Repository navigation
chore(deps-dev): bump storybook from 8.6.15 to 8.6.17 - #772
Merged
github-actions[bot] merged 1 commit intoFeb 23, 2026
Merged
Conversation
Bumps [storybook](https://github.com/storybookjs/storybook/tree/HEAD/code/core) from 8.6.15 to 8.6.17. - [Release notes](https://github.com/storybookjs/storybook/releases) - [Changelog](https://github.com/storybookjs/storybook/blob/v8.6.17/CHANGELOG.md) - [Commits](https://github.com/storybookjs/storybook/commits/v8.6.17/code/core) --- updated-dependencies: - dependency-name: storybook dependency-version: 8.6.17 dependency-type: direct:development update-type: version-update:semver-patch ... Signed-off-by: dependabot[bot] <support@github.com>
Contributor
Author
LabelsThe following labels could not be found: Please fix the above issues or remove invalid values from |
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
Contributor
❌ Console Performance Budget
📦 Bundle Size Report
Size Limits
|
os-zhuang
added a commit
that referenced
this pull request
Jul 8, 2026
…st publish errors (cloud#771/#772) (#2351) Trust-polish batch from the 2026-07-09 magic-flow acceptance run (objectstack-ai/cloud#771, #772): - The chat surface now speaks the CONVERSATION's language: a Chinese thread under an English console got English canned approval messages ("Looks good — build it as proposed."), English progress labels and English starter chips. detectConversationLanguage() reads the latest user message (CJK → zh) and wins over the UI locale; falls back to the UI language for fresh threads. - Default agent tab prefers BUILD when the catalog offers it: the build surface is the one that resumes its in-progress conversation, so landing on the ask tab hid the user's build thread behind a tab switch ("my build chat disappeared" → they rebuild from scratch — the user-side path of cloud#769). Ask-only users are unaffected. - Build persona stops masquerading as a data-Q&A box: its empty-state pitch and starter chips are now build-flavoured (zh + en), instead of the ask persona's "query and analyze your data" copy with metadata chips. - Publish failure toast carries the real error: the handler threw `new Error(String(failedCount))`, so users saw "Publish failed — 2". Now it surfaces the first failed item's type/name/error (+N more) or the server error message. - The floating panel's publish success now pulses emitMetadataRefresh() (the full-page AI route already did), so the launcher/app switcher lists the new app WITHOUT a page reload — "open it from the launcher" was a lie until F5. - Re-hydrated tool-call-only turns no longer render their internal placeholder ("(called todo_write, propose_blueprint)") as prose; they collapse to a quiet localized activity note. Build green (turbo app-shell+plugin-chatbot), vitest 1242 passed. Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
os-zhuang
added a commit
that referenced
this pull request
Jul 14, 2026
…lier handoff card (#2458) (#2496) When the `ask` agent declines a build-shaped request, the ~20s wait before the "Open in Builder →" card is the LLM's time-to-tool-call. During it the chat could show dead air: a blank bubble, or the static "执行过程" activity note when the backend streamed a `(called …)` tool-call placeholder mid-stream. `ChatbotEnhanced` now renders the existing `ThinkingDots` indicator whenever a streaming assistant turn has nothing VISIBLE yet — whitespace-only content, a `(called …)` placeholder, or reasoning hidden in `summary` mode. The static "执行过程" note is kept only for FINISHED (re-hydrated) tool-call-only turns (#772 preserved). The `builderHandoff` card already renders at `output-available` with no prose gate, so it surfaces the instant the tool result lands; the typing cursor now only paints beside real streaming prose. Tests: card renders while streaming before prose; live dots (not 执行过程) for a mid-stream placeholder; 执行过程 note preserved for a finished placeholder turn. Co-authored-by: Jack Zhuang <277994282+os-zhuang@users.noreply.github.com> Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
This was referenced Aug 8, 2026
akarma-synetal
pushed a commit
to akarma-synetal/objectui
that referenced
this pull request
Aug 10, 2026
…3837) (objectstack-ai#3897) * fix(app-shell): 方案卡的 Building… 徽标随 UI 语言取值,不再被会话语言门控 (objectstack-ai#3837) `AiChatPage` 用 `convZh`(会话语言)门控四个字符串,因为 cloud 确认门 (`service-ai-studio` `confirm-gate.ts` `APPROVAL_RE`)只认中英,发进线程的 文本必须与线程本身同语言(objectstack-ai#772/objectstack-ai#2884)。该门上方注释的后半句写明了另一半规 则:button LABELS stay on the UI locale。 `planBuildingLabel`(objectstack-ai#2632 引入)落在了这句的错误一侧,中文会话恒取硬编码的 `正在搭建…`,两个后果都已实测: - **混合语言的卡片**:英文控制台里,中文会话的方案卡 `Proposed plan` / `Build it` / `Built` / `Not yet built` 全英文,中间夹一个中文徽标。objectstack-ai#2458 第 4 条记录的是同一个病的反方向。 - **翻译永远读不到**:中文会话恒走字面量,`console.ai.planBuilding` 的 zh 值对中文读者无效。objectstack-ai#3546 切片四(PR objectstack-ai#3839)刚把该 key 补进十包,只能做围 堵 —— 把 zh 值写成与字面量逐字节相同并把两者钉在一起。 徽标现在与相邻十二个标签一样读 `t('console.ai.planBuilding', …)`:十个包全 部可达(德文控制台 + 中文会话渲染 `Wird erstellt…`),zh 包成为该中文措辞的 唯一来源(值不变,中文读者看到的字符串与改动前完全一致)。 三条出站文本(`planApproveMessage` / `planApproveDefaultsMessage` / `changesConfirmMessage`)一行未动,仍随会话语言 —— 复核确认三者都交给 `onSendMessage`(`ChatbotEnhanced.tsx:1598`、`:2395`)、由门读取,正是 `convZh` 分支存在的那一类;全文件再无其它 `*Label`/`*Title` 被该门控住 (`convZh` 全部读点:定义 1 处 + 这三条)。 钉子: - `packages/app-shell/.../AiChatPage.planCardLocale.test.tsx`(新增)真渲染 `ChatPane` + 真 `I18nProvider` + 真 `isConversationZh`,把 `ChatbotEnhanced` 换成 props 记录器:en 控制台 + 中文会话 → `Building…` 且卡上 `*Label` 无一含中日韩字符;de 控制台 + 中文会话 → `Wird erstellt…` (三方语言证明是包在应答,而非两路三元);zh 控制台 → `正在搭建…`;三条 出站文本仍随会话。 - `packages/i18n/.../console-namespace-3546.test.tsx`:切片四埋的围堵钉按其 自述翻转 —— 从「字面量与 zh 包逐字节相同」改为钉住门已移除,并把不变量放 宽一步:`convZh` 只许门控出站 `*Message`,任何新读点都会红。 反向验证(方向先判后跑,与预判一致):还原三元 → 3 红 —— i18n 钉子按自定义 消息点名、en 控制台用例、de 控制台用例;zh 控制台用例与三条出站文本保持绿 (前者按设计逐字节相同,不具区分力,已在测试头注明)。 顺带:`conversationLanguage.ts` 模块头把 progress labels 也列进「随会话」, 该句在本改动后不再为真,收窄为「只治离开控制台发给 agent 的文本」。 写钉子时量到发送半边另有一个缺陷(zh 控制台 + 英文会话会把中文确认句发进英 文线程;`planAnswerMessage` 完全没有门控),不在本单范围,已立 objectstack-ai#3896。 测试:`packages/app-shell/src/console/ai` + `packages/i18n` 47 文件 654 通过; 两包 type-check 通过,全仓 `turbo run type-check` 78/78 通过;改动文件 eslint 零告警;`check-control-bytes` / `check-i18n-call-site-keys` / `check-i18n-en-drift` 均通过(零 en 值变更)。 Claude-Session: https://claude.ai/code/session_01GTRjn8xBqp75dk7kFupVRt Co-authored-by: Claude <noreply@anthropic.com> * test(app-shell): 钉子里的 agent 桩改用具名类型断言,不用 as never 评审可读性,零行为改动:`ChatPane` 的 agents 桩从 `as never` 改成 `as unknown as AgentDescriptor`(类型导入,运行期擦除;该模块在本文件里被 vi.mock,类型不受影响)。 Claude-Session: https://claude.ai/code/session_01GTRjn8xBqp75dk7kFupVRt Co-authored-by: Claude <noreply@anthropic.com> --------- Co-authored-by: Claude <noreply@anthropic.com>
akarma-synetal
pushed a commit
to akarma-synetal/objectui
that referenced
this pull request
Aug 10, 2026
…ectstack-ai#3921) objectstack-ai#772 / objectstack-ai#2884 立下的规则写在 `AiChatPage` 的门上:出站文本随**会话**语言,渲染 标签随 UI 语言。规则只实现了一半 —— 三条出站文本写作 `convZh ? '<中文字面量>' : t('console.ai.…')`,而 `t()` 读的是 **UI 包**,于是 「非中文」那一支答的是 UI 语言、不是英文;第四条 `planAnswerMessage` 连门都没过。 两个方向各一个已实测后果: - **zh 控制台 + 英文会话发出中文。** zh 包定义了全部四个 key,所以 `t()` 是命中 而非落到英文 defaultValue:点 Build it,进英文线程的是 `确认，开始搭建。`。cloud 确认门(`service-ai-studio` `confirm-gate.ts` `APPROVAL_RE`)两种语言都认,构建 照走,agent 随后把整条线程转成中文 —— 即 objectstack-ai#2884 的症状、触发方向相反。 - **一键回答芯片两个方向都发 UI 语言。** 完全未门控:中文会话里发出 `For "…", go with: …`(objectstack-ai#772 开头那句抱怨的字面重演),zh 控制台下的英文会话里 发出中文句子。 四个站点现在统一走一个取值器 `outboundAgentText.ts`:中文会话取 `zh` 包该 key, 其余会话取 `en` 包该 key,永不读 UI 包。三个控制台 AI 界面(`/ai` 页、chat dock、 Studio copilot)都挂同一个 `ChatPane`,一并生效。 直接读两个包而不用 `t(key, { lng })` 是刻意的:i18next 的答案取决于宿主 app 装了 哪些 bundle 以及 `fallbackLng`,而 zh 查询静默回落到 en 正是上面那个错语言 bug。 取值器内的兜底表按语言分,并被钉成与两个包逐字节一致 —— 万一某个包不再定义该 key, 中文会话兜底到中文,绝不兜到英文。 标签未动:仍随 UI 语言。objectstack-ai#3837 那条「`*Label` 漂回会话门就红」的钉子现在同时守住 反向 —— 出站 `*Message` 被重新塞回 `t()` 也会红。 Claude-Session: https://claude.ai/code/session_01GTRjn8xBqp75dk7kFupVRt Co-authored-by: Claude <noreply@anthropic.com>
akarma-synetal
pushed a commit
to akarma-synetal/objectui
that referenced
this pull request
Oct 7, 2026
…8 cloud E2E (six of seven) (objectstack-ai#10924) Refs objectstack-ai#10899 — six of its seven items land here; item 4 needs a maintainer decision and the replay-ordering half of item 5 is root-caused but not changed, so the card stays open. Part of objectstack-ai/cloud#2440 Clause-②: no One commit per item, each with its own test, ablation and changeset. Session `https://claude.ai/code/session_786273a0-0246-4bb2-b026-bb37ba5d7295` (dispatch branch `claude/issue-10899-hosted-console-ux-defects`). ## Verdict per item | # | Item | Verdict | |---|---|---| | 1 | Assistant FAB covers the dock's send button | **done** — the FAB renders only while the dock is collapsed (`ConsoleLayout`) | | 2 | Failed record comment shown as sent | **done** (objectui half of cloud#2431, per the ruling that cloud mounts audit on hosted tenants) — a comment/reply lands only after its `sys_comment` write resolves; a failed write shows a localized toast and the composer keeps the draft | | 3 | Dashboard 403 renders as a raw English exception | **done** — `queryDataset` throws a typed `AnalyticsForbiddenError` for 403 `PERMISSION_DENIED`; `DatasetWidget` renders a localized no-access state. The missing grant itself is cloud#2439 | | 4 | Assistant scope chip says 「新建应用」 inside an app | **open — needs a maintainer decision** (below). Not changed here | | 5 | Empty 「执行过程」 blocks; replay orders steps differently | **empty blocks: done**. **Ordering: root-caused, producer outside this change** (below). Not changed here | | 6 | Studio edit/design offered to a non-platform-admin | **done — the producer is the console**. The server's permission answer is already right; the console ignored it | | 7 | Marketplace labels an older version as an update | **done** — "update" means higher SemVer 2.0.0 precedence, at both call sites | ## What each commit does - **Item 7** — `versionPrecedence.ts` orders two versions by SemVer 2.0.0 §11 (the grammar the package manifest declares for `version`). `cloudUpdateAvailable` and the install dialog's per-environment `hasUpdate` both use `isNewerVersion` instead of `!==`. Non-SemVer or equal pairs offer no update. - **Item 1** — the FAB is the dock's collapsed affordance (ADR-0057 open question 2). `ConsoleLayout` now renders it only while `dock.expanded` is false, which is the rule `ChatDockLauncher` already follows. - **Item 6** — the page pencil (`PageView`) and the top-bar hammer (`AppHeader`) required only the workspace-admin role, and an org owner holds that role. Both now also require `manage_metadata` through `useCanAuthorMetadata`, which reads `GET /api/v1/auth/me/permissions`. HomePage's builder CTAs already use the same check. If the server reports no `systemPermissions`, the check fails open, so older servers keep the admin-only behaviour. - Measured producer: on the PM's rig control DB (read-only), the signed-up owner holds `cloud_control_plane_member` (no system permissions) and `organization_admin_no_bypass` = `manage_org_users, setup.access, setup.write`. Framework's `organization_admin` withholds `manage_metadata` on purpose. So the server's answer was correct and the console did not read it. - **Item 2** — `RecordDetailView` no longer adds optimistic rows. It awaits the `create`, then appends the row. On failure it shows `toast.error(detail.commentFailed)` (new key, all ten packs) and rejects back to the caller. - The three composers (`RecordActivityTimeline`, `ThreadedReplies`, `RecordComments`) now handle a rejected `onAddComment` / `onAddReply` explicitly: they keep the draft. Before this, the throw escaped as an unhandled rejection. - `RecordDetailView.feedRecordScope.test.tsx` had a case that asserted the opposite of the ruling ("keeps an unpersisted comment when the write never lands"). It is rewritten to the ruled behaviour. - **Item 3** — `queryDataset` gains a code-based branch for the analytics read admission's `403` + `PERMISSION_DENIED` (`readAdmissionDeniedError` in `@objectstack/service-analytics`). The new `AnalyticsForbiddenError` carries `httpStatus: 403` and `code: 'PERMISSION_DENIED'`. `DatasetWidget` classifies the error with the shared `classifyLoadError`, the same classifier the list view uses. For `forbidden` it renders `dashboard.widgetForbiddenTitle` / `widgetForbiddenMessage` (new, all ten packs). A 403 with no code keeps the generic error. - **Item 5 (empty blocks)** — a reloaded build hydrates one assistant message per persisted step, and each step's text is the `(called …)` placeholder. `ChatbotEnhanced` now shows the 执行过程 note only when the turn shows no tool row and no build or design panel. The copy/regenerate bar renders only for a turn with visible prose. The objectstack-ai#772 fallback, where a placeholder turn with nothing else keeps its note, is preserved and pinned. ## Item 4 — why it is not changed here The in-app dock runs the build agent on an app-less, product-only conversation scope (`ChatDock.tsx`, `chatConversationScope` with no `appId`). It sends no package binding. So in ADR-0057 Amendment A1's terms the chip is telling the truth: the thread is unbound, and "New app" is the unbound label. A1.b also records the dock keeping the legacy product-only key as a decision. Making the chip name the current app means either: - **A** — bind the dock's build thread to the current app's package (`editPackageId` + `app:PKG:build` scope). The agent then gets a declared write target, and the dock, `/ai/build?package=` and the Studio copilot share one thread. This reverses the A1.b decision. - **B** — relabel the unbound chip. The agent would still guess the target app. Recommendation: A, for the build product only, and only when the active app is an authorable package. Details are in the report on the card. It changes a decided ADR line, so it waits for the maintainer. ## Item 5 — replay ordering (measured, not changed) I read the rig tenant DB (`ai_messages`, read-only). The follow-up proposal turn (EFYj) was still persisting its last `todo_write` and final prose at 04:36:42–44Z. The user's 「确认，开始搭建。」 (turn dh2L) was persisted at 04:36:39Z. The conversation store orders by `created_at, id` across turns (cloud `packages/service-ai/src/conversation/objectql-conversation-service.ts`, `MESSAGE_ORDER`). The replay therefore puts the confirm message inside the previous turn's steps, while the live run showed that turn whole. Two things contribute: - the store orders by time, not by turn (cloud); - the plan card's build button stays clickable while its own turn is still streaming (`ChatbotEnhanced`). Both are named in the report. ## Tests All runs are on HEAD `d8969d4bc` unless noted. - New pins (one per item, each ablated — revert the fix → red, restore → `git diff HEAD` empty): - item 7: `versionPrecedence-10899.test.ts`, `MarketplacePackagePage.updateIsHigherVersion-10899.test.tsx` (ablation: 1 failed) - item 1: `ConsoleLayout.fabYieldsToOpenDock-10899.test.tsx` (1 failed) - item 6: `PageView.editInStudioNeedsAuthoring-10899.test.tsx`, `AppHeader.designInStudioNeedsAuthoring-10899.test.tsx` (2 failed) - item 2: `RecordDetailView.commentWriteFailure-10899.test.tsx` + the rewritten feed-scope case (3 failed). Composer half: `composerKeepsDraftOnRejection-10899.test.tsx` — with the `catch` removed, the run exits 1 on unhandled rejections while the draft assertions still pass. Keeping the draft used to be an accident of the throw. - item 3: `queryDataset.forbidden-10899.test.ts`, `DatasetWidget.forbiddenState-10899.test.tsx` (2 failed) - item 5: `ChatbotEnhanced.hydratedStepTurns-10899.test.tsx` (2 failed) - Package suites: one `vitest run` over `packages/app-shell/ packages/plugin-chatbot/ packages/plugin-dashboard/ packages/data-objectstack/ packages/plugin-detail/ packages/i18n/` — `Test Files 1409 passed | 2 skipped (1411)`, `Tests 14850 passed | 23 skipped (14873)`, exit 0 - Build: `turbo run build` for the six touched packages plus the `@object-ui/app-shell^...` closure — 29/29 successful. - Type-check: `turbo run type-check` for the six packages — 6/6 successful. - `check:i18n-keys`, `check:i18n-drift`, `check:i18n-dead-keys`, `check:new-line-citations` (0 new), `check:control-bytes`, `check-changeset-presence` (6 changesets for 6 packages), `check-changeset-no-major` — all exit 0. - Lint, narrowed to this diff: `eslint --no-inline-config` over the 33 changed `.ts`/`.tsx` files. For every pre-existing file, the error/warning counts at HEAD equal its counts at the merge base (the only errors are 4 pre-existing `no-console` in `data-objectstack/src/index.ts`, 4/127 → 4/127). New files have 0 errors. Type-aware linting is not enabled, so this diff cannot change the verdict on any untouched file. The repo-wide lint belongs to CI. ## Acceptance notes -⚠️ The shared verify lock was **not taken**: the wrapper reported `UNLOCKED (declared) · no usable flock on this host`. Build, type-check and suites ran unserialized. - NOT MEASURED: an in-browser check of the six fixes. The PM's rig was observed through its screenshots and read-only DB reads only — no sign-in, no writes. - Out of scope, noted and not filed: - `handleToggleReaction` still swallows a failed reaction write (same optimistic shape as item 2; no carrier). - The reply input (`ThreadedReplies`) renders only under an item with `replyCount > 0`, and no feed read sets that, so a loaded discussion offers no reply affordance. - Other admin-role-only authoring affordances (view add/rename/delete in `ObjectView`, the report editor, `ObjectDataPage`) do not read `manage_metadata`. Same class as item 6, not measured on the control plane. - `ObjectChart`'s dataset path, `DatasetReportRenderer` and `DashboardFilterBar` still print the (now typed, English) refusal message instead of a localized state. - Changesets: six, all `patch`. --- _Generated by [Claude Code](https://claude.ai/code/session_786273a0-0246-4bb2-b026-bb37ba5d7295)_ --------- Co-authored-by: Claude <noreply@anthropic.com>
This branch had an error being deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Bumps storybook from 8.6.15 to 8.6.17.
Release notes
Sourced from storybook's releases.
Changelog
Sourced from storybook's changelog.
... (truncated)
Commits
c6e550aBump version from "8.6.16" to "8.6.17" [skip ci]9cf9d89Core: Require token for websocket connections7e51515Bump version from "8.6.15" to "8.6.16" [skip ci]Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting
@dependabot rebase.Dependabot commands and options
You can trigger Dependabot actions by commenting on this PR:
@dependabot rebasewill rebase this PR@dependabot recreatewill recreate this PR, overwriting any edits that have been made to it@dependabot show <dependency name> ignore conditionswill show all of the ignore conditions of the specified dependency@dependabot ignore this major versionwill close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)@dependabot ignore this minor versionwill close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)@dependabot ignore this dependencywill close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)