Skip to content

Commit b15352e

Browse files
committed
Merge remote-tracking branch 'origin/main' into claude/project-thread-v9mp90
2 parents 54efb0b + f7ff693 commit b15352e

58 files changed

Lines changed: 2801 additions & 102 deletions

File tree

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.
Lines changed: 17 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,17 @@
1+
---
2+
'@object-ui/plugin-chatbot': patch
3+
'@object-ui/app-shell': patch
4+
'@object-ui/i18n': patch
5+
---
6+
7+
fix(plugin-chatbot): a plan that overlaps an existing app asks whether to extend it or build a new one, instead of offering Build it
8+
9+
Clause-②: yes (widening)
10+
Metadata-protocol: none
11+
12+
When `propose_blueprint` finds that the plan overlaps an app the user already has, its result carries `decision: 'iterate_or_new'` and the overlapping app in `existingAppOverlap`. The chat used to draw that result as an ordinary plan card with a "Build it" button, so one click could build a duplicate of the existing app.
13+
14+
- **The question.** `detectProposedPlan` lifts the decision as `ProposedPlan.iterateOrNew`, with the app's `name` and `label` from `existingAppOverlap`. The plan card keeps its rows and, in place of "Build it" / "Adjust", asks whether to keep improving that app or build a separate one. Each of the two choices sends its answer as a chat message. A result that names no object still gets the question, never the fallback "Build it" card; an overlap that names no app gets the question without a name and a reply hint.
15+
- **The answers follow the conversation.** `ChatbotEnhanced` takes two new optional props, `planExtendExistingMessage(app)` and `planBuildNewAppMessage`. The console resolves both by the conversation's language, like the plan card's other outbound messages (`console.ai.planExtendExistingMessage` and `console.ai.planBuildNewAppMessage`, in the `en` and `zh` packs only). Neither reads as approval to the confirm gate.
16+
- **After a reload.** Server history re-parses the stored result. The message cache now writes `decision` and `existingAppOverlap` back into its compact plan envelope, so a cache-fallback reload asks the question too.
17+
- **Labels.** `chatbot.plan.iterateOrNew.*` (question, choices, reply hint) in all ten packs.
Lines changed: 12 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,12 @@
1+
---
2+
'@object-ui/app-shell': patch
3+
---
4+
5+
fix(app-shell): a server error at an RLS clause's path holds the permission-set autosave
6+
7+
Clause-②: no
8+
Metadata-protocol: none
9+
10+
In Studio's permission-set editor, an RLS `using` or `check` clause that the publish will refuse (for example a `using` that reads a formula field, refused as `rls-predicate-unenforceable`) showed the server's message on the field but kept autosaving: the message was displayed, and the save gate never counted it. Now the server's errors at the clause's own path hold the autosave, the same way a CEL syntax error does, while the clause still shows the text that verdict judged.
11+
12+
The verdict describes the stored draft and nothing else, and no door judges a text without saving it. So an edit that changes the clause's text is not held: it saves, and that save's re-read brings the next verdict, which holds or releases from there. The editor decides nothing itself: it takes the server's errors as served and restates no publish rule.
Lines changed: 12 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,12 @@
1+
---
2+
'@object-ui/app-shell': patch
3+
---
4+
5+
fix(console): a flow-backed api action shows the run's refusal, not the engine's node chain
6+
7+
Clause-②: no
8+
Metadata-protocol: none
9+
10+
A `type: 'api'` action whose target is a declared `type: 'flow'` endpoint gets the trigger door's `400 FLOW_FAILED` when the run fails. The record page's action menu and the list and toolbar actions showed that envelope's message, so a member refused on the flow's write read text like `Node 'recalc_each' failed: map … subflow … update_record(showcase_task) failed: …`. Both api handlers now read the answer by the rule a screen flow already uses: the flow's own `errorMessage`, then the producer-marked `userMessage`, then the refusal the write met (`innerRefusal`), rendered as the record form renders it (a permission denial reads the form's permission line in the console's language), and only then the message. Any other error answer reads as before.
11+
12+
A failure inside a `map` item or a `subflow` reaches this door with no `innerRefusal` until the engine carries it out of those nodes, so that chain still shows the message.
Lines changed: 14 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,14 @@
1+
---
2+
'@objectstack/objectql': patch
3+
---
4+
5+
fix(objectql): an artifact hot reload refreshes every registry entry the boot registered, so an edited email template is what the next send renders
6+
7+
Clause-②: no
8+
Metadata-protocol: none
9+
10+
On the dev loop (`os dev`, artifact watcher on), editing a declared item and letting the artifact reload used to refresh only the by-name read. The list read (`GET /api/v1/meta/<type>`) and its execution face (`getMetaItemsForExecution`) kept the boot copy until a cold restart, because the `metadata:reloaded` hook re-ingested objects alone and those reads lay the SchemaRegistry over the metadata service. The execution face is what the email plugin renders a send from and what the automation engine re-binds flows from on the same reload, so recipients kept getting the old wording and an edited flow kept its old body.
11+
12+
Now the hook re-registers each reloaded package body the registry already holds, through `registerApp` (the door the boot registered it through), under the source it was installed with. Every collection that door decomposes (email templates, flows, views, dashboards, pages, actions and the rest) holds the reloaded item, and the package's lifecycle is kept: a disabled package stays disabled. A package the boot never registered into the registry is left to the artifact door, as before. Items removed from the artifact stay registered until restart, as objects already do. A package that fails to re-register is named in a `warn` line and keeps serving what it served before.
13+
14+
Cold boots and `PUT /api/v1/meta` saves are unchanged.
Lines changed: 14 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,14 @@
1+
---
2+
'@object-ui/app-shell': patch
3+
---
4+
5+
fix(app-shell): the Studio API tab's example body leaves out `owner_id` and optional reference fields, so the copied Create and Update calls are accepted
6+
7+
Clause-②: no
8+
Metadata-protocol: none
9+
10+
The API tab of an object in Studio showed an example body that put the id `"related_record_id"` into every reference field, and added the platform's `owner_id` whenever there was room. That id names no record, so the copied Create (`POST`) and Update (`PATCH`) cURL were refused `400` with `reference_not_found` on nearly every object.
11+
12+
- **Left out of the body:** every optional reference field (`lookup`, `master_detail`, `user`, `tree`) and every field the platform adds (`system: true`, such as `owner_id`, which the platform fills in when it is left out). The note under the body names them with the other optional fields not shown.
13+
- **A required reference** is still in the body, because a create without it is refused. Its value is now a placeholder in the same form as the cURL's `<token>`, naming the object whose record id goes there, such as `<id of a showcase_project record>`.
14+
- **Unchanged:** the other fields, their sample values, the order (required fields first) and the cap of eight fields.
Lines changed: 12 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,12 @@
1+
---
2+
'@object-ui/console': patch
3+
---
4+
5+
fix(console): after a language switch, date and number cells format in the new language, not until the next page load
6+
7+
Clause-②: no
8+
Metadata-protocol: none
9+
10+
Switching the language from the avatar menu changes the UI language first and writes the signed-in user's `sys_user.locale` second. The switch remounts `LocalizationFetchProvider`, whose read of `GET /api/v1/auth/me/localization` could reach the server before that write committed and answer the previous language (or, for a user with no stored language yet, the browser's `Accept-Language`). `useDisplayLocale` ranks that answer above the UI language, so under 中文 a list grid's date cells kept reading `Oct 11` beside Chinese headers until a full page load, while a record page opened by URL read `2026年10月11日`.
11+
12+
The provider now reads again on every write to the signed-in user's own `sys_user` row that reaches the data-invalidation bus (the language switch's write and the profile page's). The re-read starts after the write has settled and opts out of joining a request still in flight from before it, so its answer is the one that lands. The single date formatter (`formatDate` through `useDisplayLocale`) is unchanged.
Lines changed: 19 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,19 @@
1+
---
2+
'@objectstack/rest': patch
3+
'@objectstack/metadata-protocol': patch
4+
---
5+
6+
fix(rest): `GET /meta/diagnostics` refuses an unknown type, severity or package instead of answering an all-clear, and keys an alias's stats by its canonical type (#27514)
7+
8+
Clause-②: no (narrowing — owes one contract-review-tier review before the queue)
9+
Metadata-protocol: none
10+
11+
<!-- adr-0087: not-required (no-migration-prescription) The narrowed face is a REQUEST to one read door: query values that were never honoured (an unknown type, severity or package) are now refused instead of answered with an empty sweep, and an alias's stats move to the canonical key the sweep already listed it under. No stored metadata, package source or object definition holds a diagnostics request or response, no schema key or export changes, and the remedy is the caller's own query string, so there is nothing for `objectstack migrate meta` to rewrite. -->
12+
13+
`/meta/diagnostics` is the door an author, an agent or CI reads to learn whether metadata is healthy, and three query values it could not honour were answered `200` with zero issues.
14+
15+
- **`?type=`** naming no metadata type (`zzz_unknown`) is refused with `400 INVALID_REQUEST`, by the same verdict `GET /meta/:type` uses: the platform's static type spelling, then the types this deployment carries live. A near-miss spelling of a declared type (`viewes`) keeps its existing refusal, which names the spellings it was reaching for.
16+
- **`?severity=`** other than `error` or `warning` is refused with `400 INVALID_REQUEST`. An empty value still means the default, `error`.
17+
- **`?package=`** naming a package that is not installed, and that no swept item is bound to, is refused with `400 INVALID_REQUEST`. An installed package with no items still answers its zeros, items bound to an uninstalled package are still swept, and `package=all` still sweeps every package.
18+
- **An alias** (`?type=views`) is swept and keyed in `stats` and `entries` under its canonical type (`view`), so it answers exactly what `?type=view` answers.
19+
- **One envelope.** Every refusal of this route now answers the nested `{ error: { code, message } }` the repeated-parameter refusal already used. The near-miss type refusal used to answer `{ error: '…', code }`.

‎content/docs/frontend/plugins/plugin-chatbot.mdx‎

Lines changed: 7 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -496,6 +496,13 @@ results declare. The plan card (`propose_blueprint`) reads the result's
496496
dashed chip for a field the agent added on its own, the record's name field,
497497
one line per view, the assumptions, the sample data the build will stage, and
498498
how the build will land (`posture`: live at once, or saved as a draft).
499+
A plan whose result says `decision: 'iterate_or_new'` overlaps an app the user
500+
already has (`existingAppOverlap`, by `name` and `label`). Its card draws no
501+
"Build it" button: in its place it asks whether to keep improving that app or
502+
build a separate one, naming the app, and each choice sends its answer as a
503+
chat message (`planExtendExistingMessage(app)` and `planBuildNewAppMessage`,
504+
which the console picks by the conversation's language). The question survives
505+
a reload from server history and from the console's message cache.
499506
The result card (`apply_blueprint`, then a later `verify_build` in the same
500507
turn) reads `checklist`: a passed/total chip, the verified lines in two
501508
columns, and the app and package that "Open app" and "View in Studio" open.

0 commit comments

Comments
 (0)