* Rename voice-control to hex everywhere, without migrations
The Cargo package and binary are now hex, every OS keeps its data in a
hex folder, and VOICE_CONTROL_* variables are HEX_*. Hex is a fresh
app: the macOS voice-control rename, its symlink, and the Windows and
Linux beta settings imports are gone. The Windows installer gets a new
AppId so it never takes over an old voice-control install's entry.
* Format
* Drop the Windows paste-last default only the beta import used
The Linux host, the single-instance socket, the HUD, and the window run in
one process. A relaunch forwards Activate with the launcher's activation
token; hex dictation, voice-action, paste-last, status, and quit forward
over the same socket, and shortcut commands go straight to the listener.
Without an instance a launch starts the managed unit (Type=notify) or
becomes Hex. The Rune window is the default on rune-ui builds and talks to
the host directly; GPUI stays behind --ui gpui. Quit Hex lives in the
sidebar.
The module map, Linux and Windows guides, roadmap, and the shared-desktop
plan describe the finished architecture: one AppWindow with per-OS modules,
one settings.json with one-time migrations, the shared mode runtime, and
interface translations. The Linux dictation check writes settings.json.
* feat(dictation): stream every dictation through one shared engine
The transcription engine now builds on macOS, Linux, and Windows. The capture
owns its live stream, so every OS feeds it the same way. macOS drives its
warm transcriber through the engine and drops pause chunking for dictation;
Linux streams on Vulkan with the 200 ms release tail. Meetings keep the pause
segmenter.
* test(linux): replay the dictation fixture again with a CPU-forced live stream
* refactor(dictation): delete helpers that only served pause chunking
* docs(dictation): describe streaming on every OS and drop pause chunking
Adds the opt-in Metal stream harness and records its first run on an M4 Air.
* refactor(settings): remove the Transcribe during pauses setting now that dictation streams
Update download links and record signed artifact verification. Explicitly select Ctrl-V for the GTK virtual-microphone fixture rather than inheriting the legacy terminal paste preference.
* feat(linux): add wlroots Wayland support and Nix packaging
* fix(linux): preserve pending status and verify native coexistence
* test(linux): verify real dictation through a virtual microphone
* test(linux): bound smoke cleanup and check microphone release
* test(linux): isolate installer XDG paths from the host
* test(linux): exercise the desktop dictation lifecycle
* refactor(linux): simplify worker completion and input snapshots