Commit Graph
728 Commits
Author SHA1 Message Date
chaochaoweb3 5bd9368ab4 fix claude oauth token exchange payload 2026-06-25 23:41:00 +08:00
feitianbubu 4567f6582b test(admin/usage): update sqlmock rows for cache breakdown columns 2026-06-24 23:59:22 +08:00
feitianbubu 063454ae96 fix(admin/usage): populate cache creation/read token breakdown in stats 2026-06-24 23:43:03 +08:00
wucm667 9f5b57fc96 fix(billing): 防止余额计费持续透支 2026-06-22 10:30:27 +08:00
Wesley Liddick 2e0ff1cfd5 Merge pull request #3258 from bwliangc/feat/channel-monitor-jitter
feat(渠道监控): 检测间隔支持正负随机抖动配置
2026-06-16 16:58:23 +08:00
Wesley Liddick 16765bde69 Merge pull request #3230 from DaydreamCoding/feat/openai-cyber-policy-passthrough
feat(openai): cyber_policy 硬阻断全链路透传、审计与计费
2026-06-16 16:55:51 +08:00
shaw f069c9ae00 fix(outbox-dedup): buildSchedulerGroupPayload typed-nil broke dedup_key consistency
#3255 introduced payload-aware dedup_key (sha256 over event_type, account_id,
group_id, payload_json). enqueueSchedulerOutbox checks "if payload != nil"
before json.Marshal — but Go interface containing a typed-nil map (returned
by buildSchedulerGroupPayload(empty)) is NOT == nil at the interface level.

So an ungrouped account's account_changed event went through this path:
  payload := buildSchedulerGroupPayload(account.GroupIDs)  // typed-nil map
  enqueueSchedulerOutbox(..., payload)                     // interface != nil
    → json.Marshal(typedNilMap) = "null"
    → dedup_key hash = sha256(... + "null")

While other call sites pass literal nil:
  enqueueSchedulerOutbox(..., nil)                         // interface == nil
    → payloadJSON stays empty
    → dedup_key hash = sha256(... + "")

The two dedup_keys differ for what should be the same logical event,
silently degrading dedup effectiveness in bursts on ungrouped accounts.

Fix: change buildSchedulerGroupPayload return type from map[string]any to
any so empty input returns true untyped-nil. All call sites pass the
result straight to enqueueSchedulerOutbox(payload any) — no inspection,
no breakage.

Adds regression test TestEnqueueSchedulerOutbox_UngroupedAccountDedupesWithLiteralNilPayload
asserting (1) typed-nil regression doesn't sneak back, (2) dedup_key for
empty-groups payload matches the literal-nil-payload key.
2026-06-16 14:21:09 +08:00
shaw acaffe29ec fix(account-repo): refresh candidates SQL excluded healthy accounts; fix CI build
Post-merge audit of #3272 found two regressions:

1. ListOAuthRefreshCandidates used "AND NOT (a AND b)" which, under PG
   3-valued logic, evaluates to NULL when both temp_unschedulable_until
   and temp_unschedulable_reason are NULL — i.e., the common healthy
   account state. Such rows were silently excluded from the background
   token refresh worker, so their OAuth access tokens would never get
   refreshed and eventually start returning 401.

   Verified empirically against PostgreSQL: only 3 of 5 test rows
   matched before the fix; after switching to "(a AND b) IS NOT TRUE"
   the expected 4 rows match.

2. The new ListOAuthRefreshCandidates method on AccountRepository was
   not implemented on stubAccountRepo in api_contract_test.go (build
   tag "unit"), breaking "make test-unit" which CI runs in
   .github/workflows/backend-ci.yml.

Tests:
- Added IS NOT TRUE and "AND NOT (" assertions to the SQL-shape unit
  test so the predicate can't regress to the broken form again.
- "go test -tags=unit ./internal/..." now passes cleanly.
2026-06-16 14:08:50 +08:00
shaw 31dc8913ac chore(outbox-cleanup): add 10s grace to defend against id-vs-commit race
PG sequences advance outside transactions, so a slow committer can hold
an id that gets surpassed by the watermark before its commit becomes
visible. Without the grace period, cleanup would delete such rows before
the snapshot poller ever sees them. 10s is a comfortable upper bound
on realistic enqueueSchedulerOutbox commit latency.
2026-06-16 11:57:32 +08:00
jjaw cb14935e9a fix: cleanup consumed scheduler outbox rows 2026-06-16 11:56:40 +08:00
shaw 1fdbe52f96 chore(migrations): renumber scheduler outbox dedup migrations 151/152 -> 152/153
#3232 (already merged) occupies 151. Renumber #3255's dedup migrations to
avoid collision and keep linear ordering. Updated runner constant + tests.
2026-06-16 11:50:25 +08:00
jjaw b3ec6288ad fix: release scheduler outbox dedup on claim 2026-06-16 11:48:13 +08:00
jjaw 60cf89ae26 fix: recover scheduler outbox invalid dedup index 2026-06-16 11:48:13 +08:00
jjaw 3ef70b045d fix: safely coalesce scheduler outbox events 2026-06-16 11:48:12 +08:00
jjaw 34e66ec0a5 fix: outbox scheduler snapshot coalesce 2026-06-16 11:48:12 +08:00
jjaw 74199b6a6e fix: reduce token refresh retry amplification 2026-06-16 11:44:39 +08:00
jjaw 8b698ff4c1 fix account list parameter limit 2026-06-16 11:39:15 +08:00
Wesley Liddick 8ddc138094 Merge pull request #3232 from jianjianai/fix1/account-expiry-autopause-index
fix:account expiry autopause index / 账号过期自动禁用索引优化
2026-06-16 11:12:57 +08:00
bwlcandClaude Fable 5 c70c6a2659 feat(渠道监控): 检测间隔支持正负随机抖动配置
新增 jitter_seconds 配置:每轮调度在 interval 基础上 ± [0, jitter]
均匀随机偏移触发,避免多个监控以固定节奏同步请求上游。

- ent schema 新增 jitter_seconds 字段(默认 0),附迁移 151
- 校验:jitter >= 0 且 interval - jitter >= 15s(创建/更新均校验)
- runner 由固定 ticker 改为每轮重新随机化的 timer,0 抖动时行为不变
- 前端监控表单新增「随机抖动 (± 秒)」输入框,上限随间隔联动

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-06-12 22:09:53 +08:00
wucm667 c1c28ac7bb fix(gateway): 解压 zstd 上游响应体 2026-06-12 15:06:01 +08:00
DaydreamCodingandClaude Opus 4.8 b62b573f7f feat(openai): cyber_policy 硬阻断全链路透传、审计与计费
上游对单次请求下发 error.code=cyber_policy 硬阻断时,网关在所有端点
(/v1/responses、/v1/chat/completions、/v1/messages、WebSocket)及流式/
非流式路径下,将该结果原样透传给客户端,绝不 failover、换号或同步拦截;
命中后异步完成审计与计费:

- 风控中心记录 cyber_policy 留痕并发送通知邮件,落库先于发信,SMTP 阻塞
  不影响留痕
- ops 错误请求记录,状态码对齐客户端实际接收(流式 200 / 非流式 400)
- 用量明细标记 request_type=cyber,按上游真实 token 计费,HTTP 与
  WebSocket 计费口径统一,零 token 命中不误扣
- 会话级自动屏蔽(管理员开关,默认关):命中的会话在可配 TTL 内本地拦截
  不再发往上游,仅屏蔽该会话不影响同 Key 其他会话
- 封号计数排除开关:可选让 cyber 命中不计入自动封号,命中当次不判定且
  历史行在违规计数中一并排除

WebSocket 多轮连接下 cyber 标记按 turn 生命周期管理,逐轮独立检测与记录;
透传的错误响应不被兜底逻辑追加内容污染。

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-12 01:47:01 +08:00
jjaw e4c255a77a fix:account expiry autopause index 2026-06-12 01:38:09 +08:00
DaydreamCodingandClaude Opus 4.8 329414ea4f feat(admin): /admin/users 新增按用户 API Key 所在分组过滤
- 支持专用/公开/订阅/已禁用四类分组:按 api_keys.group_id 精确匹配,
  排除软删除 key(EXISTS + DeletedAtIsNil);已禁用分组单独成区,
  覆盖 key 仍挂在禁用分组上的用户
- 后端:UserListFilters.APIKeyGroupID;handler 解析 api_key_group_id;
  repo HasAPIKeysWith 谓词;GetAllGroupsIncludingInactive 新接口
  (/admin/groups/all?include_inactive=true)
- 前端:下拉按类型分区单选;UsersView 独立 allGroupsForApiKeyFilter
  loader;分区标题用负数哨兵值修复 Vue :key 重复问题
- 测试:repo 集成 5/5(含软删除排除、多 key 去重、叠加 status 过滤)、
  handler 单测 5/5、前端 vitest 6/6

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-08 22:54:01 +08:00
Wesley Liddick 870cefddf5 Merge pull request #3097 from NeckBozia/fix/5h-resets-at-stale
fix(usage): sync 5h ResetsAt to SessionWindowEnd and zero expired window
2026-06-08 09:08:23 +08:00
Wesley Liddick a2bf6ad77a Merge pull request #3107 from DaydreamCoding/feat/proxy-failover-resilience
feat: 代理生命周期与故障健壮性增强
2026-06-08 08:59:15 +08:00
DaydreamCoding af19d44327 feat(proxies): 代理有效期与失败回退
- schema/迁移: 代理有效期、提醒天数、失败回退配置 + 账号 fallback 来源字段
- service/repo/DTO/handler: CRUD 透传新字段 + 校验
- fallback 目标解析纯函数(链式解析 + 环检测 + 兜底)
- SweepExpiredProxies 到期改投账号 + outbox 失效
- ProxyExpiryService 后台到期扫描任务 + wire 注册
- 账号侧手动回切原代理 + fallback 来源徽章/按钮
- 前端: 创建/编辑表单、列表到期徽章、类型/API/i18n
- ops 告警: proxy_expired_count / proxy_expiring_soon_count 指标
- 导入导出携带有效期/回退字段(备用按 name 映射)
- 补全测试 stub + 集成测试 + review 问题修复
2026-06-08 00:01:30 +08:00
CoolCoolTomato 1a86c6ce12 fix: enforce exclusive group access for api keys 2026-06-07 20:46:03 +08:00
NeckBozia 16bc876935 fix(usage): sync 5h ResetsAt to SessionWindowEnd and zero expired window
active poll 拿到新 5h ResetsAt 时只回写了 Extra.session_window_utilization,
没回写 SessionWindowEnd column;estimateSetupTokenUsage 读这个 column 作为 5h
窗口结束时间,导致被动采样模式下 dashboard 显示 utilization > 0 但 reset
时间渲染为「现在」。

- syncActiveToPassive 增加 UpdateSessionWindowEnd 回写
- estimateSetupTokenUsage 抄齐 Codex 分支的过期归零 guard,避免 active
  poll 没回写时 UI 渲染矛盾
- UsageProgressBar 区分「待刷新」/「现在」语义;i18n key 中英补齐
2026-06-07 15:23:45 +08:00
feitianbubu 029b6d61a2 feat(usage): 聚合统计拆分缓存创建与命中 token 2026-06-06 22:32:37 +08:00
Wesley Liddick 56a6f53848 Merge pull request #2999 from Pluviobyte/codex/sub2api-scheduler-snapshot-sync
fix: sync scheduler snapshots after account state updates
2026-06-06 14:27:31 +08:00
Wesley Liddick 5824359e97 Merge pull request #3062 from visa2/fix/multi-instance-bg-job-leader-lock
fix(service): gate periodic background jobs with a leader lock for multi-instance
2026-06-06 13:59:42 +08:00
Wesley Liddick b76f9524ba Merge pull request #3040 from bwliangc/feat/ops-ttft-sample-weighting
fix(ops): 运维监控首 token 延迟(TTFT)按流式样本数加权
2026-06-06 13:48:51 +08:00
Wesley Liddick 6f18f74030 Merge pull request #2970 from ghostg00/compat/redis-replicate-commands
compat(redis): make TIME-using Lua scripts work on Redis 3.2-4.x
2026-06-06 11:23:54 +08:00
shaw bba86f97d8 fix(repo): make userRepo.Delete reuse caller's transaction
userRepository.Delete used the r.client.Tx(ctx)+ErrTxStarted pattern,
which only reuses a transaction when r.client is itself a tx-bound
client. With the singleton base client (how Wire injects repos),
r.client.Tx() always starts a NEW independent transaction and ignores
any tx carried in the context.

As a result, AdminService.DeleteUser's transaction (added in #3047)
only covered the API-key deletions; the user soft-delete escaped into a
separate, immediately-committed transaction. If the outer commit failed
after the user tx had committed, the user was deleted while its API keys
stayed active, reintroducing the unrecoverable state of #3021.

Add a leading dbent.TxFromContext check (mirroring DeleteWithAudit) so
Delete joins an externally-provided transaction and lets the caller
commit/rollback. Extract the deletion body into deleteUser(ctx, exec,
id). All three original paths are preserved; callers passing a plain ctx
are unaffected.

Add user_repo_delete_atomicity_integration_test.go: builds repos with
the base client and a manual outer tx, asserting that delete-keys and
delete-user roll back together on abort and commit together on success.
2026-06-06 10:28:28 +08:00
visa2andClaude Opus 4.8 362f9e77bf fix(service): gate periodic background jobs with a leader lock for multi-instance
Three periodic background jobs ran on every instance with no cross-instance
coordination, multiplying their cost (and side effects) by the replica count:

- DashboardAggregationService.runScheduledAggregation: N× heavy GROUP BY
  aggregation queries every minute plus watermark write races.
- PaymentOrderExpiryService.runOnce: N× upstream payment-provider reconcile/
  expiry API calls per pending order.
- SubscriptionExpiryService.sendExpiryReminders: N× full active-subscription
  scans every minute and potential duplicate reminder emails.

Add a LeaderLockCache abstraction so only one instance runs each job per cycle:

- The interface lives in the service layer; the Redis-backed implementation
  (SetNX + compare-and-delete release) lives in the repository layer, so the
  service package keeps its depguard "must not import redis" boundary intact.
- tryAcquireSingletonLeaderLock prefers the cache and falls back to a Postgres
  advisory lock when Redis errors, mirroring the Ops background services. When
  neither backend is configured the job runs ungated, preserving single-instance
  and test behavior (no self-lockout: the lock is released every cycle).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-05 22:47:24 +08:00
方程 705fe7d880 fix(admin): delete user api keys with user 2026-06-05 16:23:56 +08:00
DaydreamCodingandClaude Opus 4.8 cfb195c7b2 feat(usage): 记录并展示失败请求(用户端+管理端)
- 记录失败请求并在用户端/管理端展示;分类下拉改用统一 Select 组件
- 模型过滤改后端 ILIKE 模糊匹配;新增「Key 名称」列(含已删除标记)与按 Key 过滤;时间列移至末列
- 用户可见「已删除 key 失败请求」:OpsErrorLogFilter 加 MatchDeletedKeyOwner,用户侧归属
  放宽为 (user_id OR deleted_key_owner_user_id),让 key 原所有者能看到删除 key 后继续请求
  导致的认证失败记录(他人仍 NotFound,不泄露存在性)
- 迁移 148:ops_error_logs 用户+时间索引

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-05 14:00:57 +08:00
DaydreamCodingandClaude Opus 4.8 ddf063352a feat(ops): 错误日志 key 归因与早退字段补全
让 /admin/ops 错误详情正确归因 API key 并补全早退场景字段,合并三项改动:

- 鉴权早退补全用户/分组/平台字段:引入 ops fallback key(ContextKeyOpsFallbackAPIKey),
  apiKey 一加载成功即写入,覆盖分组停用/删除、Key 停用/过期/额度、用户停用、IP 限制等早退
  路径;ops 错误日志改用 getOpsAPIKey(正式 key 优先、回退键兜底),不改「已鉴权」语义。
- 已删除 key 归因(迁移 145):删除 key 时同一事务写 deleted_api_key_audits 映射,认证失败
  时用明文反查命中原所有者,错误详情展示「已删除 Key 所有者」「尝试的 Key 前缀」。
- 有效 key 报错快照前缀(迁移 147):对绑定有效 key 的错误,落库时快照明文前 8 位到
  api_key_prefix(与 attempted_key_prefix 互斥),key 之后被删仍保留报错当时真实前缀。

均仅对上线后新产生的错误/删除生效。

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-05 14:00:57 +08:00
Wesley Liddick 57bae985a7 Merge pull request #2711 from wucm667/fix/db-pool-enforce-conn-lifetime-floor
fix(db-pool): 强制连接生命期下限,缓解 lib/pq watchCancel 协程泄漏
2026-06-05 13:55:07 +08:00
bwlcandClaude Opus 4.8 69b4654510 fix(ops): weight TTFT percentiles by streaming sample count
TTFT (first_token_ms) is only recorded for streaming requests, but the
ops dashboard weighted merged TTFT percentiles by success_count (all
successful requests, streaming + non-streaming). When non-streaming
traffic was present this diluted/skewed the merged TTFT figures shown
for longer (pre-aggregated) time ranges; the realtime path was exact.

Add a per-bucket ttft_sample_count (rows that actually recorded
first_token_ms) to ops_metrics_hourly / ops_metrics_daily and weight all
TTFT percentile merges by it instead of success_count:

- hourly/daily pre-agg upserts populate and propagate ttft_sample_count;
  daily TTFT p50/p90/avg now weighted by ttft_sample_count.
- dashboard hourly-row merge and cross-segment combine weight TTFT by
  the streaming sample count; queryUsageLatency returns it for raw
  head/tail fragments.

duration stays weighted by success_count (recorded for every request);
p95/p99/max keep the conservative MAX merge (weight-independent).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-05 10:58:44 +08:00
Pluviobyte b6c0706e30 fix: sync scheduler snapshots after account state clears 2026-06-03 14:18:12 +08:00
ghostg00 585ff09443 compat(redis): replicate effects for Lua scripts using TIME on Redis 3.2-4.x
Call redis.replicate_commands() at the top of every Lua script that
reads redis.call('TIME'), so the resulting writes (ZADD, ZREMRANGEBYSCORE,
SET, DEL, EXPIRE) replicate correctly on Redis 3.2-4.x.

This is a no-op on Redis 5.0+, where effects replication is the default.
The function remains a public API on Redis 7.0+ (deprecated but supported)
and is still the documented way to opt into effects replication on older
versions, so calling it unconditionally is safe and forward-compatible.

Affected scripts (8 in total, all already using redis.call('TIME') by
design to avoid client-side clock skew across multiple app instances):
- concurrency_cache.go: acquireScript, getCountScript,
  cleanupExpiredSlotsScript
- session_limit_cache.go: registerSessionScript, refreshSessionScript,
  getActiveSessionCountScript, isSessionActiveScript
- user_msg_queue_cache.go: releaseLockScript

Scripts that do not invoke non-deterministic commands are intentionally
left untouched.
2026-06-02 12:46:46 +08:00
Wesley Liddick 5f63fe1945 Merge pull request #2927 from moonagic/main
fix antigravity gemini rate limit and account scheduling
2026-06-01 14:50:36 +08:00
moonagic a01686c637 fix antigravity gemini rate limit and account scheduling
Squash of 4 commits:
- Fix Gemini rate limit scheduling
- fix antigravity gemini rate limit scheduling
- fix antigravity gemini limited account scheduling
- fix antigravity test stubs for default lint
2026-05-31 22:00:03 +08:00
DaydreamCodingandClaude Opus 4.8 bf24b61139 perf(usage): 优化 /admin/usage 打开速度与刷新响应
根因:页面 mount 并发 6 个请求,其中 5 个在原始 usage_logs 上 live 聚合,
且有 1 个重复 getModelStats。优化(不引入预聚合):

前端
- UsageView 经 :model-options 下传 model 列表,移除 UsageFilters 重复的
  getModelStats(mount 请求 6→5,少一次 usage_logs 全表 GROUP BY model)
- 刷新/换筛选保留旧模型数据(invalidateModelStatsCache 只失效标记不清空数据),
  图表不再闪空,刷新期间页面保持可交互

后端
- GetStatsWithFilters 4 条聚合 errgroup 并行(仅 *sql.DB 连接池路径,ent.Tx
  顺序回退以保事务内不并发),endpoint 明细 best-effort;抑制取消级联噪声日志
- /admin/usage/stats 复用 dashboard 的 newSnapshotCache 30s 处理器层缓存,按
  filters+窗口为 key;前端手动刷新带 nocache=1 强制回源(刷新=最新)

注:自合并提交 4c8396c 迁移而来,仅取"列表打开速度与刷新响应"部分;原提交的
"审计查看弹窗大 body 渲染"改动(AuditLogModal / audit-log-format / 审计 i18n)
依赖尚未迁移的审计功能(d8389ade),本次已排除。

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-05-30 18:56:08 +08:00
DaydreamCodingandClaude Opus 4.7 b60d8bb4cc feat(usage): 在 /admin/usage 支持查看已删除用户的历史使用情况
用户软删除后使用记录仍在,但身份(邮箱)被 ent 软删除拦截器隐藏。本次在
三条管理员只读路径定点穿透软删除过滤,并把删除状态传播到前端标记,零新表/
迁移/回填:

- 后端穿透:富化 usage 日志(loadUsers)、用户搜索(ListWithFilters +
  UserListFilters.IncludeDeleted)、点击详情(GetByIDIncludeDeleted /
  GetUserIncludeDeleted + getById ?include_deleted 分支)
- 状态传播:service.User / dto.User 新增 DeletedAt;SearchUsers 标记 deleted
- 前端:表格与余额弹窗展示"已删除"徽标、筛选下拉标注并排序、点击走
  include_deleted;新增 i18n admin.usage.userDeletedBadge
- 安全:普通用户 usage 仅查本人(无 PII 泄漏);主用户列表与默认 getById
  行为不变(已删用户仍 404);仅 admin 搜索设 IncludeDeleted

后端 build / 三态 vet / unit 全量 / 仓储集成全绿;前端 typecheck / vitest /
改动文件 eslint 全清。

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-30 17:52:18 +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
wucm667andClaude Opus 4.7 c9caadb378 fix(account): address second-round review on quota auto-pause
- TopK initial filter now drops quota-paused accounts: fold the quota check
  into isAccountRequestCompatible so session-hash, TopK pool, and per-candidate
  rechecks all skip paused accounts. Previously the candidate pool was built
  without the quota check, so paused accounts could fill TopK and leave the
  scheduler returning "no available accounts" even with healthy ones available.
- Add per-account explicit disable flags auto_pause_5h_disabled /
  auto_pause_7d_disabled with toggles in EditAccountModal. Without these,
  leaving the account threshold blank silently falls back to the global default,
  so admins could not exempt a single account once a global default existed.
  Disable is per-window: an account can opt out of 5h auto-pause while still
  honoring 7d. Schedule snapshot whitelist includes the new fields, i18n EN/ZH
  updated, threshold-hint text revised to explain "blank = global default".
- Move quota auto-pause settings off the request hot path: replace the per-repo
  TTL+singleflight sync DB read with a per-SettingService stale-while-revalidate
  in-memory snapshot. Get is non-blocking (atomic.Pointer load + async refresh
  on staleness); writes via UpdateOpsAdvancedSettings push directly into the
  cache through an injected sink; wire warms the cache at startup. Adds Warm
  (sync) for tests/init and SetOpenAIQuotaAutoPauseSettings (sink target).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-29 14:32:45 +08:00
wucm667andClaude Opus 4.8 8b7a822706 fix(account): address review on OpenAI quota auto-pause
- gate previous_response_id sticky path with quota auto-pause check at both
  the snapshot and DB-recheck stages (previously bypassed, #1)
- skip pausing when the usage window already reset to avoid a stale stuck-pause;
  carry codex_*_reset_at / reset_after_seconds / codex_usage_updated_at through
  the scheduler snapshot whitelist (#2)
- remove the incomplete limit mode; percentage threshold only (#3)
- add global default 5h/7d threshold inputs to the Ops settings dialog with
  validation and en/zh i18n (#4)
- downgrade account_auto_paused_by_quota log from Info to Debug; it fires
  per-candidate on the scheduling hot path (#5)

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-05-29 12:20:30 +08:00
wucm667 ead471d64b feat(account): 支持按 5h/7d 用量阈值自动暂停账号调度 2026-05-29 10:47:47 +08:00