Files
sub2api/backend/internal/service/gemini_messages_compat_service.go

3303 lines
100 KiB
Go
Raw Permalink Normal View History

package service
import (
"bufio"
"bytes"
"context"
"crypto/rand"
"encoding/hex"
"encoding/json"
"errors"
"fmt"
"io"
"log"
"math"
mathrand "math/rand"
"net/http"
"regexp"
"strings"
"time"
"github.com/Wei-Shaw/sub2api/internal/config"
"github.com/Wei-Shaw/sub2api/internal/pkg/ctxkey"
"github.com/Wei-Shaw/sub2api/internal/pkg/geminicli"
"github.com/Wei-Shaw/sub2api/internal/pkg/googleapi"
"github.com/Wei-Shaw/sub2api/internal/pkg/logger"
"github.com/Wei-Shaw/sub2api/internal/util/responseheaders"
"github.com/Wei-Shaw/sub2api/internal/util/urlvalidator"
"github.com/gin-gonic/gin"
"github.com/tidwall/gjson"
)
const geminiStickySessionTTL = time.Hour
const (
geminiMaxRetries = 5
geminiRetryBaseDelay = 1 * time.Second
geminiRetryMaxDelay = 16 * time.Second
)
// Gemini tool calling now requires `thoughtSignature` in parts that include `functionCall`.
// Many clients don't send it; we inject a known dummy signature to satisfy the validator.
// Ref: https://ai.google.dev/gemini-api/docs/thought-signatures
const geminiDummyThoughtSignature = "skip_thought_signature_validator"
type GeminiMessagesCompatService struct {
accountRepo AccountRepository
groupRepo GroupRepository
cache GatewayCache
schedulerSnapshot *SchedulerSnapshotService
tokenProvider *GeminiTokenProvider
rateLimitService *RateLimitService
httpUpstream HTTPUpstream
antigravityGatewayService *AntigravityGatewayService
cfg *config.Config
responseHeaderFilter *responseheaders.CompiledHeaderFilter
}
func NewGeminiMessagesCompatService(
accountRepo AccountRepository,
groupRepo GroupRepository,
cache GatewayCache,
schedulerSnapshot *SchedulerSnapshotService,
tokenProvider *GeminiTokenProvider,
rateLimitService *RateLimitService,
httpUpstream HTTPUpstream,
antigravityGatewayService *AntigravityGatewayService,
cfg *config.Config,
) *GeminiMessagesCompatService {
return &GeminiMessagesCompatService{
accountRepo: accountRepo,
groupRepo: groupRepo,
cache: cache,
schedulerSnapshot: schedulerSnapshot,
tokenProvider: tokenProvider,
rateLimitService: rateLimitService,
httpUpstream: httpUpstream,
antigravityGatewayService: antigravityGatewayService,
cfg: cfg,
responseHeaderFilter: compileResponseHeaderFilter(cfg),
}
}
// GetTokenProvider returns the token provider for OAuth accounts
func (s *GeminiMessagesCompatService) GetTokenProvider() *GeminiTokenProvider {
return s.tokenProvider
}
func (s *GeminiMessagesCompatService) SelectAccountForModel(ctx context.Context, groupID *int64, sessionHash string, requestedModel string) (*Account, error) {
return s.SelectAccountForModelWithExclusions(ctx, groupID, sessionHash, requestedModel, nil)
}
func (s *GeminiMessagesCompatService) SelectAccountForModelWithExclusions(ctx context.Context, groupID *int64, sessionHash string, requestedModel string, excludedIDs map[int64]struct{}) (*Account, error) {
2026-02-02 22:13:50 +08:00
// 1. 确定目标平台和调度模式
// Determine target platform and scheduling mode
platform, useMixedScheduling, hasForcePlatform, err := s.resolvePlatformAndSchedulingMode(ctx, groupID)
if err != nil {
return nil, err
}
cacheKey := "gemini:" + sessionHash
// 2. 尝试粘性会话命中
// Try sticky session hit
if account := s.tryStickySessionHit(ctx, groupID, sessionHash, cacheKey, requestedModel, excludedIDs, platform, useMixedScheduling); account != nil {
return account, nil
}
// 3. 查询可调度账户(强制平台模式:优先按分组查找,找不到再查全部)
// Query schedulable accounts (force platform mode: try group first, fallback to all)
accounts, err := s.listSchedulableAccountsOnce(ctx, groupID, platform, hasForcePlatform)
if err != nil {
return nil, fmt.Errorf("query accounts failed: %w", err)
}
// 强制平台模式下,分组中找不到账户时回退查询全部
if len(accounts) == 0 && groupID != nil && hasForcePlatform {
accounts, err = s.listSchedulableAccountsOnce(ctx, nil, platform, hasForcePlatform)
if err != nil {
return nil, fmt.Errorf("query accounts failed: %w", err)
}
}
// 4. 按优先级 + LRU 选择最佳账号
// Select best account by priority + LRU
selected := s.selectBestGeminiAccount(ctx, accounts, requestedModel, excludedIDs, platform, useMixedScheduling)
if selected == nil {
if requestedModel != "" {
return nil, fmt.Errorf("no available Gemini accounts supporting model: %s", requestedModel)
}
return nil, errors.New("no available Gemini accounts")
}
// 5. 设置粘性会话绑定
// Set sticky session binding
if sessionHash != "" {
_ = s.cache.SetSessionAccountID(ctx, derefGroupID(groupID), cacheKey, selected.ID, geminiStickySessionTTL)
}
return selected, nil
}
// resolvePlatformAndSchedulingMode 解析目标平台和调度模式。
// 返回:平台名称、是否使用混合调度、是否强制平台、错误。
//
// resolvePlatformAndSchedulingMode resolves target platform and scheduling mode.
// Returns: platform name, whether to use mixed scheduling, whether force platform, error.
func (s *GeminiMessagesCompatService) resolvePlatformAndSchedulingMode(ctx context.Context, groupID *int64) (platform string, useMixedScheduling bool, hasForcePlatform bool, err error) {
// 优先检查 context 中的强制平台(/antigravity 路由)
forcePlatform, hasForcePlatform := ctx.Value(ctxkey.ForcePlatform).(string)
if hasForcePlatform && forcePlatform != "" {
2026-02-02 22:13:50 +08:00
return forcePlatform, false, true, nil
}
if groupID != nil {
// 根据分组 platform 决定查询哪种账号
var group *Group
if ctxGroup, ok := ctx.Value(ctxkey.Group).(*Group); ok && IsGroupContextValid(ctxGroup) && ctxGroup.ID == *groupID {
group = ctxGroup
} else {
group, err = s.groupRepo.GetByIDLite(ctx, *groupID)
if err != nil {
2026-02-02 22:13:50 +08:00
return "", false, false, fmt.Errorf("get group failed: %w", err)
}
}
2026-02-02 22:13:50 +08:00
// gemini 分组支持混合调度(包含启用了 mixed_scheduling 的 antigravity 账户)
return group.Platform, group.Platform == PlatformGemini, false, nil
}
2026-02-02 22:13:50 +08:00
// 无分组时只使用原生 gemini 平台
return PlatformGemini, true, false, nil
}
2026-02-02 22:13:50 +08:00
// tryStickySessionHit 尝试从粘性会话获取账号。
// 如果命中且账号可用则返回账号;如果账号不可用则清理会话并返回 nil。
//
// tryStickySessionHit attempts to get account from sticky session.
// Returns account if hit and usable; clears session and returns nil if account unavailable.
func (s *GeminiMessagesCompatService) tryStickySessionHit(
ctx context.Context,
groupID *int64,
sessionHash, cacheKey, requestedModel string,
excludedIDs map[int64]struct{},
platform string,
useMixedScheduling bool,
) *Account {
if sessionHash == "" {
return nil
}
2026-02-02 22:13:50 +08:00
accountID, err := s.cache.GetSessionAccountID(ctx, derefGroupID(groupID), cacheKey)
if err != nil || accountID <= 0 {
return nil
}
2026-02-02 22:13:50 +08:00
if _, excluded := excludedIDs[accountID]; excluded {
return nil
}
account, err := s.getSchedulableAccount(ctx, accountID)
if err != nil {
2026-02-02 22:13:50 +08:00
return nil
}
2026-02-02 22:13:50 +08:00
// 检查账号是否需要清理粘性会话
// Check if sticky session should be cleared
if shouldClearStickySession(account, requestedModel) {
2026-02-02 22:13:50 +08:00
_ = s.cache.DeleteSessionAccountID(ctx, derefGroupID(groupID), cacheKey)
return nil
}
// 验证账号是否可用于当前请求
// Verify account is usable for current request
if !s.isAccountUsableForRequest(ctx, account, requestedModel, platform, useMixedScheduling) {
return nil
}
// 刷新会话 TTL 并返回账号
// Refresh session TTL and return account
_ = s.cache.RefreshSessionTTL(ctx, derefGroupID(groupID), cacheKey, geminiStickySessionTTL)
return account
}
// isAccountUsableForRequest 检查账号是否可用于当前请求。
// 验证:模型调度、模型支持、平台匹配、速率限制预检。
//
// isAccountUsableForRequest checks if account is usable for current request.
// Validates: model scheduling, model support, platform matching, rate limit precheck.
func (s *GeminiMessagesCompatService) isAccountUsableForRequest(
ctx context.Context,
account *Account,
requestedModel, platform string,
useMixedScheduling bool,
) bool {
return s.isAccountUsableForRequestWithPrecheck(ctx, account, requestedModel, platform, useMixedScheduling, nil)
}
func (s *GeminiMessagesCompatService) isAccountUsableForRequestWithPrecheck(
ctx context.Context,
account *Account,
requestedModel, platform string,
useMixedScheduling bool,
precheckResult map[int64]bool,
2026-02-02 22:13:50 +08:00
) bool {
// 检查模型调度能力
// Check model scheduling capability
if !account.IsSchedulableForModelWithContext(ctx, requestedModel) {
2026-02-02 22:13:50 +08:00
return false
}
// 检查模型支持
// Check model support
if requestedModel != "" && !s.isModelSupportedByAccount(account, requestedModel) {
return false
}
// 检查平台匹配
// Check platform matching
if !s.isAccountValidForPlatform(account, platform, useMixedScheduling) {
return false
}
// 速率限制预检
// Rate limit precheck
if !s.passesRateLimitPreCheckWithCache(ctx, account, requestedModel, precheckResult) {
2026-02-02 22:13:50 +08:00
return false
}
return true
}
// isAccountValidForPlatform 检查账号是否匹配目标平台。
// 原生平台直接匹配;混合调度模式下 antigravity 需要启用 mixed_scheduling。
//
// isAccountValidForPlatform checks if account matches target platform.
// Native platform matches directly; mixed scheduling mode requires antigravity to enable mixed_scheduling.
func (s *GeminiMessagesCompatService) isAccountValidForPlatform(account *Account, platform string, useMixedScheduling bool) bool {
if account.Platform == platform {
return true
}
2026-02-02 22:13:50 +08:00
if useMixedScheduling && account.Platform == PlatformAntigravity && account.IsMixedSchedulingEnabled() {
return true
}
return false
}
func (s *GeminiMessagesCompatService) passesRateLimitPreCheckWithCache(ctx context.Context, account *Account, requestedModel string, precheckResult map[int64]bool) bool {
2026-02-02 22:13:50 +08:00
if s.rateLimitService == nil || requestedModel == "" {
return true
}
if precheckResult != nil {
if ok, exists := precheckResult[account.ID]; exists {
return ok
}
}
2026-02-02 22:13:50 +08:00
ok, err := s.rateLimitService.PreCheckUsage(ctx, account, requestedModel)
if err != nil {
logger.LegacyPrintf("service.gemini_messages_compat", "[Gemini PreCheck] Account %d precheck error: %v", account.ID, err)
2026-02-02 22:13:50 +08:00
}
return ok
}
// selectBestGeminiAccount 从候选账号中选择最佳账号(优先级 + LRU + OAuth 优先)。
// 返回 nil 表示无可用账号。
//
// selectBestGeminiAccount selects best account from candidates (priority + LRU + OAuth preferred).
// Returns nil if no available account.
func (s *GeminiMessagesCompatService) selectBestGeminiAccount(
ctx context.Context,
accounts []Account,
requestedModel string,
excludedIDs map[int64]struct{},
platform string,
useMixedScheduling bool,
) *Account {
var selected *Account
precheckResult := s.buildPreCheckUsageResultMap(ctx, accounts, requestedModel)
2026-02-02 22:13:50 +08:00
for i := range accounts {
acc := &accounts[i]
2026-02-02 22:13:50 +08:00
// 跳过被排除的账号
if _, excluded := excludedIDs[acc.ID]; excluded {
continue
}
2026-02-02 22:13:50 +08:00
// 检查账号是否可用于当前请求
if !s.isAccountUsableForRequestWithPrecheck(ctx, acc, requestedModel, platform, useMixedScheduling, precheckResult) {
continue
}
2026-02-02 22:13:50 +08:00
// 选择最佳账号
if selected == nil {
selected = acc
continue
}
2026-02-02 22:13:50 +08:00
if s.isBetterGeminiAccount(acc, selected) {
selected = acc
}
}
2026-02-02 22:13:50 +08:00
return selected
}
func (s *GeminiMessagesCompatService) buildPreCheckUsageResultMap(ctx context.Context, accounts []Account, requestedModel string) map[int64]bool {
if s.rateLimitService == nil || requestedModel == "" || len(accounts) == 0 {
return nil
}
candidates := make([]*Account, 0, len(accounts))
for i := range accounts {
candidates = append(candidates, &accounts[i])
}
result, err := s.rateLimitService.PreCheckUsageBatch(ctx, candidates, requestedModel)
if err != nil {
logger.LegacyPrintf("service.gemini_messages_compat", "[Gemini PreCheckBatch] failed: %v", err)
}
return result
}
2026-02-02 22:13:50 +08:00
// isBetterGeminiAccount 判断 candidate 是否比 current 更优。
// 规则优先级更高数值更小优先同优先级时未使用过的优先OAuth > 非 OAuth其次是最久未使用的。
//
// isBetterGeminiAccount checks if candidate is better than current.
// Rules: higher priority (lower value) wins; same priority: never used (OAuth > non-OAuth) > least recently used.
func (s *GeminiMessagesCompatService) isBetterGeminiAccount(candidate, current *Account) bool {
// 优先级更高(数值更小)
if candidate.Priority < current.Priority {
return true
}
if candidate.Priority > current.Priority {
return false
}
2026-02-02 22:13:50 +08:00
// 同优先级,比较最后使用时间
switch {
case candidate.LastUsedAt == nil && current.LastUsedAt != nil:
// candidate 从未使用,优先
return true
case candidate.LastUsedAt != nil && current.LastUsedAt == nil:
// current 从未使用,保持
return false
case candidate.LastUsedAt == nil && current.LastUsedAt == nil:
// 都未使用,优先选择 OAuth 账号(更兼容 Code Assist 流程)
return candidate.Type == AccountTypeOAuth && current.Type != AccountTypeOAuth
default:
// 都使用过,选择最久未使用的
return candidate.LastUsedAt.Before(*current.LastUsedAt)
}
}
// isModelSupportedByAccount 根据账户平台检查模型支持
func (s *GeminiMessagesCompatService) isModelSupportedByAccount(account *Account, requestedModel string) bool {
if account.Platform == PlatformAntigravity {
if strings.TrimSpace(requestedModel) == "" {
return true
}
return mapAntigravityModel(account, requestedModel) != ""
}
return account.IsModelSupported(requestedModel)
}
// GetAntigravityGatewayService 返回 AntigravityGatewayService
func (s *GeminiMessagesCompatService) GetAntigravityGatewayService() *AntigravityGatewayService {
return s.antigravityGatewayService
}
func (s *GeminiMessagesCompatService) getSchedulableAccount(ctx context.Context, accountID int64) (*Account, error) {
if s.schedulerSnapshot != nil {
return s.schedulerSnapshot.GetAccount(ctx, accountID)
}
return s.accountRepo.GetByID(ctx, accountID)
}
func (s *GeminiMessagesCompatService) listSchedulableAccountsOnce(ctx context.Context, groupID *int64, platform string, hasForcePlatform bool) ([]Account, error) {
if s.schedulerSnapshot != nil {
accounts, _, err := s.schedulerSnapshot.ListSchedulableAccounts(ctx, groupID, platform, hasForcePlatform)
return accounts, err
}
useMixedScheduling := platform == PlatformGemini && !hasForcePlatform
queryPlatforms := []string{platform}
if useMixedScheduling {
queryPlatforms = []string{platform, PlatformAntigravity}
}
if groupID != nil {
return s.accountRepo.ListSchedulableByGroupIDAndPlatforms(ctx, *groupID, queryPlatforms)
}
if s.cfg != nil && s.cfg.RunMode == config.RunModeSimple {
return s.accountRepo.ListSchedulableByPlatforms(ctx, queryPlatforms)
}
return s.accountRepo.ListSchedulableUngroupedByPlatforms(ctx, queryPlatforms)
}
func (s *GeminiMessagesCompatService) validateUpstreamBaseURL(raw string) (string, error) {
if s.cfg != nil && !s.cfg.Security.URLAllowlist.Enabled {
normalized, err := urlvalidator.ValidateURLFormat(raw, s.cfg.Security.URLAllowlist.AllowInsecureHTTP)
if err != nil {
return "", fmt.Errorf("invalid base_url: %w", err)
}
return normalized, nil
}
normalized, err := urlvalidator.ValidateHTTPSURL(raw, urlvalidator.ValidationOptions{
AllowedHosts: s.cfg.Security.URLAllowlist.UpstreamHosts,
RequireAllowlist: true,
AllowPrivate: s.cfg.Security.URLAllowlist.AllowPrivateHosts,
})
if err != nil {
return "", fmt.Errorf("invalid base_url: %w", err)
}
return normalized, nil
}
// HasAntigravityAccounts 检查是否有可用的 antigravity 账户
func (s *GeminiMessagesCompatService) HasAntigravityAccounts(ctx context.Context, groupID *int64) (bool, error) {
accounts, err := s.listSchedulableAccountsOnce(ctx, groupID, PlatformAntigravity, false)
if err != nil {
return false, err
}
return len(accounts) > 0, nil
}
// SelectAccountForAIStudioEndpoints selects an account that is likely to succeed against
// generativelanguage.googleapis.com (e.g. GET /v1beta/models).
//
// Preference order:
// 1) API key accounts (AI Studio)
// 2) OAuth accounts without project_id (AI Studio OAuth)
// 3) OAuth accounts explicitly marked as ai_studio
// 4) Any remaining Gemini accounts (fallback)
func (s *GeminiMessagesCompatService) SelectAccountForAIStudioEndpoints(ctx context.Context, groupID *int64) (*Account, error) {
accounts, err := s.listSchedulableAccountsOnce(ctx, groupID, PlatformGemini, true)
if err != nil {
return nil, fmt.Errorf("query accounts failed: %w", err)
}
if len(accounts) == 0 {
return nil, errors.New("no available Gemini accounts")
}
rank := func(a *Account) int {
if a == nil {
return 999
}
switch a.Type {
case AccountTypeAPIKey:
if strings.TrimSpace(a.GetCredential("api_key")) != "" {
return 0
}
return 9
case AccountTypeOAuth:
if strings.TrimSpace(a.GetCredential("project_id")) == "" {
return 1
}
if strings.TrimSpace(a.GetCredential("oauth_type")) == "ai_studio" {
return 2
}
// Code Assist OAuth tokens often lack AI Studio scopes for models listing.
return 3
default:
return 10
}
}
var selected *Account
for i := range accounts {
acc := &accounts[i]
if selected == nil {
selected = acc
continue
}
r1, r2 := rank(acc), rank(selected)
if r1 < r2 {
selected = acc
continue
}
if r1 > r2 {
continue
}
if acc.Priority < selected.Priority {
selected = acc
} else if acc.Priority == selected.Priority {
switch {
case acc.LastUsedAt == nil && selected.LastUsedAt != nil:
selected = acc
case acc.LastUsedAt != nil && selected.LastUsedAt == nil:
// keep selected
case acc.LastUsedAt == nil && selected.LastUsedAt == nil:
if acc.Type == AccountTypeOAuth && selected.Type != AccountTypeOAuth {
selected = acc
}
default:
if acc.LastUsedAt.Before(*selected.LastUsedAt) {
selected = acc
}
}
}
}
if selected == nil {
return nil, errors.New("no available Gemini accounts")
}
return selected, nil
}
func (s *GeminiMessagesCompatService) Forward(ctx context.Context, c *gin.Context, account *Account, body []byte) (*ForwardResult, error) {
startTime := time.Now()
var req struct {
Model string `json:"model"`
Stream bool `json:"stream"`
}
if err := json.Unmarshal(body, &req); err != nil {
return nil, fmt.Errorf("parse request: %w", err)
}
if strings.TrimSpace(req.Model) == "" {
return nil, fmt.Errorf("missing model")
}
originalModel := req.Model
mappedModel := req.Model
if account.Type == AccountTypeAPIKey {
mappedModel = account.GetMappedModel(req.Model)
}
geminiReq, err := convertClaudeMessagesToGeminiGenerateContent(body)
if err != nil {
return nil, s.writeClaudeError(c, http.StatusBadRequest, "invalid_request_error", err.Error())
}
geminiReq = ensureGeminiFunctionCallThoughtSignatures(geminiReq)
originalClaudeBody := body
proxyURL := ""
if account.ProxyID != nil && account.Proxy != nil {
proxyURL = account.Proxy.URL()
}
var requestIDHeader string
var buildReq func(ctx context.Context) (*http.Request, string, error)
useUpstreamStream := req.Stream
if account.Type == AccountTypeOAuth && !req.Stream && strings.TrimSpace(account.GetCredential("project_id")) != "" {
// Code Assist's non-streaming generateContent may return no content; use streaming upstream and aggregate.
useUpstreamStream = true
}
switch account.Type {
case AccountTypeAPIKey:
buildReq = func(ctx context.Context) (*http.Request, string, error) {
apiKey := account.GetCredential("api_key")
if strings.TrimSpace(apiKey) == "" {
return nil, "", errors.New("gemini api_key not configured")
}
baseURL := account.GetGeminiBaseURL(geminicli.AIStudioBaseURL)
normalizedBaseURL, err := s.validateUpstreamBaseURL(baseURL)
if err != nil {
return nil, "", err
}
action := "generateContent"
if req.Stream {
action = "streamGenerateContent"
}
fullURL := fmt.Sprintf("%s/v1beta/models/%s:%s", strings.TrimRight(normalizedBaseURL, "/"), mappedModel, action)
if req.Stream {
fullURL += "?alt=sse"
}
upstreamReq, err := http.NewRequestWithContext(ctx, http.MethodPost, fullURL, bytes.NewReader(geminiReq))
if err != nil {
return nil, "", err
}
upstreamReq.Header.Set("Content-Type", "application/json")
upstreamReq.Header.Set("x-goog-api-key", apiKey)
return upstreamReq, "x-request-id", nil
}
requestIDHeader = "x-request-id"
case AccountTypeOAuth:
buildReq = func(ctx context.Context) (*http.Request, string, error) {
if s.tokenProvider == nil {
return nil, "", errors.New("gemini token provider not configured")
}
accessToken, err := s.tokenProvider.GetAccessToken(ctx, account)
if err != nil {
return nil, "", err
}
projectID := strings.TrimSpace(account.GetCredential("project_id"))
action := "generateContent"
if useUpstreamStream {
action = "streamGenerateContent"
}
// Two modes for OAuth:
// 1. With project_id -> Code Assist API (wrapped request)
// 2. Without project_id -> AI Studio API (direct OAuth, like API key but with Bearer token)
if projectID != "" {
// Mode 1: Code Assist API
baseURL, err := s.validateUpstreamBaseURL(geminicli.GeminiCliBaseURL)
if err != nil {
return nil, "", err
}
fullURL := fmt.Sprintf("%s/v1internal:%s", strings.TrimRight(baseURL, "/"), action)
if useUpstreamStream {
fullURL += "?alt=sse"
}
wrapped := map[string]any{
"model": mappedModel,
"project": projectID,
}
var inner any
if err := json.Unmarshal(geminiReq, &inner); err != nil {
return nil, "", fmt.Errorf("failed to parse gemini request: %w", err)
}
wrapped["request"] = inner
wrappedBytes, _ := json.Marshal(wrapped)
upstreamReq, err := http.NewRequestWithContext(ctx, http.MethodPost, fullURL, bytes.NewReader(wrappedBytes))
if err != nil {
return nil, "", err
}
upstreamReq.Header.Set("Content-Type", "application/json")
upstreamReq.Header.Set("Authorization", "Bearer "+accessToken)
upstreamReq.Header.Set("User-Agent", geminicli.GeminiCLIUserAgent)
return upstreamReq, "x-request-id", nil
} else {
// Mode 2: AI Studio API with OAuth (like API key mode, but using Bearer token)
baseURL := account.GetGeminiBaseURL(geminicli.AIStudioBaseURL)
normalizedBaseURL, err := s.validateUpstreamBaseURL(baseURL)
if err != nil {
return nil, "", err
}
fullURL := fmt.Sprintf("%s/v1beta/models/%s:%s", strings.TrimRight(normalizedBaseURL, "/"), mappedModel, action)
if useUpstreamStream {
fullURL += "?alt=sse"
}
upstreamReq, err := http.NewRequestWithContext(ctx, http.MethodPost, fullURL, bytes.NewReader(geminiReq))
if err != nil {
return nil, "", err
}
upstreamReq.Header.Set("Content-Type", "application/json")
upstreamReq.Header.Set("Authorization", "Bearer "+accessToken)
return upstreamReq, "x-request-id", nil
}
}
requestIDHeader = "x-request-id"
default:
return nil, fmt.Errorf("unsupported account type: %s", account.Type)
}
var resp *http.Response
signatureRetryStage := 0
for attempt := 1; attempt <= geminiMaxRetries; attempt++ {
upstreamReq, idHeader, err := buildReq(ctx)
if err != nil {
if errors.Is(err, context.Canceled) || errors.Is(err, context.DeadlineExceeded) {
return nil, err
}
// Local build error: don't retry.
if strings.Contains(err.Error(), "missing project_id") {
return nil, s.writeClaudeError(c, http.StatusBadRequest, "invalid_request_error", err.Error())
}
return nil, s.writeClaudeError(c, http.StatusBadGateway, "upstream_error", err.Error())
}
requestIDHeader = idHeader
// Capture upstream request body for ops retry of this attempt.
if c != nil {
// In this code path `body` is already the JSON sent to upstream.
c.Set(OpsUpstreamRequestBodyKey, string(body))
}
resp, err = s.httpUpstream.Do(upstreamReq, proxyURL, account.ID, account.Concurrency)
if err != nil {
safeErr := sanitizeUpstreamErrorMessage(err.Error())
appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
Platform: account.Platform,
AccountID: account.ID,
AccountName: account.Name,
UpstreamStatusCode: 0,
Kind: "request_error",
Message: safeErr,
})
if attempt < geminiMaxRetries {
logger.LegacyPrintf("service.gemini_messages_compat", "Gemini account %d: upstream request failed, retry %d/%d: %v", account.ID, attempt, geminiMaxRetries, err)
sleepGeminiBackoff(attempt)
continue
}
setOpsUpstreamError(c, 0, safeErr, "")
return nil, s.writeClaudeError(c, http.StatusBadGateway, "upstream_error", "Upstream request failed after retries: "+safeErr)
}
// Special-case: signature/thought_signature validation errors are not transient, but may be fixed by
// downgrading Claude thinking/tool history to plain text (conservative two-stage retry).
if resp.StatusCode == http.StatusBadRequest && signatureRetryStage < 2 {
respBody, _ := io.ReadAll(io.LimitReader(resp.Body, 2<<20))
_ = resp.Body.Close()
if isGeminiSignatureRelatedError(respBody) {
upstreamReqID := resp.Header.Get(requestIDHeader)
if upstreamReqID == "" {
upstreamReqID = resp.Header.Get("x-goog-request-id")
}
upstreamMsg := strings.TrimSpace(extractUpstreamErrorMessage(respBody))
upstreamMsg = sanitizeUpstreamErrorMessage(upstreamMsg)
upstreamDetail := ""
if s.cfg != nil && s.cfg.Gateway.LogUpstreamErrorBody {
maxBytes := s.cfg.Gateway.LogUpstreamErrorBodyMaxBytes
if maxBytes <= 0 {
maxBytes = 2048
}
upstreamDetail = truncateString(string(respBody), maxBytes)
}
appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
Platform: account.Platform,
AccountID: account.ID,
AccountName: account.Name,
UpstreamStatusCode: resp.StatusCode,
UpstreamRequestID: upstreamReqID,
Kind: "signature_error",
Message: upstreamMsg,
Detail: upstreamDetail,
})
var strippedClaudeBody []byte
stageName := ""
switch signatureRetryStage {
case 0:
// Stage 1: disable thinking + thinking->text
strippedClaudeBody = FilterThinkingBlocksForRetry(originalClaudeBody)
stageName = "thinking-only"
signatureRetryStage = 1
default:
// Stage 2: additionally downgrade tool_use/tool_result blocks to text
strippedClaudeBody = FilterSignatureSensitiveBlocksForRetry(originalClaudeBody)
stageName = "thinking+tools"
signatureRetryStage = 2
}
retryGeminiReq, txErr := convertClaudeMessagesToGeminiGenerateContent(strippedClaudeBody)
if txErr == nil {
logger.LegacyPrintf("service.gemini_messages_compat", "Gemini account %d: detected signature-related 400, retrying with downgraded Claude blocks (%s)", account.ID, stageName)
geminiReq = retryGeminiReq
// Consume one retry budget attempt and continue with the updated request payload.
sleepGeminiBackoff(1)
continue
}
}
// Restore body for downstream error handling.
resp = &http.Response{
StatusCode: http.StatusBadRequest,
Header: resp.Header.Clone(),
Body: io.NopCloser(bytes.NewReader(respBody)),
}
break
}
// 错误策略优先:匹配则跳过重试直接处理。
if matched, rebuilt := s.checkErrorPolicyInLoop(ctx, account, resp); matched {
resp = rebuilt
break
} else {
resp = rebuilt
}
if resp.StatusCode >= 400 && s.shouldRetryGeminiUpstreamError(account, resp.StatusCode) {
respBody, _ := io.ReadAll(io.LimitReader(resp.Body, 2<<20))
_ = resp.Body.Close()
// Don't treat insufficient-scope as transient.
if resp.StatusCode == 403 && isGeminiInsufficientScope(resp.Header, respBody) {
resp = &http.Response{
StatusCode: resp.StatusCode,
Header: resp.Header.Clone(),
Body: io.NopCloser(bytes.NewReader(respBody)),
}
break
}
if resp.StatusCode == 429 {
// Mark as rate-limited early so concurrent requests avoid this account.
s.handleGeminiUpstreamError(ctx, account, resp.StatusCode, resp.Header, respBody)
}
if attempt < geminiMaxRetries {
upstreamReqID := resp.Header.Get(requestIDHeader)
if upstreamReqID == "" {
upstreamReqID = resp.Header.Get("x-goog-request-id")
}
upstreamMsg := strings.TrimSpace(extractUpstreamErrorMessage(respBody))
upstreamMsg = sanitizeUpstreamErrorMessage(upstreamMsg)
upstreamDetail := ""
if s.cfg != nil && s.cfg.Gateway.LogUpstreamErrorBody {
maxBytes := s.cfg.Gateway.LogUpstreamErrorBodyMaxBytes
if maxBytes <= 0 {
maxBytes = 2048
}
upstreamDetail = truncateString(string(respBody), maxBytes)
}
appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
Platform: account.Platform,
AccountID: account.ID,
AccountName: account.Name,
UpstreamStatusCode: resp.StatusCode,
UpstreamRequestID: upstreamReqID,
Kind: "retry",
Message: upstreamMsg,
Detail: upstreamDetail,
})
logger.LegacyPrintf("service.gemini_messages_compat", "Gemini account %d: upstream status %d, retry %d/%d", account.ID, resp.StatusCode, attempt, geminiMaxRetries)
sleepGeminiBackoff(attempt)
continue
}
// Final attempt: surface the upstream error body (mapped below) instead of a generic retry error.
resp = &http.Response{
StatusCode: resp.StatusCode,
Header: resp.Header.Clone(),
Body: io.NopCloser(bytes.NewReader(respBody)),
}
break
}
break
}
defer func() { _ = resp.Body.Close() }()
if resp.StatusCode >= 400 {
respBody, _ := io.ReadAll(io.LimitReader(resp.Body, 2<<20))
// 统一错误策略:自定义错误码 + 临时不可调度
if s.rateLimitService != nil {
switch s.rateLimitService.CheckErrorPolicy(ctx, account, resp.StatusCode, respBody) {
case ErrorPolicySkipped:
upstreamReqID := resp.Header.Get(requestIDHeader)
if upstreamReqID == "" {
upstreamReqID = resp.Header.Get("x-goog-request-id")
}
return nil, s.writeGeminiMappedError(c, account, http.StatusInternalServerError, upstreamReqID, respBody)
case ErrorPolicyMatched, ErrorPolicyTempUnscheduled:
s.handleGeminiUpstreamError(ctx, account, resp.StatusCode, resp.Header, respBody)
upstreamReqID := resp.Header.Get(requestIDHeader)
if upstreamReqID == "" {
upstreamReqID = resp.Header.Get("x-goog-request-id")
}
upstreamMsg := strings.TrimSpace(extractUpstreamErrorMessage(respBody))
upstreamMsg = sanitizeUpstreamErrorMessage(upstreamMsg)
upstreamDetail := ""
if s.cfg != nil && s.cfg.Gateway.LogUpstreamErrorBody {
maxBytes := s.cfg.Gateway.LogUpstreamErrorBodyMaxBytes
if maxBytes <= 0 {
maxBytes = 2048
}
upstreamDetail = truncateString(string(respBody), maxBytes)
}
appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
Platform: account.Platform,
AccountID: account.ID,
AccountName: account.Name,
UpstreamStatusCode: resp.StatusCode,
UpstreamRequestID: upstreamReqID,
Kind: "failover",
Message: upstreamMsg,
Detail: upstreamDetail,
})
return nil, &UpstreamFailoverError{StatusCode: resp.StatusCode, ResponseBody: respBody}
}
}
// ErrorPolicyNone → 原有逻辑
s.handleGeminiUpstreamError(ctx, account, resp.StatusCode, resp.Header, respBody)
// 精确匹配服务端配置类 400 错误,触发 failover + 临时封禁
if resp.StatusCode == http.StatusBadRequest {
msg400 := strings.ToLower(strings.TrimSpace(extractUpstreamErrorMessage(respBody)))
if isGoogleProjectConfigError(msg400) {
upstreamReqID := resp.Header.Get(requestIDHeader)
if upstreamReqID == "" {
upstreamReqID = resp.Header.Get("x-goog-request-id")
}
upstreamMsg := sanitizeUpstreamErrorMessage(strings.TrimSpace(extractUpstreamErrorMessage(respBody)))
upstreamDetail := ""
if s.cfg != nil && s.cfg.Gateway.LogUpstreamErrorBody {
maxBytes := s.cfg.Gateway.LogUpstreamErrorBodyMaxBytes
if maxBytes <= 0 {
maxBytes = 2048
}
upstreamDetail = truncateString(string(respBody), maxBytes)
}
log.Printf("[Gemini] status=400 google_config_error failover=true upstream_message=%q account=%d", upstreamMsg, account.ID)
appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
Platform: account.Platform,
AccountID: account.ID,
AccountName: account.Name,
UpstreamStatusCode: resp.StatusCode,
UpstreamRequestID: upstreamReqID,
Kind: "failover",
Message: upstreamMsg,
Detail: upstreamDetail,
})
return nil, &UpstreamFailoverError{StatusCode: resp.StatusCode, ResponseBody: respBody, RetryableOnSameAccount: true}
}
}
if s.shouldFailoverGeminiUpstreamError(resp.StatusCode) {
upstreamReqID := resp.Header.Get(requestIDHeader)
if upstreamReqID == "" {
upstreamReqID = resp.Header.Get("x-goog-request-id")
}
upstreamMsg := strings.TrimSpace(extractUpstreamErrorMessage(respBody))
upstreamMsg = sanitizeUpstreamErrorMessage(upstreamMsg)
upstreamDetail := ""
if s.cfg != nil && s.cfg.Gateway.LogUpstreamErrorBody {
maxBytes := s.cfg.Gateway.LogUpstreamErrorBodyMaxBytes
if maxBytes <= 0 {
maxBytes = 2048
}
upstreamDetail = truncateString(string(respBody), maxBytes)
}
appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
Platform: account.Platform,
AccountID: account.ID,
AccountName: account.Name,
UpstreamStatusCode: resp.StatusCode,
UpstreamRequestID: upstreamReqID,
Kind: "failover",
Message: upstreamMsg,
Detail: upstreamDetail,
})
return nil, &UpstreamFailoverError{StatusCode: resp.StatusCode, ResponseBody: respBody}
}
upstreamReqID := resp.Header.Get(requestIDHeader)
if upstreamReqID == "" {
upstreamReqID = resp.Header.Get("x-goog-request-id")
}
return nil, s.writeGeminiMappedError(c, account, resp.StatusCode, upstreamReqID, respBody)
}
requestID := resp.Header.Get(requestIDHeader)
if requestID == "" {
requestID = resp.Header.Get("x-goog-request-id")
}
if requestID != "" {
c.Header("x-request-id", requestID)
}
var usage *ClaudeUsage
var firstTokenMs *int
if req.Stream {
streamRes, err := s.handleStreamingResponse(c, resp, startTime, originalModel)
if err != nil {
return nil, err
}
usage = streamRes.usage
firstTokenMs = streamRes.firstTokenMs
} else {
if useUpstreamStream {
collected, usageObj, err := collectGeminiSSE(resp.Body, true)
if err != nil {
return nil, s.writeClaudeError(c, http.StatusBadGateway, "upstream_error", "Failed to read upstream stream")
}
collectedBytes, _ := json.Marshal(collected)
claudeResp, usageObj2 := convertGeminiToClaudeMessage(collected, originalModel, collectedBytes)
c.JSON(http.StatusOK, claudeResp)
usage = usageObj2
if usageObj != nil && (usageObj.InputTokens > 0 || usageObj.OutputTokens > 0) {
usage = usageObj
}
} else {
usage, err = s.handleNonStreamingResponse(c, resp, originalModel)
if err != nil {
return nil, err
}
}
}
// 图片生成计费
imageCount := 0
imageSize := s.extractImageSize(body)
if isImageGenerationModel(originalModel) {
imageCount = 1
}
return &ForwardResult{
RequestID: requestID,
Usage: *usage,
Model: originalModel,
Stream: req.Stream,
Duration: time.Since(startTime),
FirstTokenMs: firstTokenMs,
ImageCount: imageCount,
ImageSize: imageSize,
}, nil
}
func isGeminiSignatureRelatedError(respBody []byte) bool {
msg := strings.ToLower(strings.TrimSpace(extractAntigravityErrorMessage(respBody)))
if msg == "" {
msg = strings.ToLower(string(respBody))
}
return strings.Contains(msg, "thought_signature") || strings.Contains(msg, "signature")
}
func (s *GeminiMessagesCompatService) ForwardNative(ctx context.Context, c *gin.Context, account *Account, originalModel string, action string, stream bool, body []byte) (*ForwardResult, error) {
startTime := time.Now()
if strings.TrimSpace(originalModel) == "" {
return nil, s.writeGoogleError(c, http.StatusBadRequest, "Missing model in URL")
}
if strings.TrimSpace(action) == "" {
return nil, s.writeGoogleError(c, http.StatusBadRequest, "Missing action in URL")
}
if len(body) == 0 {
return nil, s.writeGoogleError(c, http.StatusBadRequest, "Request body is empty")
}
// 过滤掉 parts 为空的消息Gemini API 不接受空 parts
if filteredBody, err := filterEmptyPartsFromGeminiRequest(body); err == nil {
body = filteredBody
}
switch action {
case "generateContent", "streamGenerateContent", "countTokens":
// ok
default:
return nil, s.writeGoogleError(c, http.StatusNotFound, "Unsupported action: "+action)
}
// Some Gemini upstreams validate tool call parts strictly; ensure any `functionCall` part includes a
// `thoughtSignature` to avoid frequent INVALID_ARGUMENT 400s.
body = ensureGeminiFunctionCallThoughtSignatures(body)
mappedModel := originalModel
if account.Type == AccountTypeAPIKey {
mappedModel = account.GetMappedModel(originalModel)
}
proxyURL := ""
if account.ProxyID != nil && account.Proxy != nil {
proxyURL = account.Proxy.URL()
}
useUpstreamStream := stream
upstreamAction := action
if account.Type == AccountTypeOAuth && !stream && action == "generateContent" && strings.TrimSpace(account.GetCredential("project_id")) != "" {
// Code Assist's non-streaming generateContent may return no content; use streaming upstream and aggregate.
useUpstreamStream = true
upstreamAction = "streamGenerateContent"
}
forceAIStudio := action == "countTokens"
var requestIDHeader string
var buildReq func(ctx context.Context) (*http.Request, string, error)
switch account.Type {
case AccountTypeAPIKey:
buildReq = func(ctx context.Context) (*http.Request, string, error) {
apiKey := account.GetCredential("api_key")
if strings.TrimSpace(apiKey) == "" {
return nil, "", errors.New("gemini api_key not configured")
}
baseURL := account.GetGeminiBaseURL(geminicli.AIStudioBaseURL)
normalizedBaseURL, err := s.validateUpstreamBaseURL(baseURL)
if err != nil {
return nil, "", err
}
fullURL := fmt.Sprintf("%s/v1beta/models/%s:%s", strings.TrimRight(normalizedBaseURL, "/"), mappedModel, upstreamAction)
if useUpstreamStream {
fullURL += "?alt=sse"
}
upstreamReq, err := http.NewRequestWithContext(ctx, http.MethodPost, fullURL, bytes.NewReader(body))
if err != nil {
return nil, "", err
}
upstreamReq.Header.Set("Content-Type", "application/json")
upstreamReq.Header.Set("x-goog-api-key", apiKey)
return upstreamReq, "x-request-id", nil
}
requestIDHeader = "x-request-id"
case AccountTypeOAuth:
buildReq = func(ctx context.Context) (*http.Request, string, error) {
if s.tokenProvider == nil {
return nil, "", errors.New("gemini token provider not configured")
}
accessToken, err := s.tokenProvider.GetAccessToken(ctx, account)
if err != nil {
return nil, "", err
}
projectID := strings.TrimSpace(account.GetCredential("project_id"))
// Two modes for OAuth:
// 1. With project_id -> Code Assist API (wrapped request)
// 2. Without project_id -> AI Studio API (direct OAuth, like API key but with Bearer token)
if projectID != "" && !forceAIStudio {
// Mode 1: Code Assist API
baseURL, err := s.validateUpstreamBaseURL(geminicli.GeminiCliBaseURL)
if err != nil {
return nil, "", err
}
fullURL := fmt.Sprintf("%s/v1internal:%s", strings.TrimRight(baseURL, "/"), upstreamAction)
if useUpstreamStream {
fullURL += "?alt=sse"
}
wrapped := map[string]any{
"model": mappedModel,
"project": projectID,
}
var inner any
if err := json.Unmarshal(body, &inner); err != nil {
return nil, "", fmt.Errorf("failed to parse gemini request: %w", err)
}
wrapped["request"] = inner
wrappedBytes, _ := json.Marshal(wrapped)
upstreamReq, err := http.NewRequestWithContext(ctx, http.MethodPost, fullURL, bytes.NewReader(wrappedBytes))
if err != nil {
return nil, "", err
}
upstreamReq.Header.Set("Content-Type", "application/json")
upstreamReq.Header.Set("Authorization", "Bearer "+accessToken)
upstreamReq.Header.Set("User-Agent", geminicli.GeminiCLIUserAgent)
return upstreamReq, "x-request-id", nil
} else {
// Mode 2: AI Studio API with OAuth (like API key mode, but using Bearer token)
baseURL := account.GetGeminiBaseURL(geminicli.AIStudioBaseURL)
normalizedBaseURL, err := s.validateUpstreamBaseURL(baseURL)
if err != nil {
return nil, "", err
}
fullURL := fmt.Sprintf("%s/v1beta/models/%s:%s", strings.TrimRight(normalizedBaseURL, "/"), mappedModel, upstreamAction)
if useUpstreamStream {
fullURL += "?alt=sse"
}
upstreamReq, err := http.NewRequestWithContext(ctx, http.MethodPost, fullURL, bytes.NewReader(body))
if err != nil {
return nil, "", err
}
upstreamReq.Header.Set("Content-Type", "application/json")
upstreamReq.Header.Set("Authorization", "Bearer "+accessToken)
return upstreamReq, "x-request-id", nil
}
}
requestIDHeader = "x-request-id"
default:
return nil, s.writeGoogleError(c, http.StatusBadGateway, "Unsupported account type: "+account.Type)
}
var resp *http.Response
for attempt := 1; attempt <= geminiMaxRetries; attempt++ {
upstreamReq, idHeader, err := buildReq(ctx)
if err != nil {
if errors.Is(err, context.Canceled) || errors.Is(err, context.DeadlineExceeded) {
return nil, err
}
// Local build error: don't retry.
if strings.Contains(err.Error(), "missing project_id") {
return nil, s.writeGoogleError(c, http.StatusBadRequest, err.Error())
}
return nil, s.writeGoogleError(c, http.StatusBadGateway, err.Error())
}
requestIDHeader = idHeader
// Capture upstream request body for ops retry of this attempt.
if c != nil {
// In this code path `body` is already the JSON sent to upstream.
c.Set(OpsUpstreamRequestBodyKey, string(body))
}
resp, err = s.httpUpstream.Do(upstreamReq, proxyURL, account.ID, account.Concurrency)
if err != nil {
safeErr := sanitizeUpstreamErrorMessage(err.Error())
appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
Platform: account.Platform,
AccountID: account.ID,
AccountName: account.Name,
UpstreamStatusCode: 0,
Kind: "request_error",
Message: safeErr,
})
if attempt < geminiMaxRetries {
logger.LegacyPrintf("service.gemini_messages_compat", "Gemini account %d: upstream request failed, retry %d/%d: %v", account.ID, attempt, geminiMaxRetries, err)
sleepGeminiBackoff(attempt)
continue
}
if action == "countTokens" {
estimated := estimateGeminiCountTokens(body)
c.JSON(http.StatusOK, map[string]any{"totalTokens": estimated})
return &ForwardResult{
RequestID: "",
Usage: ClaudeUsage{},
Model: originalModel,
Stream: false,
Duration: time.Since(startTime),
FirstTokenMs: nil,
}, nil
}
setOpsUpstreamError(c, 0, safeErr, "")
return nil, s.writeGoogleError(c, http.StatusBadGateway, "Upstream request failed after retries: "+safeErr)
}
// 错误策略优先:匹配则跳过重试直接处理。
if matched, rebuilt := s.checkErrorPolicyInLoop(ctx, account, resp); matched {
resp = rebuilt
break
} else {
resp = rebuilt
}
if resp.StatusCode >= 400 && s.shouldRetryGeminiUpstreamError(account, resp.StatusCode) {
respBody, _ := io.ReadAll(io.LimitReader(resp.Body, 2<<20))
_ = resp.Body.Close()
// Don't treat insufficient-scope as transient.
if resp.StatusCode == 403 && isGeminiInsufficientScope(resp.Header, respBody) {
resp = &http.Response{
StatusCode: resp.StatusCode,
Header: resp.Header.Clone(),
Body: io.NopCloser(bytes.NewReader(respBody)),
}
break
}
if resp.StatusCode == 429 {
s.handleGeminiUpstreamError(ctx, account, resp.StatusCode, resp.Header, respBody)
}
if attempt < geminiMaxRetries {
upstreamReqID := resp.Header.Get(requestIDHeader)
if upstreamReqID == "" {
upstreamReqID = resp.Header.Get("x-goog-request-id")
}
upstreamMsg := strings.TrimSpace(extractUpstreamErrorMessage(respBody))
upstreamMsg = sanitizeUpstreamErrorMessage(upstreamMsg)
upstreamDetail := ""
if s.cfg != nil && s.cfg.Gateway.LogUpstreamErrorBody {
maxBytes := s.cfg.Gateway.LogUpstreamErrorBodyMaxBytes
if maxBytes <= 0 {
maxBytes = 2048
}
upstreamDetail = truncateString(string(respBody), maxBytes)
}
appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
Platform: account.Platform,
AccountID: account.ID,
AccountName: account.Name,
UpstreamStatusCode: resp.StatusCode,
UpstreamRequestID: upstreamReqID,
Kind: "retry",
Message: upstreamMsg,
Detail: upstreamDetail,
})
logger.LegacyPrintf("service.gemini_messages_compat", "Gemini account %d: upstream status %d, retry %d/%d", account.ID, resp.StatusCode, attempt, geminiMaxRetries)
sleepGeminiBackoff(attempt)
continue
}
if action == "countTokens" {
estimated := estimateGeminiCountTokens(body)
c.JSON(http.StatusOK, map[string]any{"totalTokens": estimated})
return &ForwardResult{
RequestID: "",
Usage: ClaudeUsage{},
Model: originalModel,
Stream: false,
Duration: time.Since(startTime),
FirstTokenMs: nil,
}, nil
}
// Final attempt: surface the upstream error body (passed through below) instead of a generic retry error.
resp = &http.Response{
StatusCode: resp.StatusCode,
Header: resp.Header.Clone(),
Body: io.NopCloser(bytes.NewReader(respBody)),
}
break
}
break
}
defer func() { _ = resp.Body.Close() }()
requestID := resp.Header.Get(requestIDHeader)
if requestID == "" {
requestID = resp.Header.Get("x-goog-request-id")
}
if requestID != "" {
c.Header("x-request-id", requestID)
}
isOAuth := account.Type == AccountTypeOAuth
if resp.StatusCode >= 400 {
respBody, _ := io.ReadAll(io.LimitReader(resp.Body, 2<<20))
// Best-effort fallback for OAuth tokens missing AI Studio scopes when calling countTokens.
// This avoids Gemini SDKs failing hard during preflight token counting.
// Checked before error policy so it always works regardless of custom error codes.
if action == "countTokens" && isOAuth && isGeminiInsufficientScope(resp.Header, respBody) {
estimated := estimateGeminiCountTokens(body)
c.JSON(http.StatusOK, map[string]any{"totalTokens": estimated})
return &ForwardResult{
RequestID: requestID,
Usage: ClaudeUsage{},
Model: originalModel,
Stream: false,
Duration: time.Since(startTime),
FirstTokenMs: nil,
}, nil
}
// 统一错误策略:自定义错误码 + 临时不可调度
if s.rateLimitService != nil {
switch s.rateLimitService.CheckErrorPolicy(ctx, account, resp.StatusCode, respBody) {
case ErrorPolicySkipped:
respBody = unwrapIfNeeded(isOAuth, respBody)
contentType := resp.Header.Get("Content-Type")
if contentType == "" {
contentType = "application/json"
}
c.Data(http.StatusInternalServerError, contentType, respBody)
return nil, fmt.Errorf("gemini upstream error: %d (skipped by error policy)", resp.StatusCode)
case ErrorPolicyMatched, ErrorPolicyTempUnscheduled:
s.handleGeminiUpstreamError(ctx, account, resp.StatusCode, resp.Header, respBody)
evBody := unwrapIfNeeded(isOAuth, respBody)
upstreamMsg := strings.TrimSpace(extractUpstreamErrorMessage(evBody))
upstreamMsg = sanitizeUpstreamErrorMessage(upstreamMsg)
upstreamDetail := ""
if s.cfg != nil && s.cfg.Gateway.LogUpstreamErrorBody {
maxBytes := s.cfg.Gateway.LogUpstreamErrorBodyMaxBytes
if maxBytes <= 0 {
maxBytes = 2048
}
upstreamDetail = truncateString(string(evBody), maxBytes)
}
appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
Platform: account.Platform,
AccountID: account.ID,
AccountName: account.Name,
UpstreamStatusCode: resp.StatusCode,
UpstreamRequestID: requestID,
Kind: "failover",
Message: upstreamMsg,
Detail: upstreamDetail,
})
return nil, &UpstreamFailoverError{StatusCode: resp.StatusCode, ResponseBody: respBody}
}
}
// ErrorPolicyNone → 原有逻辑
s.handleGeminiUpstreamError(ctx, account, resp.StatusCode, resp.Header, respBody)
// 精确匹配服务端配置类 400 错误,触发 failover + 临时封禁
if resp.StatusCode == http.StatusBadRequest {
msg400 := strings.ToLower(strings.TrimSpace(extractUpstreamErrorMessage(respBody)))
if isGoogleProjectConfigError(msg400) {
evBody := unwrapIfNeeded(isOAuth, respBody)
upstreamMsg := sanitizeUpstreamErrorMessage(strings.TrimSpace(extractUpstreamErrorMessage(evBody)))
upstreamDetail := ""
if s.cfg != nil && s.cfg.Gateway.LogUpstreamErrorBody {
maxBytes := s.cfg.Gateway.LogUpstreamErrorBodyMaxBytes
if maxBytes <= 0 {
maxBytes = 2048
}
upstreamDetail = truncateString(string(evBody), maxBytes)
}
log.Printf("[Gemini] status=400 google_config_error failover=true upstream_message=%q account=%d", upstreamMsg, account.ID)
appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
Platform: account.Platform,
AccountID: account.ID,
AccountName: account.Name,
UpstreamStatusCode: resp.StatusCode,
UpstreamRequestID: requestID,
Kind: "failover",
Message: upstreamMsg,
Detail: upstreamDetail,
})
return nil, &UpstreamFailoverError{StatusCode: resp.StatusCode, ResponseBody: evBody, RetryableOnSameAccount: true}
}
}
if s.shouldFailoverGeminiUpstreamError(resp.StatusCode) {
evBody := unwrapIfNeeded(isOAuth, respBody)
upstreamMsg := strings.TrimSpace(extractUpstreamErrorMessage(evBody))
upstreamMsg = sanitizeUpstreamErrorMessage(upstreamMsg)
upstreamDetail := ""
if s.cfg != nil && s.cfg.Gateway.LogUpstreamErrorBody {
maxBytes := s.cfg.Gateway.LogUpstreamErrorBodyMaxBytes
if maxBytes <= 0 {
maxBytes = 2048
}
upstreamDetail = truncateString(string(evBody), maxBytes)
}
appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
Platform: account.Platform,
AccountID: account.ID,
AccountName: account.Name,
UpstreamStatusCode: resp.StatusCode,
UpstreamRequestID: requestID,
Kind: "failover",
Message: upstreamMsg,
Detail: upstreamDetail,
})
return nil, &UpstreamFailoverError{StatusCode: resp.StatusCode, ResponseBody: evBody}
}
respBody = unwrapIfNeeded(isOAuth, respBody)
upstreamMsg := strings.TrimSpace(extractUpstreamErrorMessage(respBody))
upstreamMsg = sanitizeUpstreamErrorMessage(upstreamMsg)
upstreamDetail := ""
if s.cfg != nil && s.cfg.Gateway.LogUpstreamErrorBody {
maxBytes := s.cfg.Gateway.LogUpstreamErrorBodyMaxBytes
if maxBytes <= 0 {
maxBytes = 2048
}
upstreamDetail = truncateString(string(respBody), maxBytes)
logger.LegacyPrintf("service.gemini_messages_compat", "[Gemini] native upstream error %d: %s", resp.StatusCode, truncateForLog(respBody, s.cfg.Gateway.LogUpstreamErrorBodyMaxBytes))
}
setOpsUpstreamError(c, resp.StatusCode, upstreamMsg, upstreamDetail)
appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
Platform: account.Platform,
AccountID: account.ID,
AccountName: account.Name,
UpstreamStatusCode: resp.StatusCode,
UpstreamRequestID: requestID,
Kind: "http_error",
Message: upstreamMsg,
Detail: upstreamDetail,
})
contentType := resp.Header.Get("Content-Type")
if contentType == "" {
contentType = "application/json"
}
c.Data(resp.StatusCode, contentType, respBody)
if upstreamMsg == "" {
return nil, fmt.Errorf("gemini upstream error: %d", resp.StatusCode)
}
return nil, fmt.Errorf("gemini upstream error: %d message=%s", resp.StatusCode, upstreamMsg)
}
var usage *ClaudeUsage
var firstTokenMs *int
if stream {
streamRes, err := s.handleNativeStreamingResponse(c, resp, startTime, isOAuth)
if err != nil {
return nil, err
}
usage = streamRes.usage
firstTokenMs = streamRes.firstTokenMs
} else {
if useUpstreamStream {
collected, usageObj, err := collectGeminiSSE(resp.Body, isOAuth)
if err != nil {
return nil, s.writeGoogleError(c, http.StatusBadGateway, "Failed to read upstream stream")
}
b, _ := json.Marshal(collected)
c.Data(http.StatusOK, "application/json", b)
usage = usageObj
} else {
usageResp, err := s.handleNativeNonStreamingResponse(c, resp, isOAuth)
if err != nil {
return nil, err
}
usage = usageResp
}
}
if usage == nil {
usage = &ClaudeUsage{}
}
// 图片生成计费
imageCount := 0
imageSize := s.extractImageSize(body)
if isImageGenerationModel(originalModel) {
imageCount = 1
}
return &ForwardResult{
RequestID: requestID,
Usage: *usage,
Model: originalModel,
Stream: stream,
Duration: time.Since(startTime),
FirstTokenMs: firstTokenMs,
ImageCount: imageCount,
ImageSize: imageSize,
}, nil
}
// checkErrorPolicyInLoop 在重试循环内预检查错误策略。
// 返回 true 表示策略已匹配(调用者应 breakresp 已重建可直接使用。
// 返回 false 表示 ErrorPolicyNoneresp 已重建,调用者继续走重试逻辑。
func (s *GeminiMessagesCompatService) checkErrorPolicyInLoop(
ctx context.Context, account *Account, resp *http.Response,
) (matched bool, rebuilt *http.Response) {
if resp.StatusCode < 400 || s.rateLimitService == nil {
return false, resp
}
body, _ := io.ReadAll(io.LimitReader(resp.Body, 2<<20))
_ = resp.Body.Close()
rebuilt = &http.Response{
StatusCode: resp.StatusCode,
Header: resp.Header.Clone(),
Body: io.NopCloser(bytes.NewReader(body)),
}
policy := s.rateLimitService.CheckErrorPolicy(ctx, account, resp.StatusCode, body)
return policy != ErrorPolicyNone, rebuilt
}
func (s *GeminiMessagesCompatService) shouldRetryGeminiUpstreamError(account *Account, statusCode int) bool {
switch statusCode {
case 429, 500, 502, 503, 504, 529:
return true
case 403:
// GeminiCli OAuth occasionally returns 403 transiently (activation/quota propagation); allow retry.
if account == nil || account.Type != AccountTypeOAuth {
return false
}
oauthType := strings.ToLower(strings.TrimSpace(account.GetCredential("oauth_type")))
if oauthType == "" && strings.TrimSpace(account.GetCredential("project_id")) != "" {
// Legacy/implicit Code Assist OAuth accounts.
oauthType = "code_assist"
}
return oauthType == "code_assist"
default:
return false
}
}
func (s *GeminiMessagesCompatService) shouldFailoverGeminiUpstreamError(statusCode int) bool {
switch statusCode {
case 401, 403, 429, 529:
return true
default:
return statusCode >= 500
}
}
func sleepGeminiBackoff(attempt int) {
delay := geminiRetryBaseDelay * time.Duration(1<<uint(attempt-1))
if delay > geminiRetryMaxDelay {
delay = geminiRetryMaxDelay
}
// +/- 20% jitter
r := mathrand.New(mathrand.NewSource(time.Now().UnixNano()))
jitter := time.Duration(float64(delay) * 0.2 * (r.Float64()*2 - 1))
sleepFor := delay + jitter
if sleepFor < 0 {
sleepFor = 0
}
time.Sleep(sleepFor)
}
var (
sensitiveQueryParamRegex = regexp.MustCompile(`(?i)([?&](?:key|client_secret|access_token|refresh_token)=)[^&"\s]+`)
retryInRegex = regexp.MustCompile(`Please retry in ([0-9.]+)s`)
)
func sanitizeUpstreamErrorMessage(msg string) string {
if msg == "" {
return msg
}
return sensitiveQueryParamRegex.ReplaceAllString(msg, `$1***`)
}
func (s *GeminiMessagesCompatService) writeGeminiMappedError(c *gin.Context, account *Account, upstreamStatus int, upstreamRequestID string, body []byte) error {
upstreamMsg := strings.TrimSpace(extractUpstreamErrorMessage(body))
upstreamMsg = sanitizeUpstreamErrorMessage(upstreamMsg)
upstreamDetail := ""
if s.cfg != nil && s.cfg.Gateway.LogUpstreamErrorBody {
maxBytes := s.cfg.Gateway.LogUpstreamErrorBodyMaxBytes
if maxBytes <= 0 {
maxBytes = 2048
}
upstreamDetail = truncateString(string(body), maxBytes)
}
setOpsUpstreamError(c, upstreamStatus, upstreamMsg, upstreamDetail)
appendOpsUpstreamError(c, OpsUpstreamErrorEvent{
Platform: account.Platform,
AccountID: account.ID,
AccountName: account.Name,
UpstreamStatusCode: upstreamStatus,
UpstreamRequestID: upstreamRequestID,
Kind: "http_error",
Message: upstreamMsg,
Detail: upstreamDetail,
})
if s.cfg != nil && s.cfg.Gateway.LogUpstreamErrorBody {
logger.LegacyPrintf("service.gemini_messages_compat", "[Gemini] upstream error %d: %s", upstreamStatus, truncateForLog(body, s.cfg.Gateway.LogUpstreamErrorBodyMaxBytes))
}
if status, errType, errMsg, matched := applyErrorPassthroughRule(
c,
PlatformGemini,
upstreamStatus,
body,
http.StatusBadGateway,
"upstream_error",
"Upstream request failed",
); matched {
c.JSON(status, gin.H{
"type": "error",
"error": gin.H{"type": errType, "message": errMsg},
})
if upstreamMsg == "" {
upstreamMsg = errMsg
}
if upstreamMsg == "" {
return fmt.Errorf("upstream error: %d (passthrough rule matched)", upstreamStatus)
}
return fmt.Errorf("upstream error: %d (passthrough rule matched) message=%s", upstreamStatus, upstreamMsg)
}
var statusCode int
var errType, errMsg string
if mapped := mapGeminiErrorBodyToClaudeError(body); mapped != nil {
errType = mapped.Type
if mapped.Message != "" {
errMsg = mapped.Message
}
if mapped.StatusCode > 0 {
statusCode = mapped.StatusCode
}
}
switch upstreamStatus {
case 400:
if statusCode == 0 {
statusCode = http.StatusBadRequest
}
if errType == "" {
errType = "invalid_request_error"
}
if errMsg == "" {
errMsg = "Invalid request"
}
case 401:
if statusCode == 0 {
statusCode = http.StatusBadGateway
}
if errType == "" {
errType = "authentication_error"
}
if errMsg == "" {
errMsg = "Upstream authentication failed, please contact administrator"
}
case 403:
if statusCode == 0 {
statusCode = http.StatusBadGateway
}
if errType == "" {
errType = "permission_error"
}
if errMsg == "" {
errMsg = "Upstream access forbidden, please contact administrator"
}
case 404:
if statusCode == 0 {
statusCode = http.StatusNotFound
}
if errType == "" {
errType = "not_found_error"
}
if errMsg == "" {
errMsg = "Resource not found"
}
case 429:
if statusCode == 0 {
statusCode = http.StatusTooManyRequests
}
if errType == "" {
errType = "rate_limit_error"
}
if errMsg == "" {
errMsg = "Upstream rate limit exceeded, please retry later"
}
case 529:
if statusCode == 0 {
statusCode = http.StatusServiceUnavailable
}
if errType == "" {
errType = "overloaded_error"
}
if errMsg == "" {
errMsg = "Upstream service overloaded, please retry later"
}
case 500, 502, 503, 504:
if statusCode == 0 {
statusCode = http.StatusBadGateway
}
if errType == "" {
switch upstreamStatus {
case 504:
errType = "timeout_error"
case 503:
errType = "overloaded_error"
default:
errType = "api_error"
}
}
if errMsg == "" {
errMsg = "Upstream service temporarily unavailable"
}
default:
if statusCode == 0 {
statusCode = http.StatusBadGateway
}
if errType == "" {
errType = "upstream_error"
}
if errMsg == "" {
errMsg = "Upstream request failed"
}
}
c.JSON(statusCode, gin.H{
"type": "error",
"error": gin.H{"type": errType, "message": errMsg},
})
if upstreamMsg == "" {
return fmt.Errorf("upstream error: %d", upstreamStatus)
}
return fmt.Errorf("upstream error: %d message=%s", upstreamStatus, upstreamMsg)
}
type claudeErrorMapping struct {
Type string
Message string
StatusCode int
}
func mapGeminiErrorBodyToClaudeError(body []byte) *claudeErrorMapping {
if len(body) == 0 {
return nil
}
var parsed struct {
Error struct {
Code int `json:"code"`
Message string `json:"message"`
Status string `json:"status"`
} `json:"error"`
}
if err := json.Unmarshal(body, &parsed); err != nil {
return nil
}
if strings.TrimSpace(parsed.Error.Status) == "" && parsed.Error.Code == 0 && strings.TrimSpace(parsed.Error.Message) == "" {
return nil
}
mapped := &claudeErrorMapping{
Type: mapGeminiStatusToClaudeErrorType(parsed.Error.Status),
Message: "",
}
if mapped.Type == "" {
mapped.Type = "upstream_error"
}
switch strings.ToUpper(strings.TrimSpace(parsed.Error.Status)) {
case "INVALID_ARGUMENT":
mapped.StatusCode = http.StatusBadRequest
case "NOT_FOUND":
mapped.StatusCode = http.StatusNotFound
case "RESOURCE_EXHAUSTED":
mapped.StatusCode = http.StatusTooManyRequests
default:
// Keep StatusCode unset and let HTTP status mapping decide.
}
// Keep messages generic by default; upstream error message can be long or include sensitive fragments.
return mapped
}
func mapGeminiStatusToClaudeErrorType(status string) string {
switch strings.ToUpper(strings.TrimSpace(status)) {
case "INVALID_ARGUMENT":
return "invalid_request_error"
case "PERMISSION_DENIED":
return "permission_error"
case "NOT_FOUND":
return "not_found_error"
case "RESOURCE_EXHAUSTED":
return "rate_limit_error"
case "UNAUTHENTICATED":
return "authentication_error"
case "UNAVAILABLE":
return "overloaded_error"
case "INTERNAL":
return "api_error"
case "DEADLINE_EXCEEDED":
return "timeout_error"
default:
return ""
}
}
type geminiStreamResult struct {
usage *ClaudeUsage
firstTokenMs *int
}
func (s *GeminiMessagesCompatService) handleNonStreamingResponse(c *gin.Context, resp *http.Response, originalModel string) (*ClaudeUsage, error) {
body, err := io.ReadAll(io.LimitReader(resp.Body, 8<<20))
if err != nil {
return nil, s.writeClaudeError(c, http.StatusBadGateway, "upstream_error", "Failed to read upstream response")
}
unwrappedBody, err := unwrapGeminiResponse(body)
if err != nil {
return nil, s.writeClaudeError(c, http.StatusBadGateway, "upstream_error", "Failed to parse upstream response")
}
var geminiResp map[string]any
if err := json.Unmarshal(unwrappedBody, &geminiResp); err != nil {
return nil, s.writeClaudeError(c, http.StatusBadGateway, "upstream_error", "Failed to parse upstream response")
}
claudeResp, usage := convertGeminiToClaudeMessage(geminiResp, originalModel, unwrappedBody)
c.JSON(http.StatusOK, claudeResp)
return usage, nil
}
func (s *GeminiMessagesCompatService) handleStreamingResponse(c *gin.Context, resp *http.Response, startTime time.Time, originalModel string) (*geminiStreamResult, error) {
c.Header("Content-Type", "text/event-stream")
c.Header("Cache-Control", "no-cache")
c.Header("Connection", "keep-alive")
c.Header("X-Accel-Buffering", "no")
c.Status(http.StatusOK)
flusher, ok := c.Writer.(http.Flusher)
if !ok {
return nil, errors.New("streaming not supported")
}
messageID := "msg_" + randomHex(12)
messageStart := map[string]any{
"type": "message_start",
"message": map[string]any{
"id": messageID,
"type": "message",
"role": "assistant",
"model": originalModel,
"content": []any{},
"stop_reason": nil,
"stop_sequence": nil,
"usage": map[string]any{
"input_tokens": 0,
"output_tokens": 0,
},
},
}
writeSSE(c.Writer, "message_start", messageStart)
flusher.Flush()
var firstTokenMs *int
var usage ClaudeUsage
finishReason := ""
sawToolUse := false
nextBlockIndex := 0
openBlockIndex := -1
openBlockType := ""
seenText := ""
openToolIndex := -1
openToolID := ""
openToolName := ""
seenToolJSON := ""
reader := bufio.NewReader(resp.Body)
for {
line, err := reader.ReadString('\n')
if err != nil && !errors.Is(err, io.EOF) {
return nil, fmt.Errorf("stream read error: %w", err)
}
if !strings.HasPrefix(line, "data:") {
if errors.Is(err, io.EOF) {
break
}
continue
}
payload := strings.TrimSpace(strings.TrimPrefix(line, "data:"))
if payload == "" || payload == "[DONE]" {
if errors.Is(err, io.EOF) {
break
}
continue
}
unwrappedBytes, err := unwrapGeminiResponse([]byte(payload))
if err != nil {
continue
}
var geminiResp map[string]any
if err := json.Unmarshal(unwrappedBytes, &geminiResp); err != nil {
continue
}
if fr := extractGeminiFinishReason(geminiResp); fr != "" {
finishReason = fr
}
parts := extractGeminiParts(geminiResp)
for _, part := range parts {
if text, ok := part["text"].(string); ok && text != "" {
delta, newSeen := computeGeminiTextDelta(seenText, text)
seenText = newSeen
if delta == "" {
continue
}
if openBlockType != "text" {
if openBlockIndex >= 0 {
writeSSE(c.Writer, "content_block_stop", map[string]any{
"type": "content_block_stop",
"index": openBlockIndex,
})
}
openBlockType = "text"
openBlockIndex = nextBlockIndex
nextBlockIndex++
writeSSE(c.Writer, "content_block_start", map[string]any{
"type": "content_block_start",
"index": openBlockIndex,
"content_block": map[string]any{
"type": "text",
"text": "",
},
})
}
if firstTokenMs == nil {
ms := int(time.Since(startTime).Milliseconds())
firstTokenMs = &ms
}
writeSSE(c.Writer, "content_block_delta", map[string]any{
"type": "content_block_delta",
"index": openBlockIndex,
"delta": map[string]any{
"type": "text_delta",
"text": delta,
},
})
flusher.Flush()
continue
}
if fc, ok := part["functionCall"].(map[string]any); ok && fc != nil {
name, _ := fc["name"].(string)
args := fc["args"]
if strings.TrimSpace(name) == "" {
name = "tool"
}
// Close any open text block before tool_use.
if openBlockIndex >= 0 {
writeSSE(c.Writer, "content_block_stop", map[string]any{
"type": "content_block_stop",
"index": openBlockIndex,
})
openBlockIndex = -1
openBlockType = ""
}
// If we receive streamed tool args in pieces, keep a single tool block open and emit deltas.
if openToolIndex >= 0 && openToolName != name {
writeSSE(c.Writer, "content_block_stop", map[string]any{
"type": "content_block_stop",
"index": openToolIndex,
})
openToolIndex = -1
openToolName = ""
seenToolJSON = ""
}
if openToolIndex < 0 {
openToolID = "toolu_" + randomHex(8)
openToolIndex = nextBlockIndex
openToolName = name
nextBlockIndex++
sawToolUse = true
writeSSE(c.Writer, "content_block_start", map[string]any{
"type": "content_block_start",
"index": openToolIndex,
"content_block": map[string]any{
"type": "tool_use",
"id": openToolID,
"name": name,
"input": map[string]any{},
},
})
}
argsJSONText := "{}"
switch v := args.(type) {
case nil:
// keep default "{}"
case string:
if strings.TrimSpace(v) != "" {
argsJSONText = v
}
default:
if b, err := json.Marshal(args); err == nil && len(b) > 0 {
argsJSONText = string(b)
}
}
delta, newSeen := computeGeminiTextDelta(seenToolJSON, argsJSONText)
seenToolJSON = newSeen
if delta != "" {
writeSSE(c.Writer, "content_block_delta", map[string]any{
"type": "content_block_delta",
"index": openToolIndex,
"delta": map[string]any{
"type": "input_json_delta",
"partial_json": delta,
},
})
}
flusher.Flush()
}
}
if u := extractGeminiUsage(unwrappedBytes); u != nil {
usage = *u
}
// Process the final unterminated line at EOF as well.
if errors.Is(err, io.EOF) {
break
}
}
if openBlockIndex >= 0 {
writeSSE(c.Writer, "content_block_stop", map[string]any{
"type": "content_block_stop",
"index": openBlockIndex,
})
}
if openToolIndex >= 0 {
writeSSE(c.Writer, "content_block_stop", map[string]any{
"type": "content_block_stop",
"index": openToolIndex,
})
}
stopReason := mapGeminiFinishReasonToClaudeStopReason(finishReason)
if sawToolUse {
stopReason = "tool_use"
}
usageObj := map[string]any{
"output_tokens": usage.OutputTokens,
}
if usage.InputTokens > 0 {
usageObj["input_tokens"] = usage.InputTokens
}
writeSSE(c.Writer, "message_delta", map[string]any{
"type": "message_delta",
"delta": map[string]any{
"stop_reason": stopReason,
"stop_sequence": nil,
},
"usage": usageObj,
})
writeSSE(c.Writer, "message_stop", map[string]any{
"type": "message_stop",
})
flusher.Flush()
return &geminiStreamResult{usage: &usage, firstTokenMs: firstTokenMs}, nil
}
func writeSSE(w io.Writer, event string, data any) {
if event != "" {
_, _ = fmt.Fprintf(w, "event: %s\n", event)
}
b, _ := json.Marshal(data)
_, _ = fmt.Fprintf(w, "data: %s\n\n", string(b))
}
func randomHex(nBytes int) string {
b := make([]byte, nBytes)
_, _ = rand.Read(b)
return hex.EncodeToString(b)
}
func (s *GeminiMessagesCompatService) writeClaudeError(c *gin.Context, status int, errType, message string) error {
c.JSON(status, gin.H{
"type": "error",
"error": gin.H{"type": errType, "message": message},
})
return fmt.Errorf("%s", message)
}
func (s *GeminiMessagesCompatService) writeGoogleError(c *gin.Context, status int, message string) error {
c.JSON(status, gin.H{
"error": gin.H{
"code": status,
"message": message,
"status": googleapi.HTTPStatusToGoogleStatus(status),
},
})
return fmt.Errorf("%s", message)
}
func unwrapIfNeeded(isOAuth bool, raw []byte) []byte {
if !isOAuth {
return raw
}
inner, err := unwrapGeminiResponse(raw)
if err != nil {
return raw
}
return inner
}
func collectGeminiSSE(body io.Reader, isOAuth bool) (map[string]any, *ClaudeUsage, error) {
reader := bufio.NewReader(body)
var last map[string]any
var lastWithParts map[string]any
2026-02-02 22:13:50 +08:00
var collectedTextParts []string // Collect all text parts for aggregation
usage := &ClaudeUsage{}
for {
line, err := reader.ReadString('\n')
if len(line) > 0 {
trimmed := strings.TrimRight(line, "\r\n")
if strings.HasPrefix(trimmed, "data:") {
payload := strings.TrimSpace(strings.TrimPrefix(trimmed, "data:"))
switch payload {
case "", "[DONE]":
if payload == "[DONE]" {
2026-02-02 22:13:50 +08:00
return mergeCollectedTextParts(pickGeminiCollectResult(last, lastWithParts), collectedTextParts), usage, nil
}
default:
var parsed map[string]any
var rawBytes []byte
if isOAuth {
innerBytes, err := unwrapGeminiResponse([]byte(payload))
if err == nil {
rawBytes = innerBytes
_ = json.Unmarshal(innerBytes, &parsed)
}
} else {
rawBytes = []byte(payload)
_ = json.Unmarshal(rawBytes, &parsed)
}
if parsed != nil {
last = parsed
if u := extractGeminiUsage(rawBytes); u != nil {
usage = u
}
if parts := extractGeminiParts(parsed); len(parts) > 0 {
lastWithParts = parsed
2026-02-02 22:13:50 +08:00
// Collect text from each part for aggregation
for _, part := range parts {
if text, ok := part["text"].(string); ok && text != "" {
collectedTextParts = append(collectedTextParts, text)
}
}
}
}
}
}
}
if errors.Is(err, io.EOF) {
break
}
if err != nil {
return nil, nil, err
}
}
2026-02-02 22:13:50 +08:00
return mergeCollectedTextParts(pickGeminiCollectResult(last, lastWithParts), collectedTextParts), usage, nil
}
func pickGeminiCollectResult(last map[string]any, lastWithParts map[string]any) map[string]any {
if lastWithParts != nil {
return lastWithParts
}
if last != nil {
return last
}
return map[string]any{}
}
2026-02-02 22:13:50 +08:00
// mergeCollectedTextParts merges all collected text chunks into the final response.
// This fixes the issue where non-streaming responses only returned the last chunk
// instead of the complete aggregated text.
func mergeCollectedTextParts(response map[string]any, textParts []string) map[string]any {
if len(textParts) == 0 {
return response
}
// Join all text parts
mergedText := strings.Join(textParts, "")
// Deep copy response
result := make(map[string]any)
for k, v := range response {
result[k] = v
}
// Get or create candidates
candidates, ok := result["candidates"].([]any)
if !ok || len(candidates) == 0 {
candidates = []any{map[string]any{}}
}
// Get first candidate
candidate, ok := candidates[0].(map[string]any)
if !ok {
candidate = make(map[string]any)
candidates[0] = candidate
}
// Get or create content
content, ok := candidate["content"].(map[string]any)
if !ok {
content = map[string]any{"role": "model"}
candidate["content"] = content
}
// Get existing parts
existingParts, ok := content["parts"].([]any)
if !ok {
existingParts = []any{}
}
// Find and update first text part, or create new one
newParts := make([]any, 0, len(existingParts)+1)
textUpdated := false
for _, p := range existingParts {
pm, ok := p.(map[string]any)
if !ok {
newParts = append(newParts, p)
continue
}
if _, hasText := pm["text"]; hasText && !textUpdated {
// Replace with merged text
newPart := make(map[string]any)
for k, v := range pm {
newPart[k] = v
}
newPart["text"] = mergedText
newParts = append(newParts, newPart)
textUpdated = true
} else {
newParts = append(newParts, pm)
}
}
if !textUpdated {
newParts = append([]any{map[string]any{"text": mergedText}}, newParts...)
}
content["parts"] = newParts
result["candidates"] = candidates
return result
}
type geminiNativeStreamResult struct {
usage *ClaudeUsage
firstTokenMs *int
}
func isGeminiInsufficientScope(headers http.Header, body []byte) bool {
if strings.Contains(strings.ToLower(headers.Get("Www-Authenticate")), "insufficient_scope") {
return true
}
lower := strings.ToLower(string(body))
return strings.Contains(lower, "insufficient authentication scopes") || strings.Contains(lower, "access_token_scope_insufficient")
}
func estimateGeminiCountTokens(reqBody []byte) int {
total := 0
// systemInstruction.parts[].text
gjson.GetBytes(reqBody, "systemInstruction.parts").ForEach(func(_, part gjson.Result) bool {
if t := strings.TrimSpace(part.Get("text").String()); t != "" {
total += estimateTokensForText(t)
}
return true
})
// contents[].parts[].text
gjson.GetBytes(reqBody, "contents").ForEach(func(_, content gjson.Result) bool {
content.Get("parts").ForEach(func(_, part gjson.Result) bool {
if t := strings.TrimSpace(part.Get("text").String()); t != "" {
total += estimateTokensForText(t)
}
return true
})
return true
})
if total < 0 {
return 0
}
return total
}
func estimateTokensForText(s string) int {
s = strings.TrimSpace(s)
if s == "" {
return 0
}
runes := []rune(s)
if len(runes) == 0 {
return 0
}
ascii := 0
for _, r := range runes {
if r <= 0x7f {
ascii++
}
}
asciiRatio := float64(ascii) / float64(len(runes))
if asciiRatio >= 0.8 {
// Roughly 4 chars per token for English-like text.
return (len(runes) + 3) / 4
}
// For CJK-heavy text, approximate 1 rune per token.
return len(runes)
}
type UpstreamHTTPResult struct {
StatusCode int
Headers http.Header
Body []byte
}
func (s *GeminiMessagesCompatService) handleNativeNonStreamingResponse(c *gin.Context, resp *http.Response, isOAuth bool) (*ClaudeUsage, error) {
if s.cfg != nil && s.cfg.Gateway.GeminiDebugResponseHeaders {
logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] ========== Response Headers ==========")
for key, values := range resp.Header {
if strings.HasPrefix(strings.ToLower(key), "x-ratelimit") {
logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] %s: %v", key, values)
}
}
logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] ========================================")
}
maxBytes := resolveUpstreamResponseReadLimit(s.cfg)
respBody, err := readUpstreamResponseBodyLimited(resp.Body, maxBytes)
if err != nil {
if errors.Is(err, ErrUpstreamResponseBodyTooLarge) {
setOpsUpstreamError(c, http.StatusBadGateway, "upstream response too large", "")
c.JSON(http.StatusBadGateway, gin.H{
"error": gin.H{
"type": "upstream_error",
"message": "Upstream response too large",
},
})
}
return nil, err
}
if isOAuth {
unwrappedBody, uwErr := unwrapGeminiResponse(respBody)
if uwErr == nil {
respBody = unwrappedBody
}
}
responseheaders.WriteFilteredHeaders(c.Writer.Header(), resp.Header, s.responseHeaderFilter)
contentType := resp.Header.Get("Content-Type")
if contentType == "" {
contentType = "application/json"
}
c.Data(resp.StatusCode, contentType, respBody)
if u := extractGeminiUsage(respBody); u != nil {
return u, nil
}
return &ClaudeUsage{}, nil
}
func (s *GeminiMessagesCompatService) handleNativeStreamingResponse(c *gin.Context, resp *http.Response, startTime time.Time, isOAuth bool) (*geminiNativeStreamResult, error) {
if s.cfg != nil && s.cfg.Gateway.GeminiDebugResponseHeaders {
logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] ========== Streaming Response Headers ==========")
for key, values := range resp.Header {
if strings.HasPrefix(strings.ToLower(key), "x-ratelimit") {
logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] %s: %v", key, values)
}
}
logger.LegacyPrintf("service.gemini_messages_compat", "[GeminiAPI] ====================================================")
}
if s.responseHeaderFilter != nil {
responseheaders.WriteFilteredHeaders(c.Writer.Header(), resp.Header, s.responseHeaderFilter)
}
c.Status(resp.StatusCode)
c.Header("Cache-Control", "no-cache")
c.Header("Connection", "keep-alive")
c.Header("X-Accel-Buffering", "no")
contentType := resp.Header.Get("Content-Type")
if contentType == "" {
contentType = "text/event-stream; charset=utf-8"
}
c.Header("Content-Type", contentType)
flusher, ok := c.Writer.(http.Flusher)
if !ok {
return nil, errors.New("streaming not supported")
}
reader := bufio.NewReader(resp.Body)
usage := &ClaudeUsage{}
var firstTokenMs *int
for {
line, err := reader.ReadString('\n')
if len(line) > 0 {
trimmed := strings.TrimRight(line, "\r\n")
if strings.HasPrefix(trimmed, "data:") {
payload := strings.TrimSpace(strings.TrimPrefix(trimmed, "data:"))
// Keepalive / done markers
if payload == "" || payload == "[DONE]" {
_, _ = io.WriteString(c.Writer, line)
flusher.Flush()
} else {
var rawToWrite string
rawToWrite = payload
var rawBytes []byte
if isOAuth {
innerBytes, err := unwrapGeminiResponse([]byte(payload))
if err == nil {
rawToWrite = string(innerBytes)
rawBytes = innerBytes
}
} else {
rawBytes = []byte(payload)
}
if u := extractGeminiUsage(rawBytes); u != nil {
usage = u
}
if firstTokenMs == nil {
ms := int(time.Since(startTime).Milliseconds())
firstTokenMs = &ms
}
if isOAuth {
// SSE format requires double newline (\n\n) to separate events
_, _ = fmt.Fprintf(c.Writer, "data: %s\n\n", rawToWrite)
} else {
// Pass-through for AI Studio responses.
_, _ = io.WriteString(c.Writer, line)
}
flusher.Flush()
}
} else {
_, _ = io.WriteString(c.Writer, line)
flusher.Flush()
}
}
if errors.Is(err, io.EOF) {
break
}
if err != nil {
return nil, err
}
}
return &geminiNativeStreamResult{usage: usage, firstTokenMs: firstTokenMs}, nil
}
// ForwardAIStudioGET forwards a GET request to AI Studio (generativelanguage.googleapis.com) for
// endpoints like /v1beta/models and /v1beta/models/{model}.
//
// This is used to support Gemini SDKs that call models listing endpoints before generation.
func (s *GeminiMessagesCompatService) ForwardAIStudioGET(ctx context.Context, account *Account, path string) (*UpstreamHTTPResult, error) {
if account == nil {
return nil, errors.New("account is nil")
}
path = strings.TrimSpace(path)
if path == "" || !strings.HasPrefix(path, "/") {
return nil, errors.New("invalid path")
}
baseURL := account.GetGeminiBaseURL(geminicli.AIStudioBaseURL)
normalizedBaseURL, err := s.validateUpstreamBaseURL(baseURL)
if err != nil {
return nil, err
}
fullURL := strings.TrimRight(normalizedBaseURL, "/") + path
var proxyURL string
if account.ProxyID != nil && account.Proxy != nil {
proxyURL = account.Proxy.URL()
}
req, err := http.NewRequestWithContext(ctx, http.MethodGet, fullURL, nil)
if err != nil {
return nil, err
}
switch account.Type {
case AccountTypeAPIKey:
apiKey := strings.TrimSpace(account.GetCredential("api_key"))
if apiKey == "" {
return nil, errors.New("gemini api_key not configured")
}
req.Header.Set("x-goog-api-key", apiKey)
case AccountTypeOAuth:
if s.tokenProvider == nil {
return nil, errors.New("gemini token provider not configured")
}
accessToken, err := s.tokenProvider.GetAccessToken(ctx, account)
if err != nil {
return nil, err
}
req.Header.Set("Authorization", "Bearer "+accessToken)
default:
return nil, fmt.Errorf("unsupported account type: %s", account.Type)
}
resp, err := s.httpUpstream.Do(req, proxyURL, account.ID, account.Concurrency)
if err != nil {
return nil, err
}
defer func() { _ = resp.Body.Close() }()
body, _ := io.ReadAll(io.LimitReader(resp.Body, 8<<20))
wwwAuthenticate := resp.Header.Get("Www-Authenticate")
filteredHeaders := responseheaders.FilterHeaders(resp.Header, s.responseHeaderFilter)
if wwwAuthenticate != "" {
filteredHeaders.Set("Www-Authenticate", wwwAuthenticate)
}
return &UpstreamHTTPResult{
StatusCode: resp.StatusCode,
Headers: filteredHeaders,
Body: body,
}, nil
}
// unwrapGeminiResponse 解包 Gemini OAuth 响应中的 response 字段
// 使用 gjson 零拷贝提取,避免完整 Unmarshal+Marshal
func unwrapGeminiResponse(raw []byte) ([]byte, error) {
result := gjson.GetBytes(raw, "response")
if result.Exists() && result.Type == gjson.JSON {
return []byte(result.Raw), nil
}
return raw, nil
}
func convertGeminiToClaudeMessage(geminiResp map[string]any, originalModel string, rawData []byte) (map[string]any, *ClaudeUsage) {
usage := extractGeminiUsage(rawData)
if usage == nil {
usage = &ClaudeUsage{}
}
contentBlocks := make([]any, 0)
sawToolUse := false
if candidates, ok := geminiResp["candidates"].([]any); ok && len(candidates) > 0 {
if cand, ok := candidates[0].(map[string]any); ok {
if content, ok := cand["content"].(map[string]any); ok {
if parts, ok := content["parts"].([]any); ok {
for _, part := range parts {
pm, ok := part.(map[string]any)
if !ok {
continue
}
if text, ok := pm["text"].(string); ok && text != "" {
contentBlocks = append(contentBlocks, map[string]any{
"type": "text",
"text": text,
})
}
if fc, ok := pm["functionCall"].(map[string]any); ok {
name, _ := fc["name"].(string)
if strings.TrimSpace(name) == "" {
name = "tool"
}
args := fc["args"]
sawToolUse = true
contentBlocks = append(contentBlocks, map[string]any{
"type": "tool_use",
"id": "toolu_" + randomHex(8),
"name": name,
"input": args,
})
}
}
}
}
}
}
stopReason := mapGeminiFinishReasonToClaudeStopReason(extractGeminiFinishReason(geminiResp))
if sawToolUse {
stopReason = "tool_use"
}
resp := map[string]any{
"id": "msg_" + randomHex(12),
"type": "message",
"role": "assistant",
"model": originalModel,
"content": contentBlocks,
"stop_reason": stopReason,
"stop_sequence": nil,
"usage": map[string]any{
"input_tokens": usage.InputTokens,
"output_tokens": usage.OutputTokens,
},
}
return resp, usage
}
func extractGeminiUsage(data []byte) *ClaudeUsage {
usage := gjson.GetBytes(data, "usageMetadata")
if !usage.Exists() {
return nil
}
prompt := int(usage.Get("promptTokenCount").Int())
cand := int(usage.Get("candidatesTokenCount").Int())
cached := int(usage.Get("cachedContentTokenCount").Int())
2026-02-12 23:43:47 +08:00
thoughts := int(usage.Get("thoughtsTokenCount").Int())
2026-02-02 22:13:50 +08:00
// 注意Gemini 的 promptTokenCount 包含 cachedContentTokenCount
// 但 Claude 的 input_tokens 不包含 cache_read_input_tokens需要减去
return &ClaudeUsage{
2026-02-02 22:13:50 +08:00
InputTokens: prompt - cached,
OutputTokens: cand + thoughts,
2026-02-02 22:13:50 +08:00
CacheReadInputTokens: cached,
}
}
func asInt(v any) (int, bool) {
switch t := v.(type) {
case float64:
return int(t), true
case int:
return t, true
case int64:
return int(t), true
case json.Number:
i, err := t.Int64()
if err != nil {
return 0, false
}
return int(i), true
default:
return 0, false
}
}
func (s *GeminiMessagesCompatService) handleGeminiUpstreamError(ctx context.Context, account *Account, statusCode int, headers http.Header, body []byte) {
// 遵守自定义错误码策略:未命中则跳过所有限流处理
if !account.ShouldHandleErrorCode(statusCode) {
return
}
if s.rateLimitService != nil && (statusCode == 401 || statusCode == 403 || statusCode == 529) {
s.rateLimitService.HandleUpstreamError(ctx, account, statusCode, headers, body)
return
}
if statusCode != 429 {
return
}
oauthType := account.GeminiOAuthType()
tierID := account.GeminiTierID()
projectID := strings.TrimSpace(account.GetCredential("project_id"))
isCodeAssist := account.IsGeminiCodeAssist()
resetAt := ParseGeminiRateLimitResetTime(body)
if resetAt == nil {
// 根据账号类型使用不同的默认重置时间
var ra time.Time
if isCodeAssist {
// Code Assist: fallback cooldown by tier
cooldown := geminiCooldownForTier(tierID)
if s.rateLimitService != nil {
cooldown = s.rateLimitService.GeminiCooldown(ctx, account)
}
ra = time.Now().Add(cooldown)
logger.LegacyPrintf("service.gemini_messages_compat", "[Gemini 429] Account %d (Code Assist, tier=%s, project=%s) rate limited, cooldown=%v", account.ID, tierID, projectID, time.Until(ra).Truncate(time.Second))
} else {
// API Key / AI Studio OAuth: PST 午夜
if ts := nextGeminiDailyResetUnix(); ts != nil {
ra = time.Unix(*ts, 0)
logger.LegacyPrintf("service.gemini_messages_compat", "[Gemini 429] Account %d (API Key/AI Studio, type=%s) rate limited, reset at PST midnight (%v)", account.ID, account.Type, ra)
} else {
// 兜底5 分钟
ra = time.Now().Add(5 * time.Minute)
logger.LegacyPrintf("service.gemini_messages_compat", "[Gemini 429] Account %d rate limited, fallback to 5min", account.ID)
}
}
_ = s.accountRepo.SetRateLimited(ctx, account.ID, ra)
return
}
// 使用解析到的重置时间
resetTime := time.Unix(*resetAt, 0)
_ = s.accountRepo.SetRateLimited(ctx, account.ID, resetTime)
logger.LegacyPrintf("service.gemini_messages_compat", "[Gemini 429] Account %d rate limited until %v (oauth_type=%s, tier=%s)",
account.ID, resetTime, oauthType, tierID)
}
// ParseGeminiRateLimitResetTime 解析 Gemini 格式的 429 响应,返回重置时间的 Unix 时间戳
func ParseGeminiRateLimitResetTime(body []byte) *int64 {
// 第一阶段gjson 结构化提取
errMsg := gjson.GetBytes(body, "error.message").String()
if looksLikeGeminiDailyQuota(errMsg) {
if ts := nextGeminiDailyResetUnix(); ts != nil {
return ts
}
}
// 遍历 error.details 查找 quotaResetDelay
var found *int64
gjson.GetBytes(body, "error.details").ForEach(func(_, detail gjson.Result) bool {
v := detail.Get("metadata.quotaResetDelay").String()
if v == "" {
return true
}
if dur, err := time.ParseDuration(v); err == nil {
// Use ceil to avoid undercounting fractional seconds (e.g. 10.1s should not become 10s),
// which can affect scheduling decisions around thresholds (like 10s).
ts := time.Now().Unix() + int64(math.Ceil(dur.Seconds()))
found = &ts
return false
}
return true
})
if found != nil {
return found
}
// 第二阶段regex 回退匹配 "Please retry in Xs"
matches := retryInRegex.FindStringSubmatch(string(body))
if len(matches) == 2 {
if dur, err := time.ParseDuration(matches[1] + "s"); err == nil {
ts := time.Now().Unix() + int64(math.Ceil(dur.Seconds()))
return &ts
}
}
return nil
}
func looksLikeGeminiDailyQuota(message string) bool {
m := strings.ToLower(message)
if strings.Contains(m, "per day") || strings.Contains(m, "requests per day") || strings.Contains(m, "quota") && strings.Contains(m, "per day") {
return true
}
return false
}
func nextGeminiDailyResetUnix() *int64 {
reset := geminiDailyResetTime(time.Now())
ts := reset.Unix()
return &ts
}
func ensureGeminiFunctionCallThoughtSignatures(body []byte) []byte {
// Fast path: only run when functionCall is present.
if !bytes.Contains(body, []byte(`"functionCall"`)) {
return body
}
var payload map[string]any
if err := json.Unmarshal(body, &payload); err != nil {
return body
}
contentsAny, ok := payload["contents"].([]any)
if !ok || len(contentsAny) == 0 {
return body
}
modified := false
for _, c := range contentsAny {
cm, ok := c.(map[string]any)
if !ok {
continue
}
partsAny, ok := cm["parts"].([]any)
if !ok || len(partsAny) == 0 {
continue
}
for _, p := range partsAny {
pm, ok := p.(map[string]any)
if !ok || pm == nil {
continue
}
if fc, ok := pm["functionCall"].(map[string]any); !ok || fc == nil {
continue
}
ts, _ := pm["thoughtSignature"].(string)
if strings.TrimSpace(ts) == "" {
pm["thoughtSignature"] = geminiDummyThoughtSignature
modified = true
}
}
}
if !modified {
return body
}
b, err := json.Marshal(payload)
if err != nil {
return body
}
return b
}
func extractGeminiFinishReason(geminiResp map[string]any) string {
if candidates, ok := geminiResp["candidates"].([]any); ok && len(candidates) > 0 {
if cand, ok := candidates[0].(map[string]any); ok {
if fr, ok := cand["finishReason"].(string); ok {
return fr
}
}
}
return ""
}
func extractGeminiParts(geminiResp map[string]any) []map[string]any {
if candidates, ok := geminiResp["candidates"].([]any); ok && len(candidates) > 0 {
if cand, ok := candidates[0].(map[string]any); ok {
if content, ok := cand["content"].(map[string]any); ok {
if partsAny, ok := content["parts"].([]any); ok && len(partsAny) > 0 {
out := make([]map[string]any, 0, len(partsAny))
for _, p := range partsAny {
pm, ok := p.(map[string]any)
if !ok {
continue
}
out = append(out, pm)
}
return out
}
}
}
}
return nil
}
func computeGeminiTextDelta(seen, incoming string) (delta, newSeen string) {
incoming = strings.TrimSuffix(incoming, "\u0000")
if incoming == "" {
return "", seen
}
// Cumulative mode: incoming contains full text so far.
if strings.HasPrefix(incoming, seen) {
return strings.TrimPrefix(incoming, seen), incoming
}
// Duplicate/rewind: ignore.
if strings.HasPrefix(seen, incoming) {
return "", seen
}
// Delta mode: treat incoming as incremental chunk.
return incoming, seen + incoming
}
func mapGeminiFinishReasonToClaudeStopReason(finishReason string) string {
switch strings.ToUpper(strings.TrimSpace(finishReason)) {
case "MAX_TOKENS":
return "max_tokens"
case "STOP":
return "end_turn"
default:
return "end_turn"
}
}
func convertClaudeMessagesToGeminiGenerateContent(body []byte) ([]byte, error) {
var req map[string]any
if err := json.Unmarshal(body, &req); err != nil {
return nil, err
}
toolUseIDToName := make(map[string]string)
systemText := extractClaudeSystemText(req["system"])
contents, err := convertClaudeMessagesToGeminiContents(req["messages"], toolUseIDToName)
if err != nil {
return nil, err
}
out := make(map[string]any)
if systemText != "" {
out["systemInstruction"] = map[string]any{
"parts": []any{map[string]any{"text": systemText}},
}
}
out["contents"] = contents
if tools := convertClaudeToolsToGeminiTools(req["tools"]); tools != nil {
out["tools"] = tools
}
generationConfig := convertClaudeGenerationConfig(req)
if generationConfig != nil {
out["generationConfig"] = generationConfig
}
stripGeminiFunctionIDs(out)
return json.Marshal(out)
}
func stripGeminiFunctionIDs(req map[string]any) {
// Defensive cleanup: some upstreams reject unexpected `id` fields in functionCall/functionResponse.
contents, ok := req["contents"].([]any)
if !ok {
return
}
for _, c := range contents {
cm, ok := c.(map[string]any)
if !ok {
continue
}
contentParts, ok := cm["parts"].([]any)
if !ok {
continue
}
for _, p := range contentParts {
pm, ok := p.(map[string]any)
if !ok {
continue
}
if fc, ok := pm["functionCall"].(map[string]any); ok && fc != nil {
delete(fc, "id")
}
if fr, ok := pm["functionResponse"].(map[string]any); ok && fr != nil {
delete(fr, "id")
}
}
}
}
func extractClaudeSystemText(system any) string {
switch v := system.(type) {
case string:
return strings.TrimSpace(v)
case []any:
var parts []string
for _, p := range v {
pm, ok := p.(map[string]any)
if !ok {
continue
}
if t, _ := pm["type"].(string); t != "text" {
continue
}
if text, ok := pm["text"].(string); ok && strings.TrimSpace(text) != "" {
parts = append(parts, text)
}
}
return strings.TrimSpace(strings.Join(parts, "\n"))
default:
return ""
}
}
func convertClaudeMessagesToGeminiContents(messages any, toolUseIDToName map[string]string) ([]any, error) {
arr, ok := messages.([]any)
if !ok {
return nil, errors.New("messages must be an array")
}
out := make([]any, 0, len(arr))
for _, m := range arr {
mm, ok := m.(map[string]any)
if !ok {
continue
}
role, _ := mm["role"].(string)
role = strings.ToLower(strings.TrimSpace(role))
gRole := "user"
if role == "assistant" {
gRole = "model"
}
parts := make([]any, 0)
switch content := mm["content"].(type) {
case string:
// 字符串形式的 content保留所有内容包括空白
parts = append(parts, map[string]any{"text": content})
case []any:
// 如果只有一个 block不过滤空白让上游 API 报错)
singleBlock := len(content) == 1
for _, block := range content {
bm, ok := block.(map[string]any)
if !ok {
continue
}
bt, _ := bm["type"].(string)
switch bt {
case "text":
if text, ok := bm["text"].(string); ok {
// 单个 block 时保留所有内容(包括空白)
// 多个 blocks 时过滤掉空白
if singleBlock || strings.TrimSpace(text) != "" {
parts = append(parts, map[string]any{"text": text})
}
}
case "tool_use":
id, _ := bm["id"].(string)
name, _ := bm["name"].(string)
if strings.TrimSpace(id) != "" && strings.TrimSpace(name) != "" {
toolUseIDToName[id] = name
}
signature, _ := bm["signature"].(string)
signature = strings.TrimSpace(signature)
if signature == "" {
signature = geminiDummyThoughtSignature
}
parts = append(parts, map[string]any{
"thoughtSignature": signature,
"functionCall": map[string]any{
"name": name,
"args": bm["input"],
},
})
case "tool_result":
toolUseID, _ := bm["tool_use_id"].(string)
name := toolUseIDToName[toolUseID]
if name == "" {
name = "tool"
}
parts = append(parts, map[string]any{
"functionResponse": map[string]any{
"name": name,
"response": map[string]any{
"content": extractClaudeContentText(bm["content"]),
},
},
})
case "image":
if src, ok := bm["source"].(map[string]any); ok {
if srcType, _ := src["type"].(string); srcType == "base64" {
mediaType, _ := src["media_type"].(string)
data, _ := src["data"].(string)
if mediaType != "" && data != "" {
parts = append(parts, map[string]any{
"inlineData": map[string]any{
"mimeType": mediaType,
"data": data,
},
})
}
}
}
default:
// best-effort: preserve unknown blocks as text
if b, err := json.Marshal(bm); err == nil {
parts = append(parts, map[string]any{"text": string(b)})
}
}
}
default:
// ignore
}
out = append(out, map[string]any{
"role": gRole,
"parts": parts,
})
}
return out, nil
}
func extractClaudeContentText(v any) string {
switch t := v.(type) {
case string:
return t
case []any:
var sb strings.Builder
for _, part := range t {
pm, ok := part.(map[string]any)
if !ok {
continue
}
if pm["type"] == "text" {
if text, ok := pm["text"].(string); ok {
_, _ = sb.WriteString(text)
}
}
}
return sb.String()
default:
b, _ := json.Marshal(t)
return string(b)
}
}
func convertClaudeToolsToGeminiTools(tools any) []any {
arr, ok := tools.([]any)
if !ok || len(arr) == 0 {
return nil
}
funcDecls := make([]any, 0, len(arr))
for _, t := range arr {
tm, ok := t.(map[string]any)
if !ok {
continue
}
var name, desc string
var params any
// 检查是否为 custom 类型工具 (MCP)
toolType, _ := tm["type"].(string)
if toolType == "custom" {
// Custom 格式: 从 custom 字段获取 description 和 input_schema
custom, ok := tm["custom"].(map[string]any)
if !ok {
continue
}
name, _ = tm["name"].(string)
desc, _ = custom["description"].(string)
params = custom["input_schema"]
} else {
// 标准格式: 从顶层字段获取
name, _ = tm["name"].(string)
desc, _ = tm["description"].(string)
params = tm["input_schema"]
}
if name == "" {
continue
}
// 为 nil params 提供默认值
if params == nil {
params = map[string]any{
"type": "object",
"properties": map[string]any{},
}
}
fix: 修复 /v1/messages 间歇性 400 错误 (#18) * fix(upstream): 修复上游格式兼容性问题 - 跳过Claude模型无signature的thinking block - 支持custom类型工具(MCP)格式转换 - 添加ClaudeCustomToolSpec结构体支持MCP工具 - 添加Custom字段验证,跳过无效custom工具 - 在convertClaudeToolsToGeminiTools中添加schema清理 - 完整的单元测试覆盖,包含边界情况 修复: Issue 0.1 signature缺失, Issue 0.2 custom工具格式 改进: Codex审查发现的2个重要问题 测试: - TestBuildParts_ThinkingBlockWithoutSignature: 验证thinking block处理 - TestBuildTools_CustomTypeTools: 验证custom工具转换和边界情况 - TestConvertClaudeToolsToGeminiTools_CustomType: 验证service层转换 * feat(gemini): 添加Gemini限额与TierID支持 实现PR1:Gemini限额与TierID功能 后端修改: - GeminiTokenInfo结构体添加TierID字段 - fetchProjectID函数返回(projectID, tierID, error) - 从LoadCodeAssist响应中提取tierID(优先IsDefault,回退到第一个非空tier) - ExchangeCode、RefreshAccountToken、GetAccessToken函数更新以处理tierID - BuildAccountCredentials函数保存tier_id到credentials 前端修改: - AccountStatusIndicator组件添加tier显示 - 支持LEGACY/PRO/ULTRA等tier类型的友好显示 - 使用蓝色badge展示tier信息 技术细节: - tierID提取逻辑:优先选择IsDefault的tier,否则选择第一个非空tier - 所有fetchProjectID调用点已更新以处理新的返回签名 - 前端gracefully处理missing/unknown tier_id * refactor(gemini): 优化TierID实现并添加安全验证 根据并发代码审查(code-reviewer, security-auditor, gemini, codex)的反馈进行改进: 安全改进: - 添加validateTierID函数验证tier_id格式和长度(最大64字符) - 限制tier_id字符集为字母数字、下划线、连字符和斜杠 - 在BuildAccountCredentials中验证tier_id后再存储 - 静默跳过无效tier_id,不阻塞账户创建 代码质量改进: - 提取extractTierIDFromAllowedTiers辅助函数消除重复代码 - 重构fetchProjectID函数,tierID提取逻辑只执行一次 - 改进代码可读性和可维护性 审查工具: - code-reviewer agent (a09848e) - security-auditor agent (a9a149c) - gemini CLI (bcc7c81) - codex (b5d8919) 修复问题: - HIGH: 未验证的tier_id输入 - MEDIUM: 代码重复(tierID提取逻辑重复2次) * fix(format): 修复 gofmt 格式问题 - 修复 claude_types.go 中的字段对齐问题 - 修复 gemini_messages_compat_service.go 中的缩进问题 * fix(upstream): 修复上游格式兼容性问题 (#14) * fix(upstream): 修复上游格式兼容性问题 - 跳过Claude模型无signature的thinking block - 支持custom类型工具(MCP)格式转换 - 添加ClaudeCustomToolSpec结构体支持MCP工具 - 添加Custom字段验证,跳过无效custom工具 - 在convertClaudeToolsToGeminiTools中添加schema清理 - 完整的单元测试覆盖,包含边界情况 修复: Issue 0.1 signature缺失, Issue 0.2 custom工具格式 改进: Codex审查发现的2个重要问题 测试: - TestBuildParts_ThinkingBlockWithoutSignature: 验证thinking block处理 - TestBuildTools_CustomTypeTools: 验证custom工具转换和边界情况 - TestConvertClaudeToolsToGeminiTools_CustomType: 验证service层转换 * fix(format): 修复 gofmt 格式问题 - 修复 claude_types.go 中的字段对齐问题 - 修复 gemini_messages_compat_service.go 中的缩进问题 * fix(format): 修复 claude_types.go 的 gofmt 格式问题 * feat(antigravity): 优化 thinking block 和 schema 处理 - 为 dummy thinking block 添加 ThoughtSignature - 重构 thinking block 处理逻辑,在每个条件分支内创建 part - 优化 excludedSchemaKeys,移除 Gemini 实际支持的字段 (minItems, maxItems, minimum, maximum, additionalProperties, format) - 添加详细注释说明 Gemini API 支持的 schema 字段 * fix(antigravity): 增强 schema 清理的安全性 基于 Codex review 建议: - 添加 format 字段白名单过滤,只保留 Gemini 支持的 date-time/date/time - 补充更多不支持的 schema 关键字到黑名单: * 组合 schema: oneOf, anyOf, allOf, not, if/then/else * 对象验证: minProperties, maxProperties, patternProperties 等 * 定义引用: $defs, definitions - 避免不支持的 schema 字段导致 Gemini API 校验失败 * fix(lint): 修复 gemini_messages_compat_service 空分支警告 - 在 cleanToolSchema 的 if 语句中添加 continue - 移除重复的注释 * fix(antigravity): 移除 minItems/maxItems 以兼容 Claude API - 将 minItems 和 maxItems 添加到 schema 黑名单 - Claude API (Vertex AI) 不支持这些数组验证字段 - 添加调试日志记录工具 schema 转换过程 - 修复 tools.14.custom.input_schema 验证错误 * fix(antigravity): 修复 additionalProperties schema 对象问题 - 将 additionalProperties 的 schema 对象转换为布尔值 true - Claude API 只支持 additionalProperties: false,不支持 schema 对象 - 修复 tools.14.custom.input_schema 验证错误 - 参考 Claude 官方文档的 JSON Schema 限制 * fix(antigravity): 修复 Claude 模型 thinking 块兼容性问题 - 完全跳过 Claude 模型的 thinking 块以避免 signature 验证失败 - 只在 Gemini 模型中使用 dummy thought signature - 修改 additionalProperties 默认值为 false(更安全) - 添加调试日志以便排查问题 * fix(upstream): 修复跨模型切换时的 dummy signature 问题 基于 Codex review 和用户场景分析的修复: 1. 问题场景 - Gemini (thinking) → Claude (thinking) 切换时 - Gemini 返回的 thinking 块使用 dummy signature - Claude API 会拒绝 dummy signature,导致 400 错误 2. 修复内容 - request_transformer.go:262: 跳过 dummy signature - 只保留真实的 Claude signature - 支持频繁的跨模型切换 3. 其他修复(基于 Codex review) - gateway_service.go:691: 修复 io.ReadAll 错误处理 - gateway_service.go:687: 条件日志(尊重 LogUpstreamErrorBody 配置) - gateway_service.go:915: 收紧 400 failover 启发式 - request_transformer.go:188: 移除签名成功日志 4. 新增功能(默认关闭) - 阶段 1: 上游错误日志(GATEWAY_LOG_UPSTREAM_ERROR_BODY) - 阶段 2: Antigravity thinking 修复 - 阶段 3: API-key beta 注入(GATEWAY_INJECT_BETA_FOR_APIKEY) - 阶段 3: 智能 400 failover(GATEWAY_FAILOVER_ON_400) 测试:所有测试通过 * fix(lint): 修复 golangci-lint 问题 - 应用 De Morgan 定律简化条件判断 - 修复 gofmt 格式问题 - 移除未使用的 min 函数
2026-01-01 04:21:18 +08:00
// 清理 JSON Schema
cleanedParams := cleanToolSchema(params)
funcDecls = append(funcDecls, map[string]any{
"name": name,
"description": desc,
fix: 修复 /v1/messages 间歇性 400 错误 (#18) * fix(upstream): 修复上游格式兼容性问题 - 跳过Claude模型无signature的thinking block - 支持custom类型工具(MCP)格式转换 - 添加ClaudeCustomToolSpec结构体支持MCP工具 - 添加Custom字段验证,跳过无效custom工具 - 在convertClaudeToolsToGeminiTools中添加schema清理 - 完整的单元测试覆盖,包含边界情况 修复: Issue 0.1 signature缺失, Issue 0.2 custom工具格式 改进: Codex审查发现的2个重要问题 测试: - TestBuildParts_ThinkingBlockWithoutSignature: 验证thinking block处理 - TestBuildTools_CustomTypeTools: 验证custom工具转换和边界情况 - TestConvertClaudeToolsToGeminiTools_CustomType: 验证service层转换 * feat(gemini): 添加Gemini限额与TierID支持 实现PR1:Gemini限额与TierID功能 后端修改: - GeminiTokenInfo结构体添加TierID字段 - fetchProjectID函数返回(projectID, tierID, error) - 从LoadCodeAssist响应中提取tierID(优先IsDefault,回退到第一个非空tier) - ExchangeCode、RefreshAccountToken、GetAccessToken函数更新以处理tierID - BuildAccountCredentials函数保存tier_id到credentials 前端修改: - AccountStatusIndicator组件添加tier显示 - 支持LEGACY/PRO/ULTRA等tier类型的友好显示 - 使用蓝色badge展示tier信息 技术细节: - tierID提取逻辑:优先选择IsDefault的tier,否则选择第一个非空tier - 所有fetchProjectID调用点已更新以处理新的返回签名 - 前端gracefully处理missing/unknown tier_id * refactor(gemini): 优化TierID实现并添加安全验证 根据并发代码审查(code-reviewer, security-auditor, gemini, codex)的反馈进行改进: 安全改进: - 添加validateTierID函数验证tier_id格式和长度(最大64字符) - 限制tier_id字符集为字母数字、下划线、连字符和斜杠 - 在BuildAccountCredentials中验证tier_id后再存储 - 静默跳过无效tier_id,不阻塞账户创建 代码质量改进: - 提取extractTierIDFromAllowedTiers辅助函数消除重复代码 - 重构fetchProjectID函数,tierID提取逻辑只执行一次 - 改进代码可读性和可维护性 审查工具: - code-reviewer agent (a09848e) - security-auditor agent (a9a149c) - gemini CLI (bcc7c81) - codex (b5d8919) 修复问题: - HIGH: 未验证的tier_id输入 - MEDIUM: 代码重复(tierID提取逻辑重复2次) * fix(format): 修复 gofmt 格式问题 - 修复 claude_types.go 中的字段对齐问题 - 修复 gemini_messages_compat_service.go 中的缩进问题 * fix(upstream): 修复上游格式兼容性问题 (#14) * fix(upstream): 修复上游格式兼容性问题 - 跳过Claude模型无signature的thinking block - 支持custom类型工具(MCP)格式转换 - 添加ClaudeCustomToolSpec结构体支持MCP工具 - 添加Custom字段验证,跳过无效custom工具 - 在convertClaudeToolsToGeminiTools中添加schema清理 - 完整的单元测试覆盖,包含边界情况 修复: Issue 0.1 signature缺失, Issue 0.2 custom工具格式 改进: Codex审查发现的2个重要问题 测试: - TestBuildParts_ThinkingBlockWithoutSignature: 验证thinking block处理 - TestBuildTools_CustomTypeTools: 验证custom工具转换和边界情况 - TestConvertClaudeToolsToGeminiTools_CustomType: 验证service层转换 * fix(format): 修复 gofmt 格式问题 - 修复 claude_types.go 中的字段对齐问题 - 修复 gemini_messages_compat_service.go 中的缩进问题 * fix(format): 修复 claude_types.go 的 gofmt 格式问题 * feat(antigravity): 优化 thinking block 和 schema 处理 - 为 dummy thinking block 添加 ThoughtSignature - 重构 thinking block 处理逻辑,在每个条件分支内创建 part - 优化 excludedSchemaKeys,移除 Gemini 实际支持的字段 (minItems, maxItems, minimum, maximum, additionalProperties, format) - 添加详细注释说明 Gemini API 支持的 schema 字段 * fix(antigravity): 增强 schema 清理的安全性 基于 Codex review 建议: - 添加 format 字段白名单过滤,只保留 Gemini 支持的 date-time/date/time - 补充更多不支持的 schema 关键字到黑名单: * 组合 schema: oneOf, anyOf, allOf, not, if/then/else * 对象验证: minProperties, maxProperties, patternProperties 等 * 定义引用: $defs, definitions - 避免不支持的 schema 字段导致 Gemini API 校验失败 * fix(lint): 修复 gemini_messages_compat_service 空分支警告 - 在 cleanToolSchema 的 if 语句中添加 continue - 移除重复的注释 * fix(antigravity): 移除 minItems/maxItems 以兼容 Claude API - 将 minItems 和 maxItems 添加到 schema 黑名单 - Claude API (Vertex AI) 不支持这些数组验证字段 - 添加调试日志记录工具 schema 转换过程 - 修复 tools.14.custom.input_schema 验证错误 * fix(antigravity): 修复 additionalProperties schema 对象问题 - 将 additionalProperties 的 schema 对象转换为布尔值 true - Claude API 只支持 additionalProperties: false,不支持 schema 对象 - 修复 tools.14.custom.input_schema 验证错误 - 参考 Claude 官方文档的 JSON Schema 限制 * fix(antigravity): 修复 Claude 模型 thinking 块兼容性问题 - 完全跳过 Claude 模型的 thinking 块以避免 signature 验证失败 - 只在 Gemini 模型中使用 dummy thought signature - 修改 additionalProperties 默认值为 false(更安全) - 添加调试日志以便排查问题 * fix(upstream): 修复跨模型切换时的 dummy signature 问题 基于 Codex review 和用户场景分析的修复: 1. 问题场景 - Gemini (thinking) → Claude (thinking) 切换时 - Gemini 返回的 thinking 块使用 dummy signature - Claude API 会拒绝 dummy signature,导致 400 错误 2. 修复内容 - request_transformer.go:262: 跳过 dummy signature - 只保留真实的 Claude signature - 支持频繁的跨模型切换 3. 其他修复(基于 Codex review) - gateway_service.go:691: 修复 io.ReadAll 错误处理 - gateway_service.go:687: 条件日志(尊重 LogUpstreamErrorBody 配置) - gateway_service.go:915: 收紧 400 failover 启发式 - request_transformer.go:188: 移除签名成功日志 4. 新增功能(默认关闭) - 阶段 1: 上游错误日志(GATEWAY_LOG_UPSTREAM_ERROR_BODY) - 阶段 2: Antigravity thinking 修复 - 阶段 3: API-key beta 注入(GATEWAY_INJECT_BETA_FOR_APIKEY) - 阶段 3: 智能 400 failover(GATEWAY_FAILOVER_ON_400) 测试:所有测试通过 * fix(lint): 修复 golangci-lint 问题 - 应用 De Morgan 定律简化条件判断 - 修复 gofmt 格式问题 - 移除未使用的 min 函数
2026-01-01 04:21:18 +08:00
"parameters": cleanedParams,
})
}
if len(funcDecls) == 0 {
return nil
}
return []any{
map[string]any{
"functionDeclarations": funcDecls,
},
}
}
fix: 修复 /v1/messages 间歇性 400 错误 (#18) * fix(upstream): 修复上游格式兼容性问题 - 跳过Claude模型无signature的thinking block - 支持custom类型工具(MCP)格式转换 - 添加ClaudeCustomToolSpec结构体支持MCP工具 - 添加Custom字段验证,跳过无效custom工具 - 在convertClaudeToolsToGeminiTools中添加schema清理 - 完整的单元测试覆盖,包含边界情况 修复: Issue 0.1 signature缺失, Issue 0.2 custom工具格式 改进: Codex审查发现的2个重要问题 测试: - TestBuildParts_ThinkingBlockWithoutSignature: 验证thinking block处理 - TestBuildTools_CustomTypeTools: 验证custom工具转换和边界情况 - TestConvertClaudeToolsToGeminiTools_CustomType: 验证service层转换 * feat(gemini): 添加Gemini限额与TierID支持 实现PR1:Gemini限额与TierID功能 后端修改: - GeminiTokenInfo结构体添加TierID字段 - fetchProjectID函数返回(projectID, tierID, error) - 从LoadCodeAssist响应中提取tierID(优先IsDefault,回退到第一个非空tier) - ExchangeCode、RefreshAccountToken、GetAccessToken函数更新以处理tierID - BuildAccountCredentials函数保存tier_id到credentials 前端修改: - AccountStatusIndicator组件添加tier显示 - 支持LEGACY/PRO/ULTRA等tier类型的友好显示 - 使用蓝色badge展示tier信息 技术细节: - tierID提取逻辑:优先选择IsDefault的tier,否则选择第一个非空tier - 所有fetchProjectID调用点已更新以处理新的返回签名 - 前端gracefully处理missing/unknown tier_id * refactor(gemini): 优化TierID实现并添加安全验证 根据并发代码审查(code-reviewer, security-auditor, gemini, codex)的反馈进行改进: 安全改进: - 添加validateTierID函数验证tier_id格式和长度(最大64字符) - 限制tier_id字符集为字母数字、下划线、连字符和斜杠 - 在BuildAccountCredentials中验证tier_id后再存储 - 静默跳过无效tier_id,不阻塞账户创建 代码质量改进: - 提取extractTierIDFromAllowedTiers辅助函数消除重复代码 - 重构fetchProjectID函数,tierID提取逻辑只执行一次 - 改进代码可读性和可维护性 审查工具: - code-reviewer agent (a09848e) - security-auditor agent (a9a149c) - gemini CLI (bcc7c81) - codex (b5d8919) 修复问题: - HIGH: 未验证的tier_id输入 - MEDIUM: 代码重复(tierID提取逻辑重复2次) * fix(format): 修复 gofmt 格式问题 - 修复 claude_types.go 中的字段对齐问题 - 修复 gemini_messages_compat_service.go 中的缩进问题 * fix(upstream): 修复上游格式兼容性问题 (#14) * fix(upstream): 修复上游格式兼容性问题 - 跳过Claude模型无signature的thinking block - 支持custom类型工具(MCP)格式转换 - 添加ClaudeCustomToolSpec结构体支持MCP工具 - 添加Custom字段验证,跳过无效custom工具 - 在convertClaudeToolsToGeminiTools中添加schema清理 - 完整的单元测试覆盖,包含边界情况 修复: Issue 0.1 signature缺失, Issue 0.2 custom工具格式 改进: Codex审查发现的2个重要问题 测试: - TestBuildParts_ThinkingBlockWithoutSignature: 验证thinking block处理 - TestBuildTools_CustomTypeTools: 验证custom工具转换和边界情况 - TestConvertClaudeToolsToGeminiTools_CustomType: 验证service层转换 * fix(format): 修复 gofmt 格式问题 - 修复 claude_types.go 中的字段对齐问题 - 修复 gemini_messages_compat_service.go 中的缩进问题 * fix(format): 修复 claude_types.go 的 gofmt 格式问题 * feat(antigravity): 优化 thinking block 和 schema 处理 - 为 dummy thinking block 添加 ThoughtSignature - 重构 thinking block 处理逻辑,在每个条件分支内创建 part - 优化 excludedSchemaKeys,移除 Gemini 实际支持的字段 (minItems, maxItems, minimum, maximum, additionalProperties, format) - 添加详细注释说明 Gemini API 支持的 schema 字段 * fix(antigravity): 增强 schema 清理的安全性 基于 Codex review 建议: - 添加 format 字段白名单过滤,只保留 Gemini 支持的 date-time/date/time - 补充更多不支持的 schema 关键字到黑名单: * 组合 schema: oneOf, anyOf, allOf, not, if/then/else * 对象验证: minProperties, maxProperties, patternProperties 等 * 定义引用: $defs, definitions - 避免不支持的 schema 字段导致 Gemini API 校验失败 * fix(lint): 修复 gemini_messages_compat_service 空分支警告 - 在 cleanToolSchema 的 if 语句中添加 continue - 移除重复的注释 * fix(antigravity): 移除 minItems/maxItems 以兼容 Claude API - 将 minItems 和 maxItems 添加到 schema 黑名单 - Claude API (Vertex AI) 不支持这些数组验证字段 - 添加调试日志记录工具 schema 转换过程 - 修复 tools.14.custom.input_schema 验证错误 * fix(antigravity): 修复 additionalProperties schema 对象问题 - 将 additionalProperties 的 schema 对象转换为布尔值 true - Claude API 只支持 additionalProperties: false,不支持 schema 对象 - 修复 tools.14.custom.input_schema 验证错误 - 参考 Claude 官方文档的 JSON Schema 限制 * fix(antigravity): 修复 Claude 模型 thinking 块兼容性问题 - 完全跳过 Claude 模型的 thinking 块以避免 signature 验证失败 - 只在 Gemini 模型中使用 dummy thought signature - 修改 additionalProperties 默认值为 false(更安全) - 添加调试日志以便排查问题 * fix(upstream): 修复跨模型切换时的 dummy signature 问题 基于 Codex review 和用户场景分析的修复: 1. 问题场景 - Gemini (thinking) → Claude (thinking) 切换时 - Gemini 返回的 thinking 块使用 dummy signature - Claude API 会拒绝 dummy signature,导致 400 错误 2. 修复内容 - request_transformer.go:262: 跳过 dummy signature - 只保留真实的 Claude signature - 支持频繁的跨模型切换 3. 其他修复(基于 Codex review) - gateway_service.go:691: 修复 io.ReadAll 错误处理 - gateway_service.go:687: 条件日志(尊重 LogUpstreamErrorBody 配置) - gateway_service.go:915: 收紧 400 failover 启发式 - request_transformer.go:188: 移除签名成功日志 4. 新增功能(默认关闭) - 阶段 1: 上游错误日志(GATEWAY_LOG_UPSTREAM_ERROR_BODY) - 阶段 2: Antigravity thinking 修复 - 阶段 3: API-key beta 注入(GATEWAY_INJECT_BETA_FOR_APIKEY) - 阶段 3: 智能 400 failover(GATEWAY_FAILOVER_ON_400) 测试:所有测试通过 * fix(lint): 修复 golangci-lint 问题 - 应用 De Morgan 定律简化条件判断 - 修复 gofmt 格式问题 - 移除未使用的 min 函数
2026-01-01 04:21:18 +08:00
// cleanToolSchema 清理工具的 JSON Schema移除 Gemini 不支持的字段
func cleanToolSchema(schema any) any {
if schema == nil {
return nil
}
switch v := schema.(type) {
case map[string]any:
cleaned := make(map[string]any)
for key, value := range v {
// 跳过不支持的字段
if key == "$schema" || key == "$id" || key == "$ref" ||
key == "additionalProperties" || key == "minLength" ||
key == "maxLength" || key == "minItems" || key == "maxItems" {
continue
}
// 递归清理嵌套对象
cleaned[key] = cleanToolSchema(value)
}
// 规范化 type 字段为大写
if typeVal, ok := cleaned["type"].(string); ok {
cleaned["type"] = strings.ToUpper(typeVal)
}
return cleaned
case []any:
cleaned := make([]any, len(v))
for i, item := range v {
cleaned[i] = cleanToolSchema(item)
}
return cleaned
default:
return v
}
}
func convertClaudeGenerationConfig(req map[string]any) map[string]any {
out := make(map[string]any)
if mt, ok := asInt(req["max_tokens"]); ok && mt > 0 {
out["maxOutputTokens"] = mt
}
if temp, ok := req["temperature"].(float64); ok {
out["temperature"] = temp
}
if topP, ok := req["top_p"].(float64); ok {
out["topP"] = topP
}
if stopSeq, ok := req["stop_sequences"].([]any); ok && len(stopSeq) > 0 {
out["stopSequences"] = stopSeq
}
if len(out) == 0 {
return nil
}
return out
}
// extractImageSize 从 Gemini 请求中提取 image_size 参数
func (s *GeminiMessagesCompatService) extractImageSize(body []byte) string {
var req struct {
GenerationConfig *struct {
ImageConfig *struct {
ImageSize string `json:"imageSize"`
} `json:"imageConfig"`
} `json:"generationConfig"`
}
if err := json.Unmarshal(body, &req); err != nil {
return "2K"
}
if req.GenerationConfig != nil && req.GenerationConfig.ImageConfig != nil {
size := strings.ToUpper(strings.TrimSpace(req.GenerationConfig.ImageConfig.ImageSize))
if size == "1K" || size == "2K" || size == "4K" {
return size
}
}
return "2K"
}