Release 5.20.0
A minor release that includes all changes of release 5.19.5 and the earlier 5.19.x releases. The widget setup page has been rebuilt from scratch: every widget setting with a live preview, a gallery of scenes and presets shared by the team. Operators can pin important messages that stay visible in all of a client's dialogs. SLA metrics are now created by choosing a type, and duplicates are no longer possible. Native statistics gets an "Instance Agent" tab about AI Assistant usage. Quality assurance gains auto-assessment settings in the module and a bulk assessment launched from the chats table. Some new capabilities are off by default and enabled on request — see "Please note".
Please note
- Features that are off by default. Pinned messages in Operator Line and the new AI QA auto-assessment and bulk assessment tools are enabled on the instance on request — contact your ConnectiveOne manager. Until they are enabled, the system behaves as before.
- Google Forms: the file name appears in the request only after the scenario is updated. The new variable with the original file name is available right after the update, but the request card shows the real document name only if your scenario uses this variable. Requests created earlier do not change.
Widget
New
-
Widget Lab — a new widget setup page. The page that opens from the bot's widget settings has been rebuilt from scratch. You set up the chat for your site and see the result right away:
- every widget setting in one place — a simple mode with what you need to start and an advanced mode with all settings and a field search;
- a live preview of the widget that changes as you edit, and an "Open as site" button;
- colors and themes — ready palettes, your own colors for the header, bubbles and button, and gradients;
- a gallery of scenes — the widget in typical states (dialog, scenario keyboard, inline buttons, carousel, product cards and others) so you can check how your colors and texts look everywhere at once;
- the embed code for your site, ready to copy, and the import of code you already have; in advanced mode, ready code examples for controlling the widget from your site's pages;
- interface texts of the widget, which you can translate and change.
The page uses the same light or dark theme as the Constructor, and its columns can be resized. On the first visit a short page tour opens; you can restart it at any time with the "?" button.
-
Presets shared by the team. A widget setup can be saved as a named preset of the bot. Presets are stored on the server, so all colleagues with access to the bot see them; a preset can be updated, renamed, copied as a link, exported and imported as a file. Viewing presets requires the right to view the bot settings, and saving or deleting them requires the right to save the bot settings. When a bot is deleted, its presets are deleted too.
Operator panel
New
- Pinned messages in a dialog. An operator can pin an important message — an agreement, an invoice number, a note like "VIP, call after 2 pm". Pinned messages appear in a bar under the chat header for every operator, not only in this dialog but in all later dialogs of the same client. Up to 3 messages can be pinned per client. Pinning and unpinning is available to operators whose role has the right to create internal notes. The client sees nothing. The feature is off by default and enabled on the instance on request.
Fixed
- A client message updates the chat card in the list right away. On some instances in the new operator panel design, a client message in a dialog that the operator had not opened in this browser tab did not reach its card in the "My Chats" list: the card kept the old preview, did not move up and made no sound until the operator switched folders or reloaded the page. Now the new message appears in the card at once, the card moves to the top and the unread counter grows. If the Notification Center is turned off, the usual sound and browser notification play, as on other instances.
Settings
Improved
- SLA metrics: choosing the metric type and protection against duplicates. The metric form in Settings → Metrics now has a "Metric type" field. A standard type (TFR, TTC, NRT, TIS) fills in the identifier and the description by itself — you only set the time. A custom metric keeps a free identifier, as before, and it is converted to upper case. Previously a typo in the identifier (for example, "FRT" instead of "TFR") created a metric whose timer never started, and two metrics with the same identifier could coexist, so it was unclear which of them would fire. Now the form shows "A metric with this identifier already exists" and does not let you save a duplicate. Existing metrics open with the type detected from their identifier.
Statistics
New
- The "Instance Agent" tab in native statistics. Native statistics has a new "Instance Agent" tab showing how employees use the AI Assistant: the number of requests for the period, how many employees used it, the shares of the "Agent", "Ask" and "Plan" modes, the share of failures and the response time (median and p95). Daily charts show requests by mode, by role and for the five most active employees, active users per day and failures by type. The "Users" and "Roles" tables can be exported to CSV or XLSX. The tab is visible to everyone with access to native statistics. Requests made before release 5.19.1 are counted in the totals and shown as "Unknown" in the breakdowns.
Quality assurance
New
These capabilities are available after the new AI QA auto-assessment is enabled on the instance on request and the main "All bots" setting is created. Until then, quality assurance works as before.
- Auto-assessment settings in the module. The Quality Assurance module has a new "Auto-assessment settings" tab. A quality manager can set everything without editing scenarios: skip rules (for example, the client did not reply, blocked the bot, the dialog was too short or was transferred), the threshold of a short dialog, what to do with a chat without a topic, topic routes to checklists (several topics at once, including "All other topics"), a separate setting for an individual bot and who is not auto-assessed — by role profiles and by exceptions for individual operators.
- The checklist is suggested by the topic. In both manual assessment windows, the checklist is filled in by the topic route from the auto-assessment settings. If skip rules are turned on for single assessments and a rule applies to the chat, the window shows "The chat was not assessed" with the reason, and no assessment is created.
- The reason for skipping auto-assessment in the dialog card. If auto-assessment skipped a chat, the dialog card shows the read-only "AI QA skip reason" row.
- The topic of a closed chat can be corrected automatically. During auto-assessment on chat close, Fast Line Pro can set a more accurate topic for the chat — only from the topics of this bot, only with high confidence and only if nobody changed the topic in the meantime. The change is recorded in the chat history.
- Bulk assessment from the chats table. In both versions of the operator panel chats table, "Evaluate dialogs" and the new "Bulk assessment" action open one window. Chats can be taken from the selection in the table, from a CSV file or a pasted list of IDs, or from a protected link shared by a colleague. Before the start you see which checklist each chat goes to by its topic, and you can change it for this run. A chat can be assessed again on every run.
- A protected link to a chat selection. In the bulk assessment window you can create a link with the checked chat list and share it with a colleague of the same instance. The link is valid for a limited time and starts only one run; if a run has already been started from it, the window says so and offers to open that run.
- The bulk assessment history. The Quality Assurance module shows every bulk run: who started it and when, where the chats came from, how many were assessed, skipped by rules, not assessed by topic or ended with an error. A run's page shows the result of every chat, and you can add a comment to the run.
- New role rights. The role editor has rights for the auto-assessment settings, for comments on bulk runs and for creating a selection link. The role card shows the role's AI QA profile.
- A checklist in use cannot be deleted or deactivated by accident. If a checklist is used in the auto-assessment settings, the system does not let you delete or deactivate it and shows where it is used.
- The scenario assessment node shows "Managed in the AI QA module" when its settings come from the module.
Integrations
Improved
- Google Forms: the original name of an attached file. In the "scenario" mode of Google Forms, the list of scenario variables now has
gf_<field_id>_files— a list of the files uploaded to the field with their original names and sizes. Previously the scenario received only a link to the file, and the request card showed a service identifier instead of the document name. The existing variables with the file link have not changed, so scenarios that use them keep working. To show the real document name in the request, the scenario has to use the new variable.