fix(gui): map MX Keys Mini controls to HID++ bindings

[gmpandolfo]

Summary

Applies the same fix as #1258 ("map MX Mechanical Mini controls to HID++
bindings") to the MX Keys Mini (mx_keys_mini depot, model id 0xb369).

Before, the Keys view drew a generic, evenly-spaced Esc/F1–F19 overlay (20
slots) on top of a keyboard that physically has ~14 keys in its top row, so
unbound keys looked bindable and the callouts did not line up with the real
key-caps. This rebuilds the row from the depot's semantic slot metadata and
routes the dedicated media keys through per-device HID++ bindings, exactly like
#1258 does for the Mechanical Mini.

The reconstructed row (14 hotspots):
Esc | F1–F3 (EasySwitch) | F4/F5 (backlight −/+) | Dictation | Emoji | Screen Capture | Mic Mute | Play/Pause | Mute | Volume Down | Volume Up

Backlight down/up have no HID++ control ID, so they stay native F-keys next to
Esc and the EasySwitch keys. The eight media keys already have CIDs in
KEYBOARD_KEY_CIDS and translation keys in every locale, so this needs no
openlogi-core change, no KEYBOARD_KEY_CIDS change, no locale change, and no
PROTOCOL_VERSION bump — it is contained to the desktop GUI.

Depends on #1258 — draft, do not merge yet

This is built on top of #1258, whose infrastructure it uses in full
(BindingTarget, key_definitions, commit_target, per-device editor reset).
The two commits 79d9ca80 and f615b03d on this branch are @aztkgeek's #1258,
included only so the branch compiles; once #1258 merges I will rebase this onto
master so the diff is just the MX Keys Mini commit(s).

Changes

• openlogi-desktop: mx_keys_mini_definitions() in
  features/keyboard/function_row.rs — recognises model id 0xb369, rebuilds
  the top row from device_easyswitch_image + device_keys_image slots, maps
  the eight divertable media slots to their ButtonId, keeps Esc / EasySwitch /
  backlight as global F-keys. New mx_keys_mini_control() slot→ButtonId map
  (kept separate from fix(gui): map MX Mechanical Mini controls to HID++ bindings #1258's semantic_top_row_button so the Mechanical Mini is
  untouched). Short callout labels for the newly device-targeted buttons. Three
  tests + a synthetic mx_keys_mini_asset() mirroring the shipped depot
  metadata.
• openlogi-ui: register action-icons/moon.svg in ACTION_ICONS — the file
  was vendored and referenced by Action::Sleep in the picker but never
  registered, so every picker frame showing Sleep logged
  could not find asset at path "action-icons/moon.svg". (Also on its way via a
  separate PR; harmless if it lands there first.)

Screenshots

Before — generic 20-slot Esc/F1–F19 overlay, keys don't match the hardware:

After — 14 real keys; the eight media keys are per-device HID++ controls,
Esc / EasySwitch / backlight stay native:

Testing

Local gate, on the #1258 base:

export RUSTFLAGS="-D warnings"
cargo fmt --all -- --check                                              # clean
cargo clippy -p openlogi-ui -p openlogi-desktop -p openlogi-overlay --all-targets -- -D warnings   # clean
cargo test  -p openlogi-ui -p openlogi-desktop -p openlogi-overlay      # 200 + 17 + 3 + 1 passed (incl. 3 new)
cargo clippy --workspace --all-targets -- -D warnings                   # clean
RUSTDOCFLAGS=-D warnings cargo doc --workspace --no-deps --document-private-items \
  --exclude openlogi-ui --exclude openlogi-desktop --exclude openlogi-overlay --exclude openlogi-agent   # clean

No wire-type, locale, or cfg-gated change, so those gates don't apply.

Runtime-verified on real hardware (MX Keys Mini, Bluetooth, Linux): the Keys
view renders the 14-key row aligned to the physical key-caps; binding Volume
Down / Screen Capture / etc. commits per-device and the HID++ divert takes
effect; unbinding restores native behaviour; switching devices clears stale
selection.

[greptile-apps]

Confidence Score: 4/5

The PR appears safe to merge after hardening the MX Keys Mini metadata validation; the remaining finding is a non-blocking robustness issue.

Findings

1.  Loose Slot Count Validation ▶

Summary

• Reconstructs compact keyboard hotspots from depot assignment metadata.
• Separates global Esc/F-key bindings from per-device HID++ control bindings.
• Adds five Mechanical Mini control identifiers and their HID++ capture mappings.
• Resets editor state when navigating away or switching devices.
• Registers the previously missing Sleep action icon and adds control labels to locale catalogs.
• Includes the required IPC version update for the dependency changes.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  A[Selected keyboard and depot metadata] --> B{Recognized Mini model and valid slots?}
  B -->|No| C[Generic global Esc/F-key layout]
  B -->|Yes| D[Semantic physical hotspots]
  D --> E{Hotspot type}
  E -->|Esc, EasySwitch, backlight| F[Global keyboard binding]
  E -->|Dedicated HID++ control| G[Per-device button binding]
  F --> H[OS keyboard capture path]
  G --> I[Agent HID++ keyboard session]
  H --> J[Configured action]
  I --> J

Loading

_{Reviews (1) · Last reviewed commit: "fix(gui): register moon.svg so the Sleep..."}

[greptile-apps]

Loose Slot Count Validation

The MX Keys Mini has a fixed 10-slot device_keys_image row, but this check accepts any count from 8 through 12 without validating the expected semantic slots or their uniqueness. Because every accepted assignment is rendered, incomplete metadata can omit physical controls, while duplicate or unknown slots can become duplicate HID++ targets or positional global F-key bindings. Validating the expected 10-slot semantic set before enabling this layout would avoid misleading or incorrectly targeted hotspots.

Knowledge Base Used: Desktop application shell

[AprilNEA]

Thanks for digging into this! It's superseded by #1604 (now on master), which keys keyboard controls by their HID++ control ID instead of adding one variant per key. The Keys tab now places keys by the control IDs in the device render's markers. Closing, but shout if something here isn't covered.