Adding feedback in chat for query generation

When you ask a business question in Kyvos Dialogs, there may be cases where the Large Language Model (LLM) cannot generate an SQL query. This typically occurs when specific phrases or terms in your question are missing from the semantic model or lack sufficient context for generating queries.

To address this, Kyvos supports a feedback mechanism that allows you to provide additional information and define details for any missing phrases.

From Kyvos 2025.7 onwards, by default, all feedback is preserved at the individual user level rather than the global level. This ensures that feedback provided by one user is not shown or prompted again to other users.

How does it work?

• If Kyvos Dialogs cannot generate a query, you receive a notification highlighting the specific phrases or terms causing the issue.

• You can add details on specific phrases in two ways:

  • By selecting the required column, adding a description, or providing values for that column.

  • By directly adding the value of the column.

• Kyvos stores this information and uses it to enhance future query generation for similar questions.

• Additionally, you can modify the inputs you have provided for query generation. Once a query is successfully generated, you have the option to save it for future use.

Tips for adding inputs:

• To define a calculation like “average order value,” type:
  It is the ratio of @total sales by @order count.

• To refer to a semantic model column, type @ to view a list of available columns.

• To include a specific column value, begin typing the value inside single or double quotation marks to view available options.

Open