Reading the conversation log

Updated

Plan requirement

App pluscloud AI Agent Preview
Subscription Suite Team or higher, Support Team or higher
Access Client admin, Client editor, Client user

The Log tab of the inspector shows the conversation log of the running test: its status, actions, API calls, knowledge searches and every message.

Open the log

Open the inspector and choose Log. The tab reads the same information as the conversation logs of the AI agents dashboard, for the test that is running, with your own access. When you show a conversation from the conversation logs in the preview, the tab shows the log of that conversation; see Viewing a logged conversation in the preview. Until a test conversation has started, it asks you to start one. At the top are Refresh, Copy link to conversation log, to share the conversation, and Open in conversation logs, which opens the same conversation in the dashboard. How the inspector works is in Using the inspector.

The inspector on the Log tab. The conversation overview shows the status Automated resolution, Verified resolution Yes, Generative answers and Procedures as sources, Escalation Not attempted and the session duration. Below are log entries for messages, task identification and an API integration called Order status with response code 200, its request parameter and its variable mappings, one of them masked.

Conversation overview

  • Conversation status: Session in progress while the session runs. After it ends, the status the dashboard gives it: Automated resolution, AI agent handled, Escalated, Escalation failed, Not supported or Not understood.
  • Verified resolution: yes or no, once the session has ended.
  • Information sources used: knowledge answers, generative answers, procedures and dialogues, or none.
  • Escalation: not attempted, attempted but not successful, or escalated.
  • Ticket: the Zendesk ticket of the conversation, with a link, when it has one. Test conversations get no ticket; see Viewing a logged conversation in the preview.
  • Resolution states, Labels, Sentiment and Matched segments, when the log has them.
  • Session: when it ended and how long it took, and whether the customer left.

Zendesk decides on an automated resolution after the session ends. While the tab is open, the log refreshes every 20 seconds and a moment after each new message. Refresh reloads it straight away.

The entries

Below the overview, a line counts the entries, the API calls and how many of those failed. The entries follow in the order they happened. Click an entry to see its details.

  • Messages of the customer and of the AI agent.
  • Actions, with their target, inputs and data.
  • API integrations, with the scenario, the type, the response code, the request parameters and the variable mappings. A response code outside the 200 range is marked as failed. Values set to be sanitized in the integration appear as dots, also when you copy the details. How you test those calls in the dashboard is in Testing actions.
  • Knowledge searches, with the search question, the reasoning and the articles found, and which of them were cited.
  • Task identification, with the reasoning, the tasks found and the planned answer, and the procedures that ran.
  • Session ended and the verdict on the automated resolution, once Zendesk has given it.

In a test of a dialogue, a branch or a procedure, an entry that belongs to a dialogue node has Locate in dialogue builder.

When the log stays empty

If the tab says the log is not available, click Refresh, or open the conversation logs in the dashboard. See A test tool is greyed out or does nothing. Who can read the conversation logs is in Who can use AI Agent Preview.

See also

Was this article helpful?

0

Still stuck?

Our support team will take a look with you.

Comments

0 comments

Please sign in to leave a comment.