Commit Graph
200 Commits
Author SHA1 Message Date
shaw e1b76e2245 fix(codex): normalize load-shed originators to avoid upstream capacity shedding
上游 /backend-api/codex 按 Originator 头分桶调度容量:落在降载桶的请求即使返回
HTTP 200,也会立刻推 SSE `event: error`(code=server_is_overloaded)并以
response.failed 收尾。2026-07-29 起 codex-tui 落入降载桶,codex_cli_rs 正常——
判定因子是 originator 而非 User-Agent(codex_cli_rs 配 curl UA 亦可正常返回)。

网关会把该错误判定为瞬时上游故障并冷却账号,对外表现为 Codex 账号频繁过载不可用:
server_is_overloaded → isOpenAITransientProcessingError →
shouldCooldownOpenAITransientUpstreamError → 账号冷却 → 客户端 503。

本项目有三处降载身份来源:浏览器 UA 兜底的默认 UA、客户端透传的真实 TUI 身份、
以及指纹缓存注入探针的 UA。

修复收口在 enforceCodexIdentityHeaders——HTTP / 透传 / WS 握手 / compat 桥接 /
探针 / PAT / 模型列表 / alpha-search 八条出站路径共用的唯一纯函数收口点:

- 新增 NormalizeCodexClientIdentityToCLI,把降载桶身份改写为 codex_cli_rs,
  只替换身份段并裁掉尾部 (name; version) 客户端标识组,保留版本 / OS / 架构 /
  终端指纹;改写后 originator 与 UA 首段仍然配套,不破坏 #3901 的配对不变式,
  且改写幂等。
- DefaultOpenAICodexUserAgent 从 TUI 身份改为 CLI 身份(浏览器兜底路径上最大的
  降载身份来源)。
- 管理端 Codex UA 的 placeholder / hint 原本在把管理员往降载桶引导,一并修正。

新增 gateway.disable_codex_originator_normalization(默认 false,即归一化开启),
供上游调整分桶后回滚。该开关经 NewOpenAIGatewayService 发布为进程级快照,故必须
保持反义命名:正向命名的 Go 零值 false 会让未经 viper 加载而手工构造的 Config
静默关掉全局保护,viper.SetDefault 救不了这条路径。已加用例钉住该属性。

降载桶集合是上游容量策略快照而非协议常量,上游调整分桶后需同步修订。
2026-08-02 23:00:12 +08:00
shaw da49ce3f29 fix(openai): fail open proxy stream circuit and collapse burst disconnects
The proxy stream circuit introduced in v0.1.164 (#4749) removes every
account behind a quarantined proxy from scheduling. When all schedulable
accounts share one proxy (a common deployment), two mid-stream
disconnects within a minute zeroed out capacity for 10 minutes and every
request failed with 502. One HTTP/2 connection loss also killed all
multiplexed streams at once, tripping the threshold from a single event.

- Quarantine now degrades to a preference: when the only reason no
  account is available is proxy quarantine, selection retries once with
  the quarantine bypassed, so capacity can never reach zero.
- Disconnects within 3s per proxy collapse into one failure event.
- Add gateway.openai_proxy_stream_circuit.disabled escape hatch.
- A completed stream still clears the quarantine immediately; TTL,
  thresholds and recording guards are unchanged.
2026-07-31 10:30:30 +08:00
Wesley Liddick 2e432173f7 Merge pull request #4920 from alexj11324/feat/passkey-auth
feat: add passkey authentication
2026-07-28 14:58:37 +08:00
shaw 97f44b21bb fix: keep passkey switch coupled to WebAuthn config and fix CI issues
- parseSettings now reports passkey_enabled=false whenever the WebAuthn
  deployment config is absent: a stale "true" row left behind after the
  config is removed previously made the admin update gate reject every
  settings save while the UI toggle was disabled, leaving no recovery
  path from the admin panel. Added a regression test.
- update the admin settings API contract goldens with the new
  passkey_enabled/passkey_configured/passkey_rp_id/passkey_rp_origins
  fields.
- errcheck: check rows.Close in passkey repository (repo convention).
- staticcheck QF1001: apply De Morgan's law in WebAuthn origin scheme
  validation.
2026-07-28 11:45:18 +08:00
alfadb 7e65eafbe4 feat: add Kimi K3 support 2026-07-28 10:08:39 +08:00
wucm667 5c471485ab fix(config): honor explicit CONFIG_FILE path
Make CONFIG_FILE select an explicit config for both full loading and lightweight address lookup, with regression tests.
2026-07-27 06:20:11 +08:00
Zhixuan Jiang cc62979aa7 feat: add passkey authentication 2026-07-26 09:50:28 -04:00
song 988d4b577e feat(openai): add macOS Live attestation 2026-07-25 12:50:46 +08:00
song e6eb23eaac feat(openai): add Live gateway support 2026-07-25 12:50:46 +08:00
Heatherm Huang 47ad29db3e fix(openai): quarantine proxies after stream disconnects 2026-07-23 00:12:28 +08:00
Jingru Shi 49200d4747 fix(config): support Redis ACL username 2026-07-21 01:31:46 +08:00
shaw 2518691dbc fix(config): 为 trusted_proxies 与 forwarded_client_ip_headers 注册 BindEnv,修复 env 可达性守卫测试
#4593 引入的 TestConfigKeysAreEnvReachable 与 #4604 新增的两个配置键语义相撞:
两键在 setDefaults 中无注册,守卫测试判定其环境变量不可达。

不能按守卫注释直接 SetDefault:viper 的 IsSet 会连同 defaults 一起上报,
任何已注册默认值都会让 load() 中 trustedProxiesConfigured 的显式配置探测
永远为真,摧毁 #4600 的 absent/empty 区分。改用 BindEnv 注册——绑定键同样
进入 AllKeys()(守卫测试与 Unmarshal 均可见),而变量缺席时不影响 IsSet。
两个环境变量本就由 load() 中 os.LookupEnv 手工解析,行为不变。
2026-07-20 09:41:02 +08:00
Wesley Liddick bfabfe60c8 Merge pull request #4593 from StarryKira/fix/image-storage-env-unreachable
fix: 异步生图开关配了却不生效(环境变量被静默丢弃 + 迁移到后台开关)
2026-07-20 09:20:15 +08:00
JlypxandSisyphus dd0cbe91c9 fix: 避免复制客户端 IP 原子状态
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
2026-07-20 00:22:45 +08:00
JlypxandSisyphus 041db5d824 feat: 支持自定义客户端 IP 请求头配置
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
2026-07-20 00:08:26 +08:00
JlypxandSisyphus 8a147fcc51 fix: 解析显式可信代理配置
Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
2026-07-19 21:39:24 +08:00
Jlypx 732aeef880 fix: 兼容反代和 Docker 客户端 IP 解析 2026-07-19 19:41:01 +08:00
harukaandClaude Opus 4.8 37db8d031b fix(config): 让环境变量能真正配置 image_storage 等凭证
viper.Unmarshal 只解码 AllKeys() 返回的键,而 AllKeys() 只汇总 SetDefault、
配置文件和显式 BindEnv 三个来源。AutomaticEnv 仅能覆盖已在其中的键,无法引入
新键;能兜底的 viper_bind_struct 又被 build tag 排除(我们只用 -tags embed)。

因此任何「没有注册默认值、且不在 config.yaml 里」的配置项,其环境变量会被静默
丢弃。image_storage 的 endpoint/bucket/access_key_id/secret_access_key/
public_base_url 正属此列,于是纯环境变量部署落到最坏组合:IMAGE_STORAGE_ENABLED
生效使 Enabled=true,四个凭证却为空 → Active()=false → 异步生图接口整体 404,
运维看到的却是"凭证不完整"。deploy/docker-compose.yml 默认就是纯环境变量驱动,
且自动生成的 config.yaml 从不写 image_storage 段,必然踩中(见 #4458、#4542)。

同类缺口不止于此:github_oauth、google_oauth、dingtalk_connect 三组第三方登录
配置(含 client_secret)同样完全无法用环境变量设置。

- 为这些键注册零值默认,使其进入 AllKeys() 而可被环境变量覆盖。零值与"键缺失"
  时的解码结果一致,故行为不变。
- sticky_escape_enabled 例外:它的实际默认是 true(靠 IsSet 守卫在解码后补上),
  注册 false 会让 IsSet 恒真而永久关闭该特性,故直接注册 true。
- 启动告警补上 missing_keys 字段,指明到底哪个凭证为空。
- 新增反射守卫测试:Config 结构体上每个可由环境变量表达的字段都必须已注册默认值。

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01VHreE5pzCkSYz7J45fmd2Y
2026-07-18 05:24:32 -07:00
benjamin b92bbf0299 fix: 过滤入口拒绝日志并强化鉴权边界 2026-07-18 00:11:18 +08:00
harukaandClaude Opus 4.8 0eb6e21aaa feat: 异步图片任务结果落对象存储
为异步生图任务增加 S3 兼容对象存储支持,任务结果不再把大图内联存进 Redis:

- 新增可插拔接口 service.ImageStorage(Save -> url),适配别的厂商只需实现它
- S3 实现 S3ImageStorage(AWS S3 / R2 / 阿里云 OSS / MinIO),与备份共用 S3 客户端构造
- 新增 image_storage 配置(config.yaml + IMAGE_STORAGE_* 环境变量),默认关闭
- enabled 同时作为总开关:关闭或未配置对象存储时,异步生图接口返回 404 且不写
  Redis,从根上避免几 MB 的 b64_json 结果撑爆 Redis
- 完成时把图片上传对象存储并把结果改写为短链接(公开直链或 presigned),
  上传失败则任务标记为失败

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SM1tf3CFVRzC7guuhBXvMd
2026-07-15 19:57:37 -07:00
Tian Lee 90ee85f3ef feat: 按上游计费倍率调度 OpenAI 账号 2026-07-16 00:40:46 +08:00
Wesley Liddick 0de768e8be Merge pull request #4221 from heathermhuang/codex/fix-grok-oauth-pool-health
fix(grok): refresh OAuth pools proactively
2026-07-15 16:07:09 +08:00
Wesley Liddick bac925624f Merge pull request #4289 from Tiantianr/fix/openai-ws-first-message-timeout
fix(openai-ws): make first-message timeout configurable
2026-07-15 15:45:24 +08:00
王鹏 fc4089f292 fix(openai): bound native responses first output wait
Add an opt-in first semantic output budget for native HTTP Responses, including response-header wait. Keep preamble and keepalive bytes non-semantic so a stalled account can fail over once without replaying its response IDs. Defaults remain disabled.

Related to #4201, #4185, and #4248. Complements the HTTP/2 dead-connection fix in #4207.
2026-07-15 13:01:16 +08:00
Heatherm Huang 6b25900403 fix(grok): refresh OAuth pools proactively 2026-07-15 09:40:08 +08:00
Tiantianr 74e296703a fix(openai-ws): make first-message timeout configurable
Add a dedicated client first-message timeout while preserving the legacy 30-second default.

Use the resolved value for both the WebSocket read deadline and structured timeout logs, and document tuning for large requests or slow links.

Add configuration, validation, handler, and resolver regression coverage.

Refs #4158
2026-07-14 22:40:05 +08:00
Wesley Liddick c361b0606d Merge pull request #4219 from zh239ns/codex/fix-openai-images-nonstream-keepalive
fix(images): add opt-in non-stream JSON keepalive
2026-07-14 11:30:28 +08:00
zh239ns 002c0b9fda fix(images): keep non-stream requests alive 2026-07-14 07:39:21 +08:00
bestonyandmultica-agent 54d228dda5 feat(admin): add opt-in server timing metrics
Co-authored-by: multica-agent <github@multica.ai>
2026-07-14 01:29:30 +08:00
Bestony@Homelab c8cfc93632 fix(openai-ws): bound ingress session lifecycle 2026-07-13 15:32:42 +08:00
Turtle_Li 3c43fdec11 docs: add batch image PR readiness notes 2026-07-07 03:31:39 +08:00
Turtle_Li 9703ca9d33 merge: sync batch image foundation with upstream main 2026-07-06 13:40:09 +08:00
Turtle_Li 8fab636998 feat: complete batch image workflow 2026-07-06 12:22:04 +08:00
linshuboy f26ca5661e feat: add OpenAI advanced scheduler controls
Related: #1089, #408, #123
2026-07-05 17:24:38 +08:00
Turtle_Li a994fbd77a feat: add batch image MVP 2026-07-04 05:30:50 +08:00
shaw a1b2b32e08 fix: prevent silent usage_logs drops under queue overflow (#3656)
高并发下扣费成功但 usage_logs 被双层异步队列静默丢弃(单用户实测丢失 49%),
导致消费排行与对账缺口。根因是 6a685727 为避免溢出时单行 INSERT 踩踏数据库,
把三处"队列满"改成了立即终态丢弃。

改为有界阻塞背压,兼顾数据不丢与批量写入模式不变:

- CreateBestEffort/createBatched 入队移除 default 立即丢弃分支,
  队列满时阻塞等待(受 detached 15s ctx 期限约束)
- writeUsageLogBestEffort 收到 dropped 不再直接放弃,统一走
  repo.Create 同步兜底(仍经批处理器;ctx 耗尽时换新 detached 窗口),
  重复写入由 ON CONFLICT DO NOTHING 幂等防护
- worker 池默认溢出策略 sample→sync(viper 默认 + 池常量),
  溢出时提交方内联执行(提交点在响应写出后,不阻塞客户端),
  显式配置 sample/drop 的部署不受影响
- 附带修复 ensure*Batcher 的存量数据竞争:channel nil 检查移入 sync.Once

正常负载零行为变化;突发时短暂等待替代永久丢失;数据库只见批量 INSERT。
2026-07-03 21:02:32 +08:00
zy6p 901958ba1b feat(openai-ws): add http_bridge ingress mode and account ws selector
(cherry picked from commit 58647ff63d7ff994c7c14c84c4913a8d3ed6be05)
(cherry picked from commit 9f18fb7c24e187fb20e90d1d9e89fcec91c56937)
2026-06-30 10:40:05 +08:00
Wesley Liddick 7c857bd080 Merge pull request #3441 from deqiying/feature/openai-quota-headroom-scheduler
新增 OpenAI 剩余额度调度权重
2026-06-29 09:23:32 +08:00
deqiying a2cf297d90 feat: 新增 OpenAI quota headroom 调度权重 2026-06-24 00:13:22 +08:00
wucm667 9f5b57fc96 fix(billing): 防止余额计费持续透支 2026-06-22 10:30:27 +08:00
kangjwme 510adf703c feat(scheduling): add opt-in "prefer soonest reset" account selection
Adds a use-it-or-lose-it scheduling strategy: prefer accounts whose
session window resets soonest, so near-reset accounts get drained first
instead of accounts whose reset is still far away.

Both schedulers, opt-in, default behavior unchanged:

- Anthropic (gateway_service.go): new GatewaySchedulingConfig
  .PreferSoonestReset flag. When on, the layered load-aware selection
  inserts a filterBySoonestReset stage (priority -> soonest-reset ->
  load -> LRU). Accounts with no active SessionWindowEnd are treated as
  lowest priority; ties fall through to LRU.

- OpenAI/Codex (openai_account_scheduler.go): new "reset" score weight
  in GatewayOpenAIWSSchedulerScoreWeights. Soonest-reset accounts score
  higher; weight defaults to 0 (no effect).

SessionWindowEnd (upstream 5h/quota ResetsAt) is already carried in the
scheduler snapshot, so no snapshot changes are needed.

Documented in deploy/config.example.yaml. Adds unit tests for the
Anthropic filter and the OpenAI reset factor.
2026-06-18 22:50:46 +08:00
Wesley Liddick f332e0a83c Merge pull request #2872 from wucm667/fix/scheduler-sticky-health-escape
fix(scheduler): session_hash sticky 引入健康度逃逸,慢账号不再独占用户会话
2026-06-05 13:54:37 +08:00
xlx0852 08e19bb15c fix(openai): bridge oversized websocket requests 2026-06-01 10:42:55 +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
wucm667 415d08f255 fix(scheduler): add sticky health escape 2026-05-29 10:25:26 +08:00
Wesley Liddick bebc082306 Merge pull request #2766 from DaydreamCoding/feat/user-platform-quota
feat(quota): 用户 × 平台 USD 配额
2026-05-26 14:13:18 +08:00
mt21625457 33ac8eb27d fix openai http2 response header timeout 2026-05-26 13:57:59 +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
shaw 1e406fed52 fix: optimize OpenAI account cooldown scheduling 2026-05-23 10:18:43 +08:00