Bug - Select migration display value fixes

[SpicyPete]

• Fixes the display value of Selects
• Fixes from previous migration work

[SpicyPete]

@claude review

[github-actions]

Claude finished @SpicyPete's task in 4m 16s —— View job

---

Claude PR Review — Select migration display value fixes

• Read the full diff against origin/master
• Understand the Select migration fixes
• Review for bugs, correctness & consistency
• Post review feedback

---

Summary

The change is sound and addresses a real bug. Base UI's Select only resolves an item's label after the popup first mounts; until then it stringifies the raw value, so a controlled, enum-backed select flashes the underlying value (e.g. 1, TOPIC, a model id) instead of its label. Passing an items map to <Select> (Base UI's Select.Root) lets it resolve the label up front. This matches the documented contract in select.tsx:56-59.

I verified the items payloads against the actually-rendered SelectItems in every changed file:

• Single-source-of-truth (excellent): select-slot (slot.options), create-topic-dialog (options), advanced-topic-settings (units), transcript-filter-bar (TIME_RANGES/OPERATOR_OPTIONS), observability-toolbar & pipeline-throughput-card (TIME_RANGES), ai-agent-transcripts-tab (STATUS_FILTER_OPTIONS), resource-tier-select, secret-selector, llm-config provider select, install-rpk, remote-mcp, schema-create references (contextSelectOptions). These reuse the same array that renders the items, so they can't drift. 👍
• Refactors that de-duplicated branching logic: claude-code.tsx / gemini.tsx — collapsing the selectedScope === 'x' && ... chains into a SCOPE_OPTIONS lookup is a nice cleanup, and configFile for local/user both stay ~/.claude.json (no regression).
• Hand-written maps that match their SelectItems exactly: schema-create Strategy, all three create-acl selects (including the dynamic getResourceName(...) labels).

No correctness bugs found in the changed code.

Main finding — a missed sibling select (not in this PR's diff)

frontend/src/components/ui/ai-agent/llm-config-section.tsx:288 — the model select is enum-backed: value={model.value} (a model id) but it displays {model.name}. It has neither an items map nor a render-prop SelectValue, so it still exhibits exactly the bug this PR fixes — when editing an agent with a pre-selected model, the trigger flashes the raw model id until the popup opens. Notably, the provider select in the same file (line 201) was fixed in this PR, but this one right below it was not.

<Select
  disabled={hasNoProviders || hasNoModels}
  items={filteredModels.map((m) => ({ value: m.value, label: m.name }))}
  onValueChange={field.onChange}
  value={field.value ?? ''}
>

Fix this →

Other controlled selects I checked (no action needed)

For completeness I swept every remaining <Select value=…> without items. All are already correct because they either use a render-prop SelectValue (the other documented escape hatch) or have value === label:

• ui/ai/chat-model-select, embedding-model-select, reranker-model-select — render-prop child showing {value}.
• redpanda-ui/.../auto-form/fields/select.tsx — render-prop via renderStaticSelectedValue.
• template-gallery/slot-fields/secret-slot.tsx — render-prop child.
• ui/user/user-selector, add-user-step (SASL mechanism), schema-create topic-name select — value === label.

Minor / nits (non-blocking)

• DRY in schema-create & create-acl: the items={{…}} literals duplicate the hardcoded <SelectItem> children. The create-acl Selector case is the most fragile — both the items map and the SelectItems independently call getResourceName(rule.resourceType) / getPluralResourceName(...), so a future label tweak must be made in two places or the trigger and dropdown will disagree. Consider deriving one from the other (or mapping over a shared array, as the other files do).
• I did not run bun run type:check / lint / test (no bun install in this review environment), but the items shapes all conform to Base UI's { value, label }[] / Record<value,label> contract and reuse existing typed arrays, so I'd expect them to pass.
  · bug/select-items