Release 6.1.3
Release 6.1.3 is a patch release. It adds a setting for which message the minutes in a dialog row are counted from, and fixes the collapsing of earlier emails in replies from Outlook and AI evaluations in Quality Control.
Operator Panel
New
- Minutes in a dialog row from the last message. Under "Settings" → "Operator Panel" → "Timers" there are two new fields. The "Where to count the minutes in a dialog row from" field has two values: "As before: while the client is waiting, from their first message" (default) and "Always from the last message (client's or operator's)". The "Show minutes up to, min" field sets after how many minutes a date is shown instead (from 1 to 1440, empty means 60). The row color and the list order do not change. Until the setting is changed, the list looks as before. Operators see the change after reloading the panel.
Fixed
- Replies from desktop Outlook are collapsed. Previously the whole earlier email, including the "From / Sent / To / Subject" header, was shown open under the client's new reply in the dialog. Now only the new text is visible, and earlier emails are under "Earlier emails in the thread", where you can see who wrote them and when. The standalone email window works the same way. If there is also a "… wrote:" quote under the header, the header no longer stays visible above the collapsed part.
AI QA
Fixed
- An evaluation is credited to the operator profile of the person who started it. An AI evaluation started manually, in bulk or approved is now credited to the operator profile of the person who started it. Agents and automations working through an access key can assess chats: the evaluation is credited to the operator profile of the key's author. Without an operator profile the system shows a clear message instead of the "Failed to create assessment" error.
- Automatic evaluations pass again after a checklist edit. For checklists on newer models, evaluations could fail with a model parameter error after the checklist was edited. Such evaluations now pass. For checklists that already broke, evaluations recover after the next edit of the checklist.