All interactions between users and the bot are captured and available in the Conversation History workspace in the Bot Builder interface. Using the Conversation History, Administrators, and Skill Developers can view the bot and user interactions grouped as conversations and understand the details of underlying system events. This is particularly useful when troubleshooting problems that occur with the bot, configured channels, or integrations. A conversation begins when a user enters a phrase and starts a skill. The conversation ends when the skill fulfilled by executing all steps as part of the conversation flow, the user ends the conversation by //abort command, due to idle time out in the conversation flow and current skill branches to the next skill. All the system skills and small talk conversations are also captured as part of conversation history. The Conversation history is stored at the bot level within the tenant.
On the Bot Builder, click “Conversation History” from the Bots menu on the left navigation pane as shown below.
Clicking Conversation History opens the following workspace.
The history of chat interactions is grouped as Conversations. The Conversations History has the following filters:
Channels
Users
Skills Category
Skills
Date Range - filters by the conversation start date and time
Events
The default filter applied is All for all the fields and the Last 24 hours data will be shown for the Date Range filter. In the filters section, you can select the checkbox to see the list of conversations that resulted in Platform Errors/System Errors. This will help to quickly identify the conversations that errored.
The most recent Conversation is shown on the top of the list as the default sort order is the Start Date based on the Date Range filter. A user can alter the sort order by selecting the field from the dropdown next to the field name. Sorting can be performed on the selected field. You may apply the filters on the results for the columns Channels, Users, Skills Category, Skills, Start Date and Time. You can hide the filters section to have more space for the conversations.
The conversation log contains the list of events that occurred during the specific conversation between the bot and the user. Only one conversation's events can be viewed at a time. The date in the log for each Conversation is displayed in chronological order of the events that occurred with the oldest event on top and newest on the bottom. Therefore scrolling down through the events follows the order of the conversation from start to finish.
The Conversation’s events will be shown expanded by default. A user can unselect the Expand Events checkbox in the details view of the Conversation to have a minimized view.
Bot Response is highlighted in orange color and user messages are in grey color. If there is a Date change in between an ongoing conversation then the date will be mentioned in the conversation Timeline.
The following are the list of events that are captured as part of the conversation and their associated meaning:
Channel Event Received: The Virtual Agent received an event from a channel which it will process and respond.
User Message Received: The Virtual Agent received a user message from a channel which it will process and respond.
Message Validated: The Virtual Agent has validated the authenticity of the incoming message in order to verify and confirm that the message came from the stated sender.
Access Permissions Checked: The Virtual Agent performed the pre-requisite permission checks to determine whether the user is allowed and permitted to interact with the Virtual Agent.
User Authenticated: The Virtual Agent has auto-verified and confirmed the identity of the user
User Auto Registered: The Virtual Agent has auto-registered the user for the Virtual Agent services.
User Status Checked: The Virtual Agent performed the user status checks to determine whether the user is active and allowed to interact with the Virtual Agent.
User Manually Registered: The Virtual Agent has registered the user via the manual registration process for the Virtual Agent Services.
User Registration Cancelled: The Virtual Agent has canceled the user registration via the manual registration process for the Luma bot services.
Fallback Detected: A Virtual Agent is attempting to parse the user message and recognized that the skill is not identified or skill with less confidence score identifier causing the fallback actions to be triggered.
Skill Conflict Detected: A Virtual Agent is attempting to disambiguate the user's intention so that it can capture & infer the human intent confidently and accurately.
Skill Recognized: The Virtual Agent recognized the user's intention based on the user given phrase/invocation name/skill identifier.
Skill Conflict Resolved: A Virtual Agent has successfully and confidently disambiguated the user's intention.
Rephrase Requested: The Virtual Agent asked the user to rephrase the request to identify and identify the correct/right skill with greater confidence.
Skill Permissions Checked: The Virtual Agent has verified the user's access permissions on the skill that was triggered by the user from channel.
Response sent: The Virtual Agent delivered a response to the user with an informational message, clarification prompt requesting input from user, or a hangup phrase.
Conversation Flow Started: The conversation flow of identified skill is started to collect information in a specific order from the user and guide the user through a series of configured steps as part of the skill.
Attribute Prompted: An Attribute value was requested in the conversation asking the user to provide a value for it.
Attribute Value Filled: An Attribute was successfully filled with a value that obtained either from the user or a third-party fulfillment system during the conversation.
Attribute Value Validated: A validation was performed on the Attribute’s value obtained from the user against the pre-configured validation rules.
Operation Executed: An operation was performed by the Virtual Agent to collect, modify, or post information in a pre-configured Application Instance.
Error Rules Evaluated: The operation error rule set comprising of a collection of ordered rules with a set of execution semantics was evaluated and executed the rule’s actions based on the results of the rule’s condition evaluation.
No Data Rules Evaluated: The operation no data rule with a set of execution semantics was evaluated and executed the rule’s actions based on the results of the rule’s condition evaluation.
Success Rules Evaluated: The operation success rule set comprising of a collection of ordered rules with a set of execution semantics was evaluated and executed the rule’s actions based on the results of the rule’s condition evaluation.
Rule Set Evaluated: A rule set comprising of a collection of ordered rules with a set of execution semantics was evaluated and executed the rule’s actions based on the results of the rule’s condition evaluation.
Conversation Branched: Conversational branched to a (sub)-skills based on the conversation routing rules.
Conversation Aborted: Conversation was interrupted.
Conversation Errored: Conversation was abruptly terminated due to an irrecoverable system error.
The context at each event of the conversation is also provided as part of the events data in the conversation log. It could be accessed by clicking on the icon to the event. Please see the below picture.
The context variables pop up will be shown as below. This includes the data for the global and local variables used inside the skill and operation.
The details about the execution results of the operation, ruleset could be viewed clicking on the events. Please see the following examples.
