Mandata 2.8.0

A cleaner Settings screen, and a full review of the app: steadier reading, a safer database, a sturdier server.

• Settings: every tab is now laid out the same way. Settings are grouped in cards, and each one is a row: what it does on the left, its control on the right, all aligned. Long explanations sit under the label instead of in it.
• Task reading: a short name the AI writes ("Ann") no longer matches a longer one ("Joanna"). "ASAP" or "urgent" never becomes a deadline unless a day is named. "By Friday" counts from the email that says it, and "Friday next week" is read as next week. A task is never dropped because a thread holds many.
• Duplicates: two different pieces of work in one thread are no longer folded into one task when the AI is unsure.
• Review when idle: new mail arriving during a review stops it from applying stale verdicts. A task is closed only for a stated reason, and never by the email that asked for it. A split task's parts no longer copy a deadline their own sentence does not set.
• Database: faster thread lists and name lookups. Backups are written whole or not at all. A damaged backup is refused before anything is replaced.
• Server and mail: an email is never sent twice after a network time-out. Microsoft 365 waits as asked when it is busy. The app recovers when Outlook's helper fails to start. Turning phone access off now disconnects the phone at once.
• Screens: dialogs work fully from the keyboard, and a dialog that opens by itself never takes the keys you are typing. A failed question in Ask Mandata goes back in the box. The weekly report prints without tinted fills.
• The AI on the processor: on computers without a separate graphics card, Mandata now sees that the model runs on the processor and uses the lighter limits, so readings stop timing out.