The first analysis is slow: that is normal
On the first click of Analyze after launching the app, the local detection engine loads into memory. That takes about fifteen seconds, and the indicator says so while you wait.
Every subsequent analysis takes one to two seconds.
The initial wait comes back after each restart of the app. It has an upside: it is proof the engine runs on your machine, not in the cloud. A remote service would have nothing to load.
Tip: at the start of your workday, type anything in the composer and analyze once. The engine stays warm for the rest of the day.
An analysis really never finishes
If an analysis runs well past a minute:
- Check the text length. A very long document or a massive paste takes longer: the engine splits the text into windows and processes them one by one.
- Restart the app. Your conversations, files and vaults are persisted and encrypted: you lose nothing (except an unsent draft, so copy it first). After restarting, analyze again; allow the fifteen-second initial load.
- Still stuck? Write to us describing the approximate text length and your machine. Never include the client content itself in your message.
The analysis fails with an error
A failed analysis blocks the send: that is the fail-closed principle, never an accidental clear-text send. Try the analysis again; if it keeps failing, restart the app, then write to us. The failure is logged in the audit log with a technical code, without your content.