[AfonsoNeto]

What

Refreshed onto the restructured omacast tree (flattened layout + the flea-search backend) after the Nov 2 merges — same feature, rebased and re-verified.

Type a message in the launcher, press Ctrl+L, and it appears in very large characters across the whole screen — Alfred's Large Type, which Raycast only covers through a third-party extension. Esc, a click, or Ctrl+L again returns to the launcher with the query and focus intact, ready to edit.

The full interaction loop (type → Ctrl+L → Esc → edit continues → Ctrl+L again):

A second entry point: a "Show in Large Type" row under the existing "Use "" with…" fallbacks (the same query-level mechanism as Search Google / Search DuckDuckGo). Happy to drop that row if you'd rather keep the root list untouched and shortcut-only.

How

• core/LargeType.js (new): the testable policy — displayText() (blank/whitespace input becomes a no-op; content preserved exactly, markup included) and fitRange() (clamped pixel-size bounds so fitting can never run away).
• ui/LargeTypeOverlay.qml (new): an opaque, theme-following surface inside the existing launcher window — no second layer surface, so dismissal can't orphan anything. Text uses Qt's bounded Text.Fit between the core bounds, Text.PlainText (user input is never markup), wrapping with WrapAtWordBoundaryOrAnywhere so long unbroken strings degrade gracefully.
• ui/LauncherWindow.qml: the Ctrl+L binding (verified unused — no plugin binding, and no Hyprland bind on L in Omarchy defaults), a largeType outcome handler for the fallback row, and state reset in closePopups() so hide/show cycles never leave the overlay stale.
• core/RootSearch.js / Backend.qml: the fallback row and its { largeType: query } outcome; getActions/getItemInfo work through the existing fallback paths.

Blank or whitespace-only queries are a no-op — never a blank takeover.

Testing

• node --test tests/*.test.js: 146 pass, 0 fail on the new tree (3 skips are the Rust-helper integration tests, identical without a built omacast-search); new tests/largetype.test.js covers the display policy and fit bounds, and the root-search tests cover the row's placement, actions, and absence on the empty root and in submenus.
• CI parity: bash -n on all four scripts, manifest JSON check; also bin/build-search + the search-integration suite locally.
• Live on Omarchy (1920×1080, wtype + grim screenshots, re-verified on this tree): short text, accented Portuguese, literal <b>Hello</b>, a 300-character unbroken string (renders wrapped within margins, no crash), blank/whitespace no-ops, Esc restoring query + focus, Ctrl+L toggling, hyprctl layers showing exactly one omacast surface while shown and zero when hidden. Overlay follows the live theme.
• Security pass on the diff: the overlay renders PlainText only, performs no exec/network/eval, and Escape stays reachable in every state.

Notes

• No new dependencies; QML + plain JS per the ground rules.
• Follow-ups deliberately out of scope: clipboard/selected-text sources, calculator result, configurable styling.
• Pre-existing observation, not from this change: queries over FleaSearch.MAX_QUERY_LENGTH (256) bypass the native matcher and re-run the full synchronous JS search on every keystroke, so very long queries queue seconds of work. Measured live; worth a debounce discussion in a separate issue.

[AfonsoNeto]

Hi — just a friendly follow-up on this one. I saw the big refactor that landed this week (the omacast rename, flattened layout, and the flea-search backend). This PR predates all of that and won't apply cleanly to the new tree anymore.

The feature itself still ports cleanly — the fallback mechanism in core/RootSearch.js survived the restructure, so it's a straightforward move onto the new layout (~250 lines, 3 commits). Happy to rebase and push a refreshed branch if you'd still like the feature; equally fine to close this if Large Type isn't a fit for omacast's direction. Thanks!