Add Windows support and Korean localization

[sioaeko]

Description

This adds a first Windows version of OpenLogi. The goal is to make the app usable on Windows without needing Logi Options+ running in the background.

The Windows path now discovers Logitech HID++ devices through the Windows HID API, supports the main GUI flow, installs a Windows mouse hook for button remapping, and adds Windows-specific pairing and startup handling. I also added Korean translations and fixed a few layout issues that showed up while testing the app in Korean, including the settings language row layout.

I tested this with an MX Master 4 connected through a Logi Bolt receiver after stopping the Logi Options+ agent.

Screenshots

English:

Korean:

What changed

• Added Windows HID++ device discovery and transport support
• Added Windows mouse hook support for button remapping
• Added DPI and SmartShift support on Windows
• Added Windows Bluetooth pairing support through the Windows pairing API
• Added Windows startup-at-login handling
• Added Korean language support
• Added a Windows executable icon using the existing OpenLogi icon asset
• Fixed several Windows/Korean UI layout issues in the main window and settings window
• Avoid selecting receiver slots that claim to be online but do not actually respond to HID++ probes

Tested

Tested on Windows with an MX Master 4 connected through a Logi Bolt receiver, with the Logi Options+ agent stopped.

• openlogi list detects the mouse and reads model/battery info
• openlogi diag features dumps the HID++ feature table
• openlogi diag dpi reads, writes, reads back, and restores DPI
• openlogi diag smartshift toggles SmartShift and restores the previous mode
• cargo fmt --all
• cargo test -p openlogi-cli -p openlogi-hid -p openlogi-core
• cargo clippy -p openlogi-cli -p openlogi-hid -p openlogi-core --all-targets -- -D warnings
• cargo build -p openlogi --release
• cargo build -p openlogi-gui --release

There is still more hardware coverage needed across other Logitech models, Unifying receivers, and direct Bluetooth devices, but the core Windows path is working on real hardware now.

[pullfrog]

ℹ️ No critical issues — minor suggestions inline.

Reviewed changes — a substantial first cut at Windows support (HID transport, mouse hook, Bluetooth pairing, startup handling, input synthesis) with a full Korean localization and CJK-aware layout fixes. 42 files changed across every major subsystem.

• Windows HID++ transport — new WindowsHidppChannel composites short+long collections; NativeHidWriter provides a Win32 fallback when async-hid writes fail
• Windows mouse hook — WH_MOUSE_LL implementation in openlogi-hook/src/windows.rs with foreground process path query
• Windows Bluetooth pairing — WinRT-based BLE enumeration and pairing via DeviceWatcher/PairAsync
• Windows input synthesis — SendInput-based helpers in openlogi-core/src/binding.rs with macOS→Windows VK code mapping
• Windows startup-at-login — reg.exe-based Run registry key management
• Device enumeration fix — receiver slots that claim online but don't respond to HID++ probes are now marked offline
• Korean localization — full ko column in locales/app.yml, locale detection, and test assertions
• CJK layout fixes — wider panels, line_height on text, vertical language row, centered accessibility footer, platform-aware launch descriptions
• cfg migration — all #[cfg(not(target_os = "macos"))] stubs updated to #[cfg(not(any(target_os = "macos", target_os = "windows")))]

---

Note: 1 inline comment(s) dropped because they did not anchor to lines inside the PR diff:

• crates/openlogi-hook/src/tests.rs:56 (RIGHT) — line 56 (RIGHT) is not inside a diff hunk

  ^{｜ Fix all ➔ ｜ Fix 👍s ➔ ｜ View workflow run ｜ Using Big Pickle (free) (credentials for Anthropic not configured) ｜ 𝕏}

[pullfrog]

Pullfrog stalled

The agent stopped emitting events for 302s and was killed by the activity-timeout watchdog. 45507 events were processed before the failure.

Recent agent stderr

Performing one time database migration, may take a few minutes...
sqlite-migration:done
Database migration complete.

  ^{｜ Rerun failed job ➔ ｜ View workflow run ｜ via Pullfrog ｜ Using Big Pickle (free) (credentials for Anthropic not configured) ｜ 𝕏}

[pullfrog]

Pullfrog stalled

The agent stopped emitting events for 305s and was killed by the activity-timeout watchdog. 1834 events were processed before the failure.

Recent agent stderr

Performing one time database migration, may take a few minutes...
sqlite-migration:done
Database migration complete.

  ^{｜ Rerun failed job ➔ ｜ View workflow run ｜ via Pullfrog ｜ Using Big Pickle (free) (credentials for Anthropic not configured) ｜ 𝕏}

[greptile-apps]

Greptile Summary

This PR adds Windows platform support to OpenLogi, including HID++ device discovery via native Win32 HID APIs, a WH_MOUSE_LL mouse hook for button remapping, Windows Bluetooth pairing through the WinRT device enumeration API, SendInput-based action synthesis, registry-based launch-at-login, and Korean (ko) localization. All previously-flagged issues have been resolved in this revision.

Confidence Score: 5/5

The Windows platform layer is well-structured and all previously-identified concerns have been correctly resolved in this revision.

Every formerly-flagged issue has been addressed: injected synthetic events are now filtered via LLMHF_INJECTED preventing hook re-entrancy, WinRT apartment initialization is balanced by a WinrtApartment RAII drop guard, NTSTATUS codes are formatted as raw hex integers, the Windows Bluetooth button is gated behind cfg!(target_os = "windows"), and error detail in the add-device window now pattern-matches on enum variants directly. The new SendInput dispatch, NativeHidWriter write-method fallback chain, and registry-based launch-at-login all follow correct Win32 idioms. No logic errors or resource leaks were identified.

No files require special attention.

Important Files Changed

Filename	Overview
crates/openlogi-hook/src/windows.rs	New WH_MOUSE_LL hook implementation; correctly filters injected events with LLMHF_INJECTED and uses RAII + UnhookWindowsHookEx to clean up; message-loop and callback lifecycle look sound.
crates/openlogi-hid/src/windows_pairing.rs	WinRT Bluetooth pairing; WinrtApartment RAII guard correctly balances RoInitialize/RoUninitialize on each blocking call; device watcher lifecycle and token cleanup look correct.
crates/openlogi-hid/src/windows_hid.rs	Native HID report writer via CreateFileW; tries three access levels and multiple write APIs; HidHandle/PreparsedData RAII guards prevent handle/data leaks; NTSTATUS formatted as hex integer correctly.
crates/openlogi-gui/src/platform/launch_agent.rs	Windows startup-at-login via reg.exe; args passed as array (no shell injection); delete-nonexistent silently debugs; best-effort semantics match macOS path.
crates/openlogi-gui/src/windows/add_device.rs	Windows Bluetooth button is now correctly guarded with cfg!(target_os = "windows"); windows_pairing_error_detail matches enum variants directly instead of parsing Display strings.
crates/openlogi-core/src/binding.rs	Windows action dispatch via SendInput; exhaustive match on Action variants; mac_virtual_key_to_windows warns and drops on unmapped keys; modifier dedup logic for Cmd+Ctrl is handled.
crates/openlogi-hid/src/transport.rs	Windows path opens a NativeHidWriter alongside async-hid for report writes; is_long_only_collection correctly suppressed with allow(dead_code) on Windows where the Windows path bypasses it.
crates/openlogi-gui/src/watchers/pairing.rs	New Windows pairing session path; correct event-driven state machine; 120s idle deadline; stray controls handled correctly in both receiver and Windows sessions.

_{Reviews (4): Last reviewed commit: "fix: ignore injected Windows mouse event..." | Re-trigger Greptile}