Email Assist

Simulation Mode is ON – emails are drafted and approved but never actually sent. Disable in Settings when ready.

Settings

Configure Email Assist behaviour and AI preferences

LLM / Ollama Settings

Ollama Base URL

Main Model

Classifier Model

Behaviour

Simulation Mode (Dry Run)

Enabled by default. Emails are fetched, filtered, and drafted but never actually sent via SMTP. Turn this off only when you are satisfied with the generated drafts and tone profile.

Freeze Tone Profile

Stop updating the tone profile from new emails. Use the current profile only.

Auto-Send on Approval

Automatically send replies as soon as they are approved. Use with caution. Has no effect while Simulation Mode is enabled.

Draft Safety (Auto-Draft Limits)

Prevent overloading the local AI with stale or excessive drafts.

Recency Window (hours)

Only generate drafts for emails received within this many hours. Older emails are stored but not drafted.

Max Drafts Per Run (per account)

If more qualifying emails are found in a single run, only the most recent N receive drafts. Protects against generating dozens of drafts you won't use.

Privacy & Data Security

Control what email content is stored locally in the database.

Privacy Mode – Do Not Store Email Bodies

When enabled, email body text is used for AI classification and draft generation during the fetch run, but is not persisted to the local database. Only metadata (sender, subject, timestamps) is stored. The body is permanently discarded after the run, so it will not be available for re-generating drafts, viewing original emails in the UI, or any future processing.

Additional Blocked Domains

One domain per line. These are added to the built-in block list.