Activity

The top of the page is four stat cards: total sessions, credits used, last 7 days of activity, and average messages per session. Quick glance — if any number looks wrong, drill into the filters.

Two filters above the session list:

• Channel — widget, WhatsApp, Slack, Messenger, Instagram, Email, Zapier, API
• Date range — calendar popover, presets for 7/14/30/90 days

On desktop, the page splits into two panes: session list on the left, the open conversation on the right. Click any session to load its full message history. On mobile, the conversation opens in a modal.

Each message shows its role (user / assistant / operator / system) and a timestamp. Operator messages — sent by a human takeover — and navigation messages (the user clicked a link inside the widget) have their own badges.

Hover any AI message to reveal an edit affordance. Open it, fix the response, save. Two things happen:

• The conversation log shows your edit (with an ‘edited’ marker)
• The edit feeds back into training — next time a similar question comes in, the agent leans toward your corrected version

This is the highest-leverage feedback loop in the product. A few edits a week measurably lifts answer quality.

A world map at the bottom of the page with the top eight cities by session count. Useful for sanity-checking timezone coverage and spotting unexpected geographies.

• Free — Activity page is blocked entirely; upgrade prompt instead
• Go+ — basic activity view, filters, replay, edit
• Plus / Enterprise — adds ‘advanced’ metadata: which model + which sources + which prompt version handled each response