Automation V3: Utterance Generator

An Utterance is the human side of an AI conversation.
This is usually a distinct input that AI must comprehend the intent of in order to give an appropriate response using the entities (or details) within the request to provide the correct context.

Example:

Utterance: “What’s the weather like today in San Francisco?”
Intent: Weather Inquiry
Entities: Location-San Francisco, Date-Today

UseCase:

This step is best used to generate prompts that a user might submit to your Agents.

Below are the key details and behaviors found in the Step Settings:

Step Settings

Inputs:

Intent (Text, Mandatory)
The intent must be clearly defined, appropriate, and actionable. Providing well-phrased intent text helps ensure accurate utterance generation.

Count (Integer, Mandatory)
Specify the approximate number of utterances to generate. This field defaults to 10 and is limited by the result output type.

Result:

Result Name
Choose a name for the results to be stored locally as a list.

Result Scope
Select a scope to avoid data collisions in various testing scenarios by limiting the access of the Results Name variable.

Generate CSV File
If the box is unchecked, generated utterances will be stored in the Result Name locally as a list with a limit of 200 utterances.

If the box is checked, generated utterances will be stored in a CSV file in the templates folder on the Files tab. Each row contains a unique utterance generated for the specified intent. This output format has a limit of generating 1000 utterances.

• File naming convention: Utterance_Data_<8DigitUniqueId>_<Timestamp>.csv
• The CSV always includes a header row with: Utterances
• The file name is stored in a test case variable, using the “Result Name” parameter provided by the user. This allows further automation within Provar.

The output format is designed to be easily consumed in Provar tests without additional transformation by calling the “Result Name” variable.

Troubleshooting

If the AI is unable to generate utterances, there may be a couple of reasons.

For example, due to an unclear or inappropriate intent—the output CSV will still be created with the same structure but with the content below:
Sorry, I can only generate utterances for clear, appropriate, and  actionable intent topics. Please rephrase your request.

You may also see error messages for too many or too few utterances as below:
Utterance count must be more than 0 and less than 200.

To improve success rates, users should rephrase vague or overly broad intents to be more specific.

---

Feedback

Was this article helpful for you?

We really appreciate it!

Thank you for your feedback.

Tell us how we can improve this article.

No Yes