Skip 400ms search throttle for in-memory autocomplete filtering

[GeorgiZhelev]

WHY are these changes introduced?

Filtering inside renderAutocompletePrompt (e.g. the theme selector that appears for shopify theme push) feels noticeably laggy: there's a ~400ms gap between typing and seeing matching results update.

The cause is the throttle on the search callback in AutocompletePrompt (packages/cli-kit/src/private/node/ui/components/AutocompletePrompt.tsx). The 400ms window makes sense for callers that supply a remote/paginated search (we don't want to hammer an API on every keystroke), but it also gates the default in-memory filter that renderAutocompletePrompt injects when the caller doesn't pass one — and that filter is fully synchronous (Promise.resolve(choices.filter(...))), so there's no I/O to throttle.

WHAT is this pull request doing?

• Adds an optional searchDebounceMs prop to AutocompletePrompt, default 400 (backward-compatible).
• renderAutocompletePrompt sets searchDebounceMs: 0 only when it's injecting its own default in-memory filter; if the caller supplied their own search, the default 400ms throttle still applies.
• Adds a test asserting that searchDebounceMs={0} invokes the search callback on every keystroke (vs. coalescing to leading + trailing edges).

Net effect: prompts that filter against in-memory choices feel instant; remote/paginated consumers see no behavior change unless they opt in.

How to test your changes?

1. Build the CLI: pnpm nx build cli
2. From any theme directory, run a command that opens the theme selector — e.g. node packages/cli/bin/dev.js theme push
3. Type into the search box; results should update on each keystroke instead of in ~400ms bursts.

Unit tests:

pnpm -F @shopify/cli-kit vitest run src/private/node/ui/components/AutocompletePrompt.test.tsx

Post-release steps

None.

Checklist

• I've considered possible cross-platform impacts (Mac, Linux, Windows) — pure JS timing change, platform-agnostic.
• I've considered possible documentation changes — JSDoc on the new prop covers it.
• I've considered analytics changes to measure impact — none required.
• The change is user-facing — I've added a patch changeset for @shopify/cli-kit.

[gonzaloriestra]

Thanks for your contribution! Could you please rebase? The failing check should pass after that

[GeorgiZhelev]

Hey, done!

[gonzaloriestra]

Nice work, thanks!! It really feels much faster ⚡

[gonzaloriestra]

Would you mind simplifying here a bit?

Suggested change

Make the autocomplete prompt feel instant when filtering against in-memory choices. The 400ms throttle on the search callback was designed for remote/paginated backends, but it also gated the default in-memory filter used by callers that don't supply their own `search` (e.g. the theme selector), producing a noticeable lag while typing. The prompt now exposes a `searchDebounceMs` prop, and `renderAutocompletePrompt` sets it to `0` when it injects its own synchronous filter. Custom remote-search consumers keep the existing 400ms throttle unless they opt out.

Make the autocomplete prompt feel instant when filtering against in-memory choices.