Commit Graph
395 Commits
Author SHA1 Message Date
Wesley Liddick 427d591212 Merge pull request #2930 from touwaeriol/feat/image-token-billing
fix(billing): channel pricing override for image generation + display image_output_tokens
2026-06-06 09:28:46 +08:00
shaw 650981f2e4 fix(lint): 修复 golangci-lint errcheck 与 gofmt 报错
- gateway_service.go buildStableSessionSeed: strings.Builder.WriteString
  返回值用 _, _ = 显式忽略(errcheck)
- openai_codex_transform_additions_test.go: client_metadata 类型断言改为
  双值形式 cm3, _ := ...(errcheck check-type-assertions)
- openai_gateway_service.go: codexCLIUserAgent 常量单独成组,收敛 = 前多余
  空格对齐(gofmt)
2026-06-06 09:21:32 +08:00
eyre 219da4b9e5 feat(claude-mimicry): align Claude Code fingerprint with CLI 2.1.161
- bump impersonated CLI version 2.1.92 -> 2.1.161; derive User-Agent from
  CLICurrentVersion so the two hardcoded copies can no longer drift apart
- fix x-stainless headers to real 2.1.161 values: package-version
  0.70.0 -> 0.94.0, runtime-version v24.13.0 -> v24.3.0 (verified against
  the installed Bun-compiled binary)
- expand the disguise-path system prompt from a 2-block identity skeleton
  to a 3-block layout (billing + identity + tool-agnostic prose), matching
  real CC's multi-block shape; cache breakpoint moved to the last static
  block. Deliberately excludes # Doing tasks / # Using your tools /
  # Executing actions to avoid polluting proxied-client behavior
- stabilize the synthesized metadata.user_id session_id across conversation
  turns: derive it from (account + client discriminator + first user
  message) instead of a per-turn content/body hash. Sticky-routing
  GenerateSessionHash is intentionally left untouched; remove now-dead
  hashBodyForSessionSeed

Tests: update the 3-block system assertions in gateway_prompt_test and
gateway_anthropic_apikey_passthrough_test; add a session_id cross-turn
stability test in gateway_oauth_metadata_test.
2026-06-06 01:01:58 +08:00
erio 1e2e8b1d6b fix(billing): channel pricing fully overrides image output token price
When channel pricing is configured, image output token price now
correctly overrides all other sources (LiteLLM, group config).
Add ImageOutputPriceExplicit flag to distinguish intentional zero
price from missing price, preventing fallback to output price.

Affects Gemini, OpenAI inline image gen, and Images API paths.
2026-05-31 22:57:56 +08:00
name 34de99ee0e refactor(gateway): cap upstream error body reads 2026-05-30 20:01:40 +08:00
name 2caee9d884 refactor(gateway): snapshot usage worker inputs 2026-05-30 20:01:39 +08:00
name 619e5ae619 refactor(gateway): isolate anthropic body rewrites
Keep Anthropic request body rewrites attempt-local and synchronize the accepted wire body only after upstream success so failover and retry paths do not reuse stale parsed state.
2026-05-30 20:01:39 +08:00
name b1c4be4ac8 refactor(gateway): remove parsed request object graphs
Keep large gateway payloads as raw body ranges and bind OpenAI parsed-body caches to the body bytes so failover and mapping do not reuse stale mutable state.
2026-05-30 20:01:38 +08:00
name d8cbf9ab5c refactor(gateway): introduce request body refs 2026-05-30 20:01:38 +08:00
DaydreamCodingandClaude Opus 4.8 f7f5e33830 feat(quota): user×platform 配额 DB 写聚合 flusher
Redis 同步权威 + DB 镜像,不在进程内维护 delta:
- 写入点 HasUserPlatformQuotaLimit 守卫:无 limit 跳过 Redis 写与持久化
- 累加 usage 的 Lua 在 flusher_enabled 时 SADD 脏集 billing:upq:dirty
- UserPlatformQuotaUsageFlusher 定时 SPOP 脏集 → 批量 HGETALL 读当前窗口 usage 快照
  → BatchSnapshotUsage 绝对值 UPSERT 覆盖 DB(去 SELECT FOR UPDATE 行锁)
  → 失败 SADD 回 / FK(23503)整批丢弃
- flusher 单批 clamp 到 ≤6000,保证一次 flush 只生成一条 UPSERT(单事务原子)
- flusher_enabled 默认 false(降级=旧异步直写 DB)

效果:DB 写连接从 O(QPS) 收敛到 O(副本)。
循环依赖:service 层独立 Snapshot/FK 类型,repository adapter 转换 + %w 映射 FK error。
admin reset/upsert 后失效 cache(脏残留被 flusher 当 MISS 跳过)。

健壮性与可观测性:
- flusher_enabled=false 时 Start 不注册定时器;flush_interval_ms 非法回退 2s
- Readd 回填失败单独计 dirty_lost(不再误记 dirty_readd)并 ALERT;脏集 Readd 补兜底 TTL
- 单 tick 达 max batches 上限仍有积压时记 log
- admin 失效 cache 失败升级为 ALERT(提示 enforcement 可能延迟至 sentinel TTL)
- BatchGet 单条命令失败 / usage 字段损坏均记 log,避免静默以 0 覆写 DB

三态 go vet + 单测/集成测全绿。
已知取舍(默认 flusher_enabled=false 不触发):
- FK 整批丢弃牵连同批正常 key(活跃 key 靠下次 SADD+绝对值快照自愈;Redis 仍权威)
- admin reset/upsert 直写 DB 与 flusher 异步刷存在覆盖竞态:flusher 持旧快照在途时可能覆盖
  admin 刚写值(limit 列不受影响;usage 有 preflight windowExpired 兜底;低频)。彻底消除需 version OCC。

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-05-29 17:19:15 +08:00
DaydreamCodingandClaude Opus 4.8 06fca66273 feat(quota): sentinel 回填消除无配额行用户 preflight 每请求回源 DB
无 user×platform 配额行的用户,preflight 每次 cache MISS 后回源 DB 查得"无行"
却不缓存该结论,导致每请求一次 DB 往返。本 PR 回填 sentinel 占位 entry,使后续
请求命中 Redis 后稳定判"无 limit",TTL 内不再查 DB。

- config: 加 UserPlatformQuotaSentinelTTLSeconds(默认 3600s,短于普通 quota
  cache 的 86400s 以控 Redis 内存)
- metrics: 加 userPlatformQuotaSentinelSetCacheErrorTotal,并入
  GatewayUserPlatformQuotaIncrStats 暴露
- billing_cache: checkUserPlatformQuotaEligibility 在 cache MISS + DB 无行且
  cacheErr==nil 时回填 sentinel(三 limit nil、三 window_start non-nil、SchemaV1);
  TTL<=0 fallback 1h 防 EXPIRE 立即删 key 击穿;SET 失败 fail-open + 计 metric
- billing_cache: HIT 路径对 sentinel(三 limit nil)跳过 windowExpired refresh,
  避免短 sentinel TTL 被误升级为 86400s

有配额 limit 的用户 enforcement 行为不变(rec!=nil 不回填、isSentinel=false 不跳过 refresh)。

测试:扩展 fakeFullCache 夹具(setCalls/lastSetTTL/getErr/setErr);新增回填正确性 /
Redis-GET-故障不回填 / SET-失败 fail-open / sentinel 跨窗口不 refresh 四个单测。
go build、quota+billing unit、三态 go vet 全绿。

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-05-29 17:19:15 +08:00
Wesley Liddick 52292741cb Merge pull request #2849 from Pluviobyte/fix/count-tokens-payload-filter
fix(gateway): filter count_tokens generation fields
2026-05-29 16:11:01 +08:00
PluviobyteandCursor 27600b1d2c fix(gateway): filter count_tokens generation fields
Anthropic count_tokens rejects generation-only fields such as temperature, top_p, top_k, stream, and stop sequences. Passing the original messages payload through unchanged can turn otherwise valid requests into upstream 400 errors.

Sanitize only the count_tokens upstream payload after the gateway's existing request normalization, preserving fields that existing compatibility paths rely on while removing parameters the count_tokens endpoint does not accept.

Fixes #2764

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-05-28 05:40:50 +00:00
alfadb ddf91e9a7f fix(gateway): 按最终 anthropic-beta header 对 body.context_management 做能力维度 sanitize
上游 Anthropic 在 body 含 `context_management` 但最终发出去的
`anthropic-beta` header 不含 `context-management-2025-06-27` 时会拒收:

    {
      "type": "invalid_request_error",
      "message": "context_management: Extra inputs are not permitted"
    }
    (HTTP 400, request_id 形如 req_011C...)

该 400 在 haiku 路径上触发,因为三个 beta header 构造器有意排除了
context-management beta:

  - HaikuBetaHeader          (messages,     OAuth / mimic CC)
  - APIKeyHaikuBetaHeader    (messages,     API-key)
  - CountTokensBetaHeader    (count_tokens, 所有认证类型)

但 body 中仍然带着 `context_management` 字段,原因有二:

  1. normalizeClaudeOAuthRequestBody 在 thinking_enabled / thinking_adaptive
     打开时为 `clear_thinking_20251015` 主动注入;
  2. 客户端 (Claude Code CLI >= 2.1.87) 原样发送, 网关透传时一并转发。

修复方案: 能力维度对称约束
==========================

对齐已有的 Bedrock 模式
(`backend/internal/service/bedrock_request.go` 中的
`sanitizeBedrockFieldsForBetaTokens`):
根据 **最终** 发出的 `anthropic-beta` header 决定是否保留
`body.context_management`, 而不是按 model 名或路由分类来决定。

新增纯函数:

    sanitizeAnthropicBodyForBetaTokens(body, betaHeader) (body, changed)

如果 `betaHeader` 不含 `context-management-2025-06-27`, 用 sjson 把 body
字段 strip 掉; 否则原样返回。

在所有 Anthropic / Anthropic-兼容 上游出口都接入:

| 路径                                       | sanitize 接入点                                       |
|--------------------------------------------|-------------------------------------------------------|
| /v1/messages OAuth mimic CC                | buildUpstreamRequest                                  |
| /v1/messages OAuth 真 CC 透传              | buildUpstreamRequest                                  |
| /v1/messages API-key                       | buildUpstreamRequest                                  |
| /v1/messages API-key passthrough           | buildUpstreamRequestAnthropicAPIKeyPassthrough        |
| /v1/messages Vertex / service-account      | buildUpstreamRequestAnthropicVertex                   |
| /v1/messages/count_tokens (全部 4 条路径)  | buildCountTokensRequest,                              |
|                                            | buildCountTokensRequestAnthropicAPIKeyPassthrough     |
| Antigravity Anthropic-兼容 上游            | AntigravityGatewayService.ForwardUpstream             |
| Bedrock                                    | (已由 sanitizeBedrockFieldsForBetaTokens 处理)        |

为什么要重排 (而不是加一行调用)
================================

sanitize 必须 **在** `signBillingHeaderCCH` 之前运行。CCH 对整个 body
取 xxHash64 摘要后写入 billing header 里 5 位十六进制的 `cch` 字段;
如果先签名再 strip, 上游对发出去的 body 重算 hash 会和 `cch` 不一致,
请求被判为 third-party。这就要求在 `http.NewRequest` 之前算出最终的
`anthropic-beta` header, 所以把原本内联在 builder 里的 beta 计算逻辑
抽成了两个纯函数:

  - computeFinalAnthropicBeta              (messages 路径: mimic 不透传
                                            客户端 beta)
  - computeFinalCountTokensAnthropicBeta   (count_tokens 路径: mimic 不
                                            跳过白名单透传)

两者逐位保留原行为:

  - mimic 路径在 messages 上跳过客户端 beta, 在 count_tokens 上合并
  - API-key 路径尊重 `InjectBetaForAPIKey` 开关
  - dropSet (`defaultDroppedBetasSet` + BetaPolicy filter) 应用在主路径,
    passthrough / Vertex 路径有意不应用 —— 这条原有的不对称行为本 PR
    不动。

一条语义测试 (`TestSanitizeMustBeBeforeCCHSigning_HashConsistency`) 把
顺序约束文档化并强制守住: 它证明 `sanitize -> signBillingHeaderCCH`
产生的 `cch` 与最终 body 一致, 而 `signBillingHeaderCCH -> sanitize`
产生的 `cch` 会被上游 hash 重算判失败。

为什么是能力维度 (而不是 haiku 模型名匹配)
==========================================

最朴素的"按 model 名 strip"方案
(`strings.Contains(modelID, "haiku") -> DeleteBytes "context_management"`)
有四个真实失败模式:

  1. 过度删除。CLI >= 2.1.87 的真 Claude Code 客户端在 haiku 上同时
     发送 body 字段 **和** `anthropic-beta: context-management-2025-06-27`。
     一律 strip 会让该用户的 `clear_thinking_20251015` 静默失效。
  2. 别名漂移。未来的 haiku 别名 (`claude-3-haiku-...`,
     `claude-haiku-...` 等) 改变匹配面; 任何新别名都会悄悄绕过 strip。
  3. count_tokens 漏覆盖。count_tokens 有自己的 builder 和不同的 beta
     header 集合; 在一个地方做 model 名检查会漏掉这条路径。
  4. API-key passthrough 早退。passthrough builder 在 model 名 strip
     之前就 return 了, strip 根本不执行。

能力维度沿着 header 端到端走, 上述 4 个 case 都由构造方式保证正确,
不依赖任何 modelID 匹配。

防御项
======

  - 当 `sjson.DeleteBytes` 在 `gjson` 刚验证过字段存在的 body 上失败时,
    `sanitizeAnthropicBodyForBetaTokens` 会记 warning 日志 —— 这种情况
    现实中仅在请求中途被破坏时发生, 日志把此前会静默发生的 body / header
    不一致暴露出来。
  - `header_util.go` 新增 `deleteHeaderAllForms`: 在白名单透传已经写入
    canonical 大小写的 `Anthropic-Beta` 之后再覆盖, 否则会同时留下两条。

测试
====

`backend/internal/service` 下新增 44 个测试:

  - 纯函数: anthropicBetaTokensContains x 5, sanitize keep/strip x 6,
    computeFinal{Anthropic,CountTokens}AnthropicBeta x 12
  - normalize 回归 x 5
  - buildUpstreamRequest 端到端 x 4
      (OAuth mimic haiku strip / mimic sonnet preserve /
       真 CC haiku 带客户端 beta preserve / API-key haiku strip)
  - buildCountTokensRequest 端到端 x 2
  - buildUpstreamRequestAnthropicAPIKeyPassthrough x 2 (strip / preserve)
  - buildCountTokensRequestAnthropicAPIKeyPassthrough x 2 (strip / preserve)
  - buildUpstreamRequestAnthropicVertex x 2 (strip / preserve, 含 outgoing
    `anthropic-beta` header 对称断言)
  - CCH 顺序语义测试 x 1

unit 套件全过 (本机 88s), `golangci-lint` 0 issues。

已知局限 (本 PR 范围外)
========================

  - Vertex 路径用透传过来的客户端 `anthropic-beta` header 作为 sanitize
    依据, 而不是 Vertex 侧的能力矩阵。最坏情况是过度 strip (= 当前 main
    的行为, 主路径本来什么都不 strip); 不是 regression。完整的 Vertex
    能力模型属于单独的 PR。
  - Vertex builder 仍然不应用 BetaPolicy filter / dropSet。这是该 builder
    早 return 的既有架构决策, 本 PR 不动。
  - count_tokens mimic 在 haiku 上仍然注入 `context-management-2025-06-27`
    (因为原 count_tokens mimic 逻辑并不像 messages mimic 那样排除它)。
    本 PR 逐位保留 main 的行为; 是否要让它与 messages mimic 的排除策略
    统一是另一个问题。
  - `sanitizeAnthropicBodyForBetaTokens` 目前只处理
    `context_management <-> context-management-2025-06-27` 这一对。如果
    Anthropic 后续推出更多 beta-gated body 字段, 可以在后续 PR 重构为
    `{body 路径 -> required beta token}` 注册表的形式。
2026-05-28 00:02:50 +08:00
Wesley Liddick bbe847ed3e Merge pull request #2805 from StarryKira/feat/configurable-pool-retry-status-codes
feat(account): configurable pool-mode same-account retry status codes
2026-05-27 22:09:55 +08:00
StarryKiraandClaude Opus 4.7 21033dceb9 feat(account): configurable pool-mode same-account retry status codes
Pool mode currently retries the same account for a fixed set of
upstream HTTP statuses: 401, 403, 429. Some upstream pool deployments
also need same-account retry for transient provider/proxy statuses
such as 502, 503, 520, 529, but hard-coding more statuses changes
behavior for everyone.

Add a per-account credentials option `pool_mode_retry_status_codes`
that lets admins choose which upstream HTTP status codes trigger
same-account retry in pool mode:

- Unset (default): preserve the current 401/403/429 default
- Explicit list: override the defaults with the configured codes
- Codes normalized to the 100-599 range, deduplicated, sorted

The standalone `isPoolModeRetryableStatus` helper is kept as the
default-only fallback. All 15 gateway call sites switch to the new
`Account.IsPoolModeRetryableStatus` method so behavior is preserved
for accounts that do not configure the new field.

Frontend admin UI gains a "Retry Status Codes" comma-separated input
under the pool-mode section in both Create/Edit account modals
(en + zh i18n).

Fixes #2731

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-26 11:24:25 -07:00
wucm667 a31b507484 fix(scheduler): 模型404仅冷却账号模型组合 2026-05-26 20:29:48 +08:00
DaydreamCodingandClaude Opus 4.7 6b39b344d8 feat(quota): 用户 × 平台 USD 配额
为用户在 anthropic/openai/gemini/antigravity 四个平台上提供日/周/月
三个窗口的 USD 配额管控。配额语义:未设置=不限制,0=禁用,>0=美元上限。

两层模型:
- 配置层:系统默认配额,以及 email/linuxdo/oidc/wechat/github/google/
  dingtalk 七个鉴权来源的默认配额,存于 settings,以嵌套 JSON 整体读写
  (系统 1 个 key + 每个来源 1 个 key),整体替换语义。
- 运行时层:user_platform_quota 表按用户记录实际配额,与配置层解耦。

后端:新增 ent schema 与 140_user_platform_quotas.sql 迁移、repository
与 service 端口、计费链路集成、管理端与用户端读写接口。
前端:管理端设置页配额编辑、用户配额管理 Modal、用户 Dashboard 展示、
中英文案。

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-26 10:49:20 +08:00
erio fe1c6c958b feat(bedrock): add Claude Code compatibility for AWS Bedrock
- Export ApplyBedrockCCCompat() in GatewayService, called after channel
  model mapping to ensure mapped model ID is used for Opus 4.7+ detection
- Add sanitizeBedrockCCFields(): remove service_tier/interface_geo/
  context_management, inject max_tokens/anthropic_version defaults
- Add sanitizeBedrockCCBetaTokens(): filter anthropic_beta to keep only
  Bedrock-supported tokens, reusing autoInjectBedrockBetaTokens and
  filterBedrockBetaTokens for consistent rules
- Remove unsupported beta tokens (interleaved-thinking, context-management)
  from whitelist based on AWS official docs
- Simplify IsBedrockCCCompatEnabled() to check boolean toggle directly,
  applying CC compat to all accounts regardless of platform
- Add unit tests for IsBedrockCCCompatEnabled (8 cases),
  sanitizeBedrockCCFields (8 cases), sanitizeBedrockCCBetaTokens (7 cases)
- Update bedrock beta policy tests for removed auto-injection
2026-05-21 11:46:24 +08:00
erio 4fd21994c5 feat(bedrock): add Claude Code compatibility transformations for Bedrock accounts
Add channel-level Bedrock CC compatibility toggle (similar to web_search_emulation)
that fixes 4 types of Bedrock 400 errors seen with Claude Code:

1. thinking.type "enabled" → "adaptive" for Opus 4.7+ (only supports adaptive)
2. Add default budget_tokens when missing for older models
3. Replace illegal characters in tool_use IDs to match Bedrock's ^[a-zA-Z0-9_-]+$ pattern
4. anthropic_version / invalid beta flag (already handled elsewhere)

Transformations run in Forward() before any forwarding path, so both native Bedrock
accounts and apikey passthrough accounts pointing to Bedrock relays benefit.

Includes channel-level toggle UI and unit tests.
2026-05-20 21:47:38 +08:00
name 8211aa7066 fix: retry on "thinking block must contain thinking" upstream error
Some clients reuse assistant history from other models when switching to
claude with extended thinking enabled. If a prior thinking block lacks the
thinking text field, upstream returns:
  messages.X.content.Y.thinking: each thinking block must contain thinking

Add this pattern to isThinkingBlockSignatureError so the existing
FilterThinkingBlocksForRetry retry path triggers and rewrites/drops the
offending blocks.
2026-05-20 18:46:50 +08:00
name 2eb622f2f6 Remove ops retry replay storage 2026-05-19 19:37:41 +08:00
DaydreamCodingandClaude Opus 4.7 b19da9c7fe feat(dingtalk): 钉钉 OAuth 登录接入与 internal_only 用户属性同步
⚠️ 应用类型约束:当前实现仅支持「钉钉登录-企业内部应用」(DingTalk 开放平台
internal_app 类型)。第三方个人应用、第三方企业应用类型暂不支持——OAuth 流程
相同但 corp 校验、跨企业行为不同。backend 通过 DingTalkAppKind 校验对非
internal_app 类型 fail-closed(硬约束)。

钉钉 OAuth 登录主链
- 4 步 OAuth 链:ExchangeCodeForUserToken / GetUnionIdByUserToken /
  GetUserIdByUnionId / GetStaffInfoByUserId;app token 缓存
- pending session 机制持久化 OAuth 中间态;cookie-only token 持久化
- 三种分流:bind_login_required / email_completion / choose_account_action
- corp_restriction_policy 支持 none + internal_only;stale "whitelist" 在
  加载层与写入层均静默 coerce 为 none + slog.Warn
- bypass_registration 开关:企业内部模式豁免全局 REGISTRATION_DISABLED
- isReservedEmail / signup_source / canUnbindProvider / OAuth pending flow
  等横切点支持 dingtalk provider
- migration 136:4 表 CHECK 约束加入 'dingtalk' provider 值

internal_only 模式同步企业邮箱/姓名/部门到用户属性
- SyncCorpEmail / SyncDisplayName / SyncDept 三个独立开关 + 对应
  SyncXxxAttrKey 目标属性 key(默认 dingtalk_email / dingtalk_name /
  dingtalk_department);非 internal_only policy 在写入层与加载层均
  coerce 为 false,admin handler 与 setting_service 双层兜底
- 同步语义:首次注册写 users.username(昵称优先 → 企业姓名 fallback),
  之后每次登录刷新 3 个属性;空值也写入以覆盖旧值
- 邮箱三级 fallback:org_email > email > extension["企业邮箱"]
  (钉钉自定义字段 JSON)
- 部门路径递归向上拼接,跳过 dept_id=1 选首个真实子部门,剥离根组织名
- GetUnionIdByUserToken 同时返回 OIDC /contact/users/me 的 nick 字段;
  新增 GetDeptInfo 调用 OAPI /topapi/v2/department/get
- AuthHandler 注入 UserAttributeService;OAuth pending flow 在
  createPendingOAuthAccount / bindPendingOAuthLogin 分别派发到
  AfterRegistration(syncUsername=true)/ AfterLogin
- migration 137 seed dingtalk_email/name/department 三个用户属性定义

附带修复(同集成路径暴露的两个 OAuth 注册回归)
- LoginOrRegisterOAuthWithTokenPair 新建用户分支用 inferLegacySignupSource
  覆写 caller 显式传入的 signupSource,导致 dingtalk/linuxdo/oidc/wechat
  渠道授权按 email 渠道读取;改为只在 caller 未显式传入时回退邮箱推断
- mergeProviderDefaultGrantSettings 把 parse fallback 默认值
  (Concurrency=5 / Balance=0) 当作"未配置"哨兵,admin 显式设 5 时被误判
  退回全局默认(复现:全局默认 1 + 渠道默认并发 5 + grant_on_signup → 新
  用户实际 concurrency=1);去掉哨兵,admin 任何 >=0 值都覆盖 globalDefaults

前端
- DingTalk Login / Callback / EmailCompletion / ChoiceAccount / Error
  视图;router + auth API client
- admin SettingsView:corp policy radio(none / internal_only)+ bypass
  注册开关 + i18n;internal_only 下展示三同步开关 + 目标 attr key 下拉
  (拉取 user attribute definitions),展示 fieldEmail /
  qyapi_get_department_list 钉钉权限申请提示
- Profile:S1 主动绑定 / S5 解绑钉钉按钮 + 合成邮箱防自锁

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-19 15:27:47 +08:00
Wesley Liddick a340002c6d Merge pull request #2401 from 2ue/fix/normalize-image-billing-size
修复图片计费尺寸归一化与使用记录展示
2026-05-19 14:00:24 +08:00
lyen1688 164e2f610c fix: add keepalive for Anthropic passthrough streams 2026-05-18 18:41:25 +08:00
2ue bb4c1abe28 Fix image billing size normalization 2026-05-12 15:21:31 +08:00
shaw 9377c96746 fix: 让消息 cache_control 改写默认关闭 2026-05-11 21:26:41 +08:00
shaw 501b7f2772 fix: stabilize anthropic passthrough timeout error 2026-05-07 10:24:29 +08:00
2ue 6faa344916 feat: add OpenAI image generation controls 2026-05-05 03:26:54 +08:00
shaw 72d5ee4cd1 fix: drain OpenAI compat streams for usage 2026-05-03 17:11:27 +08:00
shaw 73b872998e feat: 添加 Anthropic 缓存 TTL 注入开关 2026-04-30 13:38:22 +08:00
shaw 733627cf9d fix: improve sticky session scheduling 2026-04-30 11:38:11 +08:00
Wesley Liddick 4d676dddd1 Merge pull request #2066 from alfadb/fix/anthropic-stream-eof-failover
fix(gateway): Anthropic 流式 EOF 失败移交 + SSE error 帧标准化
2026-04-29 17:09:47 +08:00
alfadb d78478e866 fix(gateway): sanitize stream errors to avoid leaking infrastructure topology
(*net.OpError).Error() concatenates Source/Addr fields, so the previous
disconnectMsg surfaced internal source IP/port and upstream server address
to clients via SSE error frames and UpstreamFailoverError.ResponseBody
(reported by @Wei-Shaw on PR #2066).

- Add sanitizeStreamError that maps known errors (io.ErrUnexpectedEOF,
  context.Canceled, syscall.ECONNRESET/EPIPE/ETIMEDOUT/...) to fixed
  descriptions and falls back to a generic placeholder, with an explicit
  *net.OpError branch that drops Source/Addr fields entirely.
- Use sanitized message in client-facing disconnectMsg; full ev.err is
  still preserved in the existing operator log line for diagnosis.
- Tests cover net.OpError redaction, the failover ResponseBody path, and
  every known sanitized error mapping.
2026-04-29 15:44:54 +08:00
alfadbandClaude Opus 4.7 4c474616b9 fix(gateway): emit Anthropic-standard SSE error events and failover body
Two follow-ups to PR #2066's failover-wrap fix:

1. Failover ResponseBody (`UpstreamFailoverError.ResponseBody`) was encoded
   as `{"error": "<msg>"}` (string field). `ExtractUpstreamErrorMessage`
   probes for `error.message`, `detail`, or top-level `message` only — so
   `handleFailoverExhausted` and downstream passthrough rules saw an empty
   message, losing the EOF root cause in ops logs. Re-encode as the
   Anthropic standard shape `{"type":"error","error":{"type":"upstream_disconnected","message":"..."}}`.
   (Addresses the inline review comment from copilot-pull-request-reviewer
   on Wei-Shaw/sub2api#2066.)

2. The streaming `event: error` SSE frame for `response_too_large`,
   `stream_read_error`, and `stream_timeout` was non-standard
   (`{"error":"<reason>"}`). Anthropic SDKs (and Claude Code) expect
   `{"type":"error","error":{"type":"...","message":"..."}}` and parse
   `error.type`/`error.message` accordingly. Refactor `sendErrorEvent` to
   take both reason and message, and emit the standard frame so client
   SDKs surface a real diagnostic message instead of a generic stream error.

This does not by itself prevent task interruption on long-stream EOF
(SSE has no resume; client-side retry remains the only complete fix), but
it gives both server-side ops logs and client-side error UIs a meaningful
upstream message so users know the next step is to retry.

Tests updated to assert the new body shape on both branches plus a new
assertion that `ExtractUpstreamErrorMessage` returns a non-empty string.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 20:24:17 +08:00
alfadbandClaude Opus 4.7 6327573534 fix(gateway): wrap Anthropic stream EOF as failover error before client output
Anthropic streaming path (gateway_service.go) returned a plain error on
upstream SSE read failure, so the handler-level UpstreamFailoverError check
never fired and the client received a bare `stream_read_error` event,
breaking long-running tasks even when no bytes had been written yet.

The most common trigger is HTTP/2 GOAWAY from api.anthropic.com edge
backends doing graceful rotation: Go's http.Transport surfaces this as
`unexpected EOF` and never auto-retries.

Mirror what the OpenAI and antigravity gateways already do: when the read
error happens before any byte has reached the client (`!c.Writer.Written()`),
return `*UpstreamFailoverError{StatusCode: 502, RetryableOnSameAccount: true}`
so the handler can retry on the same or another account. After client
output has begun, SSE has no resume protocol — keep the existing passthrough
behavior.

Tests cover both branches via streamReadCloser-based fixtures.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-28 19:12:48 +08:00
Oliver 6d11f9ed77 Add Vertex service account support 2026-04-25 20:39:58 -04:00
shaw 496469ac4e fix(gateway): skip body mimicry for real Claude Code clients to restore prompt caching
PR #1914 unconditionally applied the full mimicry pipeline to all OAuth
accounts, including real Claude Code CLI clients. This replaced the
client's long system prompt (~10K+ tokens with stable cache_control
breakpoints) with a short ~45 token [billing, CC prompt] pair, which
falls below Anthropic's 1024-token minimum cacheable prefix threshold.
The result: every request created a new cache but never hit an existing
one.

Fix: restore the Claude Code client detection gate so that real CC
clients bypass body-level mimicry (system rewrite, message cache
management, tool name obfuscation). Non-CC third-party clients
(opencode, etc.) continue to receive full mimicry.

Also harden the detection logic:
- Make UA regex case-insensitive (align with claude_code_validator.go)
- Validate metadata.user_id format via ParseMetadataUserID() instead of
  just checking non-empty, preventing third-party tools from spoofing
  a claude-cli/* UA with an arbitrary user_id string to bypass mimicry
2026-04-25 22:50:35 +08:00
hungryboy1025 8987e0ba67 fix(openai): tighten responses stream account tests 2026-04-25 16:56:50 +08:00
shaw 732d6495ea chore(gateway): fix lint issues from cc-mimicry-parity merge
- staticcheck QF1001: apply De Morgan's law to the OAuth-mimic header
  passthrough guard (`!(a && b)` → `a != ... || !b`).
- unused: drop `isClaudeCodeRequest`, which became dead after PR #1914
  switched both `/v1/messages` and `/count_tokens` paths to unconditional
  `account.IsOAuth()` mimicry. The lowercase helper `isClaudeCodeClient`
  is kept (still referenced by `TestIsClaudeCodeClient`).
2026-04-25 08:58:57 +08:00
keh4l bdbd2916f5 fix(gateway): skip client header passthrough on OAuth mimicry path
Root cause of persistent third-party detection: sub2api's
buildUpstreamRequest transparently forwards client headers via
allowedHeaders whitelist (addHeaderRaw) before applying mimicry
overrides. When third-party clients (opencode, etc.) send their own
anthropic-beta / user-agent / x-stainless-* / x-claude-code-session-id
values, these get appended to the request alongside our injected
headers, creating an inconsistent header set that Anthropic detects.

Parrot's build_upstream_headers constructs exactly 9 headers from
scratch and never forwards anything from the client. This is why
'same opencode version, some users work some don't' — different
opencode configs/versions send different header combinations.

Fix: when tokenType=oauth and mimicClaudeCode=true, skip the
client header passthrough loop entirely. The subsequent
applyClaudeCodeMimicHeaders + ApplyFingerprint + beta merge
pipeline constructs all necessary headers from our controlled values.

Also: remove systemIncludesClaudeCodePrompt gate — OAuth accounts
now unconditionally rewrite system (even if client already sent a
Claude Code-style prompt), ensuring billing attribution block is
always present.
2026-04-25 00:43:38 +08:00
keh4l 6dc89765fd fix(gateway): always apply full mimicry for OAuth accounts regardless of client identity
Before: isClaudeCodeRequest() checked whether the client looks like a
real Claude Code CLI (UA, system prompt, X-App header, metadata format).
If it looked like Claude Code, all mimicry was skipped — the assumption
being that a real CLI needs no help.

Problem: third-party tools like opencode partially impersonate Claude
Code (sending claude-cli UA + claude-code beta + CC system prompt) but
miss critical details (billing attribution block, tool-name obfuscation,
cache breakpoints, full beta set). Some users' opencode instances pass
the isClaudeCodeRequest check, causing sub2api to skip mimicry entirely,
while Anthropic still detects the request as third-party.

This explains why 'same opencode version, some users work, some don't'
— it depends on which opencode features/config trigger the validator.

Fix: OAuth accounts now unconditionally run the full mimicry pipeline,
matching Parrot's behavior (Parrot never checks client identity).
This is safe because our mimicry is strictly more complete than any
third-party client's partial impersonation.

Changed:
  - /v1/messages path: remove isClaudeCode gate
  - /v1/messages/count_tokens path: same
2026-04-25 00:26:37 +08:00
keh4l f3233db01f fix(gateway): apply D/E/F mimicry to native /v1/messages and count_tokens paths
The previous commit only wired stripMessageCacheControl,
addMessageCacheBreakpoints, and tool-name obfuscation into
applyClaudeCodeOAuthMimicryToBody (used by /chat/completions and
/responses). The native /v1/messages path and count_tokens path
have their own independent mimicry code blocks and were missed.

Now all three entry points share the same D/E/F pipeline:
  - /v1/messages (gateway_service.go forwardAnthropic)
  - /v1/messages/count_tokens (gateway_service.go countTokens)
  - OpenAI compat (applyClaudeCodeOAuthMimicryToBody)
2026-04-24 23:16:32 +08:00
keh4l 6e12578bc5 feat(gateway): port Parrot tool-name obfuscation + message cache breakpoints
Implements the remaining three parity items with Parrot cc_mimicry:

  D) Tool-name obfuscation
     - Dynamic mapping when tools.length > 5 (matches Parrot threshold).
       Fake names follow {prefix}{name[:3]}{i:02d} (e.g. 'manage_bas00').
       Go port of random.Random(hash(tuple(names))) uses fnv64a seed +
       math/rand; byte-exact reproduction is impossible (Python hash vs
       Go hash), but the two invariants that matter are preserved:
         * same input tool_names yield identical mapping (cache hit)
         * prefix pool is shuffled (names look distributed)
     - Static prefix map (sessions_ -> cc_sess_, session_ -> cc_ses_)
       applied as fallback, matching Parrot TOOL_NAME_REWRITES verbatim.
     - Server tools (web_search_20250305, computer_*, etc.) are NOT
       renamed; only type=='function' and type=='custom' tools are.
     - tool_choice.name is rewritten in sync (only when type=='tool').
     - Response side: bytes-level replace on every SSE chunk / JSON
       body at 6 injection points (standard stream/non-stream,
       passthrough stream/non-stream, chat_completions stream +
       non-stream, responses stream + non-stream). Reverse mapping
       applied longest-fake-name-first to prevent substring conflicts
       (parity with Parrot _restore_tool_names_in_chunk).
     - tool_choice is no longer unconditionally deleted in
       normalizeClaudeOAuthRequestBody — Parrot passes it through.

  E) tools[-1] cache_control breakpoint
     - Injected as {type:ephemeral, ttl:<DefaultCacheControlTTL>} when
       the last tool has no cache_control. Client-provided ttl is
       passed through unchanged (repo-wide policy).

  F) messages cache_control strategy
     - stripMessageCacheControl removes every client-provided
       messages[*].content[*].cache_control (multi-turn stability).
     - addMessageCacheBreakpoints then injects two stable breakpoints:
       (1) last message, and (2) second-to-last user turn when
       messages.length >= 4.
     - Combined with the system block breakpoint and tools[-1]
       breakpoint, this gives exactly the 4 breakpoints Anthropic
       allows per request.

Non-trivial implementation details to be aware of when rebasing:

  * Two new files, no upstream collision:
      gateway_tool_rewrite.go       (D + E algorithms)
      gateway_messages_cache.go     (F strip + breakpoints)
  * Two new feature calls bolted onto the tail of
    applyClaudeCodeOAuthMimicryToBody in gateway_service.go — rebase
    conflicts will be ~10 lines maximum.
  * Response-side injection points all wrap their existing write with
    reverseToolNamesIfPresent(c, ...), preserving original behavior
    when no mapping is stored (static prefix rollback still runs).
  * Non-stream chat/responses switched from c.JSON to
    json.Marshal + c.Data so bytes-level replace is possible.
  * Retry bodies (FilterThinkingBlocksForRetry,
    FilterSignatureSensitiveBlocksForRetry, RectifyThinkingBudget)
    only prune blocks — they preserve the already-obfuscated tool
    names, so no extra mapping re-application is needed.

Manual QA: end-to-end scenario verified with 6 tools (above threshold)
and tool_choice.type=='tool'. Obfuscation + restore roundtrip shown
in test logs; then removed the temp test file.

Tests (16 new):
  - buildDynamicToolMap stability + below-threshold guard
  - sanitizeToolName precedence (dynamic > static)
  - restoreToolNamesInBytes longest-first + static rollback
  - applyToolNameRewriteToBody skips server tools + syncs tool_choice
  - applyToolsLastCacheBreakpoint defaults to 5m + passes client ttl
  - stripMessageCacheControl + addMessageCacheBreakpoints in the
    1/4/string-content cases + second-to-last user turn selection
  - buildToolNameRewriteFromBody ReverseOrdered is desc-by-fake-length
  - fake name shape follows Parrot {prefix}{head3}{i:02d}
2026-04-24 23:16:32 +08:00
keh4l a25faecadd feat(gateway): align body shape with real Claude Code CLI defaults
Three field-level alignments in normalizeClaudeOAuthRequestBody to
match real Claude Code CLI traffic byte-for-byte:

  1. temperature: previously deleted unconditionally; now passes
     through client value, defaults to 1 when absent (real CLI
     always sends temperature, default 1).

  2. max_tokens: defaults to 128000 when absent (real CLI default).

  3. context_management: when thinking.type is enabled/adaptive
     and the client did not provide context_management, inject
     {"edits":[{"type":"clear_thinking_20251015","keep":"all"}]}
     to mirror real CLI behavior.

tool_choice removal is unchanged (Claude Code OAuth credentials
do not allow client-supplied tool_choice).

Tests updated:
  - gateway_body_order_test.go: temperature/max_tokens are now
    expected in output; tool_choice still removed.
  - gateway_prompt_test.go: system array is now 2 blocks
    (billing + cc prompt), assertions adjusted.
  - gateway_anthropic_apikey_passthrough_test.go: same 2-block
    assertion.
2026-04-24 23:16:32 +08:00
keh4l 5862e2d8d9 feat(gateway): add billing attribution block with cc_version fingerprint
Real Claude Code CLI always sends a 2-block system array:

  [0] {"type":"text", "text":"x-anthropic-billing-header: cc_version=X.Y.Z.{fp}; cc_entrypoint=cli; cch=00000;"}
  [1] {"type":"text", "text":"You are Claude Code...", "cache_control":{...}}

Before this commit, sub2api's mimicry path only produced block [1].
The missing billing block is one of the primary third-party detection
signals Anthropic uses for Claude-Code-scoped OAuth tokens.

New file gateway_billing_block.go ports the fingerprint algorithm
(byte-for-byte from Parrot cc_mimicry.py:compute_fingerprint):
pick chars at positions [4,7,20] of the first user text, then
`sha256(SALT + chars + cc_version)[:3]`.

  - claude/constants.go: CLICurrentVersion = "2.1.92" (must match UA)
  - gateway_billing_block.go: computeClaudeCodeFingerprint +
    buildBillingAttributionBlockJSON + extractFirstUserText
  - gateway_service.go: rewriteSystemForNonClaudeCode now emits both
    blocks in order; cch=00000 is filled in later by
    signBillingHeaderCCH in buildUpstreamRequest.

Downstream compat note: syncBillingHeaderVersion's regex
`cc_version=\d+\.\d+\.\d+` only matches the semver triple,
leaving the `.{fp}` suffix intact when rewriting in buildUpstreamRequest.
2026-04-24 23:16:32 +08:00
keh4l 66d6454535 feat(claude): add ttl to cache_control with default 5m
Real Claude CLI traffic sends cache_control as
`{"type":"ephemeral","ttl":"1h"}`. Our previous payload only
sent `{"type":"ephemeral"}`, which is a bytewise mismatch with
the official CLI and one more third-party detection signal.

Policy: client-provided ttl is always passed through unchanged.
Proxy-generated cache_control blocks default to 5m (vs Parrot's 1h)
to avoid burning the 1h cache budget on automatic breakpoints while
still aligning with the `ttl` field being present.

  - claude/constants.go: DefaultCacheControlTTL = "5m"
  - apicompat/types.go: new AnthropicCacheControl type with TTL field;
    AnthropicTool gains optional CacheControl pointer so the mimicry
    path can attach a cache breakpoint to tools[-1] later.
  - service/gateway_service.go: anthropicCacheControlPayload gains TTL;
    marshalAnthropicSystemTextBlock and rewriteSystemForNonClaudeCode
    emit ttl=5m by default.
2026-04-24 23:16:32 +08:00
keh4l 165553cfb0 fix(gateway): use full beta list in buildUpstreamRequest mimicry path
The previous commit added FullClaudeCodeMimicryBetas() but the two
call sites in buildUpstreamRequest still hardcoded the old 3-token
subset. Anthropic now checks the complete set of beta tokens to
decide if a request qualifies as Claude Code. Wire them up:

  - /v1/messages mimic path: requiredBetas = FullClaudeCodeMimicryBetas()
  - /v1/messages/count_tokens mimic path: same + BetaTokenCounting

Haiku models keep the 2-token exemption (BetaOAuth + InterleaveThinking).
2026-04-24 23:16:32 +08:00
keh4l b5467d610a fix(gateway): apply full Claude Code mimicry on /chat/completions and /responses
Before: the OpenAI-compat forwarders only called injectClaudeCodePrompt,
which prepends the Claude Code banner but leaves the rest of the body
in its original non-Claude-Code shape. The codebase already admits this
is insufficient (see the comment on rewriteSystemForNonClaudeCode in
gateway_service.go: "仅前置追加 Claude Code 提示词无法通过检测").

Effect: OAuth accounts served through /v1/chat/completions or /v1/responses
were detected as third-party apps and bled plan quota with:

    Third-party apps now draw from your extra usage, not your plan limits.

Fix:
  - apicompat.AnthropicRequest: add Metadata json.RawMessage so metadata
    survives the OpenAI->Anthropic->Marshal round trip; without it the
    downstream rewrite has no user_id to work with.
  - service: extract applyClaudeCodeOAuthMimicryToBody, a ParsedRequest-free
    variant of the /v1/messages mimicry pipeline
    (rewriteSystemForNonClaudeCode + normalizeClaudeOAuthRequestBody +
    metadata.user_id injection) so the OpenAI-compat forwarders can reuse it.
  - service: add buildOAuthMetadataUserIDFromBody + hashBodyForSessionSeed
    for the same reason (no ParsedRequest at the call site).
  - ForwardAsChatCompletions / ForwardAsResponses: replace the 3-line
    prompt-prepend with the full mimicry pipeline.
  - applyClaudeCodeMimicHeaders: set x-client-request-id per-request
    (real Claude CLI always does); missing/duplicated values are one more
    third-party fingerprint signal.

No change to the native /v1/messages path: it already called the full
pipeline, we only lift those helpers into a reusable function.

Tests:
  - go build ./... passes
  - go test ./internal/service/... ./internal/pkg/apicompat/... passes
  - lsp_diagnostics clean on all touched files
  - pre-existing failures in internal/config are unrelated (env-sensitive
    tests that also fail on upstream main)
2026-04-24 23:16:32 +08:00
erio 258fd145ff fix(account): prevent quota-exceeded API key/Bedrock accounts from being scheduled
Add quota exceeded check to IsSchedulable() and refactor
shouldClearStickySession to delegate to IsSchedulable(), eliminating
duplicated logic and fixing missed overload/rate-limit/expired checks.
Frontend displays quota exceeded status independently via quota fields.
2026-04-19 18:45:04 +08:00