Know-Me profile

What it collects

Dashboard ▸ Know-Me ▸ Run interview asks five questions, one per screen. Every one is skippable.

Question	Example answer	What it fixes
What should we call you?	Alex Chen	Your own name being misspelled in your own writing.
What do you do?	iOS engineer at a small startup	Gives cleanup domain context for ambiguous words.
Who do you mention most?	Priya, Sam, Dr. Okafor	Colleagues' names, the single most common complaint.
Any words it keeps getting wrong?	WhisperKit, TestFlight, Kubernetes	Your jargon.
How do you like to sound?	concise, no exclamation marks, Oxford comma	Stops cleanup from making you sound like someone else.

Onboarding asks only for your name; the rest live in the dashboard interview so first-run setup stays short.

How it is used

Your answers become a short personal-context block supplied to the cleanup model alongside your dictionary — who you are, who you mention, which terms to spell correctly, and how you like to sound. It is why cleanup can pick the right spelling of a colleague's name from an ambiguous transcript.

Running the interview also seeds your dictionary with the names and technical terms you list, without disturbing entries you added yourself.

Reviewing, re-running, clearing

The Know-Me pane shows what is currently stored. Re-run interview walks the five questions again; Clear empties the profile entirely. Both take effect on your next dictation.

Where it lives

~/Library/Application Support/OpenVoiceFlow/profile.json on your Mac. If you enable cloud cleanup, this context is sent with each cleanup request to the provider you chose, under your own API key — that is what makes it work. If that trade is not one you want, use local Ollama cleanup, which keeps the same behaviour entirely on your machine, or leave cleanup off and skip the interview.