This repository was archived by the owner on Nov 4, 2026. It is now read-only.
Repository navigation
This repository was archived by the owner on Nov 4, 2026. It is now read-only.
AiUsageIndicator：识别免费版新的 resetKind 'weekly' 与 resetsAt（cloud#1843 周额度的消费侧跟进） #7371
Description
背景（cloud#1843 / PR cloud#1852 的消费侧跟进）
免费版 AI 额度改为 7 天滚动窗口后，objectos 运行时的用量端点与 429 信封有两处形状增量：
/api/v1/ai/usage：resetKind新增值'weekly'（原只有'daily' | 'monthly'），并带resetsAt（ISO，= 窗口内最早一笔计入用量的时间 + 7 天，即用量开始滚出的时刻）。- 429
AI_DESIGN_QUOTA_EXHAUSTED的error.details：新增resetsAt（ISO）；免费用户resetsTonight=false。
cloud 侧 dev 核对：chatbotQuota.* 四个键与窗口无关（「明天会重置」那句一直是服务端下发的 message/messageEn，已在服务端改成带日期的文案）；需要改的是 AiUsageIndicator（packages/app-shell/src/layout/AiUsageIndicator.tsx），它按 resetKind 渲染「今晚重置 / 下个周期」，收到 'weekly' 会落到未知分支。
要求
AiUsageIndicator识别resetKind: 'weekly'：显示「N 天后重置」（由resetsAt算，≤1 天时显示小时），中英文都走@object-ui/i18n；D5：不显示 token 数。- 类型：usage 响应类型里
resetKind联合加'weekly'，resetsAt?: string；未知 kind 回落到不显示重置时间（而不是崩或显示错的话）。 - 测试：weekly + resetsAt 渲染；未知 kind 回落。
前提 re-check
gh pr diff 1852 --repo objectstack-ai/cloud | grep -n "weekly\|resetsAt"；objectui grep -rn "resetKind" packages/app-shell/src/layout/AiUsageIndicator.tsx。
Activity
hotlong commentedon Sep 2, 2026
Claim: PM loop round 3 (epic objectstack-ai/cloud#1841)
Session:
session_c5c0ce54-bb9c-478c-9e5b-cf44b80d4569Branch:
claude/issue-7371-usage-indicator-weeklyWorktree:
objectui-issue-7371Domain: domain:ui
File surface:
packages/app-shell/src/layout/AiUsageIndicator.tsx+ its tests, the usage response type,packages/i18n/**zh/en keys for the new sentence (stop on breach; explain in the report)Container & model: S, mode:subagent, model: sonnet (PM judgement)
Clause-②: no
Serial constraints cleared: objectui#7273 (merged) touched AiUsageIndicator — base on current main; #7256 (in flight) owns console home/logo, not this file; #7366 (enqueued) touches i18n packs — different keys, rebase before pushing
hotlong commentedon Sep 2, 2026
os-dev-report
{
"issue": 7371,
"status": "done",
"branch": "claude/issue-7371-usage-indicator-weekly",
"pr": "#7374",
"premise_still_valid": true,
"summary": "AiUsageIndicator (packages/app-shell/src/layout/AiUsageIndicator.tsx) now recognizes resetKind: 'weekly' from the free plan's rolling 7-day AI quota window (cloud PR #1852), rendering 'Resets in N days' (or N hours inside the final day) computed from resetsAt via new @object-ui/i18n keys (real plural family, base + _one/_other, all ten locale packs). AiUsageResetKind gained 'weekly'; resetsAt was already string | null on this side and needed no shape change. Contract-first: resetsAt is read verbatim, never guessed client-side, so a weekly meter with resetsAt: null and any unrecognized resetKind both render no reset line (fail-soft, no crash, no stale copy). Tests cover both weekly branches, the resetsAt: null fallback, and the unknown-resetKind fallback.",
"tests": "All from repo root against merge-base 20c04b2 (rebased after #7366/#7367 landed touching i18n), commit 1b7793e:
pnpm --filter '@object-ui/app-shell^...' buildexit 0;pnpm exec vitest run packages/app-shell/src/layout/ packages/i18n/ --maxWorkers=2-> 'Test Files 91 passed (91) / Tests 1152 passed (1152)';pnpm --filter @object-ui/app-shell type-checkexit 0;node scripts/check-i18n-call-site-keys.mjsexit 0 -> 'Every in-scope call-site key resolves against the en pack (2959 keys)...';node scripts/check-i18n-en-drift.mjsexit 0 -> '0 en value(s) changed (6 key(s) added...)';node scripts/check-i18n-dead-keys.mjs(report-only) exit 0, new keys absent from its dead-key report;node scripts/check-changeset-presence.mjsexit 0 -> '13 source file(s) of 2 released package(s) changed, and this change declares 1 changeset(s)';node scripts/check-changeset-no-major.mjsandnode scripts/check-control-bytes.mjsexit 0. No ablation applicable to this change.","mcp_calls": 0,
"open_questions": [],
"out_of_scope_findings": []
}
hotlong commentedon Sep 2, 2026
ACCEPT — reviewer of record: PM (epic objectstack-ai/cloud#1841), verified against GitHub.
Fixes #7371, 14 files:AiUsageIndicator.tsx(+30/−5),useAiUsage.tstype union, test (+64), the 10 i18n packs (a real plural family, en +11 / others +6), changeset — inside the surface.resetKind: 'weekly'renders "Resets in N days" (hours inside the final day) computed fromresetsAtonly — never guessed client-side;weeklywithresetsAt: nulland any unknown kind render no reset line (fail-soft); no token numbers. Contract matches cloud PR perf(studio): lazy-mount live block previews in the page designer canvas #1852 (UsageResetKind+resetsAt).type-checkexit 0; i18n call-site keys (2959) / en-drift (0 changed, 6 added) / dead-keys / changeset / control-bytes gates quoted OK; rebased onto main after fix(studio,chatbot,i18n): workbench labels over internal ids, localized property panel, one tool-card status #7366 landed on the same packs. No ablation — additive rendering branch with direct positive+negative tests; acceptable at S tier.Generated by Claude Code
hotlong commentedon Sep 2, 2026
os-dev-report
{
"issue": 7371,
"status": "done",
"branch": "claude/issue-7371-usage-indicator-weekly",
"pr": "#7374",
"premise_still_valid": true,
"summary": "Patch round on PM's report: CI's package-level
pnpm lint(unlike my earlier targeted run) hit react-hooks/purity —weeklyResetLabelcalledDate.now()directly in AiUsageIndicator's render body. Moved the clock read out of render: auseState<number | null>(null)populated by a mount effect (refreshed every 60s), with the render body only reading that state and doing pure day/hour math from resetsAt/now. now === null (the one frame before the effect fires) renders no reset line, same treatment as any other not-yet-known input. No behavior change otherwise; existing weekly tests pass unmodified since RTL flushes the mount effect before assertions run.","tests": "New head 3003f23 (on top of 1b7793e, same branch, fast-forward push, no force). All from repo root:
pnpm --filter @object-ui/app-shell lint-> before the fix reproduced CI's exact failure ('133:51 error Cannot call impure function during render' / react-hooks/purity, 1 error + 2876 pre-existing warnings, exit 1); after the fix -> exit 0, '0 errors, 2877 warnings' (warning count pre-existing/unrelated, not gated by --max-warnings).pnpm --filter '@object-ui/app-shell^...' buildexit 0.pnpm --filter @object-ui/app-shell type-check(tsc --noEmit && tsc -p tsconfig.test.json) exit 0.pnpm exec vitest run packages/app-shell/src/layout/ --maxWorkers=2-> 'Test Files 33 passed (33) / Tests 225 passed (225)', including the weekly-reset tests added in the prior round, unmodified. Re-rannode scripts/check-i18n-call-site-keys.mjs,node scripts/check-changeset-presence.mjs,node scripts/check-control-bytes.mjs— all exit 0, unaffected (this round touched only AiUsageIndicator.tsx, no i18n/changeset files). Remote CI on 3003f23 was still in_progress at report time (Lint/Type Check/Test shards running, everything else already green) — not polled further per the no-CI-polling rule; local reproduction of the exact CI failure signature going 1 error -> 0 is the evidence.","mcp_calls": 0,
"open_questions": [],
"out_of_scope_findings": []
}
hotlong commentedon Sep 2, 2026
Patch round accepted (head 3003f23): the clock read moved out of render into a mount effect refreshed every 60s, render does pure math from
resetsAt/state; the CIreact-hooks/puritysignature reproduced locally (1 error → 0), layout suite 33 files / 225 tests,type-checkclean. Enqueuing when every check on the new head is green. Generated by Claude Code