Files and data
Everything ozen keeps is in your ozen folder, ~/ozen, on your Mac.
| Path | What it holds | Shared? |
|---|---|---|
transcript.txt |
The readable transcript, one line per utterance | No |
lines.jsonl |
Every line with its id, time, source, speaker guess, text and voiceprint | No |
tags.json |
Speakers you set by tagging | No |
fixes.json |
Text corrections you made | No |
learned.json |
Hint words and automatic corrections learned from fixes | No |
fixes/ |
Audio of fixed lines with the right text, for future fine-tuning | No |
ignore.json |
Voiceprints of ignored voices | No |
vocab.txt |
Terms the transcriber should spell right; edit freely | No (in the ozen repo) |
places.json |
Your places, including where you live and work | No |
voices/ |
Names, voiceprints and accuracy history. Its own local git repository; keep it one | No |
context/ |
Folders written for AI agents | No |
chunks/ |
Audio waiting to be transcribed; deleted once transcribed | No |
recent/ |
The last 20 transcribed chunks, for eval.py. Set OZEN_KEEP_AUDIO=0 to keep none |
No |
screen.png, screen-small.png |
Last ozen look screenshot |
No |
start.log |
Logs from the recorder and transcriber | No |
Personal files (places.json, ignore.json, fixes*, context/) are gitignored in the ozen repo, so they
aren’t committed by accident.
Deleting data
Section titled “Deleting data”- A meeting:
delete_meetingover MCP, which removes its lines, fixes and tags. - Specific lines:
delete_linesover MCP. - A person:
delete_personover MCP. - Everything: stop ozen and delete
~/ozen.