Messages

In the messages tab, you can view all conversations in a table format. Every row in the table indicates an interaction, also known as a chat log. This view is useful for

• Filtering Chat Logs

• Exporting to CSV or UAT

• Viewing the Sources of the bot's responses

Filtering Chat Logs

Filter chat logs (on the top of the page) based on several fields. Each field has a specific filter use case.

Message Tab Filter Options

Field	Use Case
ID	Finds all chat logs from a specific ID
User ID	Finds all chat logs from a specific user ID
Story ID	Find all chat logs that triggered a specific story ID
Recorded At	Find all chat logs before or after a specific date and time
Image	Find all chat logs that were triggered with an image message
Button Click	Find all chat logs that were triggered by a button click or postback event
Message Text	Find all chat logs containing a specific keyword
User Name	Find all chat logs from a specific user, filterable by the user's name
Channel	Find all chat logs from a specific channel
Widget ID	Find all chat logs with a specific widget ID
Story Name	Find all chat logs that triggered a specific story, filterable by the 's name
Sentiment	Find all chat logs from message with a positive, neutral, or negative sentiment

Filtering and Exporting Chat Logs

Filter chat logs by clicking on add filter and selecting the appropriate attribute and function for the search parameter. To export filtered logs, click on Select Action and Export rows to CSV.

Example of using Message View

In the Example above, we are filtering all messages before 27th June 2024 containing message text 'gelato' then exporting those messages to a UAT.

Export filtered messages to CSV can be done in two ways by using Action > Export to CSV or Export to UAT. Export to CSV will generate a .csv file with the following fields:

Fields	Description
ID	An identifier based on the unique Message ID.
Date [Timezone]	This column reflects the timezone of the user exporting the data. For example, if the export is generated in New York, the column will show 'Date [GMT-05:00]'.
From	The source or type of user initiating the interaction such as user or bot.
From ID	A unique identifier for the sender such as user ID or bot ID.
From Name	Displays only the sender's display name.
Message	The content of the message exchanged during the interaction.
Reached Fallback	Indicates whether the interaction triggered a fallback response where "Y" stands for yes or "N" for no.
Story ID	A unique identifier for the story associated with the interaction.
Story Name	The name or title of the story.
Sentiment Value	A numerical score representing the sentiment of the user's message ranging from -1 for negative, 0 for neutral, and 1 for positive.
Sentiment	The sentiment category derived from the sentiment value such as positive, neutral, or negative.
Channel	The platform or medium through which the interaction occurred.
Widget ID	Only applies to Website Bot Widgets and cannot be used for Messaging Apps.
Chat History Link	A URL or link to access the complete chat history for the interaction.

Export to UAT will generate a .csv file with the following fields:

Field	Description
ID	An identifier based on the unique Message ID.
Description	The name or title of the story.
Message	The content of the message exchanged during the interaction.
Memory	This refers to the bot’s memory context and can be used to simulate data that the bot remembers from the user’s conversation.
User	This refers to the user context, enabling simulation of an existing user on BotDistrikt or a new user with attributes defined via variables.
Refresh	This acts as a caching identifier. When a response is retrieved, it is cached for 10 minutes. Changing this value forces the cache to clear and recomputes the response.
Expected Response	The anticipated output or behaviour from the bot for the given input and context.
Actual Response	The actual output or behaviour observed from the system during the test.
Actual = Expected? (Y/N):	A validation field to confirm if the actual response matches the expected response.
Developer Comments	Additional notes or observations from the developer regarding the test case or discrepancies found.

This is useful for generating extensive end-to-end automated regression tests for the chatbot using the BotDistrikt Auto-UAT Feature, available to Enterprise customers only. Please get in touch with [e-mail].

Viewing the Sources of the bot's responses

The viewing of sources would provides context-aware answers to user queries. To enable this feature, you must first connect an LLM integration, such as OpenAI or other supported models followed by training the LLM using relevant Sources.

Each generated response includes a confidence level, which indicates how certain the model is about the accuracy of the response based on the provided sources.

• Higher confidence: Indicates a high similarity score between the user’s message and retrieved sources. However, this does not mean the response is relevant.

• Lower confidence: Indicates a low similarity score between the user’s message and retrieved sources. However, this does not mean the response is irrelevant.

You can view generated response from the Messages tab by hovering over the generated response text and clicking view responses.

Viewing Generated Responses

You can also view the sources to determine how the information is being derived from. This will bring you to the source material.