Habit Guard 0.1.0

First public version (alpha).

Added

• Watching for four habits: nail and finger biting, mustache, beard and lip picking, brow, lash and
  hair pulling, and face touching, each with its own zone, dwell time and zone size, and an
  optional wider area (chin and beard line, scalp).
• Face zones measured in eye distances, so they follow distance and head tilt.
• Hand finding and 21-point hand landmarks with OpenCV's DNN module (MediaPipe's models, ONNX
  conversions by the OpenCV Zoo), with hand following between pictures and face memory while a
  hand covers the face.
• A low-cost analysis cadence: motion gate, three speeds, three performance profiles. Measured
  0.34 % of a 16-thread machine with the face in view and the hands away.
• Alarms: sound (three escalating built-in tones, or your own WAV, MP3 or similar file), screen
  curtain (dim or pulsing red frame, click-through), spoken phrase with the system's offline voice,
  desktop notification.
• Escalation while the hand stays, and a timed pause from the tray (15 minutes, 1 hour, 3 hours).
• Statistics: daily counts per habit, 7-day chart, clean streak. Stored locally as plain numbers.
• Tray app with settings, camera preview and statistics windows, in English and Turkish.
• Installers: a per-user Windows installer (optional PATH entry and start at sign-in, upgrades over
  a running app, English and Turkish) and a portable ZIP, a macOS disk image (Apple silicon) and a
  Linux AppImage and tarball, built and smoke-tested by CI, with checksums and build-provenance
  attestations.
• Command line: doctor, bench, stats, reset, autostart, selftest, ctl (control a
  running app: status, pause, resume, toggle, test, settings, preview, stats, quit), --camera,
  --config-dir. ctl talks through small files in the settings folder, not a socket.
• A project website in English and Turkish with an interactive zone demo drawn from the real zone
  geometry, and interface screenshots in both languages in the READMEs and on the site. The camera
  picture in the preview is an illustration, never a photograph.
• Privacy scan (scripts/check_privacy.py), bundle check (scripts/check_bundle.py) and model
  verification (scripts/fetch_models.py) in CI.

Fixed

• Models failed to load from a folder with letters outside the Windows ANSI code page (for example
  a Turkish user name): they are now read in Python and handed to OpenCV as bytes.
• The alarm sound never played in early builds: Qt Multimedia is not part of PySide6-Essentials.
  Sound now uses what the system has (winsound, afplay, paplay, pw-play, aplay or play).
• A sound that is still playing is no longer restarted by another alarm of the same or a lower
  level, so a long custom sound does not stutter when two habits raise their alarms together.
• A raised alarm is released when the camera is lost, the camera is released again when the preview
  is closed while paused, statistics are saved on logoff, and the camera thread survives driver
  errors.

Download

System	File
Windows 10/11 x64	HabitGuard-0.1.0-windows-x64-setup.exe (installer) or HabitGuard-0.1.0-windows-x64-portable.zip
macOS 14+ Apple silicon	HabitGuard-0.1.0-macos-arm64.dmg
Linux x86_64	HabitGuard-0.1.0-linux-x86_64.AppImage or HabitGuard-0.1.0-linux-x86_64.tar.gz

SHA256SUMS.txt lists the checksums; every file also has a build-provenance attestation
(gh attestation verify <file> --repo bugraskl/habit-guard).

Builds are not code-signed yet. Windows SmartScreen may warn: choose More info, then Run
anyway. On macOS, right-click the app and choose Open the first time. The app is signed ad hoc: macOS asks for the camera permission again after each update. Intel Macs can
run from source. More in the
platform notes.

Habit Guard is alpha software: please report false alarms and misses in the
issue tracker.