Skip to content

refactor(dashboard): name the decision model by its role in the UI - #37185

Merged
bolichen97 merged 2 commits into
mainfrom
feat/decision-model-neutral-wording
Nov 14, 2026
Merged

bolichen97 merged 2 commits into
mainfrom
feat/decision-model-neutral-wording

Conversation

@bolichen97

@bolichen97 bolichen97 commented Nov 14, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: The chat UI, settings card and docs name the decision feature by its role ("the decision model"), with Jev appearing only as one provider among several.

Today the model picker says "Auto (Jev)", the transcript says "Jev chose: run after this turn" and "Jev picked the wrong skills", and the settings card is titled "Decisions (Jev)". A person who runs a local model or a SageMaker endpoint still reads a hosted product's name on every surface. The governance PR (#36271) made Jev one provider entry in the picker; this PR makes the rest of the wording agree.

Why it matters

Every user of the decision features reads this wording, including people who never turn Jev on and administrators who disable it. The product name on neutral surfaces says the feature needs one vendor, which is no longer true.

Not a goal

  • Renaming code identifiers, catalog keys, config values or the jev provider id (auto_jev, jev_label, provider: "jev"). Only user-visible text changes.
  • Changing the Jev provider's own entry: its option, API key, accuracy note and judge option still say Jev.
  • Touching historical RFCs under docs/request-for-change/ and CHANGELOG.md.

What changed (motivation → approach → change)

Approach: say "the decision model" wherever the text is about the feature, and keep "Jev" only where it is about the provider.

  • 61 English catalog strings now say "the decision model" (for example Auto (decision model), The decision model chose: run after this turn, card title Decisions). Seven strings keep the name, all of them the Jev provider's own entry. The same 61 keys are translated in the eleven other languages, each using the term that language already uses for "decision model".
  • The translator notes in en.context.json no longer say the name is a Latin-script product name.
  • The settings card's registry id follows its label: developer.decisions-jev becomes developer.decisions, and the old id is added to LEGACY_ID_EXACT so saved ?highlight= links and command-bar history still land on the switch. The generated registry, the UI index and the pseudolocale are regenerated.
  • A catalog test (decisionModelNeutralWording.test.ts) fails when any language names Jev outside the seven provider-entry keys, and when the allow list names a key that no longer says Jev.
  • Second commit: the shipped decisions guide, the feature map and specs, the two owner-only refusal messages, and the help text of two decision settings. config-baseline.json is regenerated for that help text.

Backwards compatibility

Compatible: no catalog key, config value or wire id changes; only displayed text and the settings registry id of one card.

Reader: website/src/components/commandPalette/settingsSearchCore.ts:DECISIONS_SETTING_ID -- crew page (settings search) -- updated to developer.decisions in this PR; settingsSearchGovernance.test.ts and settingsCoverage.test.ts pin it against the generated registry.
Reader: website/src/hooks/useSettingHighlight.ts:LEGACY_ID_EXACT -- crew page (?highlight= links) and command bar (frecency) -- maps the old id to the new one; settingHighlightAliases.test.ts pins the mapping and that no legacy key is a live id.
Reader: website/src/components/commandPalette/settingsRegistry.gen.ts and src/kiro_crew/docs/settings-registry.generated.json -- crew page, agent-facing settings docs -- regenerated with npm run gen:settings; the freshness gate compares them with the live extraction.
Reader: src/kiro_crew/docs/ui-index.generated.json -- agent-facing UI index -- regenerated with npm run gen:ui.

Tests

  • website/src/i18n/decisionModelNeutralWording.test.ts (new): one case per language plus the stale allow-list case. With one string put back to "Jev latency", the English case fails.
  • settingHighlightAliases.test.ts: the old Decisions id resolves to developer.decisions.
  • Existing decision-strip, memory-recall, steer-line, tool-risk, settings-card and search-governance tests updated to the new wording: 1,370 tests across 47 files pass locally.
  • npm run i18n:check passes with I18N_BASE_REF=origin/main (19 checks, 0 changed-value findings). Python: test_config_baseline.py, the model-route and session-control suites (182 tests) pass.

Proof: not provable -- copy-only change

Manual verification

Built the SPA and ran the capture scripts for the decision strip, the send-mode picker and the settings card (images below). Not run: a real gateway with a hosted Jev key.

Screenshots / video

decision strip, expanded
send mode picker
steered receipt
Decisions settings card

Checklist

  • At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)
  • Existing tests pass and new tests added for new functionality
  • Self-review completed; code follows project style guidelines
  • Documentation updated (if applicable)
  • No secrets, credentials, or internal references in the diff

Contribution License Agreement

@bolichen97
bolichen97 requested review from a team and KiroCrewCaptain as code owners November 14, 2026 10:36
@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Nov 14, 2026
@github-actions

github-actions Bot commented Nov 14, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5) — ✅ PASS

Design-level review of 48beff7b3a0a5bd6b7d71d64a046255c09e914dc — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

Design-Verdict: PASS

Nothing blocks. This changes copy only. The one renamed id (developer.decisions-jev) keeps a legacy alias, and no reader of the old id is left in the tree. The refusal path accepts both the old and new picker labels.

[DESIGN-REVIEWED] 48beff7

Description read

[DESCRIPTION-READ] 2af6b0b4629f5dd2d9075dcab18d2383ec6227cdc336b6881b1d8d5e0664825f

That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Nov 14, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5) — ✅ PASS

Premise-level review of 48beff7b3a0a5bd6b7d71d64a046255c09e914dc — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

First-Principles-Verdict: PASS

This is a copy change that follows the Goal. Only one user-visible refusal still names the provider.

Notes

  • Item 6 — symptom-level: the description says it changed "the two owner-only refusal messages", but the 403 refusal on the arming route still names the provider. It reads "arming Jev routing is owner-only", once in the denial response and once in the JSON error, at src/kiro_crew/dashboard/chat_api/slot_model.py:471 and :473. That is 2 sites left of the pattern "[^"]*\bJev\b[^"]*" across non-provider backend literals. The other matches are in the provider's own key and loopback text, or in comments.

What this change ships

Inventory (10 items) — 9 justified

Intent: people who use a decision provider other than Jev stop seeing a hosted product's name across the decision features. ADDITION (a relabel).

  1. The model picker, transcript strips and steer line say "decision model" instead of Jev (61 English strings) — justified
  2. The same 61 strings changed in the 11 other languages — justified
  3. The settings card title changed from "Decisions (Jev)" to "Decisions" — justified
  4. The card's deep-link id is now developer.decisions, and old links are redirected through LEGACY_ID_EXACT — justified
  5. session_set_model refuses the new "Auto (decision model)" label as well as the old one — justified
  6. The owner-only refusal and the MCP tool description use the new wording — symptom-level: the arming refusal in slot_model.py still says Jev
  7. Help text for two decision settings and config-baseline.json — justified
  8. The shipped decisions guide, the feature map and the specs use the new wording — justified
  9. A new catalog test fails when Jev appears outside the seven provider-entry keys — justified
  10. The translator notes no longer call the name a Latin-script product name — justified

[FIRST-PRINCIPLES-REVIEWED] 48beff7

Description read

[DESCRIPTION-READ] b36a0742407c86eea3009080a0e321fea6e44325d7ae3640968f568942a4e2d1

That sha256 names the PR description this verdict read. Recompute it with .github/scripts/pr-description-capture.sh; if it differs, the description changed after this verdict was formed and any finding drawn from the description is unproven.

@github-actions

github-actions Bot commented Nov 14, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5) — ✅ PASS (a BLOCK with no Trigger and no Outcome does not block)

UX-level review of 48beff7b3a0a5bd6b7d71d64a046255c09e914dc — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS does not.

UX-Verdict: PASS

cannot evaluate: missing screenshots of the model picker's "Auto (decision model)" label and of the memory-recall, tool-risk and model-route strips, whose copy this PR rewrites.

Evidence gaps

  • Model picker chip and dropdown entry, "Auto (decision model)" (components.modelDropdownList.auto_jev, shown by ContextShelf.tsx:540 and ModelDropdownList.tsx:45), plus its description, "Let the decision model judge each turn…". The PR's Problem section uses this picker as its first example, but no screenshot shows it. shot-02 is the busy-send menu, a different control. To close: a screenshot of the composer's model chip with the dropdown open.
  • Memory-recall strip: "Decision model kept: {{count}}", "Kept by the decision model", "Rate the decision model's narrowing". No screenshot. To close: a screenshot of an expanded memory-recall strip.
  • Tool-risk badge: "Decision model: {{tier}}" and "The decision model read the risk right/wrong". No screenshot. To close: a screenshot of a flagged tool call with its badge.
  • Model-route strip: "Decision model: {{tier}}", "The decision model picked the right model". No screenshot. To close: a screenshot of the decision strip for a routed turn.
  • Settings opt-in toggles (tool-call arguments, compaction, memory snippets, watching-loop evidence) and their rewritten descriptions. shot-04 does not show them. To close: a screenshot of the Decisions card scrolled to those toggles.

Notes

  • The skills strip's one-line summary in shot-01 now cuts off at "Decision model:…", so the pick it names is lost (blind reader: "It is cut off at 'Decision model:…', so I can't see what it says there"). The longer phrase pushes the pick out of view; shorten the label in jev_named, or let the names win the truncation.
  • In shot-01 the blind reader could not tell the two thumbs rows apart ("Decision model" vs "Word match"): "I'd be afraid to rate the wrong thing". Put each rating row next to the result it rates.

[UX-REVIEWED] 48beff7

Description read

[DESCRIPTION-READ] 7b6284c7abf57b689b1c819a802a3498061ab313430ce951d062418d84325e97

That sha256 names everything this verdict read: the PR description, plus the 7 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Nov 14, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review — ✅ no blocking findings

Reviewed 48beff7b3a0a5bd6b7d71d64a046255c09e914dc — this comment is updated in place on each push.

Review details

No findings. Both candidates were falsified against the code.

[OPUS-REVIEWED] 48beff7

Verdict parsed from the review's SHA-scoped output markers for commit 48beff7b3a0a5bd6b7d71d64a046255c09e914dc.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 48beff7b3a0a5bd6b7d71d64a046255c09e914dc false-positive: <reason, 20 to 500 characters>

@github-actions

github-actions Bot commented Nov 14, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review — ✅ no blocking findings

GPT 6.1 completed its review of 48beff7b3a0a5bd6b7d71d64a046255c09e914dc and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 48beff7

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 48beff7b3a0a5bd6b7d71d64a046255c09e914dc false-positive: <reason, 20 to 500 characters>

@bolichen97
bolichen97 force-pushed the feat/decision-model-neutral-wording branch from ab04b50 to ce387d8 Compare November 14, 2026 10:47
@bolichen97
bolichen97 enabled auto-merge (squash) November 14, 2026 10:50
@KiroCrewCaptain

KiroCrewCaptain commented Nov 14, 2026 •

Copy link
Copy Markdown
Collaborator

Code change approver recommendation: A person should decide

Why: the decider estimates a 12% chance that this change introduces a defect, within the 12% limit for approving on its own, but it changes 2342 lines, more than the 400 it may approve on its own.

Copy-only rename of Jev to "the decision model" across the UI, catalogs and docs. The one behavioural surface, session_set_model's owner-only refusal, now refuses the new label "Auto (decision model)" as well as the old spellings, and the parametrized test covers it. The settings id rename developer.decisions-jev → developer.decisions has a LEGACY_ID_EXACT redirect. The backend highlight targets and DECISIONS_SETTING_ID are updated, and nothing else uses the old id. Settings rows whose ids still contain "jev" are hand-written entries with fixed ids. The renamed doc headings have no inbound anchors, and the tests asserting doc wording were updated. A new catalog test enforces neutral wording. Linux backend and frontend CI are green; the Windows failure is unrelated. Remaining gaps (one Jev refusal in slot_model.py, doc index drift) don't block.

An automated recommendation, not a review: a maintainer still decides. It approves on its own only when the estimated chance of a defect is at most 12% and the change has at most 400 changed lines. Decider v3c (217fe607) · commit 48beff7b3a · 2342 changed lines. If this recommendation is wrong, reply here: replies are how the approver is evaluated.

@bolichen97
bolichen97 force-pushed the feat/decision-model-neutral-wording branch from ce387d8 to b2bcaa0 Compare November 14, 2026 11:15
@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Nov 14, 2026
@bolichen97
bolichen97 force-pushed the feat/decision-model-neutral-wording branch from b2bcaa0 to 26b0a67 Compare November 14, 2026 11:44
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Nov 14, 2026
@bolichen97
bolichen97 force-pushed the feat/decision-model-neutral-wording branch from 26b0a67 to 22ea773 Compare November 14, 2026 12:48
@github-actions github-actions Bot added the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Nov 14, 2026
@bolichen97
bolichen97 force-pushed the feat/decision-model-neutral-wording branch from 22ea773 to b383d6e Compare November 14, 2026 13:41
The decision feature read as one product: the model picker said Auto
(Jev), the transcript said Jev chose, the settings card said Decisions
(Jev). Jev is one provider among several, so name the role instead.

Sixty-one catalog strings, in English and the eleven other languages,
now say the decision model. Seven strings stay: the ones about the Jev
provider's own entry (its option, API key, accuracy note and judge
option). A catalog test pins that list, so a new string cannot leak the
provider name back into neutral wording.

The owner-only refusal of session_set_model now also catches the new
label, so an agent cannot store it as a pending model. The capture
scripts use the new labels.

The settings card's registry id moves from developer.decisions-jev to
developer.decisions with its label, and the old id is added to the
legacy highlight table so saved links and command-bar history still
land on the switch. The generated registry, UI index and pseudolocale
are regenerated.
Follow the UI rename into the places a reader meets the same words: the
shipped decisions guide, the feature map and specs, two owner-only
refusal messages, and the help text of two decision settings. The Jev
provider keeps its own name where the text is about that provider.

config-baseline.json is regenerated for the changed help text.
@bolichen97
bolichen97 force-pushed the feat/decision-model-neutral-wording branch from b383d6e to 48beff7 Compare November 14, 2026 13:49
@github-actions github-actions Bot removed the merge conflict Branch has merge conflicts with its base — author must resolve before merge label Nov 14, 2026
@bolichen97
bolichen97 merged commit 0f3e09c into main Nov 14, 2026
153 of 157 checks passed
@bolichen97
bolichen97 deleted the feat/decision-model-neutral-wording branch November 14, 2026 15:15
@github-actions github-actions Bot removed the readiness: checking Automated validation is still running label Nov 14, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants