Description

The frontend app currently renders intent confidence scores based on the API response from Ludwig NLU, even when the user hasn't explicitly defined any intents within their Hexabot instance for a particular input.

This behavior stems from Ludwig NLU's ability to classify user input and infer intents based on its training data, even when no specific intent values are provided by the user.

Example:

Consider the following API response from Ludwig NLU:

{"entities":[{"entity":"language","value":"fr","confidence":0.987273097038269},
{"entity":"intent","value":"greetings_goodevening","confidence":0.9968143105506897}]}

In this case, Ludwig NLU has identified the intent as greetings_goodevening with a high confidence score (0.9968), even though the user might not have configured any intent with this specific value within their Hexabot setup.

Problem:

The UI directly displays this confidence score for the inferred intent. This can be confusing for users because:

• Misleading Confidence: Users might see confidence scores for intents they haven't explicitly created or intended to use in their Hexabot instance.
• Lack of Context: It's unclear from the UI where these intents are originating from (Ludwig's general training vs. user-defined intents).

Reproduction steps

- Select Ludwig NLU engine as your default NLU engine.
- Go to the NLU view in your Hexabot instance.
- Verify that no intents are currently defined.
- Input some text in the Frond UI (e.g., "Hello").
- Check the Frond UI for displayed intent confidence scores.