath()
// 计算本地文件哈希
localHash, err := s.computeFileHash(pricingFile)
if err != nil {
logger.LegacyPrintf("service.pricing", "[Pricing] Failed to compute local hash: %v", err)
return s.downloadPricingData()
}
// 如果配置了哈希URL,从远程获取哈希进行比对
if s.cfg.Pricing.HashURL != "" {
remoteHash, err := s.fetchRemoteHash()
@@ -230,8 +248,13 @@ func (s *PricingService) syncWithRemote() error {
return nil // 哈希获取失败不影响正常使用
}
if remoteHash != localHash {
logger.LegacyPrintf("service.pricing", "%s", "[Pricing] Remote hash differs, downloading new version...")
s.mu.RLock()
localHash := s.localHash
s.mu.RUnlock()
if localHash == "" || remoteHash != localHash {
logger.LegacyPrintf("service.pricing", "[Pricing] Remote hash differs (local=%s remote=%s), downloading new version...",
localHash[:min(8, len(localHash))], remoteHash[:min(8, len(remoteHash))])
return s.downloadPricingData()
}
logger.LegacyPrintf("service.pricing", "%s", "[Pricing] Hash check passed, no update needed")
@@ -239,6 +262,7 @@ func (s *PricingService) syncWithRemote() error {
}
// 没有哈希URL时,基于时间检查
pricingFile := s.getPricingFilePath()
info, err := os.Stat(pricingFile)
if err != nil {
return s.downloadPricingData()
@@ -266,11 +290,12 @@ func (s *PricingService) downloadPricingData() error {
ctx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
defer cancel()
var expectedHash string
// 获取远程哈希(用于同步锚点,不作为完整性校验)
var remoteHash string
if strings.TrimSpace(s.cfg.Pricing.HashURL) != "" {
expectedHash, err = s.fetchRemoteHash()
remoteHash, err = s.fetchRemoteHash()
if err != nil {
return fmt.Errorf("fetch remote hash: %w", err)
logger.LegacyPrintf("service.pricing", "[Pricing] Failed to fetch remote hash (continuing): %v", err)
}
}
@@ -279,11 +304,13 @@ func (s *PricingService) downloadPricingData() error {
return fmt.Errorf("download failed: %w", err)
}
if expectedHash != "" {
actualHash := sha256.Sum256(body)
if !strings.EqualFold(expectedHash, hex.EncodeToString(actualHash[:])) {
return fmt.Errorf("pricing hash mismatch")
}
// 哈希校验:不匹配时仅告警,不阻止更新
// 远程哈希文件可能与数据文件不同步(如维护者更新了数据但未更新哈希文件)
dataHash := sha256.Sum256(body)
dataHashStr := hex.EncodeToString(dataHash[:])
if remoteHash != "" && !strings.EqualFold(remoteHash, dataHashStr) {
logger.LegacyPrintf("service.pricing", "[Pricing] Hash mismatch warning: remote=%s data=%s (hash file may be out of sync)",
remoteHash[:min(8, len(remoteHash))], dataHashStr[:8])
}
// 解析JSON数据(使用灵活的解析方式)
@@ -298,11 +325,14 @@ func (s *PricingService) downloadPricingData() error {
logger.LegacyPrintf("service.pricing", "[Pricing] Failed to save file: %v", err)
}
// 保存哈希
hash := sha256.Sum256(body)
hashStr := hex.EncodeToString(hash[:])
// 使用远程哈希作为同步锚点,防止重复下载
// 当远程哈希不可用时,回退到数据本身的哈希
syncHash := dataHashStr
if remoteHash != "" {
syncHash = remoteHash
}
hashFile := s.getHashFilePath()
if err := os.WriteFile(hashFile, []byte(hashStr+"\n"), 0644); err != nil {
if err := os.WriteFile(hashFile, []byte(syncHash+"\n"), 0644); err != nil {
logger.LegacyPrintf("service.pricing", "[Pricing] Failed to save hash: %v", err)
}
@@ -310,7 +340,7 @@ func (s *PricingService) downloadPricingData() error {
s.mu.Lock()
s.pricingData = data
s.lastUpdated = time.Now()
s.localHash = hashStr
s.localHash = syncHash
s.mu.Unlock()
logger.LegacyPrintf("service.pricing", "[Pricing] Downloaded %d models successfully", len(data))
@@ -491,16 +521,6 @@ func (s *PricingService) validatePricingURL(raw string) (string, error) {
return normalized, nil
}
// computeFileHash 计算文件哈希
func (s *PricingService) computeFileHash(filePath string) (string, error) {
data, err := os.ReadFile(filePath)
if err != nil {
return "", err
}
hash := sha256.Sum256(data)
return hex.EncodeToString(hash[:]), nil
}
// GetModelPricing 获取模型价格(带模糊匹配)
func (s *PricingService) GetModelPricing(modelName string) *LiteLLMModelPricing {
s.mu.RLock()
@@ -32,8 +32,9 @@ type TokenRefreshService struct {
privacyClientFactory PrivacyClientFactory
proxyRepo ProxyRepository
stopCh chan struct{}
wg sync.WaitGroup
stopCh chan struct{}
stopOnce sync.Once
wg sync.WaitGroup
}
// NewTokenRefreshService 创建token刷新服务
@@ -130,7 +131,9 @@ func (s *TokenRefreshService) Start() {
// Stop 停止刷新服务(可安全多次调用)
func (s *TokenRefreshService) Stop() {
close(s.stopCh)
s.stopOnce.Do(func() {
close(s.stopCh)
})
s.wg.Wait()
slog.Info("token_refresh.service_stopped")
}
@@ -430,6 +433,7 @@ func isNonRetryableRefreshError(err error) bool {
"unauthorized_client", // 客户端未授权
"access_denied", // 访问被拒绝
"missing_project_id", // 缺少 project_id
"no refresh token available",
}
for _, needle := range nonRetryable {
if strings.Contains(msg, needle) {
@@ -19,6 +19,7 @@ type tokenRefreshAccountRepo struct {
updateCredentialsCalls int
setErrorCalls int
clearTempCalls int
setTempUnschedCalls int
lastAccount *Account
updateErr error
}
@@ -58,6 +59,11 @@ func (r *tokenRefreshAccountRepo) ClearTempUnschedulable(ctx context.Context, id
return nil
}
func (r *tokenRefreshAccountRepo) SetTempUnschedulable(ctx context.Context, id int64, until time.Time, reason string) error {
r.setTempUnschedCalls++
return nil
}
type tokenCacheInvalidatorStub struct {
calls int
err error
@@ -490,6 +496,31 @@ func TestTokenRefreshService_RefreshWithRetry_NonRetryableErrorAllPlatforms(t *t
}
}
func TestTokenRefreshService_RefreshWithRetry_NoRefreshTokenDoesNotTempUnschedule(t *testing.T) {
repo := &tokenRefreshAccountRepo{}
cfg := &config.Config{
TokenRefresh: config.TokenRefreshConfig{
MaxRetries: 2,
RetryBackoffSeconds: 0,
},
}
service := NewTokenRefreshService(repo, nil, nil, nil, nil, nil, nil, cfg, nil)
account := &Account{
ID: 18,
Platform: PlatformOpenAI,
Type: AccountTypeOAuth,
}
refresher := &tokenRefresherStub{
err: errors.New("no refresh token available"),
}
err := service.refreshWithRetry(context.Background(), account, refresher, refresher, time.Hour)
require.Error(t, err)
require.Equal(t, 0, repo.updateCalls)
require.Equal(t, 0, repo.setTempUnschedCalls, "missing refresh token should not mark the account temp unschedulable")
require.Equal(t, 1, repo.setErrorCalls, "missing refresh token should be treated as a non-retryable credential state")
}
// TestIsNonRetryableRefreshError 测试不可重试错误判断
func TestIsNonRetryableRefreshError(t *testing.T) {
tests := []struct {
@@ -503,6 +534,7 @@ func TestIsNonRetryableRefreshError(t *testing.T) {
{name: "invalid_client", err: errors.New("invalid_client"), expected: true},
{name: "unauthorized_client", err: errors.New("unauthorized_client"), expected: true},
{name: "access_denied", err: errors.New("access_denied"), expected: true},
{name: "no_refresh_token", err: errors.New("no refresh token available"), expected: true},
{name: "invalid_grant_with_desc", err: errors.New("Error: invalid_grant - token revoked"), expected: true},
{name: "case_insensitive", err: errors.New("INVALID_GRANT"), expected: true},
}
@@ -21,10 +21,10 @@ func optionalNonEqualStringPtr(value, compare string) *string {
}
func forwardResultBillingModel(requestedModel, upstreamModel string) string {
if trimmedUpstream := strings.TrimSpace(upstreamModel); trimmedUpstream != "" {
return trimmedUpstream
if trimmed := strings.TrimSpace(requestedModel); trimmed != "" {
return trimmed
}
return strings.TrimSpace(requestedModel)
return strings.TrimSpace(upstreamModel)
}
func optionalInt64Ptr(v int64) *int64 {