debouncing-strategies

Store the last-shown timestamp in DataStore and compare elapsed hours against a minimum interval before emitting. A decision function returns true on first run, then only after the configured interval (e.g., 72 hours) has passed.

Read the current hour from a Calendar instance and check it against user-configured start and end hours, handling ranges that wrap past midnight. If the current time falls inside quiet hours, suppress the insight or notification regardless of other rules.

Keep a daily counter and date string in DataStore, incrementing on each send and resetting when the date changes. Block new notifications once the daily cap (for example 3) is reached, and prioritize competing types such as journal reminders over insights.

Use WorkManager or coroutines instead of Handler.postDelayed, since Handler timers are less robust across process death and app restarts. The skill explicitly lists Handler.postDelayed as an anti-pattern for this use case.

Comparing raw Long millisecond values avoids timezone and locale errors that Date comparisons can introduce. Convert elapsed milliseconds to hours or days with TimeUnit for interval checks.

Record 'not useful' feedback per insight type in a local database and disable a rule after repeated negative marks, such as three within 30 days. Also reduce frequency automatically when users consistently ignore insights.