Commit Graph
34 Commits
Author SHA1 Message Date
Wesley Liddick 49ad7021d7 Merge pull request #6164 from jimbirthday/fix/openai-sticky-capacity-spillover
fix(openai): preserve sticky binding on capacity spillover
2026-08-25 21:28:38 +08:00
hank 8e60d57475 fix(openai): honor Codex session-id header 2026-08-25 17:08:27 +08:00
超级英雄吉姆 e55727d4c8 fix(openai): preserve sticky binding on capacity spillover 2026-08-25 10:27:48 +08:00
NellPoi 6f972145b7 feat: 支持 OpenAI 重置卡按用量阈值自动使用 2026-08-24 13:28:33 +08:00
anguobao123 3fd66a33be fix(scheduler): diagnose load-batch OpenAI exclusions 2026-08-24 01:25:54 +08:00
Jarl fa4587041c fix(openai): keep auto-review on parent account 2026-08-22 12:14:19 +08:00
IanShaw 1429e8f714 修复 PR 5888 剩余审计问题 2026-08-20 22:01:20 -07:00
IanShaw acce29af27 补齐 OpenAI 与 Grok 协议兼容处理 2026-08-19 23:13:29 -07:00
Randark 901a0439f1 feat: 国产供应商一等支持(Kimi/Zhipu/DeepSeek 多协议 + 配额/余额监控)
后端:
- 协议凭证维度 credentials[api_protocol] ∈ chat_completions(默认)/anthropic/responses(deepseek)
- /v1/messages 零转换直通原生 Anthropic 端点(kimi/zhipu/deepseek),CC/Responses
  入站交叉组合走 apicompat 双向转换链(responses/chat_completions anthropic-native 转发器)
- count_tokens:anthropic 协议透传原生端点;其余 CN 协议本地 tiktoken 估算
- Coding Plan 额度探测(5h/weekly 滚动窗口)+ payg 余额探测(kimi/deepseek),
  deepseek 双币种 CNY+USD 明细,任一币种达标不停调
- 周期任务 [CNBalance] 并发探测 + 预算随工作量放大;响应式 429 冷却到最早窗口
  重置点;余额不足可恢复临时停调;智谱 CREDIT_LIMIT 不污染窗口解析
- CC→anthropic 流式客户端断开后继续排水上游保住 usage 计量

前端:
- 创建/编辑弹窗 account_mode + api_protocol + base_url 联动预设(含 watcher 竞态防护)
- 用量单元格:kimi/zhipu coding 显示 5h/weekly 窗口,kimi/deepseek payg 显示余额,
  多币种并列展示;探测失败保留快照;挂载自动探测 5min 去抖
- 调度阈值设置面板补 kimi/zhipu 平台(对齐后端 AllowedSchedulingThresholdPlatforms)
2026-08-15 10:37:51 +00:00
InCerryGit a8b9ea22b7 fix(openai): separate native and legacy compaction routing 2026-08-14 18:18:53 +08:00
InCerryGit 9662cff2e7 fix(openai): preserve remote compaction v2 responses endpoint 2026-08-14 16:23:52 +08:00
Wesley Liddick f19e3816d0 Merge pull request #5316 from wucm667/fix/issue-5313-legacy-scheduler-diagnostics
fix(scheduler): diagnose legacy OpenAI exclusions
2026-08-10 10:52:14 +08:00
IanShaw027 3a415e6d04 fix(grok): 五轮评审 — messages SearchCount 与 OpenAI free-gate
- /v1/messages Grok:SSE/buffered 路径累计 SearchCount
- OpenAI 遗留 sticky/list 应用 free soft-gate(advanced 关闭时)
- 单测:OpenAI getSchedulableAccount free-gate + messages 计数器
2026-08-07 18:23:28 +08:00
IanShaw027 7c62382d04 feat(grok): 吸收调度阈值、配额解析、批量用量与 CLI 身份
从 personal-dev 取优并接线到 feat/grok-complete-integration:

- 调度阈值:grok_sched_* 写入、ApplyAccountSchedulingThreshold、
  account_scheduling_thresholds 设置与 Settings UI、网关过滤
- 配额头:x-rate-limit 别名、相对秒 reset、tier/entitlement 扩展
- 批量 usage:GetUsageBatch + POST /usage/batch + AccountsView 合并加载
- CLI 身份:xai.cli_identity 统一 pin/UA;service 层 applyGrokCLIHeaders
  与 transport 最终改写对齐;media eligibility 模块落地
- UsageCell:月度 cents→USD 与 30d 进度条展示

保留既有 failure taxonomy / sticky / free recovery 等 HEAD cools。
2026-08-07 14:53:18 +08:00
IanShaw027 2413441b5f feat(grok): P2 模型额度软封、spending reauth、粘连分轨与观测模型
- free-usage 带模型名时按账号+模型软封,其它模型仍可调度
- spending-limit 使用 24h 冷却并标记 needs_reauth(extra + SetError)
- sticky affinity 与 upstream prompt_cache_key 分轨,按 model 隔离粘连
- free-usage recovery 轻量提示窗口;配额查询后异步同步 /v1/models 快照
2026-08-07 14:06:21 +08:00
IanShaw027 e12e0dc1a6 feat(grok): 对齐 free-usage/empty 失败分类与 sticky/ReAuth 体验
从 grokcli 吸收 body-first 失败分类(free-usage/empty/billing),
接到现有 temp_unsched 与 failover,保留 content-policy 与 pool_mode。
Grok 粘连/缓存在无显式 session 时可用 previous_response_id;
ReAuth 预填 email---- 并默认密码 tab(密码不落库)。
2026-08-07 13:47:09 +08:00
shaw dbb42881c0 fix(openai): default OAuth identity to codex-tui 2026-08-07 09:56:03 +08:00
wucm667 df249ee5d5 fix(scheduler): diagnose legacy OpenAI exclusions 2026-08-06 02:54:16 +08:00
Brisbanehuang fad2f215e8 fix(profit-control): decouple gate from image intent, propagate gate via selection, restore eager sticky fallback
Review fixes for the profit-control feature commit:

- Image intent no longer disables the profit gate. The shared /v1/responses
  handler previously skipped the pricing context (and therefore the gate)
  whenever the platform-wide image intent predicate matched, which includes
  Codex's passive image_gen namespace declaration: any client could disable
  admission control for anthropic/gemini/antigravity groups by declaring a
  namespace tool in the request body. Both /v1/responses paths now always
  install the token pricing context; image intent only drives capability
  routing and image billing. Mixed token+image requests stay token-gated;
  only dedicated media endpoints remain out of scope.
- Out-of-scope paths are now explicitly suppressed instead of implicitly
  ungated: Grok media (billed by media multipliers; also prevents in-flight
  video lookups from turning into spurious 404s), OpenAI-group count_tokens
  (unbilled), and Live calls (duration billed) carry a suppress marker that
  every install point honors, including the defensive scheduler-entry
  install.
- The gate resolved during selection now travels back to handlers on the
  AccountSelectionResult. The shared gateway installed the gate only on a
  scheduler-local context, so handler-side post-slot terminal rechecks and
  post-admission sticky binding were no-ops for anthropic/gemini/
  antigravity/shared-grok requests (and for composite-routed member groups
  on the OpenAI path). Handlers re-apply the carried gate via
  ContextWithSelectionProfitGate before the terminal recheck and binding;
  the WS acquired-selection branch gained the previously missing recheck.
- Sticky binding semantics restored for ungated traffic:
  BindStickySessionAfterProfitAdmission falls back to the official eager
  bind when no gate is installed (wait paths lost their only binding point
  otherwise), reads the pre-existing binding at bind time only when gated
  (removes the unconditional per-request Redis read the feature added to
  the shared handlers), and the legacy engine's three selection-time
  binding writes are skipped under a gate so a terminally vetoed account
  can no longer become the new sticky target.
- Responses WS connections re-freeze pricingAt and re-resolve the gate at
  every turn (BeforeTurn) and bill each turn with its own instant, closing
  the connect-at-valley/bill-at-valley window; a turn that fails the
  recheck closes the connection so the client reselects on reconnect.
- Terminal recheck no longer swaps a DB-fresh selected account for a
  staler snapshot object (UpdatedAt guard), and observer counters are
  documented as per-evaluation.
- Migrations renumbered 191/192 -> 192/193 after the passkey migration
  landed upstream as 191.

New regressions: selection-carried gate propagation (control group proves
the pre-fix no-op), image intent not disabling the shared gate, eager
binding fallback without a gate (both services), gated
read-failure/sentinel-miss binding semantics, legacy-engine deferred
binding under a gate with eager behavior preserved ungated, turn-level
pricing refresh (config re-resolution, scheduled-group precedence,
suppress, mid-connection disable), and suppress-marker coverage for the
request pricing context.
2026-08-01 22:39:32 +08:00
Brisbanehuang 20ad5ec506 feat(scheduler): per-group profit control for token account admission
Group pricing (rate multiplier, peak windows, per-user overrides) and
account cost (accounts.rate_multiplier) already live side by side, but
nothing stops the scheduler from handing a request to an account whose
cost multiplier exceeds what the group's pricing can profitably serve.
Add an opt-in per-group profit gate that filters scheduling candidates
by a margin rule, while ordering, scoring, stickiness and breakers keep
working unchanged among qualified accounts.

Admission rule: an account qualifies iff U <= D * (1 - min_margin -
safety_buffer) within a small relative epsilon, where U is
accounts.rate_multiplier (0 is legal; missing/negative/NaN/Inf are
conservatively rejected as invalid) and D is the requester's effective
downstream multiplier (user-group override ?? group default, times the
group peak factor) frozen at the request's pricing instant.

- groups gain profit_control_enabled / profit_min_margin /
  profit_safety_buffer (migration 191); the durable auth-cache
  invalidation trigger additionally watches the profit and pricing
  columns (migration 192) so out-of-band group edits cannot leave
  stale auth snapshots; GetByKeyForAuth explicitly projects the new
  columns and the API-key auth snapshot version is bumped to force a
  refresh of pre-existing snapshots
- request-level pricing instant: token entry points install pricingAt
  into ctx; the profit threshold D and the RecordUsage peak factor
  read the same instant, so one request never changes price mid-flight
  across waits/retries/failover (media and unwired paths keep the
  existing record-time semantics)
- the gate covers token requests on openai, anthropic, gemini, grok
  and antigravity groups: OpenAI-family handlers via
  WithOpenAIRequestPricingContext (responses incl. WS bridge, chat
  completions, messages, embeddings, alpha search), the shared gateway
  via WithGatewayTokenRequestPricing (messages, chat completions,
  responses, gemini model actions); composite groups cannot enable it
  directly; image/video/models/usage/count_tokens stay ungated and an
  explicit image-generation intent suppresses the gate end to end
- post-slot recheck: after a slot is acquired the account is re-read
  via SchedulerSnapshotService.GetAccount (scheduler cache, then DB;
  only when both fail the check fails open with WARN + metric); a
  vetoed account releases its slot and joins the request's exclusion
  set for reselection; sticky bindings are written only after the
  final check passes, and an over-threshold sticky account is skipped,
  not deleted, so it comes back once its rate recovers
- sticky-session cache contract: GatewayCache.GetSessionAccountID now
  returns ErrStickySessionNotFound on a miss (mapped from redis.Nil in
  the repository implementation, mirroring ErrRefreshTokenNotFound) so
  the profit sticky path can distinguish "no binding yet" from a real
  read failure without importing the cache driver in service code
- cross-group re-entry (composite parent -> member group) resolves the
  gate against the member group and clears a stale parent gate instead
  of letting a foreign threshold veto accounts
- per-platform/group activity counters (installs, threshold vetoes,
  invalid-rate vetoes, refresh failures) for observability
- admin UI: profit-control section on the five platforms' group forms
  with percent input, validation and platform-switch reset; group
  create/update/duplicate normalize and validate the config at a
  single choke point
- cmd/profit-preview: offline what-if tool that replays the production
  admission semantics over an exported config/account/override/model
  dump, reports per-model admitted-account counts under the default
  and the worst-case (lowest user override) D, and surfaces probe-sync
  staleness as warnings without affecting admission

Tests: service unit coverage for gate resolution/veto/threshold
epsilon/pricing instant/suppress marker/scheduler filtering and
post-slot recheck (incl. -race on the profit surface), unit-tagged
handler slot-recheck and capability-mapping regressions, sqlmock and
real-PostgreSQL integration regressions for the GetByKeyForAuth
projection and the migration-192 trigger watch list, API contract
update, and frontend specs for the five-platform form helpers.
2026-08-01 22:39:31 +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
Edison42 1c0cb24c7e feat(usage): persist client session identifiers 2026-07-24 01:22:34 +08:00
Heatherm Huang 47ad29db3e fix(openai): quarantine proxies after stream disconnects 2026-07-23 00:12:28 +08:00
superman2003 e31783e443 fix(grok): preserve OpenCode and CodeBuddy cache sessions 2026-07-21 15:22:12 +08:00
Wesley Liddick 1f9eac4fbd Merge pull request #4660 from superman2003/fix/openai-scheduler-exclusion-observability-20260720
fix(scheduler): 高级调度器无可用账号错误补充排除原因统计,可自诊断静默剔除 (#4599)
2026-07-21 10:43:21 +08:00
superman2003andCursor 29bea0a758 fix(scheduler): 高级调度器无可用账号错误补充排除原因统计
问题背景(Refs #4599):OpenAI OAuth 账号请求 /v1/responses 全模型 503,
日志仅有 no available OpenAI accounts supporting model 且
excluded_account_count=0。selectByLoadBalance 初筛循环的每个过滤分支
(配额自动暂停、渠道上游模型限制、运行时熔断等)都是静默 continue,
最多只有 Debug 级日志,用户与维护者都无法定位候选账号被哪个隐式
过滤点剔除。

修改:
- selectByLoadBalance 初筛循环为每个 continue 分支记录原因计数
  (excluded/not_schedulable/platform_mismatch/runtime_blocked/
  privacy_not_set/transport_incompatible 及请求兼容性细分原因)
- isAccountRequestCompatible 新增 isAccountRequestCompatibleReason
  变体,细分 runtime_blocked/quota_auto_pause_<窗口>/
  shadow_parent_unhealthy/model_not_supported/
  channel_upstream_restricted/capability_mismatch;原 bool 函数改为
  包装变体,其余调用点行为不变
- noAvailableOpenAISelectionError 增加 details 参数,初筛全灭时输出
  如 (pool=3, filtered: excluded=1 model_not_supported=1
  quota_auto_pause_7d=1),原因按字典序排序保证输出确定;快照池为空
  时输出 (pool=0);选择序列耗尽路径追加 selection_order_exhausted
  标注;legacy 调用点传空保持原消息不变

兼容性:错误仍通过 Unwrap 匹配 ErrNoAvailableAccounts,分类与 ops
标记逻辑不变;客户端可见错误仍为通用分类消息,统计仅进服务端日志;
happy path 原因 map 惰性分配,无额外开销。

Refs #4599

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-07-20 21:44:42 +08:00
abbzbb 2ae61f3da0 fix(grok): Codex compact 适配与链式视频 content 代理
- #4223/#4554: Grok 模拟 /responses/compact,调度允许 Grok 账号
- #4494/#4626: 支持链式中继的受保护视频 /content 下载
2026-07-20 16:12:56 +08:00
Heatherm Huang d2ac6b2c97 fix(grok): harden media quarantine edge cases 2026-07-17 12:08:50 +08:00
Heatherm Huang 8bc0a277f2 fix(grok): quarantine ineligible media accounts 2026-07-17 11:51:16 +08:00
shaw 0408bdb34f Merge remote-tracking branch 'origin/main' into feat/upstream-rate-scheduling
# Conflicts:
#	backend/internal/handler/openai_gateway_handler.go
2026-07-16 10:31:21 +08:00
Tian Lee 90ee85f3ef feat: 按上游计费倍率调度 OpenAI 账号 2026-07-16 00:40:46 +08:00
王鹏 40b8f04a6a fix(openai): scope transient cooldowns by model 2026-07-15 22:52:35 +08:00
superman2003 0478fd3668 feat(grok): enable free OAuth prompt caching 2026-07-11 19:34:01 +08:00
shaw d754be0d8e refactor(gateway): 抽取 CC forwarder 公共管线并拆分两大 service 文件
PR #3802 遗留项:三个 CC forwarder(raw 直转 / responses 回退 / messages
回退)间约 85% 重复的 HTTP 管线与 SSE 循环骨架收敛到新文件
openai_gateway_cc_pipeline.go,messages / chat_completions 两条主路径中
逐字相同的错误处理块一并接入。各路径有意保留的行为差异(GLM effort
归一化、fast policy 适用范围、ClientDisconnect 语义、Grok 分支等)留在
调用方,未强行统一。

同时对两个最臃肿的网关文件做纯移动拆分(零语义变化,逐字节校验):

- openai_gateway_service.go 7821→4872 行:
  调度 → openai_gateway_scheduling.go
  passthrough → openai_gateway_passthrough.go
  用量/计费/codex 快照 → openai_gateway_usage.go
- gateway_service.go 10912→7294 行:
  调度 → gateway_scheduling.go
  Anthropic APIKey 直通 → gateway_anthropic_passthrough.go
  Bedrock → gateway_bedrock.go

回归保障:全部搬迁块与 HEAD 逐字节 diff 一致;留存文件经"HEAD 减去搬迁
范围"重构比对,差异仅为 goimports 移除的孤儿 import;定向单测
(fallback/raw/cyber/grok/transport/调度/用量/广域 Forward-Handle 扫描)
全绿;另经独立对抗审计确认零行为差异。
2026-07-07 22:14:46 +08:00