Skip to content
DevEmperor edited this page Oct 31, 2026 · 2 revisions

SayOVA can remember things about you across conversations — but only what you ask it to. It doesn't quietly collect facts from what you say; you decide what it knows, and you can see and delete every single fact.

In a conversation: Remember that I'm vegetarian. Remembered: Vegetarian Settings → Memory: Use memory, only what you ask for The remembered facts: Vegetarian, Lives in Hamburg, Has a dog called Bruno

Remembering

"Remember that I'm vegetarian." · "Remember that my sister's birthday is on 3 May." · "Keep in mind that I work night shifts." · "Remember that I prefer metric units."

SayOVA confirms with Remembered and a note in the transcript (Remembered: Vegetarian). From now on every conversation knows: ask "What could I cook tonight?" and it suggests something without meat.

When something changes, just say so — "I moved to Munich" — and SayOVA replaces the old fact instead of keeping both.

Forgetting

"Forget that I'm vegetarian." · "Forget where I live." · "What do you know about me?"

SayOVA confirms with Forgotten. Ask what it knows, and it tells you.

Settings → Memory

Use memory switched on; Remembered · 5/30
  • Use memory — on by default. Switched off, SayOVA neither remembers nor uses any facts; the facts already saved are kept until you turn it on again or delete them.
  • Remembered · 5/30 — every fact, oldest first. Tap one to forget it.
  • Forget everything — deletes all facts at once.

SayOVA keeps up to 30 facts of up to 150 characters each. When the memory is full, SayOVA asks you to let go of one first.


Where the facts are

The facts are stored only on the watch, never in a cloud and never in a backup. To be useful, they are given to your AI provider at the start of every conversation, as part of SayOVA's instructions — so don't ask SayOVA to remember anything you wouldn't tell your provider. Deleting a fact on the watch removes it from every later conversation.

See also Privacy and security.

Clone this wiki locally