Settings
Tune how sensitive Guardian Line is, and set up your family's safety net.
Language
Call language
Language the caller is speaking, used only to match scam phrases. French, Mandarin, Vietnamese, and Tagalog cover core scam patterns but the app's own interface stays in English or Spanish.
Appearance
Theme, color mode, and readability.
Swaps red/green for blue/orange in the Trust Meter.
Sensitivity
Higher sensitivity flags risk earlier, but may raise more false positives.
Privacy & accessibility
Voice analysis always runs on-device. When enabled, no data of any kind is transmitted off this device.
Speak alerts aloud for anyone who can't look at the screen mid-call.
Get a system notification if a call reaches the danger band while this tab isn't focused.
Safe Word Check
A private phrase only your family knows. Register one per relative if you'd like — Guardian Line will prompt for it if a call turns risky.
Family Circle
With your consent, a trusted contact can be notified in real time if a call crosses your chosen threshold.
Community patterns
This build has no server, so there's no live community feed — but you can export the phrases you've flagged as a plain JSON file (no names or transcripts, just category + phrase) and import a file someone else shares with you.
Backup
Export your settings to a file, or restore them on another device.
LLM classifier (experimental)
Bring your own OpenAI-compatible API key to compare an LLM classification against the built-in offline heuristic. Never used in a live call unless you wire it in yourself — this is a side-by-side tester.