Conversation History

The Conversation History section provides a comprehensive log of all interactions between users and AI Agents. This tool is essential for auditing conversations, troubleshooting flow logic, and gaining insights into user behavior.

Access Conversation History

To view the conversation history of an AI Agent:

  1. Select the AI Agent.
  2. From the History & Logs menu, click Conversation History. The Conversation History page appears.

Conversation History Overview

The Conversation History page displays a list of conversation sessions, which you can organize and filter to find specific data:

  • Filters. Narrow down the history by Date range, Activity Events (AE), Correlation Id, and Conversation Id. Click More filters to access advanced criteria like specific channels or users.

  • Group by Interaction. Bundles messages into distinct sessions based on user activity. Use it to view the Conversation transcript, the conversation path in a tree view and detailed conversation activity insights. Grouping activities by conversations enables you to view per conversation the start and end time of the conversation, duration time and how many turns (message exchanges) the conversation had. It also provides you with an overview of all activities registered throughout the conversation. This option allows debugging messages directly in the flow editor.

  • Show all Messages. Provides a flat list of every individual message sent within the selected parameters. Use it to debug the flows and flows steps in the Flow Diagram by clicking on the desired flow name or flow step name.

Filter conversation history data

Druid provides you with various filtering options. The following filters are available to debug your conversations:

When you access the Conversion History, only 4 filtering options are shown on the page.

Filter option

Description

Date range

The time interval from which you want to show conversation history data.

Activity Events (AE)

This filter allows multiple selection. The following activity events are available:

  • No error – There were no errors during the conversation.
  • Utterance not found – The bot did not match what the user said to none of the exiting flows utterance.
  • Conversation path not found - The number of inputs that don't match the flow step validation criteria, hero or choice steps. The AI Agent tries to understand what the user says and convert into expected format. If it fails, it responds the user with “I do not understand, please repeat”.
  • Invalid input – The number of messages where inside a dialog the user input was not the one the AI Agent expected. If for example, we have a prompt step that expects in input mapping a date of birth (type DateTime) and the user inputs the name (type string), the AI Agent will trigger the special message InputValueInterpretationError.
  • Security violation – The client tried performing an action for specific permissions are required (e.g., a user who doesn’t have the Manager security role assigned tries accessing a flow which has set the Manager security role set).
  • Connected with Agent – The client connected with an agent in a live chat conversation.
  • Integration error – There were integration errors during the conversation.
  • Transfer to agent failed – During the conversation, transferring a waiting client to an agent has failed.
  • Outside working hours - The message was sent outside working hours. This event is available in Druid 8.3 and higher.

If you want to select see all activity events, select All Activity events.

NOTE: Connected with Agent and Transfer to agent failed are available only for activities grouped by conversations.

Correlation Identifier

Each message sent by the bot or by the user, has a corresponding correlation ID. You can use the correlation ID to debug a specific message.

Conversation Id

Shows details of the specific conversation associated with the provided Id.

NOTE: On a conversation Id there can be more activity events.

 

To see more filtering options, click the More filtersbutton. The following additional filters appear:

Filter option

Description

User

The name of the user.

Flow

The name of the flow.

NOTE: You can filter the conversation history of a Druid Conductor by a specific flow that belongs to an orchestrated AI Agent, first select the child bot from Orchestrated AI Agents and then, select the desired flow.

Channel

The channel used by the user to AI Agent with the bot.

Agent Name

The name of the agent.

Context Data

Filter by information stored in the entity [[ChatActivityData]]. This filter applies only to activities grouped by conversation.

Origin URL

The base URL of the webpage where the web chat is hosted.

Analytics tag

The tag set on a flow in section Analytics.

Path Name

The full path to the web page from which the user initiates the chat.
Connected AI Agents

To debug only the conversations that were triggered from a Druid Conductor, from the Orchestrated AI Agents filter, select the Druid Conductor and click Apply.

Flow-level Conversation History

In addition to the global conversation history, you can access logs specific to a single flow. This is particularly useful for analyzing how users interact with a specific piece of logic or identifying where a particular flow might be failing.

To access the conversation history from a flow:

  1. Select the AI Agent and solution.

  2. Go to Flows and select the specific flow in the Flow Designer (diagram view).

  3. From the Actions menu, select Analytics.

The system will display the Conversation History filtered automatically for that specific flow, allowing you to see every interaction that triggered those specific steps.

Exporting Conversation History

You can extract conversation data for external analysis, reporting, or auditing by using the Export Excel button from both global and flow-level Conversation History.

NOTE: Starting with version 9.20, the ability to export data is restricted. To use the Export Excel feature, a Druid administrator must grant you the 'Export conversation history' permission.

To export the conversation history:

  1. Use the filters to isolate the data you wish to extract.
  2. Click the Export Excel button located at the top right of the page.
    3. HINT: If the log file is big, it might take a couple of seconds for the log file to be ready for download.

An Excel file containing the logs will be downloaded automatically to the default download folder on your PC.

The exported Excel file contains two sheets:

  • The first sheet provides a granular view of the interaction data, including:
    • AI Agent Name & User Name. Identifies the AI Agent involved and the user (e.g., admin or anonymous).
    • Channel Id. Specifies the communication platform (e.g., directline, webchat).
    • Message & Language. The exact user input and the NLU-detected language.
    • Status & Score. Technical match details (e.g., SingleMatch) and the confidence score.
    • Flow Details. Identification strings for the specific flows triggered (e.g., Flow1 Id, Flow1 Name).
  • The second sheet provides details on who and when downloaded the log file, from which tenant and for AI Agent.