Repository navigation
omna 0.6.15
Smart masking is now on by default. It catches names and addresses written as ordinary sentences, on top of the rules. Every machine gets a compact open-source model (Apache-2.0, ~64 MB); machines with 16 GB of memory or more also get the OpenAI privacy filter, and both read every text. Models download in the background — starting Omna never waits, and every message gets fast masking until a model is ready.
omna smart on|off— saved;omna status, the installer and a one-time Mac notification say what smart masking does, whether it is on or off.omna categories off PHONE(andon) — let a category through: names, emails, phones, addresses, government IDs, money, health, work, devices and links, dates of birth. Secrets are always masked and cannot be switched off.- India: fast masking now finds Indian names and GSTIN, UPI IDs, driving licences, vehicle numbers, ABHA, UAN, PIN codes and mobiles written with a leading 0 — alongside Aadhaar, PAN, Voter ID and passport.
- Names without a model: a first name and surname written together ("Emma Clarke") are masked by fast masking, from public name lists. One word alone never counts.
- Smart masking no longer cuts a word in half (a model once marked the "U" of "UPI").
Needs the masking engine omna-pii-mask 0.2.18 (installed automatically).
Install or upgrade: curl -fsSL https://omna.dev/cli/install.sh | sh · brew upgrade --cask gaurjin/omna/omna