What's new in v0.0.4
- On-device AI on macOS — Sentinel Monitor can now run its vision model entirely on your Mac (Apple Silicon GPU). No cloud account, no Docker, no setup: the local AI engine ships inside the app.
- Model picker on first launch — choose the AI model size that fits your machine (2B / 4B / 8B); the app downloads it for you after install.
- Editable camera addresses per tile — type or fix an RTSP URL directly on any camera tile; changes are saved automatically.
- Built-in user guide — a new Help button opens a condensed guide in-app, plus a full bilingual (繁體中文 / English) operator manual.
- Cleaner monitoring screen — the top control area and the inference-backend panel now collapse into a slim bar, giving the camera grid more room.
- Camera limits by plan — the number of simultaneous cameras now follows your license tier.
Requirements & known notes
- macOS: Apple Silicon (M1 or later) is required for the built-in local AI engine; macOS on Intel is not supported for local inference.
- The installers are not yet code-signed: macOS Gatekeeper and Windows SmartScreen will show a one-time warning on first launch (right-click → Open on macOS, or "More info → Run anyway" on Windows).
- The local AI model is downloaded on first run (roughly 2–6 GB depending on the size you pick); allow time and disk space for it.
- Cloud (Gemini) and custom local endpoints remain available as alternative inference backends.
- Linux (AppImage) and Windows builds use cloud or custom-endpoint inference; the bundled local engine is macOS-only in this release.