Console field designer still offers an "Indexed" toggle for a key FieldSchema rejects by name

[yinlianghui]

Found while measuring the Field indexing claim on reference/performance-and-limits.mdx (#1119, PR #1121). Filed unassigned as an observation — nothing in HotCRM hits it today, because this repo authors every field in src/objects/*.object.ts and never through the designer.

The measurement

@objectstack/console 17.0.0-rc.6 renders a field-designer toggle labelled Indexed (designer.field.indexed, translated in the shipped locale bundles) whose commit handler writes { indexed: value || undefined } onto the field:

(0,Y.jsx)(q,{label:s(`designer.field.indexed`),value:!!g.indexed,
  onCommit:e=>b({indexed:e||void 0}),disabled:a})

FieldSchema in @objectstack/spec 17 rejects that key by name:

FieldSchema.safeParse({ name:'foo', label:'Foo', type:'text', indexed:true })
→ unrecognized_keys: "Unrecognized key(s) on this field: `indexed`.
     • never a FieldSchema key; a field-level index flag built no index (#2377).
       Declare the index in the object's `indexes[]`. Until #4001 closed this shape
       these were dropped silently — the field was still created, minus whatever the
       key was meant to constrain, protect or compute."

So the designer offers a control whose value cannot survive a parse. FieldSchema has no index / indexed key at all; the real surface is the object-level indexes: [{ name, fields, unique }] block, materialised by SqlDriver.syncDeclaredIndexes.

Why it is worth recording

This is the #2377 / #4001 family seen from the other side. The spec closed the silent-drop hole and now refuses the key loudly — but the UI that produces it was not retired with it, so an admin who ticks Indexed in Studio either gets a save rejected for a key they did not know was retired, or (if some path still strips unknown keys before parsing) a toggle that stays ticked and builds no index. Either way the control promises an index nobody creates, which is exactly what #2377 removed the field-level flag for.

Where it lives

Platform-side, in @objectstack/console — not something HotCRM can fix in src/. Recording it here because HotCRM is where it was measured and HotCRM admins are who would meet it; it wants forwarding to the platform repo rather than a change in this one.

Refs #1119 · PR #1121 · #2377 · #4001 · #5248 (the sibling removals of indexes[].type / indexes[].partial)

---

Generated by Claude Code

[hotlong]

GA re-test on 17.0.0 — REPRODUCES, and the round trip is now measured end to end · stays OPEN

Version — @objectstack/console 17.0.0 GA and @objectstack/spec 17.0.0 GA (this card measured rc.6). hotcrm @ d4ddee0.
Probe — both halves: the shipped console asset re-read statically, then the full round trip driven in a real browser (Chromium 141 / Playwright) against a booted GA app on port 4003.

1. Both ends of the contradiction survive, unchanged

The toggle is still shipped. dist/assets/metadata-admin-DqAvOmNf.js in console 17.0.0 GA, byte-identical in shape to the rc.6 reading in this card's body:

label:c(`designer.field.indexed`),value:!!C.indexed,onCommit:e=>D({indexed:e||void 0}),disabled:o

The spec still refuses the key. Run against the installed 17.0.0 @objectstack/spec:

FieldSchema.safeParse({ name:'foo', label:'Foo', type:'text', indexed:true })
  → success = false
  → code = unrecognized_keys
  → "Unrecognized key(s) on this field: `indexed`.
     • never a FieldSchema key; a field-level index flag built no index (#2377) …"

control — identical field WITHOUT `indexed` → success = true

The control line is the point: the rejection is caused by indexed alone, not by anything else in the fixture.

2. The round trip — which of the card's two predicted branches actually happens

This card left it open: "an admin who ticks Indexed either gets a save rejected for a key they did not know was retired, or a toggle that stays ticked and builds no index." Measured on GA: the first branch.

HotCRM itself cannot exercise it — a code-loaded package is Read-only in Studio and its field designer offers no editable properties panel at all. So I created a throwaway writable package (com.probe.p1154, namespace probe) with one object, purely to reach an editable designer. Steps and results:

1. Studio → Data → object → Form · Layout → click a field → Field properties panel opens.
2. The ADVANCED section renders exactly eight controls — Read-only, Hidden, Indexed, External ID, Track history (plus Required, Unique, Allow quick-create above). Indexed is offered on a plain field, unchecked, fully enabled.
3. Tick it → aria-checked flips false → true.
4. Save draft →

PUT /api/v1/meta/object/probe_widget?mode=draft&package=com.probe.p1154
→ HTTP 422
{"error":"[invalid_metadata] object/probe_widget failed spec validation:
  fields.owner_id: Unrecognized key(s) on this field: `indexed`.
   • never a FieldSchema key; a field-level index flag built no index (#2377).
     Declare the index in the object's `indexes[]`. …",
 "code":"INVALID_METADATA",
 "issues":[{"path":"fields.owner_id","message":"Unrecognized key(s) on this field: `indexed`. …"}]}

5. The draft does not save. The header still reads Unpublished draft, the sub-header flips to Draft layout — your unsaved changes, and the toggle stays ticked.

A correction to my own first reading — the failure is NOT silent

My scripted pass collected the toast elements and got [""], which reads like a silent rejection. The screenshot says otherwise and the screenshot is right: the console renders a red inline banner above the layout carrying the full spec text —

• fields.owner_id — Unrecognized key(s) on this field: indexed.
• never a FieldSchema key; a field-level index flag built no index (#2377). Declare the index in the object's indexes[]. Until #4001 closed this shape these were dropped silently — the field was still created, minus whatever the key was meant to constrain, protect or compute.

Recording this because "the toast was empty" would have been a wrong and more alarming claim than the truth. The rejection is loud and well-worded; #4001's work is visibly holding.

Why the card still stands

The defect is not the error message — it is that the designer offers a control whose value can never be saved. On GA an admin can tick Indexed, and the only possible outcome is a hard-blocked save on a key the UI itself just wrote. Every save of that object is blocked until the toggle is found and un-ticked. The control promises an index nobody creates, which is precisely what #2377 removed the field-level flag for; the UI simply was not retired with it.

Screenshots — 1122-field-props-panel.png (ADVANCED section, Indexed offered), 1122-indexed-ticked.png (ticked), 1122-save-422.png (red banner + still-ticked Indexed + your unsaved changes).

Routing — the fix is in the console field designer (metadata-admin bundle), so the mirror goes to objectui, not objectstack. Filed as a follow-up comment below. Throwaway package and probe scaffolding deleted; no product code changed.

Probed under #1154 (parent #1150).

---

Generated by Claude Code

[hotlong]

Mirror filed: objectstack-ai/objectui#4644 — "console: field designer still offers an Indexed toggle for a key FieldSchema rejects by name — ticking it hard-blocks the save".

Repo choice: objectui, not objectstack. The fix is in the console field designer (metadata-admin bundle), which lives in objectui; objectstack/packages/console is a publish wrapper with no source. This also follows the maintainer's file-at-destination ruling (objectstack-ai/objectstack#7167, 2026-08-10), under which objectstack#5084 was migrated to objectui#4024 rather than worked in place, and matches where every prior console-chrome fix in this family actually landed (objectui PR #3379, PR #4602).

Duplicate check before filing: searched objectui open+closed for the designer/indexed/FieldSchema cluster — nearest hits were unrelated (#3951 grid column key spellings, #2324/#2322/#2325 missing createSchema). No existing card.

Nominated for target:v17 on objectstack-ai/objectstack#8668 (Seat B — console). Not #8667: that epic's scope section explicitly assigns repo:objectui cards to the Seat B sibling, so nominating a console card there would be the wrong desk. No labels applied by me — the maintainer approves nominations.

---

Generated by Claude Code

[hotlong]

Closed — moved upstream to objectstack-ai/objectui#4644

Dispositioned under the transfer sweep #1156 (part 2). Maintainer ruling, 2026-08-14, quoted verbatim and untranslated:

hotcrm 席位的原则是用平台的能力做元数据应用的开发，平台的需求应该转给平台

and for the hotcrm side of a transferred card, the same ruling: 「关掉,只留指向镜像的指针」. Authority: #1156.

The mirror — verified before this close, not assumed

objectstack-ai/objectui#4644 — "console: field designer still offers an 'Indexed' toggle for a key FieldSchema rejects by name — ticking it hard-blocks the save". Confirmed open, labelled bug, and carrying this card's defect: the same shipped-asset line, the same unrecognized_keys rejection with its control, and the same browser round trip ending in the 422.

Note the mirror is in objectui, not objectstack — the fix is in the console field designer (metadata-admin bundle), which lives in objectui; objectstack/packages/console is a publish wrapper with no source. That follows the maintainer's file-at-destination ruling (objectstack#7167, 2026-08-10) and matches where every prior console fix in this family actually landed (objectui PR #3379, PR #4602).

The reading this close rests on

The GA re-test already on this card — comment of 2026-08-14, both halves: the shipped console asset re-read statically, then the full round trip driven in a real browser against a booted GA app.

Both ends of the contradiction survive. The designer still ships the Indexed toggle, fully enabled on a plain field; FieldSchema still refuses indexed by name, with a control showing the identical field parses without it. This card had left two branches open — a rejected save, or a toggle that sticks and builds no index — and the round trip settles it as the first: PUT … ?mode=draft answers 422 INVALID_METADATA, the draft does not save, and the toggle stays ticked.

That reading also carries a correction the probe made against itself: a scripted pass collected empty toast text and would have reported a silent rejection; the screenshot showed a red inline banner carrying the full spec guidance. The rejection is loud and well-worded, and #4001's work is visibly holding. The defect is narrower and truer than "silent" — the designer offers a control whose value can never be persisted, and every save of that object stays blocked until the user finds and un-ticks it.

Why it closes here

The remedy is to retire the toggle from the field designer (and, if an index-authoring affordance is wanted, put it on the object, writing indexes[]). That is console source. HotCRM cannot even reach the control in its own packages — a code-loaded package is read-only in Studio, which is why the probe had to create a throwaway writable package to observe it at all.

Nothing is lost by this close: objectui#4644 is open, carries both ends of the contradiction and the round trip, and is readable without hotcrm context.

---

Generated by Claude Code

---

Generated by Claude Code