fix(service-analytics): translate select dimension option labels on dataset charts

[os-trump]

Fixes #16773

Clause-②: yes

DimensionLabelDeps is re-exported wholesale from packages/services/service-analytics/src/index.ts,
so translateSelectOptions is a new key on a published exported type — the wire shape and
the accept/refuse behaviour are otherwise unchanged, and a host wiring no i18n service falls
back to today's authored-label behaviour, but the published-type surface itself grew and that
is the mechanical floor for this clause regardless of the key being optional. Corrected from
this PR's own provisional no — see "Clause-② correction" below for the measurement.

Reproduction

Re-driven at code level (no browser/hotclm access from this seat): a unit-level reproduction
against resolveDimensionLabels on origin/main confirmed both halves of the correlation:

• Same-object select dimension (field: 'direction'): the raw stored value (purchase) is
  unconditionally overwritten with the field's AUTHORED options[].label (Purchase) —
  locale-blind, every time.
• Dotted cross-object dimension (field: 'contract.direction'): resolveDimensionLabels looks
  up fields[dim.field] against the BASE object's own field map, keyed by plain field names —
  a dotted relationship path never matches a key there, so meta is undefined and the whole
  per-dimension body (if (!meta) continue) is skipped, on both strategies (the field name
  reaching this function is the dataset-authored string verbatim, unaffected by which strategy
  resolved the query). The row's raw value passes through completely untouched.

This is measured, not assumed — see the ablation below and the new
select option i18n (#16773) describe block in dimension-labels.test.ts, including a
dedicated control test for the dotted-field-name mismatch.

Which path the dotted arm takes

No path inside this package. The dotted arm's translated rendering (per the card) is not
produced by anything in service-analytics — this file's select AND lookup branches both
require a fields[dim.field] hit, which a dotted field name structurally cannot produce
against the base object's field map. Whatever renders 采购/销售 for the dotted arm does so
downstream of this package, off the untouched raw machine value this file leaves behind.

Routing vs. teaching

Routing the same-object arm through "leave it raw, like the dotted arm" was rejected: this
file's own docblock states the select/lookup resolution exists because raw values are not
human-readable at all (not merely un-translated), and removing it would regress every
locale — including the plain default/English case, and any consumer that isn't a smart
client capable of its own value→label mapping. That is a wire-contract downgrade with no
verifiable safety net (objectui is a separate repo, not available from this seat to confirm),
well past "no behaviour change beyond label resolution."

Instead this PR routes the same-object arm through the repo's one existing "translate a
select option label" implementation — translateObject (@objectstack/spec/system), the
same function GET /meta/object/:name already calls, which is where the console's list grid
gets its translated option labels. DimensionLabelDeps gains one new optional capability,
translateSelectOptions(objectName, fieldName, options, locale), implemented in plugin.ts by
building a TranslationBundle from the registered i18n service (mirroring
RestServer.buildTranslationBundle, packages/rest) and calling translateObject on a
one-field ObjectLike doc. No new spec export, no new wire key: translateObject,
ObjectLike, ResolveOptions and TranslationBundle are all already-published
@objectstack/spec/system exports (confirmed against packages/spec/api-surface/system.json).
A kernel with no i18n service registered — or nothing for the requested locale — degrades to
exactly today's authored-label behaviour (regression tests pin this fallback).

The dotted arm is untouched by construction: it never reaches the modified branch (if (!meta) continue fires first), not merely by intent — pinned by
a dotted cross-object field name never matches the base object field map — left untouched, translateSelectOptions never consulted in dimension-labels.test.ts.

How #16390's future LOOKUP_TYPES widening inherits this

Not automatically, and that is stated plainly rather than left for the next seat to find:
lookup/master_detail labels resolve through the separate fetchRecordLabels capability (a
related RECORD's display name, read live off the referenced object), which this PR does not
touch. translateSelectOptions only ever applies to a field's authored options[] — a
different bundle address (objects dot OBJECT dot fields dot FIELD dot options dot
VALUE) than a record's display name has no translation bundle entry at all today.

What #16390 does inherit for free: the i18n service bridge this PR adds to plugin.ts
(i18nService() / buildTranslationBundle(), both private helpers scoped to this file) is
already wired to ctx.getService('i18n'). Adding translated lookup/master_detail labels later
is a ctx.getService('i18n') away rather than a fresh integration — but it is not free today,
and #16390's card should say so when it lands.

Ablation (mutation reached disk, proven; restored, proven)

Predicted BEFORE running: stripping the deps.translateSelectOptions?.(...) consultation out
of the select branch turns exactly 2 tests red (the two asserting a translated result) and
leaves all other tests — including both explicit fallback tests and the dotted-arm control —
green, since the mutation only removes a conditional read that those tests don't exercise.

HEAD blob: 497953a21bbfffb6562fbbd5ed41642eab39d3fd
translateSelectOptions occurrences BEFORE: 4
mutation applied
working-tree blob AFTER mutation: 528ce947f554175ed25349cc4c2447c13178d580   (differs from HEAD blob)
translateSelectOptions occurrences AFTER: 3 (expect 3)

 Test Files  1 failed (1)
      Tests  2 failed | 25 passed (27)
 FAIL  ... > routes a select dimension through translateSelectOptions ...
 FAIL  ... > #16773 — a same-object select dimension renders the LOCALIZED option label end to end ...

== RESTORING ... from HEAD ==
restored blob: 497953a21bbfffb6562fbbd5ed41642eab39d3fd   (matches HEAD blob)
RESTORE OK: blob matches HEAD
git diff HEAD -- $ABSPATH (must be empty):
(end of diff)

Measured exactly the predicted 2 red / 25 green. Restoration proven by blob equality AND an
empty git diff HEAD, both under a trap ... EXIT INT TERM, never by reading an exit code.
Full suite re-confirmed green after restore (98 files / 2195 tests, post-merge).

Test resolution path

packages/services/service-analytics/vitest.config.ts declares only
disableConsoleIntercept: true — no resolve.alias, no test.projects[], no root-level
vitest.workspace.ts. Test files import the fix by relative path (../dimension-labels.js),
which Vite/vitest transforms straight from src/*.ts on the fly. Resolution is through src,
not dist — confirmed by absence of the aliasing shape a sibling delivery found elsewhere
today (packages/qa/dogfood's inert top-level resolve.alias), not assumed.

Clause-② correction

This PR's own claim carried a provisional Clause-②: no, reasoned from its own fence text
("a new exported symbol, a new key on a published payload, or any packages/spec/src/**
path") — none of which translateSelectOptions is: it is a new MEMBER on an EXISTING exported
symbol, not a wire payload key, and touches no packages/spec path. That reading missed a
broader rule this repo already ships under (#16778): a new key on a published exported
TYPE is the clause-② floor on its own, optionality included, because DimensionLabelDeps
is re-exported wholesale from index.ts (confirmed: export type { DimensionLabelDeps, ... } from './dimension-labels.js') — so any downstream package implementing it now sees a wider
published shape, whether or not the wire ever carries it.

Measured, not argued: node scripts/pm/check-widening-tells.mjs --declaration no --diff - on
this PR's full diff exits 0 ("no widening tell on any declared surface") — the script's
T1/T2 tells are scoped to packages/spec/src/** and its T3 tell to
packages/spec/api-surface/**, so a service-analytics interface member is outside every
surface it mechanically checks; this is a known, accepted gap in the tool (documented false
negatives), not evidence against the yes reading. #16778's own check-widening-tells --declaration no exit-4 came from an unrelated T2 hit on packages/spec/src/migrations/registry.ts
in that same diff — its DatasetCompileOptions.declaredFieldType precedent was argued by hand
in that PR's body too, under the identical rule, never caught by this script. Declaring yes
here is the same hand-argued rule applied consistently, not a mechanically-forced outcome.

Scope discipline

• The dotted/cross-object arm is unchanged — see "routing vs. teaching" above; a dedicated
  control test pins it.
• No behaviour change beyond label resolution: AnalyticsResult's wire shape (rows,
  fields[]) is unchanged; only a select dimension's rendered VALUE for a row can differ, and
  only when an i18n service is registered.
• No packages/spec/src/** change, no new payload key, no wire-shape change — but the
  published DimensionLabelDeps type does grow by one optional key, which is what flips
  Clause-② to yes (see above).
• content/docs/releases/** untouched.

Changeset

minor on @objectstack/service-analytics — additive, backward-compatible (no removed or
renamed key, no wire-shape change), but the published DimensionLabelDeps type gained a key,
which this repo grades at least minor regardless of the new key being optional (#16778).
Not major: nothing an existing implementor of DimensionLabelDeps wrote stops compiling or
behaving as before. No ADR-0087 disposition: the changeset declares no breaking change.

Gate reconciliation

node scripts/pm/dispatch-gates.mjs --commands --repo objectstack-ai/objectstack derived 58
command(s) from the diff (.changeset/dataset-select-dimension-option-i18n.md,
dimension-labels.ts, dimension-labels.test.ts, plugin.ts) against a freshly-fetched
origin/main. All 58 ran: 55 passed; 3 (check:dual-build-cjs-loads, check:lean-entry-closure,
check:type-check-debt) exited 3 PREREQUISITE NOT MET — each explicitly needs a FULL
monorepo pnpm build (dozens of unrelated packages with no dist/, e.g. @objectstack/hono,
@objectstack/account, @objectstack/client), which is CI's job, not local scope. NOT
MEASURED, not a finding — named rather than guessed at. check:route-envelope (Silent-bucket
per #16828) was run explicitly: PASS, unaffected (no REST route touched).

Local verification performed

• Dependency closure: pnpm --filter '@objectstack/service-analytics^...' build — green, both
  before and after merging origin/main in.
• pnpm --filter @objectstack/service-analytics build && typecheck && test — green (98 test
  files / 2195 tests) after the merge.
• No downstream package imports DimensionLabelDeps (grepped); the six consumers of
  @objectstack/service-analytics (rest, runtime, qa/dogfood, cli, client, verify)
  reference only AnalyticsServicePlugin's registration identity, not this interface, so no
  consumer-side typecheck is owed for an additive optional member.

🤖 Generated with Claude Code

https://claude.ai/code/session_012zTkyNHJ7TkuN2oXtP5x37

[github-actions]

📓 Docs Drift Check

This PR changes 1 package(s): @objectstack/service-analytics, touching 6 documentable anchor(s).

8 hand-written doc(s) NAME something this change touched and may need an implementation-accuracy re-verification:

• content/docs/api/wire-format.mdx (via /meta/object/:name (route, a path literal in AnalyticsServicePlugin; a path literal in DimensionLabelDeps; a path literal in resolveDimensionLabels))
• content/docs/concepts/metadata-lifecycle.mdx (via /meta/object/:name (route, a path literal in AnalyticsServicePlugin; a path literal in DimensionLabelDeps; a path literal in resolveDimensionLabels))
• content/docs/deployment/production-readiness.mdx (via /meta/object/:name (route, a path literal in AnalyticsServicePlugin; a path literal in DimensionLabelDeps; a path literal in resolveDimensionLabels))
• content/docs/plugins/packages.mdx (via AnalyticsServicePlugin (symbol, a top-level class))
• content/docs/protocol/kernel/error-handling.mdx (via /meta/object/:name (route, a path literal in AnalyticsServicePlugin; a path literal in DimensionLabelDeps; a path literal in resolveDimensionLabels))
• content/docs/protocol/kernel/http-protocol.mdx (via /meta/object/:name (route, a path literal in AnalyticsServicePlugin; a path literal in DimensionLabelDeps; a path literal in resolveDimensionLabels))
• content/docs/protocol/objectql/state-machine.mdx (via /meta/object/:name (route, a path literal in AnalyticsServicePlugin; a path literal in DimensionLabelDeps; a path literal in resolveDimensionLabels))
• content/docs/ui/forms.mdx (via /meta/object/:name (route, a path literal in AnalyticsServicePlugin; a path literal in DimensionLabelDeps; a path literal in resolveDimensionLabels))

⛔ 1 release-owned page(s) also name something this change touched. These are read-only:

• content/docs/releases/v17.mdx (via /meta/object/:name (route, a path literal in AnalyticsServicePlugin; a path literal in DimensionLabelDeps; a path literal in resolveDimensionLabels))

content/docs/releases/ is RELEASE-OWNED (AGENTS.md "Documentation Guardrails"): release
notes are written centrally at release time, and a code PR that edits them is the exact PR
that guardrail exists to stop. They are still audited — read-only. If one of them is actually
wrong, file an issue or open a dedicated docs-only PR; do not edit it here.

What this run could not see

• the SDK route bridge reached 60 of 216 client-bound route-ledger rows — the other 156 have no registrar path: tail to select them, so pages documenting THEIR client methods cannot appear above, on this or any run. Of those 156: 0 are remediable by widening that discovery convention (an in-repo file declares the path; the convention did not scan it); 56 are structural — on a ledger where NOT ONE row is declared in-repo, so no discovery change reaches them at any price; 100 are undecided (no in-repo declaration, on a ledger that has other in-repo registrars — absence and an unreadable spelling are not distinguishable here). The rows themselves: node scripts/docs-audit/affected-docs.mjs --bridge-coverage
• a page that states a rule by its inputs shares no identifier with the emitter that implements the rule, so an emitter-only diff cannot list it — not on this run and not on any run. Measured on fix(driver-sql): emit varchar(maxLength) for a text field a declared index keys on #11430: content/docs/protocol/objectql/types.mdx documents the text-family column mapping by the ObjectQL type names it maps FROM (text / textarea / html) while the diff changed createColumn; it went unlisted, and it was the page that diff falsified, in four places. No shared token exists to detect this on, so a rule your change carries has to be re-read by hand in the pages that restate it.

Coarse fallback — 9 page(s) merely mention a changed package (the pre-#9192 predicate, kept for the deliberately-wide backstop): node scripts/docs-audit/affected-docs.mjs --json edf59e3599a3324598ae6cabbc72be322fe0d287 → packageMentionDocs.

Which tree this was computed on

This run read content/docs from 95c586b741b72ce8533db91f944cf08acb10417c — the merge of head 9b4c0af7f4164d3f6a99c7fb42c80785559995b6 into base edf59e3599a3324598ae6cabbc72be322fe0d287, which is what actions/checkout gives a pull_request run. Not the PR head.

A worktree cut from an older main holds a different content/docs, so re-deriving there can legitimately return a different list — that is a different tree, not a wrong row. To answer on the same tree:

# while this PR is open — GitHub drops the merge commit once it closes
git fetch origin 95c586b741b72ce8533db91f944cf08acb10417c && git checkout 95c586b741b72ce8533db91f944cf08acb10417c
# afterwards, rebuild it from the two parents, which stay fetchable
git fetch origin edf59e3599a3324598ae6cabbc72be322fe0d287 9b4c0af7f4164d3f6a99c7fb42c80785559995b6 && git checkout -B drift-repro edf59e3599a3324598ae6cabbc72be322fe0d287 && git merge --no-ff 9b4c0af7f4164d3f6a99c7fb42c80785559995b6

node scripts/docs-audit/affected-docs.mjs --json edf59e3599a3324598ae6cabbc72be322fe0d287

⚠️ That checkout carried uncommitted changes, so the commit above does not fully identify what was read.

Advisory only, and a precision-first one (#9192): a page is listed because it names a
symbol, wire route or SDK method this diff touched — not because it mentions a changed
package. Each row says which anchor put it there, so a wrong row is reportable rather than
merely annoying. To re-verify, run the docs-accuracy-audit workflow scoped to these files:
node scripts/docs-audit/affected-docs.mjs edf59e3599a3324598ae6cabbc72be322fe0d287 → pass the list as
args.docs, on the commit named under Which tree this was computed on.

[os-bill]

Director seat adoption record — summon #20, session_01Tep4AYXZvyBA7jsvne5KZV (os-bill), 2026-09-09T07:18Z. The verdict below is adopted verbatim from an isolated contract-review subagent (explicit model = CONTRACT_REVIEW_TIER). Transcript tier check before adoption: every harness-stamped model field in the subagent transcript reads claude-fable-5-1 (184 stamps, no other value). Head re-read at posting time = 9b4c0af7f4, unchanged since the review. ⛔ This seat takes no release action on this carrier (no ready flip, no auto-merge, no enqueue, no label write): the owning seat (domain:services, claim 5595841999) adopts this verdict verbatim or discards it, and acts per the state machine.

---

Contract review (CONTRACT_REVIEW_TIER, isolated seat) — PR #17061 @ 9b4c0af7f4164d3f6a99c7fb42c80785559995b6

Verdict: PASS WITH FINDINGS

Head unchanged since briefing (9b4c0af7f4…, 3 commits: 59a383d fix, bedf733 merge of main, 9b4c0af changeset regrade). Increment = git diff origin/main...refs/pr-review/17061: 4 files, +266/−4 — .changeset/dataset-select-dimension-option-i18n.md, packages/services/service-analytics/src/{dimension-labels.ts,plugin.ts,__tests__/dimension-labels.test.ts}.

Ruling conformance

• Triage / ruling on the card: the card body (analytics: chart category labels drawn from a select field's options render untranslated on a SAME-object dataset dimension, while the identical values translate on a dotted cross-object one #16773, split by the triage seat from i18n: _views.VIEW.description is declared, walked and counted as covered but never rendered — plus three surfaces a bundle cannot address at all #16735 finding 4) is the triage; there is no maintainer ruling comment. Claim = 5595841999 (Clause-②: no — provisional, and it can flip; fences: measure the dotted arm first, prefer routing over teaching, dotted arm is the control and stays unchanged, no behaviour change beyond labels, state how analytics: 数据集维度只解析 lookup 字段,user 字段维度返回原始用户 id —— 以人为轴的图表/报表标签是一串 id #16390's LOOKUP_TYPES widening inherits, re-run dedup, content/docs/releases/** untouched, a Clause-② flip is STOP-and-report, not a mid-delivery judgement). Dev report = 5596229619. Seat ACCEPT = 5596243524.
• Met: dotted arm measured (it never reaches the select branch: fields[dim.field] misses on a dotted name → if (!meta) continue, dimension-labels.ts:353) and pinned by a control test; routing not teaching (translateObject from @objectstack/spec/system reused, plugin.ts:780); dotted arm byte-unchanged by construction; no behaviour beyond labels (only rows[dim] text for a select dimension, only under a registered i18n service and a non-empty context.locale); analytics: 数据集维度只解析 lookup 字段,user 字段维度返回原始用户 id —— 以人为轴的图表/报表标签是一串 id #16390 inheritance stated in the PR body ("not automatically", correct — lookup labels go through fetchRecordLabels, a different address); dedup re-run with a firing control (report mcp_calls); content/docs/releases/** untouched.
• Not met to the letter (F1): the Clause-② flip no → yes was made mid-delivery — PR body edit + changeset regrade patch → minor in 9b4c0af at 06:30Z — with no STOP-and-report on the card. The card's claim carrier still reads Clause-②: no, and the seat's ACCEPT (05:21Z, before the flip) asserts "Clause-②: no holds as declared" using a spec-exports-only rule that this lane's own feat(service-analytics)!: refuse an aggregate a datetime measure's field type cannot carry, and reconcile the storage-form annotations to one measured statement #16778 precedent already overrode. The direction of the flip is the conservative one and every gate passes under it, so this is a record defect, not a merit defect.
• Seat claim "the card's hypothesis was wrong" — partially overstated (F2). Verified against objectui: the dotted arm is translated through the object-metadata pipeline, exactly as the card hypothesised — GET /api/v1/meta/object/<target> (translated server-side by translateObject via translateMetadataDocument, rest-server.ts:3647) → objectui useDatasetDimensionMeta (packages/react/src/hooks/useDatasetDimensionLabels.ts:130-139) → loadDimensionFieldMeta / resolveDimensionFieldMeta walks contract.direction to contract (packages/core/src/utils/chart-series.ts:1186-1230) → deriveDimensionLabelMaps → relabelDimensions (chart-series.ts:709). What is true is that this happens in objectui, downstream, while service-analytics passes the raw value through. The PR body states this correctly ("downstream of this package"); the ACCEPT's framing should not be inherited.

Derived judgments

(a) What moves, and the mechanism. One package: @objectstack/service-analytics (17.4.0, released, not private). Functions: DimensionLabelDeps gains optional translateSelectOptions(objectName, fieldName, options, locale) (dimension-labels.ts:94-99); withLabelFetchCache passes it through (:244-247, and the end-to-end test pins that — queryDataset always wraps the resolver in this cache, analytics-service.ts:1261); resolveDimensionLabels select branch consults it first and falls back to meta.options (:363-364); AnalyticsServicePlugin.init implements it (plugin.ts:669-791). Exact production path: REST POST dataset door → resolveExecCtx (rest-server.ts:10757) → resolveExecutionContext → assembleExecutionContextOrGuest sets locale: anonymous ? undefined : (requestLocale ?? localization?.locale) (core/security/assemble-execution-context.ts:353; requestLocale = preferredLocaleFromHeader(accept-language)) → svc.queryDataset(dataset, selection, context) (:10857) → resolveDimensionLabels(..., labelDeps, resolveScope, context) (analytics-service.ts:1480/1486) → deps.translateSelectOptions(baseObject, dim.field, meta.options, context.locale) → plugin: ctx.getService('i18n') → buildTranslationBundle (every getLocales() × getTranslations(), cached references — file-i18n-adapter.ts:162-168) → translateObject({ name, fields: { [field]: { name, options } } }, bundle, { locale, fallbackChain: [getFallbackLocale()], defaultLocale: getDefaultLocale() }) → lookupObjectFieldOption at objects.<obj>.fields.<field>.options.<value> (i18n-resolver.ts:2153-2170). The {locale, fallbackChain, defaultLocale} derivation is byte-equivalent to RestServer.translateOptionsFor (rest-server.ts:3560-3570), so the chart now reads the same bundle by the same rules as /meta/object/:name. I executed translateObject on the plugin's exact doc shape (spec source unchanged by the PR — blob 8162eca7… identical on both sides): zh-CN → 采购/销售 with an untranslated option keeping Other; unknown locale → authored labels; fallbackChain honoured; defaultLocale short-circuit correct. Mechanism proven, not inherited.

(b) Direction. Accept set unchanged (no request previously accepted is refused, none previously refused is accepted). AnalyticsResult wire shape unchanged — no key gained or lost; only the text of rows[<select dim>] differs, and only under a registered i18n service with a non-empty context.locale; anonymous/no-i18n/no-locale paths are byte-identical to before. Exported symbols: none appear/disappear; the published type DimensionLabelDeps (re-exported at index.ts:30) gains one optional member. No new error code. No packages/spec/src/** path.

(c) Acceptance criteria, as written.

• "A same-object select dimension on a dataset-backed chart renders its option labels in the console's locale, matching what the list grid renders for the identical values." — Met for translations in the server bundle: same translator, same bundle, same locale/fallback derivation as the /meta/object/:name read the grid consumes. Caveat F5.
• "The dotted cross-object arm is unchanged (it is the control)." — Met, by construction (if (!meta) continue precedes the modified branch) and pinned by the control test (translateSelectOptions never consulted, raw value survives).
• "One copy of 'translate a select option label', not two." — Met: the lookup is translateObject only; translateDataset (spec) translates dataset/member labels, not options, so there was no closer existing copy. The ~10-line buildTranslationBundle glue is duplicated from REST (F6) — glue, not the translation.

(d) Tests. Controls present: translator declines → authored label; capability absent → authored label; dotted field name → translator never consulted; end-to-end through AnalyticsService.queryDataset (covers the cache pass-through). Missing: any test of the plugin bridge itself — every test injects a fake translateSelectOptions; no test in the package registers an i18n service (F3). No lookup-typed-dimension control with a translator wired (branch order makes it unreachable; minor).

Semver / changeset

• .changeset/dataset-select-dimension-option-i18n.md: "@objectstack/service-analytics": minor — the only released package whose packages/*/src/** moves. First commit 59a383d declared patch; regraded to minor in 9b4c0af.
• node scripts/check-changeset-no-major.mjs --base 041d9fdc6 --head refs/pr-review/17061 --event <pr.json> → exit 0: "no major bump"; "LEVEL AXIS: declares clause-② yes, and no package whose packages/**/src/** it moves is graded patch" (carrier needs:contract-review IS on the PR; declaration line Clause-②: yes).
• No BREAKING banner, no ADR-0087 marker owed: nothing previously accepted is refused. ! title correctly absent.
• Docs: no page in content/docs names DimensionLabelDeps / fetchRecordLabels / labelResolver; the drift bot's 8 rows are path-literal hits on /meta/object/:name in the new comments, not a route change. No doc owed.

Boundary flags

• Clause-②: yes by the directional rule — public surface widened (new optional member on the published exported type DimensionLabelDeps, the same floor feat(service-analytics)!: refuse an aggregate a datetime measure's field type cannot carry, and reconcile the storage-form annotations to one measured statement #16778 applied to DatasetCompileOptions.declaredFieldType). Not widened: accept set, wire payload, error codes. Path arm: not hit (no packages/spec/src/**). check-widening-tells.mjs --declaration no|yes → exit 0 both (its tells are scoped to packages/spec/**; documented false-negative, as the PR body says).
• Declaration carriers disagree: card claim 5595841999 reads no (what check-clause2-carriers.mjs needsWideningRead reads from the claim comment); PR body reads yes (what the level axis reads). See F1.
• Governed: no — register docs/adr/** · .claude/** · skills/** · AGENTS.md · CLAUDE.md; none of the 4 paths hit. Governed Surface Queue Guard green.
• Single-writer: .objectui-sha untouched.

Findings

• F1 — non-blocking (ruling conformance / record). Clause-② flipped no → yes mid-delivery (PR body + 9b4c0af changeset regrade, 06:30Z) without the STOP-and-report the claim (5595841999) required; the card's claim carrier still reads no, and ACCEPT 5596243524 asserts no on a rule this lane's feat(service-analytics)!: refuse an aggregate a datetime measure's field type cannot carry, and reconcile the storage-form annotations to one measured statement #16778 already overrode. Remedy: the domain:services seat posts the corrected Clause-②: yes on the card (claim-shaped, so the carriers script reads it) and acknowledges the minor regrade before enqueue. No code change.
• F2 — non-blocking (record correction). ACCEPT's "the card's hypothesis was wrong" overstates: the dotted arm is translated through the object-metadata pipeline (/meta/object/<target> → objectui useDatasetDimensionLabels.ts:130-139 → chart-series.ts:1186-1230 → :709), downstream of this package. Adopt the PR body's wording, not the ACCEPT's.
• F3 — non-blocking (test gap). plugin.ts:669-791 — i18nService(), buildTranslationBundle(), and the translateSelectOptions → translateObject call — is the only place the real mechanism lives and has zero coverage; all PR tests fake the hook. Mechanism independently proven by execution (above), so the gap is regression protection. Cheap remedy in the existing pattern (record-label-read-scope-vacancy.test.ts:135-173: new AnalyticsServicePlugin().init(ctx) with getService('i18n') returning a stub) asserting queryDataset under locale: 'zh-CN' returns the bundle label.
• F4 — non-blocking (coherence; answers the report's open question). createOrderLabelResolver select branch (dimension-labels.ts:179-184) still sorts by the authored label while display now shows the translated one — for a non-default locale, an order on a select dimension recreates exactly the "order that presents as arbitrary once the labels render" Dataset order sorts a select/lookup dimension by its stored value, not the label the user reads #3680 fixed (:116-126). Answer: A is acceptable for this PR under the "no behaviour change beyond labels" fence only with a follow-up card filed — the report says "noted, not filed", so the seat files it. B is the correct end state and is a one-line consult of the same hook (keeps one copy).
• F5 — non-blocking (parity caveat). The grid's option labels are translateObject over the server bundle plus objectui's client-side i18next fieldOptionLabel (packages/i18n/src/useObjectLabel.ts:323). A translation present only client-side shows in the grid but not in the chart's same-object arm. Out of this card's lane (objectui's own copy); the acceptance holds for bundle-resident translations, which is the card's provenance.
• F6 — non-blocking (duplication note). buildTranslationBundle (plugin.ts:684-693) duplicates RestServer.buildTranslationBundle (rest-server.ts:3438-3450). Glue, not the translation; a shared helper would remove the drift risk. Not owed by this card.

Report deviations: none declared; none found beyond F1. out_of_scope_findings: the order-resolver note — F4.

CI at read time

Head 9b4c0af7f4164d3f6a99c7fb42c80785559995b6: 39 check runs, 33 latest-per-name — 28 success, 5 skipped (Auto Label, Build Docs, Check PR Size, Console Pin Gate, Packed-tarball smoke (opt-in) — all conditional), 0 failures, 0 in progress. Green: Check Changeset, Governed Surface Queue Guard, Lint & Repo Gates, Type Check ×5, Test Core ×7, Dogfood Regression Gate ×4, Dogfood Verify CLI, Temporal Conformance, both claim/single-writer guards. PR is mergeable: true / clean, still draft.

Implemented-by: branch claude/issue-16773-select-dimension-label-translation
Reviewed-by: director seat summon #20 (isolated fable subagent, transcript-verified before adoption)

{"pr":17061,"head":"9b4c0af7f4164d3f6a99c7fb42c80785559995b6","verdict":"PASS WITH FINDINGS","blocking":[],"clause2":"yes","semver_ok":true,"governed":false,"ci":"green — 33 latest-per-name: 28 success, 5 skipped (conditional), 0 failures"}

---

Generated by Claude Code

[huangyiirene]

Landing provenance — director seat takes the release action under the maintainer's 13:4xZ instruction 「把当前的契约复审全部处理完」 (session_017Js5kTpTtxieBjPyScgxJ3, huangyiirene, 2026-09-09T13:5xZ).

• Verdict of record on this exact head 9b4c0af7f4164d3f6a99c7fb42c80785559995b6: PASS WITH FINDINGS (director seat summon chore(ci)(deps): bump actions/configure-pages from 4 to 5 #20 adoption, os-bill, 07:18Z) — fix(service-analytics): translate select dimension option labels on dataset charts #17061 (comment), adopted verbatim.
• Read at posting: head unchanged; CI 34 success / 5 skipped / 0 failure / 0 in progress; mergeable_state: unknown; check-clause2-carriers --pair 17061 exit 0.
• Chain: needs:contract-review off this PR and card analytics: chart category labels drawn from a select field's options render untranslated on a SAME-object dataset dimension, while the identical values translate on a dotted cross-object one #16773 in this stroke (read-modify-write, assignees kept, read back) → ready for review → auto-merge (squash). Card closes by its Fixes keyword on merge; the non-blocking findings of the verdict are follow-up card candidates, not riders.
• The owning lane seat's carrier was idle since the verdict; this is a maintainer-ordered takeover of the release action, recorded on seat post [PM seat] director(项目总监) — 🟢 os-zhuang（第 15 场）· ⏹️ 第 16–20 场已收班（第 20 场 = 总监契约复审 session_01Tep4AYXZvyBA7jsvne5KZV，收班 2026-09-09T07:20Z） #12708, not a change to the charter.

---

Generated by Claude Code